| 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 | |
| 26 | #include <linux/module.h> |
| 27 | |
| 28 | #include <asm/uaccess.h> |
| 29 | #include <net/sock.h> |
| 30 | |
| 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> |
| 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> |
| 44 | #include <linux/string.h> |
| 45 | #include <linux/scatterlist.h> |
| 46 | #include "drbd_int.h" |
| 47 | #include "drbd_protocol.h" |
| 48 | #include "drbd_req.h" |
| 49 | #include "drbd_vli.h" |
| 50 | |
| 51 | #define PRO_FEATURES (FF_TRIM) |
| 52 | |
| 53 | struct packet_info { |
| 54 | enum drbd_packet cmd; |
| 55 | unsigned int size; |
| 56 | unsigned int vnr; |
| 57 | void *data; |
| 58 | }; |
| 59 | |
| 60 | enum finish_epoch { |
| 61 | FE_STILL_LIVE, |
| 62 | FE_DESTROYED, |
| 63 | FE_RECYCLED, |
| 64 | }; |
| 65 | |
| 66 | static int drbd_do_features(struct drbd_connection *connection); |
| 67 | static int drbd_do_auth(struct drbd_connection *connection); |
| 68 | static int drbd_disconnected(struct drbd_peer_device *); |
| 69 | static void conn_wait_active_ee_empty(struct drbd_connection *connection); |
| 70 | static enum finish_epoch drbd_may_finish_epoch(struct drbd_connection *, struct drbd_epoch *, enum epoch_event); |
| 71 | static int e_end_block(struct drbd_work *, int); |
| 72 | |
| 73 | |
| 74 | #define GFP_TRY (__GFP_HIGHMEM | __GFP_NOWARN) |
| 75 | |
| 76 | /* |
| 77 | * some helper functions to deal with single linked page lists, |
| 78 | * page->private being our "next" pointer. |
| 79 | */ |
| 80 | |
| 81 | /* If at least n pages are linked at head, get n pages off. |
| 82 | * Otherwise, don't modify head, and return NULL. |
| 83 | * Locking is the responsibility of the caller. |
| 84 | */ |
| 85 | static struct page *page_chain_del(struct page **head, int n) |
| 86 | { |
| 87 | struct page *page; |
| 88 | struct page *tmp; |
| 89 | |
| 90 | BUG_ON(!n); |
| 91 | BUG_ON(!head); |
| 92 | |
| 93 | page = *head; |
| 94 | |
| 95 | if (!page) |
| 96 | return NULL; |
| 97 | |
| 98 | while (page) { |
| 99 | tmp = page_chain_next(page); |
| 100 | if (--n == 0) |
| 101 | break; /* found sufficient pages */ |
| 102 | if (tmp == NULL) |
| 103 | /* insufficient pages, don't use any of them. */ |
| 104 | return NULL; |
| 105 | page = tmp; |
| 106 | } |
| 107 | |
| 108 | /* add end of list marker for the returned list */ |
| 109 | set_page_private(page, 0); |
| 110 | /* actual return value, and adjustment of head */ |
| 111 | page = *head; |
| 112 | *head = tmp; |
| 113 | return page; |
| 114 | } |
| 115 | |
| 116 | /* may be used outside of locks to find the tail of a (usually short) |
| 117 | * "private" page chain, before adding it back to a global chain head |
| 118 | * with page_chain_add() under a spinlock. */ |
| 119 | static struct page *page_chain_tail(struct page *page, int *len) |
| 120 | { |
| 121 | struct page *tmp; |
| 122 | int i = 1; |
| 123 | while ((tmp = page_chain_next(page))) |
| 124 | ++i, page = tmp; |
| 125 | if (len) |
| 126 | *len = i; |
| 127 | return page; |
| 128 | } |
| 129 | |
| 130 | static int page_chain_free(struct page *page) |
| 131 | { |
| 132 | struct page *tmp; |
| 133 | int i = 0; |
| 134 | page_chain_for_each_safe(page, tmp) { |
| 135 | put_page(page); |
| 136 | ++i; |
| 137 | } |
| 138 | return i; |
| 139 | } |
| 140 | |
| 141 | static void page_chain_add(struct page **head, |
| 142 | struct page *chain_first, struct page *chain_last) |
| 143 | { |
| 144 | #if 1 |
| 145 | struct page *tmp; |
| 146 | tmp = page_chain_tail(chain_first, NULL); |
| 147 | BUG_ON(tmp != chain_last); |
| 148 | #endif |
| 149 | |
| 150 | /* add chain to head */ |
| 151 | set_page_private(chain_last, (unsigned long)*head); |
| 152 | *head = chain_first; |
| 153 | } |
| 154 | |
| 155 | static struct page *__drbd_alloc_pages(struct drbd_device *device, |
| 156 | unsigned int number) |
| 157 | { |
| 158 | struct page *page = NULL; |
| 159 | struct page *tmp = NULL; |
| 160 | unsigned int i = 0; |
| 161 | |
| 162 | /* Yes, testing drbd_pp_vacant outside the lock is racy. |
| 163 | * So what. It saves a spin_lock. */ |
| 164 | if (drbd_pp_vacant >= number) { |
| 165 | spin_lock(&drbd_pp_lock); |
| 166 | page = page_chain_del(&drbd_pp_pool, number); |
| 167 | if (page) |
| 168 | drbd_pp_vacant -= number; |
| 169 | spin_unlock(&drbd_pp_lock); |
| 170 | if (page) |
| 171 | return page; |
| 172 | } |
| 173 | |
| 174 | /* GFP_TRY, because we must not cause arbitrary write-out: in a DRBD |
| 175 | * "criss-cross" setup, that might cause write-out on some other DRBD, |
| 176 | * which in turn might block on the other node at this very place. */ |
| 177 | for (i = 0; i < number; i++) { |
| 178 | tmp = alloc_page(GFP_TRY); |
| 179 | if (!tmp) |
| 180 | break; |
| 181 | set_page_private(tmp, (unsigned long)page); |
| 182 | page = tmp; |
| 183 | } |
| 184 | |
| 185 | if (i == number) |
| 186 | return page; |
| 187 | |
| 188 | /* Not enough pages immediately available this time. |
| 189 | * No need to jump around here, drbd_alloc_pages will retry this |
| 190 | * function "soon". */ |
| 191 | if (page) { |
| 192 | tmp = page_chain_tail(page, NULL); |
| 193 | spin_lock(&drbd_pp_lock); |
| 194 | page_chain_add(&drbd_pp_pool, page, tmp); |
| 195 | drbd_pp_vacant += i; |
| 196 | spin_unlock(&drbd_pp_lock); |
| 197 | } |
| 198 | return NULL; |
| 199 | } |
| 200 | |
| 201 | static void reclaim_finished_net_peer_reqs(struct drbd_device *device, |
| 202 | struct list_head *to_be_freed) |
| 203 | { |
| 204 | struct drbd_peer_request *peer_req, *tmp; |
| 205 | |
| 206 | /* The EEs are always appended to the end of the list. Since |
| 207 | they are sent in order over the wire, they have to finish |
| 208 | in order. As soon as we see the first not finished we can |
| 209 | stop to examine the list... */ |
| 210 | |
| 211 | list_for_each_entry_safe(peer_req, tmp, &device->net_ee, w.list) { |
| 212 | if (drbd_peer_req_has_active_page(peer_req)) |
| 213 | break; |
| 214 | list_move(&peer_req->w.list, to_be_freed); |
| 215 | } |
| 216 | } |
| 217 | |
| 218 | static void drbd_reclaim_net_peer_reqs(struct drbd_device *device) |
| 219 | { |
| 220 | LIST_HEAD(reclaimed); |
| 221 | struct drbd_peer_request *peer_req, *t; |
| 222 | |
| 223 | spin_lock_irq(&device->resource->req_lock); |
| 224 | reclaim_finished_net_peer_reqs(device, &reclaimed); |
| 225 | spin_unlock_irq(&device->resource->req_lock); |
| 226 | list_for_each_entry_safe(peer_req, t, &reclaimed, w.list) |
| 227 | drbd_free_net_peer_req(device, peer_req); |
| 228 | } |
| 229 | |
| 230 | static void conn_reclaim_net_peer_reqs(struct drbd_connection *connection) |
| 231 | { |
| 232 | struct drbd_peer_device *peer_device; |
| 233 | int vnr; |
| 234 | |
| 235 | rcu_read_lock(); |
| 236 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
| 237 | struct drbd_device *device = peer_device->device; |
| 238 | if (!atomic_read(&device->pp_in_use_by_net)) |
| 239 | continue; |
| 240 | |
| 241 | kref_get(&device->kref); |
| 242 | rcu_read_unlock(); |
| 243 | drbd_reclaim_net_peer_reqs(device); |
| 244 | kref_put(&device->kref, drbd_destroy_device); |
| 245 | rcu_read_lock(); |
| 246 | } |
| 247 | rcu_read_unlock(); |
| 248 | } |
| 249 | |
| 250 | /** |
| 251 | * drbd_alloc_pages() - Returns @number pages, retries forever (or until signalled) |
| 252 | * @device: DRBD device. |
| 253 | * @number: number of pages requested |
| 254 | * @retry: whether to retry, if not enough pages are available right now |
| 255 | * |
| 256 | * Tries to allocate number pages, first from our own page pool, then from |
| 257 | * the kernel. |
| 258 | * Possibly retry until DRBD frees sufficient pages somewhere else. |
| 259 | * |
| 260 | * If this allocation would exceed the max_buffers setting, we throttle |
| 261 | * allocation (schedule_timeout) to give the system some room to breathe. |
| 262 | * |
| 263 | * We do not use max-buffers as hard limit, because it could lead to |
| 264 | * congestion and further to a distributed deadlock during online-verify or |
| 265 | * (checksum based) resync, if the max-buffers, socket buffer sizes and |
| 266 | * resync-rate settings are mis-configured. |
| 267 | * |
| 268 | * Returns a page chain linked via page->private. |
| 269 | */ |
| 270 | struct page *drbd_alloc_pages(struct drbd_peer_device *peer_device, unsigned int number, |
| 271 | bool retry) |
| 272 | { |
| 273 | struct drbd_device *device = peer_device->device; |
| 274 | struct page *page = NULL; |
| 275 | struct net_conf *nc; |
| 276 | DEFINE_WAIT(wait); |
| 277 | unsigned int mxb; |
| 278 | |
| 279 | rcu_read_lock(); |
| 280 | nc = rcu_dereference(peer_device->connection->net_conf); |
| 281 | mxb = nc ? nc->max_buffers : 1000000; |
| 282 | rcu_read_unlock(); |
| 283 | |
| 284 | if (atomic_read(&device->pp_in_use) < mxb) |
| 285 | page = __drbd_alloc_pages(device, number); |
| 286 | |
| 287 | /* Try to keep the fast path fast, but occasionally we need |
| 288 | * to reclaim the pages we lended to the network stack. */ |
| 289 | if (page && atomic_read(&device->pp_in_use_by_net) > 512) |
| 290 | drbd_reclaim_net_peer_reqs(device); |
| 291 | |
| 292 | while (page == NULL) { |
| 293 | prepare_to_wait(&drbd_pp_wait, &wait, TASK_INTERRUPTIBLE); |
| 294 | |
| 295 | drbd_reclaim_net_peer_reqs(device); |
| 296 | |
| 297 | if (atomic_read(&device->pp_in_use) < mxb) { |
| 298 | page = __drbd_alloc_pages(device, number); |
| 299 | if (page) |
| 300 | break; |
| 301 | } |
| 302 | |
| 303 | if (!retry) |
| 304 | break; |
| 305 | |
| 306 | if (signal_pending(current)) { |
| 307 | drbd_warn(device, "drbd_alloc_pages interrupted!\n"); |
| 308 | break; |
| 309 | } |
| 310 | |
| 311 | if (schedule_timeout(HZ/10) == 0) |
| 312 | mxb = UINT_MAX; |
| 313 | } |
| 314 | finish_wait(&drbd_pp_wait, &wait); |
| 315 | |
| 316 | if (page) |
| 317 | atomic_add(number, &device->pp_in_use); |
| 318 | return page; |
| 319 | } |
| 320 | |
| 321 | /* Must not be used from irq, as that may deadlock: see drbd_alloc_pages. |
| 322 | * Is also used from inside an other spin_lock_irq(&resource->req_lock); |
| 323 | * Either links the page chain back to the global pool, |
| 324 | * or returns all pages to the system. */ |
| 325 | static void drbd_free_pages(struct drbd_device *device, struct page *page, int is_net) |
| 326 | { |
| 327 | atomic_t *a = is_net ? &device->pp_in_use_by_net : &device->pp_in_use; |
| 328 | int i; |
| 329 | |
| 330 | if (page == NULL) |
| 331 | return; |
| 332 | |
| 333 | if (drbd_pp_vacant > (DRBD_MAX_BIO_SIZE/PAGE_SIZE) * minor_count) |
| 334 | i = page_chain_free(page); |
| 335 | else { |
| 336 | struct page *tmp; |
| 337 | tmp = page_chain_tail(page, &i); |
| 338 | spin_lock(&drbd_pp_lock); |
| 339 | page_chain_add(&drbd_pp_pool, page, tmp); |
| 340 | drbd_pp_vacant += i; |
| 341 | spin_unlock(&drbd_pp_lock); |
| 342 | } |
| 343 | i = atomic_sub_return(i, a); |
| 344 | if (i < 0) |
| 345 | drbd_warn(device, "ASSERTION FAILED: %s: %d < 0\n", |
| 346 | is_net ? "pp_in_use_by_net" : "pp_in_use", i); |
| 347 | wake_up(&drbd_pp_wait); |
| 348 | } |
| 349 | |
| 350 | /* |
| 351 | You need to hold the req_lock: |
| 352 | _drbd_wait_ee_list_empty() |
| 353 | |
| 354 | You must not have the req_lock: |
| 355 | drbd_free_peer_req() |
| 356 | drbd_alloc_peer_req() |
| 357 | drbd_free_peer_reqs() |
| 358 | drbd_ee_fix_bhs() |
| 359 | drbd_finish_peer_reqs() |
| 360 | drbd_clear_done_ee() |
| 361 | drbd_wait_ee_list_empty() |
| 362 | */ |
| 363 | |
| 364 | struct drbd_peer_request * |
| 365 | drbd_alloc_peer_req(struct drbd_peer_device *peer_device, u64 id, sector_t sector, |
| 366 | unsigned int data_size, bool has_payload, gfp_t gfp_mask) __must_hold(local) |
| 367 | { |
| 368 | struct drbd_device *device = peer_device->device; |
| 369 | struct drbd_peer_request *peer_req; |
| 370 | struct page *page = NULL; |
| 371 | unsigned nr_pages = (data_size + PAGE_SIZE -1) >> PAGE_SHIFT; |
| 372 | |
| 373 | if (drbd_insert_fault(device, DRBD_FAULT_AL_EE)) |
| 374 | return NULL; |
| 375 | |
| 376 | peer_req = mempool_alloc(drbd_ee_mempool, gfp_mask & ~__GFP_HIGHMEM); |
| 377 | if (!peer_req) { |
| 378 | if (!(gfp_mask & __GFP_NOWARN)) |
| 379 | drbd_err(device, "%s: allocation failed\n", __func__); |
| 380 | return NULL; |
| 381 | } |
| 382 | |
| 383 | if (has_payload && data_size) { |
| 384 | page = drbd_alloc_pages(peer_device, nr_pages, |
| 385 | gfpflags_allow_blocking(gfp_mask)); |
| 386 | if (!page) |
| 387 | goto fail; |
| 388 | } |
| 389 | |
| 390 | memset(peer_req, 0, sizeof(*peer_req)); |
| 391 | INIT_LIST_HEAD(&peer_req->w.list); |
| 392 | drbd_clear_interval(&peer_req->i); |
| 393 | peer_req->i.size = data_size; |
| 394 | peer_req->i.sector = sector; |
| 395 | peer_req->submit_jif = jiffies; |
| 396 | peer_req->peer_device = peer_device; |
| 397 | peer_req->pages = page; |
| 398 | /* |
| 399 | * The block_id is opaque to the receiver. It is not endianness |
| 400 | * converted, and sent back to the sender unchanged. |
| 401 | */ |
| 402 | peer_req->block_id = id; |
| 403 | |
| 404 | return peer_req; |
| 405 | |
| 406 | fail: |
| 407 | mempool_free(peer_req, drbd_ee_mempool); |
| 408 | return NULL; |
| 409 | } |
| 410 | |
| 411 | void __drbd_free_peer_req(struct drbd_device *device, struct drbd_peer_request *peer_req, |
| 412 | int is_net) |
| 413 | { |
| 414 | might_sleep(); |
| 415 | if (peer_req->flags & EE_HAS_DIGEST) |
| 416 | kfree(peer_req->digest); |
| 417 | drbd_free_pages(device, peer_req->pages, is_net); |
| 418 | D_ASSERT(device, atomic_read(&peer_req->pending_bios) == 0); |
| 419 | D_ASSERT(device, drbd_interval_empty(&peer_req->i)); |
| 420 | if (!expect(!(peer_req->flags & EE_CALL_AL_COMPLETE_IO))) { |
| 421 | peer_req->flags &= ~EE_CALL_AL_COMPLETE_IO; |
| 422 | drbd_al_complete_io(device, &peer_req->i); |
| 423 | } |
| 424 | mempool_free(peer_req, drbd_ee_mempool); |
| 425 | } |
| 426 | |
| 427 | int drbd_free_peer_reqs(struct drbd_device *device, struct list_head *list) |
| 428 | { |
| 429 | LIST_HEAD(work_list); |
| 430 | struct drbd_peer_request *peer_req, *t; |
| 431 | int count = 0; |
| 432 | int is_net = list == &device->net_ee; |
| 433 | |
| 434 | spin_lock_irq(&device->resource->req_lock); |
| 435 | list_splice_init(list, &work_list); |
| 436 | spin_unlock_irq(&device->resource->req_lock); |
| 437 | |
| 438 | list_for_each_entry_safe(peer_req, t, &work_list, w.list) { |
| 439 | __drbd_free_peer_req(device, peer_req, is_net); |
| 440 | count++; |
| 441 | } |
| 442 | return count; |
| 443 | } |
| 444 | |
| 445 | /* |
| 446 | * See also comments in _req_mod(,BARRIER_ACKED) and receive_Barrier. |
| 447 | */ |
| 448 | static int drbd_finish_peer_reqs(struct drbd_device *device) |
| 449 | { |
| 450 | LIST_HEAD(work_list); |
| 451 | LIST_HEAD(reclaimed); |
| 452 | struct drbd_peer_request *peer_req, *t; |
| 453 | int err = 0; |
| 454 | |
| 455 | spin_lock_irq(&device->resource->req_lock); |
| 456 | reclaim_finished_net_peer_reqs(device, &reclaimed); |
| 457 | list_splice_init(&device->done_ee, &work_list); |
| 458 | spin_unlock_irq(&device->resource->req_lock); |
| 459 | |
| 460 | list_for_each_entry_safe(peer_req, t, &reclaimed, w.list) |
| 461 | drbd_free_net_peer_req(device, peer_req); |
| 462 | |
| 463 | /* possible callbacks here: |
| 464 | * e_end_block, and e_end_resync_block, e_send_superseded. |
| 465 | * all ignore the last argument. |
| 466 | */ |
| 467 | list_for_each_entry_safe(peer_req, t, &work_list, w.list) { |
| 468 | int err2; |
| 469 | |
| 470 | /* list_del not necessary, next/prev members not touched */ |
| 471 | err2 = peer_req->w.cb(&peer_req->w, !!err); |
| 472 | if (!err) |
| 473 | err = err2; |
| 474 | drbd_free_peer_req(device, peer_req); |
| 475 | } |
| 476 | wake_up(&device->ee_wait); |
| 477 | |
| 478 | return err; |
| 479 | } |
| 480 | |
| 481 | static void _drbd_wait_ee_list_empty(struct drbd_device *device, |
| 482 | struct list_head *head) |
| 483 | { |
| 484 | DEFINE_WAIT(wait); |
| 485 | |
| 486 | /* avoids spin_lock/unlock |
| 487 | * and calling prepare_to_wait in the fast path */ |
| 488 | while (!list_empty(head)) { |
| 489 | prepare_to_wait(&device->ee_wait, &wait, TASK_UNINTERRUPTIBLE); |
| 490 | spin_unlock_irq(&device->resource->req_lock); |
| 491 | io_schedule(); |
| 492 | finish_wait(&device->ee_wait, &wait); |
| 493 | spin_lock_irq(&device->resource->req_lock); |
| 494 | } |
| 495 | } |
| 496 | |
| 497 | static void drbd_wait_ee_list_empty(struct drbd_device *device, |
| 498 | struct list_head *head) |
| 499 | { |
| 500 | spin_lock_irq(&device->resource->req_lock); |
| 501 | _drbd_wait_ee_list_empty(device, head); |
| 502 | spin_unlock_irq(&device->resource->req_lock); |
| 503 | } |
| 504 | |
| 505 | static int drbd_recv_short(struct socket *sock, void *buf, size_t size, int flags) |
| 506 | { |
| 507 | struct kvec iov = { |
| 508 | .iov_base = buf, |
| 509 | .iov_len = size, |
| 510 | }; |
| 511 | struct msghdr msg = { |
| 512 | .msg_flags = (flags ? flags : MSG_WAITALL | MSG_NOSIGNAL) |
| 513 | }; |
| 514 | return kernel_recvmsg(sock, &msg, &iov, 1, size, msg.msg_flags); |
| 515 | } |
| 516 | |
| 517 | static int drbd_recv(struct drbd_connection *connection, void *buf, size_t size) |
| 518 | { |
| 519 | int rv; |
| 520 | |
| 521 | rv = drbd_recv_short(connection->data.socket, buf, size, 0); |
| 522 | |
| 523 | if (rv < 0) { |
| 524 | if (rv == -ECONNRESET) |
| 525 | drbd_info(connection, "sock was reset by peer\n"); |
| 526 | else if (rv != -ERESTARTSYS) |
| 527 | drbd_err(connection, "sock_recvmsg returned %d\n", rv); |
| 528 | } else if (rv == 0) { |
| 529 | if (test_bit(DISCONNECT_SENT, &connection->flags)) { |
| 530 | long t; |
| 531 | rcu_read_lock(); |
| 532 | t = rcu_dereference(connection->net_conf)->ping_timeo * HZ/10; |
| 533 | rcu_read_unlock(); |
| 534 | |
| 535 | t = wait_event_timeout(connection->ping_wait, connection->cstate < C_WF_REPORT_PARAMS, t); |
| 536 | |
| 537 | if (t) |
| 538 | goto out; |
| 539 | } |
| 540 | drbd_info(connection, "sock was shut down by peer\n"); |
| 541 | } |
| 542 | |
| 543 | if (rv != size) |
| 544 | conn_request_state(connection, NS(conn, C_BROKEN_PIPE), CS_HARD); |
| 545 | |
| 546 | out: |
| 547 | return rv; |
| 548 | } |
| 549 | |
| 550 | static int drbd_recv_all(struct drbd_connection *connection, void *buf, size_t size) |
| 551 | { |
| 552 | int err; |
| 553 | |
| 554 | err = drbd_recv(connection, buf, size); |
| 555 | if (err != size) { |
| 556 | if (err >= 0) |
| 557 | err = -EIO; |
| 558 | } else |
| 559 | err = 0; |
| 560 | return err; |
| 561 | } |
| 562 | |
| 563 | static int drbd_recv_all_warn(struct drbd_connection *connection, void *buf, size_t size) |
| 564 | { |
| 565 | int err; |
| 566 | |
| 567 | err = drbd_recv_all(connection, buf, size); |
| 568 | if (err && !signal_pending(current)) |
| 569 | drbd_warn(connection, "short read (expected size %d)\n", (int)size); |
| 570 | return err; |
| 571 | } |
| 572 | |
| 573 | /* quoting tcp(7): |
| 574 | * On individual connections, the socket buffer size must be set prior to the |
| 575 | * listen(2) or connect(2) calls in order to have it take effect. |
| 576 | * This is our wrapper to do so. |
| 577 | */ |
| 578 | static void drbd_setbufsize(struct socket *sock, unsigned int snd, |
| 579 | unsigned int rcv) |
| 580 | { |
| 581 | /* open coded SO_SNDBUF, SO_RCVBUF */ |
| 582 | if (snd) { |
| 583 | sock->sk->sk_sndbuf = snd; |
| 584 | sock->sk->sk_userlocks |= SOCK_SNDBUF_LOCK; |
| 585 | } |
| 586 | if (rcv) { |
| 587 | sock->sk->sk_rcvbuf = rcv; |
| 588 | sock->sk->sk_userlocks |= SOCK_RCVBUF_LOCK; |
| 589 | } |
| 590 | } |
| 591 | |
| 592 | static struct socket *drbd_try_connect(struct drbd_connection *connection) |
| 593 | { |
| 594 | const char *what; |
| 595 | struct socket *sock; |
| 596 | struct sockaddr_in6 src_in6; |
| 597 | struct sockaddr_in6 peer_in6; |
| 598 | struct net_conf *nc; |
| 599 | int err, peer_addr_len, my_addr_len; |
| 600 | int sndbuf_size, rcvbuf_size, connect_int; |
| 601 | int disconnect_on_error = 1; |
| 602 | |
| 603 | rcu_read_lock(); |
| 604 | nc = rcu_dereference(connection->net_conf); |
| 605 | if (!nc) { |
| 606 | rcu_read_unlock(); |
| 607 | return NULL; |
| 608 | } |
| 609 | sndbuf_size = nc->sndbuf_size; |
| 610 | rcvbuf_size = nc->rcvbuf_size; |
| 611 | connect_int = nc->connect_int; |
| 612 | rcu_read_unlock(); |
| 613 | |
| 614 | my_addr_len = min_t(int, connection->my_addr_len, sizeof(src_in6)); |
| 615 | memcpy(&src_in6, &connection->my_addr, my_addr_len); |
| 616 | |
| 617 | if (((struct sockaddr *)&connection->my_addr)->sa_family == AF_INET6) |
| 618 | src_in6.sin6_port = 0; |
| 619 | else |
| 620 | ((struct sockaddr_in *)&src_in6)->sin_port = 0; /* AF_INET & AF_SCI */ |
| 621 | |
| 622 | peer_addr_len = min_t(int, connection->peer_addr_len, sizeof(src_in6)); |
| 623 | memcpy(&peer_in6, &connection->peer_addr, peer_addr_len); |
| 624 | |
| 625 | what = "sock_create_kern"; |
| 626 | err = sock_create_kern(&init_net, ((struct sockaddr *)&src_in6)->sa_family, |
| 627 | SOCK_STREAM, IPPROTO_TCP, &sock); |
| 628 | if (err < 0) { |
| 629 | sock = NULL; |
| 630 | goto out; |
| 631 | } |
| 632 | |
| 633 | sock->sk->sk_rcvtimeo = |
| 634 | sock->sk->sk_sndtimeo = connect_int * HZ; |
| 635 | drbd_setbufsize(sock, sndbuf_size, rcvbuf_size); |
| 636 | |
| 637 | /* explicitly bind to the configured IP as source IP |
| 638 | * for the outgoing connections. |
| 639 | * This is needed for multihomed hosts and to be |
| 640 | * able to use lo: interfaces for drbd. |
| 641 | * Make sure to use 0 as port number, so linux selects |
| 642 | * a free one dynamically. |
| 643 | */ |
| 644 | what = "bind before connect"; |
| 645 | err = sock->ops->bind(sock, (struct sockaddr *) &src_in6, my_addr_len); |
| 646 | if (err < 0) |
| 647 | goto out; |
| 648 | |
| 649 | /* connect may fail, peer not yet available. |
| 650 | * stay C_WF_CONNECTION, don't go Disconnecting! */ |
| 651 | disconnect_on_error = 0; |
| 652 | what = "connect"; |
| 653 | err = sock->ops->connect(sock, (struct sockaddr *) &peer_in6, peer_addr_len, 0); |
| 654 | |
| 655 | out: |
| 656 | if (err < 0) { |
| 657 | if (sock) { |
| 658 | sock_release(sock); |
| 659 | sock = NULL; |
| 660 | } |
| 661 | switch (-err) { |
| 662 | /* timeout, busy, signal pending */ |
| 663 | case ETIMEDOUT: case EAGAIN: case EINPROGRESS: |
| 664 | case EINTR: case ERESTARTSYS: |
| 665 | /* peer not (yet) available, network problem */ |
| 666 | case ECONNREFUSED: case ENETUNREACH: |
| 667 | case EHOSTDOWN: case EHOSTUNREACH: |
| 668 | disconnect_on_error = 0; |
| 669 | break; |
| 670 | default: |
| 671 | drbd_err(connection, "%s failed, err = %d\n", what, err); |
| 672 | } |
| 673 | if (disconnect_on_error) |
| 674 | conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_HARD); |
| 675 | } |
| 676 | |
| 677 | return sock; |
| 678 | } |
| 679 | |
| 680 | struct accept_wait_data { |
| 681 | struct drbd_connection *connection; |
| 682 | struct socket *s_listen; |
| 683 | struct completion door_bell; |
| 684 | void (*original_sk_state_change)(struct sock *sk); |
| 685 | |
| 686 | }; |
| 687 | |
| 688 | static void drbd_incoming_connection(struct sock *sk) |
| 689 | { |
| 690 | struct accept_wait_data *ad = sk->sk_user_data; |
| 691 | void (*state_change)(struct sock *sk); |
| 692 | |
| 693 | state_change = ad->original_sk_state_change; |
| 694 | if (sk->sk_state == TCP_ESTABLISHED) |
| 695 | complete(&ad->door_bell); |
| 696 | state_change(sk); |
| 697 | } |
| 698 | |
| 699 | static int prepare_listen_socket(struct drbd_connection *connection, struct accept_wait_data *ad) |
| 700 | { |
| 701 | int err, sndbuf_size, rcvbuf_size, my_addr_len; |
| 702 | struct sockaddr_in6 my_addr; |
| 703 | struct socket *s_listen; |
| 704 | struct net_conf *nc; |
| 705 | const char *what; |
| 706 | |
| 707 | rcu_read_lock(); |
| 708 | nc = rcu_dereference(connection->net_conf); |
| 709 | if (!nc) { |
| 710 | rcu_read_unlock(); |
| 711 | return -EIO; |
| 712 | } |
| 713 | sndbuf_size = nc->sndbuf_size; |
| 714 | rcvbuf_size = nc->rcvbuf_size; |
| 715 | rcu_read_unlock(); |
| 716 | |
| 717 | my_addr_len = min_t(int, connection->my_addr_len, sizeof(struct sockaddr_in6)); |
| 718 | memcpy(&my_addr, &connection->my_addr, my_addr_len); |
| 719 | |
| 720 | what = "sock_create_kern"; |
| 721 | err = sock_create_kern(&init_net, ((struct sockaddr *)&my_addr)->sa_family, |
| 722 | SOCK_STREAM, IPPROTO_TCP, &s_listen); |
| 723 | if (err) { |
| 724 | s_listen = NULL; |
| 725 | goto out; |
| 726 | } |
| 727 | |
| 728 | s_listen->sk->sk_reuse = SK_CAN_REUSE; /* SO_REUSEADDR */ |
| 729 | drbd_setbufsize(s_listen, sndbuf_size, rcvbuf_size); |
| 730 | |
| 731 | what = "bind before listen"; |
| 732 | err = s_listen->ops->bind(s_listen, (struct sockaddr *)&my_addr, my_addr_len); |
| 733 | if (err < 0) |
| 734 | goto out; |
| 735 | |
| 736 | ad->s_listen = s_listen; |
| 737 | write_lock_bh(&s_listen->sk->sk_callback_lock); |
| 738 | ad->original_sk_state_change = s_listen->sk->sk_state_change; |
| 739 | s_listen->sk->sk_state_change = drbd_incoming_connection; |
| 740 | s_listen->sk->sk_user_data = ad; |
| 741 | write_unlock_bh(&s_listen->sk->sk_callback_lock); |
| 742 | |
| 743 | what = "listen"; |
| 744 | err = s_listen->ops->listen(s_listen, 5); |
| 745 | if (err < 0) |
| 746 | goto out; |
| 747 | |
| 748 | return 0; |
| 749 | out: |
| 750 | if (s_listen) |
| 751 | sock_release(s_listen); |
| 752 | if (err < 0) { |
| 753 | if (err != -EAGAIN && err != -EINTR && err != -ERESTARTSYS) { |
| 754 | drbd_err(connection, "%s failed, err = %d\n", what, err); |
| 755 | conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_HARD); |
| 756 | } |
| 757 | } |
| 758 | |
| 759 | return -EIO; |
| 760 | } |
| 761 | |
| 762 | static void unregister_state_change(struct sock *sk, struct accept_wait_data *ad) |
| 763 | { |
| 764 | write_lock_bh(&sk->sk_callback_lock); |
| 765 | sk->sk_state_change = ad->original_sk_state_change; |
| 766 | sk->sk_user_data = NULL; |
| 767 | write_unlock_bh(&sk->sk_callback_lock); |
| 768 | } |
| 769 | |
| 770 | static struct socket *drbd_wait_for_connect(struct drbd_connection *connection, struct accept_wait_data *ad) |
| 771 | { |
| 772 | int timeo, connect_int, err = 0; |
| 773 | struct socket *s_estab = NULL; |
| 774 | struct net_conf *nc; |
| 775 | |
| 776 | rcu_read_lock(); |
| 777 | nc = rcu_dereference(connection->net_conf); |
| 778 | if (!nc) { |
| 779 | rcu_read_unlock(); |
| 780 | return NULL; |
| 781 | } |
| 782 | connect_int = nc->connect_int; |
| 783 | rcu_read_unlock(); |
| 784 | |
| 785 | timeo = connect_int * HZ; |
| 786 | /* 28.5% random jitter */ |
| 787 | timeo += (prandom_u32() & 1) ? timeo / 7 : -timeo / 7; |
| 788 | |
| 789 | err = wait_for_completion_interruptible_timeout(&ad->door_bell, timeo); |
| 790 | if (err <= 0) |
| 791 | return NULL; |
| 792 | |
| 793 | err = kernel_accept(ad->s_listen, &s_estab, 0); |
| 794 | if (err < 0) { |
| 795 | if (err != -EAGAIN && err != -EINTR && err != -ERESTARTSYS) { |
| 796 | drbd_err(connection, "accept failed, err = %d\n", err); |
| 797 | conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_HARD); |
| 798 | } |
| 799 | } |
| 800 | |
| 801 | if (s_estab) |
| 802 | unregister_state_change(s_estab->sk, ad); |
| 803 | |
| 804 | return s_estab; |
| 805 | } |
| 806 | |
| 807 | static int decode_header(struct drbd_connection *, void *, struct packet_info *); |
| 808 | |
| 809 | static int send_first_packet(struct drbd_connection *connection, struct drbd_socket *sock, |
| 810 | enum drbd_packet cmd) |
| 811 | { |
| 812 | if (!conn_prepare_command(connection, sock)) |
| 813 | return -EIO; |
| 814 | return conn_send_command(connection, sock, cmd, 0, NULL, 0); |
| 815 | } |
| 816 | |
| 817 | static int receive_first_packet(struct drbd_connection *connection, struct socket *sock) |
| 818 | { |
| 819 | unsigned int header_size = drbd_header_size(connection); |
| 820 | struct packet_info pi; |
| 821 | struct net_conf *nc; |
| 822 | int err; |
| 823 | |
| 824 | rcu_read_lock(); |
| 825 | nc = rcu_dereference(connection->net_conf); |
| 826 | if (!nc) { |
| 827 | rcu_read_unlock(); |
| 828 | return -EIO; |
| 829 | } |
| 830 | sock->sk->sk_rcvtimeo = nc->ping_timeo * 4 * HZ / 10; |
| 831 | rcu_read_unlock(); |
| 832 | |
| 833 | err = drbd_recv_short(sock, connection->data.rbuf, header_size, 0); |
| 834 | if (err != header_size) { |
| 835 | if (err >= 0) |
| 836 | err = -EIO; |
| 837 | return err; |
| 838 | } |
| 839 | err = decode_header(connection, connection->data.rbuf, &pi); |
| 840 | if (err) |
| 841 | return err; |
| 842 | return pi.cmd; |
| 843 | } |
| 844 | |
| 845 | /** |
| 846 | * drbd_socket_okay() - Free the socket if its connection is not okay |
| 847 | * @sock: pointer to the pointer to the socket. |
| 848 | */ |
| 849 | static bool drbd_socket_okay(struct socket **sock) |
| 850 | { |
| 851 | int rr; |
| 852 | char tb[4]; |
| 853 | |
| 854 | if (!*sock) |
| 855 | return false; |
| 856 | |
| 857 | rr = drbd_recv_short(*sock, tb, 4, MSG_DONTWAIT | MSG_PEEK); |
| 858 | |
| 859 | if (rr > 0 || rr == -EAGAIN) { |
| 860 | return true; |
| 861 | } else { |
| 862 | sock_release(*sock); |
| 863 | *sock = NULL; |
| 864 | return false; |
| 865 | } |
| 866 | } |
| 867 | |
| 868 | static bool connection_established(struct drbd_connection *connection, |
| 869 | struct socket **sock1, |
| 870 | struct socket **sock2) |
| 871 | { |
| 872 | struct net_conf *nc; |
| 873 | int timeout; |
| 874 | bool ok; |
| 875 | |
| 876 | if (!*sock1 || !*sock2) |
| 877 | return false; |
| 878 | |
| 879 | rcu_read_lock(); |
| 880 | nc = rcu_dereference(connection->net_conf); |
| 881 | timeout = (nc->sock_check_timeo ?: nc->ping_timeo) * HZ / 10; |
| 882 | rcu_read_unlock(); |
| 883 | schedule_timeout_interruptible(timeout); |
| 884 | |
| 885 | ok = drbd_socket_okay(sock1); |
| 886 | ok = drbd_socket_okay(sock2) && ok; |
| 887 | |
| 888 | return ok; |
| 889 | } |
| 890 | |
| 891 | /* Gets called if a connection is established, or if a new minor gets created |
| 892 | in a connection */ |
| 893 | int drbd_connected(struct drbd_peer_device *peer_device) |
| 894 | { |
| 895 | struct drbd_device *device = peer_device->device; |
| 896 | int err; |
| 897 | |
| 898 | atomic_set(&device->packet_seq, 0); |
| 899 | device->peer_seq = 0; |
| 900 | |
| 901 | device->state_mutex = peer_device->connection->agreed_pro_version < 100 ? |
| 902 | &peer_device->connection->cstate_mutex : |
| 903 | &device->own_state_mutex; |
| 904 | |
| 905 | err = drbd_send_sync_param(peer_device); |
| 906 | if (!err) |
| 907 | err = drbd_send_sizes(peer_device, 0, 0); |
| 908 | if (!err) |
| 909 | err = drbd_send_uuids(peer_device); |
| 910 | if (!err) |
| 911 | err = drbd_send_current_state(peer_device); |
| 912 | clear_bit(USE_DEGR_WFC_T, &device->flags); |
| 913 | clear_bit(RESIZE_PENDING, &device->flags); |
| 914 | atomic_set(&device->ap_in_flight, 0); |
| 915 | mod_timer(&device->request_timer, jiffies + HZ); /* just start it here. */ |
| 916 | return err; |
| 917 | } |
| 918 | |
| 919 | /* |
| 920 | * return values: |
| 921 | * 1 yes, we have a valid connection |
| 922 | * 0 oops, did not work out, please try again |
| 923 | * -1 peer talks different language, |
| 924 | * no point in trying again, please go standalone. |
| 925 | * -2 We do not have a network config... |
| 926 | */ |
| 927 | static int conn_connect(struct drbd_connection *connection) |
| 928 | { |
| 929 | struct drbd_socket sock, msock; |
| 930 | struct drbd_peer_device *peer_device; |
| 931 | struct net_conf *nc; |
| 932 | int vnr, timeout, h; |
| 933 | bool discard_my_data, ok; |
| 934 | enum drbd_state_rv rv; |
| 935 | struct accept_wait_data ad = { |
| 936 | .connection = connection, |
| 937 | .door_bell = COMPLETION_INITIALIZER_ONSTACK(ad.door_bell), |
| 938 | }; |
| 939 | |
| 940 | clear_bit(DISCONNECT_SENT, &connection->flags); |
| 941 | if (conn_request_state(connection, NS(conn, C_WF_CONNECTION), CS_VERBOSE) < SS_SUCCESS) |
| 942 | return -2; |
| 943 | |
| 944 | mutex_init(&sock.mutex); |
| 945 | sock.sbuf = connection->data.sbuf; |
| 946 | sock.rbuf = connection->data.rbuf; |
| 947 | sock.socket = NULL; |
| 948 | mutex_init(&msock.mutex); |
| 949 | msock.sbuf = connection->meta.sbuf; |
| 950 | msock.rbuf = connection->meta.rbuf; |
| 951 | msock.socket = NULL; |
| 952 | |
| 953 | /* Assume that the peer only understands protocol 80 until we know better. */ |
| 954 | connection->agreed_pro_version = 80; |
| 955 | |
| 956 | if (prepare_listen_socket(connection, &ad)) |
| 957 | return 0; |
| 958 | |
| 959 | do { |
| 960 | struct socket *s; |
| 961 | |
| 962 | s = drbd_try_connect(connection); |
| 963 | if (s) { |
| 964 | if (!sock.socket) { |
| 965 | sock.socket = s; |
| 966 | send_first_packet(connection, &sock, P_INITIAL_DATA); |
| 967 | } else if (!msock.socket) { |
| 968 | clear_bit(RESOLVE_CONFLICTS, &connection->flags); |
| 969 | msock.socket = s; |
| 970 | send_first_packet(connection, &msock, P_INITIAL_META); |
| 971 | } else { |
| 972 | drbd_err(connection, "Logic error in conn_connect()\n"); |
| 973 | goto out_release_sockets; |
| 974 | } |
| 975 | } |
| 976 | |
| 977 | if (connection_established(connection, &sock.socket, &msock.socket)) |
| 978 | break; |
| 979 | |
| 980 | retry: |
| 981 | s = drbd_wait_for_connect(connection, &ad); |
| 982 | if (s) { |
| 983 | int fp = receive_first_packet(connection, s); |
| 984 | drbd_socket_okay(&sock.socket); |
| 985 | drbd_socket_okay(&msock.socket); |
| 986 | switch (fp) { |
| 987 | case P_INITIAL_DATA: |
| 988 | if (sock.socket) { |
| 989 | drbd_warn(connection, "initial packet S crossed\n"); |
| 990 | sock_release(sock.socket); |
| 991 | sock.socket = s; |
| 992 | goto randomize; |
| 993 | } |
| 994 | sock.socket = s; |
| 995 | break; |
| 996 | case P_INITIAL_META: |
| 997 | set_bit(RESOLVE_CONFLICTS, &connection->flags); |
| 998 | if (msock.socket) { |
| 999 | drbd_warn(connection, "initial packet M crossed\n"); |
| 1000 | sock_release(msock.socket); |
| 1001 | msock.socket = s; |
| 1002 | goto randomize; |
| 1003 | } |
| 1004 | msock.socket = s; |
| 1005 | break; |
| 1006 | default: |
| 1007 | drbd_warn(connection, "Error receiving initial packet\n"); |
| 1008 | sock_release(s); |
| 1009 | randomize: |
| 1010 | if (prandom_u32() & 1) |
| 1011 | goto retry; |
| 1012 | } |
| 1013 | } |
| 1014 | |
| 1015 | if (connection->cstate <= C_DISCONNECTING) |
| 1016 | goto out_release_sockets; |
| 1017 | if (signal_pending(current)) { |
| 1018 | flush_signals(current); |
| 1019 | smp_rmb(); |
| 1020 | if (get_t_state(&connection->receiver) == EXITING) |
| 1021 | goto out_release_sockets; |
| 1022 | } |
| 1023 | |
| 1024 | ok = connection_established(connection, &sock.socket, &msock.socket); |
| 1025 | } while (!ok); |
| 1026 | |
| 1027 | if (ad.s_listen) |
| 1028 | sock_release(ad.s_listen); |
| 1029 | |
| 1030 | sock.socket->sk->sk_reuse = SK_CAN_REUSE; /* SO_REUSEADDR */ |
| 1031 | msock.socket->sk->sk_reuse = SK_CAN_REUSE; /* SO_REUSEADDR */ |
| 1032 | |
| 1033 | sock.socket->sk->sk_allocation = GFP_NOIO; |
| 1034 | msock.socket->sk->sk_allocation = GFP_NOIO; |
| 1035 | |
| 1036 | sock.socket->sk->sk_priority = TC_PRIO_INTERACTIVE_BULK; |
| 1037 | msock.socket->sk->sk_priority = TC_PRIO_INTERACTIVE; |
| 1038 | |
| 1039 | /* NOT YET ... |
| 1040 | * sock.socket->sk->sk_sndtimeo = connection->net_conf->timeout*HZ/10; |
| 1041 | * sock.socket->sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT; |
| 1042 | * first set it to the P_CONNECTION_FEATURES timeout, |
| 1043 | * which we set to 4x the configured ping_timeout. */ |
| 1044 | rcu_read_lock(); |
| 1045 | nc = rcu_dereference(connection->net_conf); |
| 1046 | |
| 1047 | sock.socket->sk->sk_sndtimeo = |
| 1048 | sock.socket->sk->sk_rcvtimeo = nc->ping_timeo*4*HZ/10; |
| 1049 | |
| 1050 | msock.socket->sk->sk_rcvtimeo = nc->ping_int*HZ; |
| 1051 | timeout = nc->timeout * HZ / 10; |
| 1052 | discard_my_data = nc->discard_my_data; |
| 1053 | rcu_read_unlock(); |
| 1054 | |
| 1055 | msock.socket->sk->sk_sndtimeo = timeout; |
| 1056 | |
| 1057 | /* we don't want delays. |
| 1058 | * we use TCP_CORK where appropriate, though */ |
| 1059 | drbd_tcp_nodelay(sock.socket); |
| 1060 | drbd_tcp_nodelay(msock.socket); |
| 1061 | |
| 1062 | connection->data.socket = sock.socket; |
| 1063 | connection->meta.socket = msock.socket; |
| 1064 | connection->last_received = jiffies; |
| 1065 | |
| 1066 | h = drbd_do_features(connection); |
| 1067 | if (h <= 0) |
| 1068 | return h; |
| 1069 | |
| 1070 | if (connection->cram_hmac_tfm) { |
| 1071 | /* drbd_request_state(device, NS(conn, WFAuth)); */ |
| 1072 | switch (drbd_do_auth(connection)) { |
| 1073 | case -1: |
| 1074 | drbd_err(connection, "Authentication of peer failed\n"); |
| 1075 | return -1; |
| 1076 | case 0: |
| 1077 | drbd_err(connection, "Authentication of peer failed, trying again.\n"); |
| 1078 | return 0; |
| 1079 | } |
| 1080 | } |
| 1081 | |
| 1082 | connection->data.socket->sk->sk_sndtimeo = timeout; |
| 1083 | connection->data.socket->sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT; |
| 1084 | |
| 1085 | if (drbd_send_protocol(connection) == -EOPNOTSUPP) |
| 1086 | return -1; |
| 1087 | |
| 1088 | /* Prevent a race between resync-handshake and |
| 1089 | * being promoted to Primary. |
| 1090 | * |
| 1091 | * Grab and release the state mutex, so we know that any current |
| 1092 | * drbd_set_role() is finished, and any incoming drbd_set_role |
| 1093 | * will see the STATE_SENT flag, and wait for it to be cleared. |
| 1094 | */ |
| 1095 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) |
| 1096 | mutex_lock(peer_device->device->state_mutex); |
| 1097 | |
| 1098 | set_bit(STATE_SENT, &connection->flags); |
| 1099 | |
| 1100 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) |
| 1101 | mutex_unlock(peer_device->device->state_mutex); |
| 1102 | |
| 1103 | rcu_read_lock(); |
| 1104 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
| 1105 | struct drbd_device *device = peer_device->device; |
| 1106 | kref_get(&device->kref); |
| 1107 | rcu_read_unlock(); |
| 1108 | |
| 1109 | if (discard_my_data) |
| 1110 | set_bit(DISCARD_MY_DATA, &device->flags); |
| 1111 | else |
| 1112 | clear_bit(DISCARD_MY_DATA, &device->flags); |
| 1113 | |
| 1114 | drbd_connected(peer_device); |
| 1115 | kref_put(&device->kref, drbd_destroy_device); |
| 1116 | rcu_read_lock(); |
| 1117 | } |
| 1118 | rcu_read_unlock(); |
| 1119 | |
| 1120 | rv = conn_request_state(connection, NS(conn, C_WF_REPORT_PARAMS), CS_VERBOSE); |
| 1121 | if (rv < SS_SUCCESS || connection->cstate != C_WF_REPORT_PARAMS) { |
| 1122 | clear_bit(STATE_SENT, &connection->flags); |
| 1123 | return 0; |
| 1124 | } |
| 1125 | |
| 1126 | drbd_thread_start(&connection->ack_receiver); |
| 1127 | /* opencoded create_singlethread_workqueue(), |
| 1128 | * to be able to use format string arguments */ |
| 1129 | connection->ack_sender = |
| 1130 | alloc_ordered_workqueue("drbd_as_%s", WQ_MEM_RECLAIM, connection->resource->name); |
| 1131 | if (!connection->ack_sender) { |
| 1132 | drbd_err(connection, "Failed to create workqueue ack_sender\n"); |
| 1133 | return 0; |
| 1134 | } |
| 1135 | |
| 1136 | mutex_lock(&connection->resource->conf_update); |
| 1137 | /* The discard_my_data flag is a single-shot modifier to the next |
| 1138 | * connection attempt, the handshake of which is now well underway. |
| 1139 | * No need for rcu style copying of the whole struct |
| 1140 | * just to clear a single value. */ |
| 1141 | connection->net_conf->discard_my_data = 0; |
| 1142 | mutex_unlock(&connection->resource->conf_update); |
| 1143 | |
| 1144 | return h; |
| 1145 | |
| 1146 | out_release_sockets: |
| 1147 | if (ad.s_listen) |
| 1148 | sock_release(ad.s_listen); |
| 1149 | if (sock.socket) |
| 1150 | sock_release(sock.socket); |
| 1151 | if (msock.socket) |
| 1152 | sock_release(msock.socket); |
| 1153 | return -1; |
| 1154 | } |
| 1155 | |
| 1156 | static int decode_header(struct drbd_connection *connection, void *header, struct packet_info *pi) |
| 1157 | { |
| 1158 | unsigned int header_size = drbd_header_size(connection); |
| 1159 | |
| 1160 | if (header_size == sizeof(struct p_header100) && |
| 1161 | *(__be32 *)header == cpu_to_be32(DRBD_MAGIC_100)) { |
| 1162 | struct p_header100 *h = header; |
| 1163 | if (h->pad != 0) { |
| 1164 | drbd_err(connection, "Header padding is not zero\n"); |
| 1165 | return -EINVAL; |
| 1166 | } |
| 1167 | pi->vnr = be16_to_cpu(h->volume); |
| 1168 | pi->cmd = be16_to_cpu(h->command); |
| 1169 | pi->size = be32_to_cpu(h->length); |
| 1170 | } else if (header_size == sizeof(struct p_header95) && |
| 1171 | *(__be16 *)header == cpu_to_be16(DRBD_MAGIC_BIG)) { |
| 1172 | struct p_header95 *h = header; |
| 1173 | pi->cmd = be16_to_cpu(h->command); |
| 1174 | pi->size = be32_to_cpu(h->length); |
| 1175 | pi->vnr = 0; |
| 1176 | } else if (header_size == sizeof(struct p_header80) && |
| 1177 | *(__be32 *)header == cpu_to_be32(DRBD_MAGIC)) { |
| 1178 | struct p_header80 *h = header; |
| 1179 | pi->cmd = be16_to_cpu(h->command); |
| 1180 | pi->size = be16_to_cpu(h->length); |
| 1181 | pi->vnr = 0; |
| 1182 | } else { |
| 1183 | drbd_err(connection, "Wrong magic value 0x%08x in protocol version %d\n", |
| 1184 | be32_to_cpu(*(__be32 *)header), |
| 1185 | connection->agreed_pro_version); |
| 1186 | return -EINVAL; |
| 1187 | } |
| 1188 | pi->data = header + header_size; |
| 1189 | return 0; |
| 1190 | } |
| 1191 | |
| 1192 | static int drbd_recv_header(struct drbd_connection *connection, struct packet_info *pi) |
| 1193 | { |
| 1194 | void *buffer = connection->data.rbuf; |
| 1195 | int err; |
| 1196 | |
| 1197 | err = drbd_recv_all_warn(connection, buffer, drbd_header_size(connection)); |
| 1198 | if (err) |
| 1199 | return err; |
| 1200 | |
| 1201 | err = decode_header(connection, buffer, pi); |
| 1202 | connection->last_received = jiffies; |
| 1203 | |
| 1204 | return err; |
| 1205 | } |
| 1206 | |
| 1207 | static void drbd_flush(struct drbd_connection *connection) |
| 1208 | { |
| 1209 | int rv; |
| 1210 | struct drbd_peer_device *peer_device; |
| 1211 | int vnr; |
| 1212 | |
| 1213 | if (connection->resource->write_ordering >= WO_BDEV_FLUSH) { |
| 1214 | rcu_read_lock(); |
| 1215 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
| 1216 | struct drbd_device *device = peer_device->device; |
| 1217 | |
| 1218 | if (!get_ldev(device)) |
| 1219 | continue; |
| 1220 | kref_get(&device->kref); |
| 1221 | rcu_read_unlock(); |
| 1222 | |
| 1223 | /* Right now, we have only this one synchronous code path |
| 1224 | * for flushes between request epochs. |
| 1225 | * We may want to make those asynchronous, |
| 1226 | * or at least parallelize the flushes to the volume devices. |
| 1227 | */ |
| 1228 | device->flush_jif = jiffies; |
| 1229 | set_bit(FLUSH_PENDING, &device->flags); |
| 1230 | rv = blkdev_issue_flush(device->ldev->backing_bdev, |
| 1231 | GFP_NOIO, NULL); |
| 1232 | clear_bit(FLUSH_PENDING, &device->flags); |
| 1233 | if (rv) { |
| 1234 | drbd_info(device, "local disk flush failed with status %d\n", rv); |
| 1235 | /* would rather check on EOPNOTSUPP, but that is not reliable. |
| 1236 | * don't try again for ANY return value != 0 |
| 1237 | * if (rv == -EOPNOTSUPP) */ |
| 1238 | drbd_bump_write_ordering(connection->resource, NULL, WO_DRAIN_IO); |
| 1239 | } |
| 1240 | put_ldev(device); |
| 1241 | kref_put(&device->kref, drbd_destroy_device); |
| 1242 | |
| 1243 | rcu_read_lock(); |
| 1244 | if (rv) |
| 1245 | break; |
| 1246 | } |
| 1247 | rcu_read_unlock(); |
| 1248 | } |
| 1249 | } |
| 1250 | |
| 1251 | /** |
| 1252 | * drbd_may_finish_epoch() - Applies an epoch_event to the epoch's state, eventually finishes it. |
| 1253 | * @device: DRBD device. |
| 1254 | * @epoch: Epoch object. |
| 1255 | * @ev: Epoch event. |
| 1256 | */ |
| 1257 | static enum finish_epoch drbd_may_finish_epoch(struct drbd_connection *connection, |
| 1258 | struct drbd_epoch *epoch, |
| 1259 | enum epoch_event ev) |
| 1260 | { |
| 1261 | int epoch_size; |
| 1262 | struct drbd_epoch *next_epoch; |
| 1263 | enum finish_epoch rv = FE_STILL_LIVE; |
| 1264 | |
| 1265 | spin_lock(&connection->epoch_lock); |
| 1266 | do { |
| 1267 | next_epoch = NULL; |
| 1268 | |
| 1269 | epoch_size = atomic_read(&epoch->epoch_size); |
| 1270 | |
| 1271 | switch (ev & ~EV_CLEANUP) { |
| 1272 | case EV_PUT: |
| 1273 | atomic_dec(&epoch->active); |
| 1274 | break; |
| 1275 | case EV_GOT_BARRIER_NR: |
| 1276 | set_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags); |
| 1277 | break; |
| 1278 | case EV_BECAME_LAST: |
| 1279 | /* nothing to do*/ |
| 1280 | break; |
| 1281 | } |
| 1282 | |
| 1283 | if (epoch_size != 0 && |
| 1284 | atomic_read(&epoch->active) == 0 && |
| 1285 | (test_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags) || ev & EV_CLEANUP)) { |
| 1286 | if (!(ev & EV_CLEANUP)) { |
| 1287 | spin_unlock(&connection->epoch_lock); |
| 1288 | drbd_send_b_ack(epoch->connection, epoch->barrier_nr, epoch_size); |
| 1289 | spin_lock(&connection->epoch_lock); |
| 1290 | } |
| 1291 | #if 0 |
| 1292 | /* FIXME: dec unacked on connection, once we have |
| 1293 | * something to count pending connection packets in. */ |
| 1294 | if (test_bit(DE_HAVE_BARRIER_NUMBER, &epoch->flags)) |
| 1295 | dec_unacked(epoch->connection); |
| 1296 | #endif |
| 1297 | |
| 1298 | if (connection->current_epoch != epoch) { |
| 1299 | next_epoch = list_entry(epoch->list.next, struct drbd_epoch, list); |
| 1300 | list_del(&epoch->list); |
| 1301 | ev = EV_BECAME_LAST | (ev & EV_CLEANUP); |
| 1302 | connection->epochs--; |
| 1303 | kfree(epoch); |
| 1304 | |
| 1305 | if (rv == FE_STILL_LIVE) |
| 1306 | rv = FE_DESTROYED; |
| 1307 | } else { |
| 1308 | epoch->flags = 0; |
| 1309 | atomic_set(&epoch->epoch_size, 0); |
| 1310 | /* atomic_set(&epoch->active, 0); is already zero */ |
| 1311 | if (rv == FE_STILL_LIVE) |
| 1312 | rv = FE_RECYCLED; |
| 1313 | } |
| 1314 | } |
| 1315 | |
| 1316 | if (!next_epoch) |
| 1317 | break; |
| 1318 | |
| 1319 | epoch = next_epoch; |
| 1320 | } while (1); |
| 1321 | |
| 1322 | spin_unlock(&connection->epoch_lock); |
| 1323 | |
| 1324 | return rv; |
| 1325 | } |
| 1326 | |
| 1327 | static enum write_ordering_e |
| 1328 | max_allowed_wo(struct drbd_backing_dev *bdev, enum write_ordering_e wo) |
| 1329 | { |
| 1330 | struct disk_conf *dc; |
| 1331 | |
| 1332 | dc = rcu_dereference(bdev->disk_conf); |
| 1333 | |
| 1334 | if (wo == WO_BDEV_FLUSH && !dc->disk_flushes) |
| 1335 | wo = WO_DRAIN_IO; |
| 1336 | if (wo == WO_DRAIN_IO && !dc->disk_drain) |
| 1337 | wo = WO_NONE; |
| 1338 | |
| 1339 | return wo; |
| 1340 | } |
| 1341 | |
| 1342 | /** |
| 1343 | * drbd_bump_write_ordering() - Fall back to an other write ordering method |
| 1344 | * @connection: DRBD connection. |
| 1345 | * @wo: Write ordering method to try. |
| 1346 | */ |
| 1347 | void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev, |
| 1348 | enum write_ordering_e wo) |
| 1349 | { |
| 1350 | struct drbd_device *device; |
| 1351 | enum write_ordering_e pwo; |
| 1352 | int vnr; |
| 1353 | static char *write_ordering_str[] = { |
| 1354 | [WO_NONE] = "none", |
| 1355 | [WO_DRAIN_IO] = "drain", |
| 1356 | [WO_BDEV_FLUSH] = "flush", |
| 1357 | }; |
| 1358 | |
| 1359 | pwo = resource->write_ordering; |
| 1360 | if (wo != WO_BDEV_FLUSH) |
| 1361 | wo = min(pwo, wo); |
| 1362 | rcu_read_lock(); |
| 1363 | idr_for_each_entry(&resource->devices, device, vnr) { |
| 1364 | if (get_ldev(device)) { |
| 1365 | wo = max_allowed_wo(device->ldev, wo); |
| 1366 | if (device->ldev == bdev) |
| 1367 | bdev = NULL; |
| 1368 | put_ldev(device); |
| 1369 | } |
| 1370 | } |
| 1371 | |
| 1372 | if (bdev) |
| 1373 | wo = max_allowed_wo(bdev, wo); |
| 1374 | |
| 1375 | rcu_read_unlock(); |
| 1376 | |
| 1377 | resource->write_ordering = wo; |
| 1378 | if (pwo != resource->write_ordering || wo == WO_BDEV_FLUSH) |
| 1379 | drbd_info(resource, "Method to ensure write ordering: %s\n", write_ordering_str[resource->write_ordering]); |
| 1380 | } |
| 1381 | |
| 1382 | /** |
| 1383 | * drbd_submit_peer_request() |
| 1384 | * @device: DRBD device. |
| 1385 | * @peer_req: peer request |
| 1386 | * @rw: flag field, see bio->bi_rw |
| 1387 | * |
| 1388 | * May spread the pages to multiple bios, |
| 1389 | * depending on bio_add_page restrictions. |
| 1390 | * |
| 1391 | * Returns 0 if all bios have been submitted, |
| 1392 | * -ENOMEM if we could not allocate enough bios, |
| 1393 | * -ENOSPC (any better suggestion?) if we have not been able to bio_add_page a |
| 1394 | * single page to an empty bio (which should never happen and likely indicates |
| 1395 | * that the lower level IO stack is in some way broken). This has been observed |
| 1396 | * on certain Xen deployments. |
| 1397 | */ |
| 1398 | /* TODO allocate from our own bio_set. */ |
| 1399 | int drbd_submit_peer_request(struct drbd_device *device, |
| 1400 | struct drbd_peer_request *peer_req, |
| 1401 | const unsigned rw, const int fault_type) |
| 1402 | { |
| 1403 | struct bio *bios = NULL; |
| 1404 | struct bio *bio; |
| 1405 | struct page *page = peer_req->pages; |
| 1406 | sector_t sector = peer_req->i.sector; |
| 1407 | unsigned data_size = peer_req->i.size; |
| 1408 | unsigned n_bios = 0; |
| 1409 | unsigned nr_pages = (data_size + PAGE_SIZE -1) >> PAGE_SHIFT; |
| 1410 | int err = -ENOMEM; |
| 1411 | |
| 1412 | if (peer_req->flags & EE_IS_TRIM_USE_ZEROOUT) { |
| 1413 | /* wait for all pending IO completions, before we start |
| 1414 | * zeroing things out. */ |
| 1415 | conn_wait_active_ee_empty(peer_req->peer_device->connection); |
| 1416 | /* add it to the active list now, |
| 1417 | * so we can find it to present it in debugfs */ |
| 1418 | peer_req->submit_jif = jiffies; |
| 1419 | peer_req->flags |= EE_SUBMITTED; |
| 1420 | spin_lock_irq(&device->resource->req_lock); |
| 1421 | list_add_tail(&peer_req->w.list, &device->active_ee); |
| 1422 | spin_unlock_irq(&device->resource->req_lock); |
| 1423 | if (blkdev_issue_zeroout(device->ldev->backing_bdev, |
| 1424 | sector, data_size >> 9, GFP_NOIO, false)) |
| 1425 | peer_req->flags |= EE_WAS_ERROR; |
| 1426 | drbd_endio_write_sec_final(peer_req); |
| 1427 | return 0; |
| 1428 | } |
| 1429 | |
| 1430 | /* Discards don't have any payload. |
| 1431 | * But the scsi layer still expects a bio_vec it can use internally, |
| 1432 | * see sd_setup_discard_cmnd() and blk_add_request_payload(). */ |
| 1433 | if (peer_req->flags & EE_IS_TRIM) |
| 1434 | nr_pages = 1; |
| 1435 | |
| 1436 | /* In most cases, we will only need one bio. But in case the lower |
| 1437 | * level restrictions happen to be different at this offset on this |
| 1438 | * side than those of the sending peer, we may need to submit the |
| 1439 | * request in more than one bio. |
| 1440 | * |
| 1441 | * Plain bio_alloc is good enough here, this is no DRBD internally |
| 1442 | * generated bio, but a bio allocated on behalf of the peer. |
| 1443 | */ |
| 1444 | next_bio: |
| 1445 | bio = bio_alloc(GFP_NOIO, nr_pages); |
| 1446 | if (!bio) { |
| 1447 | drbd_err(device, "submit_ee: Allocation of a bio failed (nr_pages=%u)\n", nr_pages); |
| 1448 | goto fail; |
| 1449 | } |
| 1450 | /* > peer_req->i.sector, unless this is the first bio */ |
| 1451 | bio->bi_iter.bi_sector = sector; |
| 1452 | bio->bi_bdev = device->ldev->backing_bdev; |
| 1453 | bio->bi_rw = rw; |
| 1454 | bio->bi_private = peer_req; |
| 1455 | bio->bi_end_io = drbd_peer_request_endio; |
| 1456 | |
| 1457 | bio->bi_next = bios; |
| 1458 | bios = bio; |
| 1459 | ++n_bios; |
| 1460 | |
| 1461 | if (rw & REQ_DISCARD) { |
| 1462 | bio->bi_iter.bi_size = data_size; |
| 1463 | goto submit; |
| 1464 | } |
| 1465 | |
| 1466 | page_chain_for_each(page) { |
| 1467 | unsigned len = min_t(unsigned, data_size, PAGE_SIZE); |
| 1468 | if (!bio_add_page(bio, page, len, 0)) { |
| 1469 | /* A single page must always be possible! |
| 1470 | * But in case it fails anyways, |
| 1471 | * we deal with it, and complain (below). */ |
| 1472 | if (bio->bi_vcnt == 0) { |
| 1473 | drbd_err(device, |
| 1474 | "bio_add_page failed for len=%u, " |
| 1475 | "bi_vcnt=0 (bi_sector=%llu)\n", |
| 1476 | len, (uint64_t)bio->bi_iter.bi_sector); |
| 1477 | err = -ENOSPC; |
| 1478 | goto fail; |
| 1479 | } |
| 1480 | goto next_bio; |
| 1481 | } |
| 1482 | data_size -= len; |
| 1483 | sector += len >> 9; |
| 1484 | --nr_pages; |
| 1485 | } |
| 1486 | D_ASSERT(device, data_size == 0); |
| 1487 | submit: |
| 1488 | D_ASSERT(device, page == NULL); |
| 1489 | |
| 1490 | atomic_set(&peer_req->pending_bios, n_bios); |
| 1491 | /* for debugfs: update timestamp, mark as submitted */ |
| 1492 | peer_req->submit_jif = jiffies; |
| 1493 | peer_req->flags |= EE_SUBMITTED; |
| 1494 | do { |
| 1495 | bio = bios; |
| 1496 | bios = bios->bi_next; |
| 1497 | bio->bi_next = NULL; |
| 1498 | |
| 1499 | drbd_generic_make_request(device, fault_type, bio); |
| 1500 | } while (bios); |
| 1501 | return 0; |
| 1502 | |
| 1503 | fail: |
| 1504 | while (bios) { |
| 1505 | bio = bios; |
| 1506 | bios = bios->bi_next; |
| 1507 | bio_put(bio); |
| 1508 | } |
| 1509 | return err; |
| 1510 | } |
| 1511 | |
| 1512 | static void drbd_remove_epoch_entry_interval(struct drbd_device *device, |
| 1513 | struct drbd_peer_request *peer_req) |
| 1514 | { |
| 1515 | struct drbd_interval *i = &peer_req->i; |
| 1516 | |
| 1517 | drbd_remove_interval(&device->write_requests, i); |
| 1518 | drbd_clear_interval(i); |
| 1519 | |
| 1520 | /* Wake up any processes waiting for this peer request to complete. */ |
| 1521 | if (i->waiting) |
| 1522 | wake_up(&device->misc_wait); |
| 1523 | } |
| 1524 | |
| 1525 | static void conn_wait_active_ee_empty(struct drbd_connection *connection) |
| 1526 | { |
| 1527 | struct drbd_peer_device *peer_device; |
| 1528 | int vnr; |
| 1529 | |
| 1530 | rcu_read_lock(); |
| 1531 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
| 1532 | struct drbd_device *device = peer_device->device; |
| 1533 | |
| 1534 | kref_get(&device->kref); |
| 1535 | rcu_read_unlock(); |
| 1536 | drbd_wait_ee_list_empty(device, &device->active_ee); |
| 1537 | kref_put(&device->kref, drbd_destroy_device); |
| 1538 | rcu_read_lock(); |
| 1539 | } |
| 1540 | rcu_read_unlock(); |
| 1541 | } |
| 1542 | |
| 1543 | static int receive_Barrier(struct drbd_connection *connection, struct packet_info *pi) |
| 1544 | { |
| 1545 | int rv; |
| 1546 | struct p_barrier *p = pi->data; |
| 1547 | struct drbd_epoch *epoch; |
| 1548 | |
| 1549 | /* FIXME these are unacked on connection, |
| 1550 | * not a specific (peer)device. |
| 1551 | */ |
| 1552 | connection->current_epoch->barrier_nr = p->barrier; |
| 1553 | connection->current_epoch->connection = connection; |
| 1554 | rv = drbd_may_finish_epoch(connection, connection->current_epoch, EV_GOT_BARRIER_NR); |
| 1555 | |
| 1556 | /* P_BARRIER_ACK may imply that the corresponding extent is dropped from |
| 1557 | * the activity log, which means it would not be resynced in case the |
| 1558 | * R_PRIMARY crashes now. |
| 1559 | * Therefore we must send the barrier_ack after the barrier request was |
| 1560 | * completed. */ |
| 1561 | switch (connection->resource->write_ordering) { |
| 1562 | case WO_NONE: |
| 1563 | if (rv == FE_RECYCLED) |
| 1564 | return 0; |
| 1565 | |
| 1566 | /* receiver context, in the writeout path of the other node. |
| 1567 | * avoid potential distributed deadlock */ |
| 1568 | epoch = kmalloc(sizeof(struct drbd_epoch), GFP_NOIO); |
| 1569 | if (epoch) |
| 1570 | break; |
| 1571 | else |
| 1572 | drbd_warn(connection, "Allocation of an epoch failed, slowing down\n"); |
| 1573 | /* Fall through */ |
| 1574 | |
| 1575 | case WO_BDEV_FLUSH: |
| 1576 | case WO_DRAIN_IO: |
| 1577 | conn_wait_active_ee_empty(connection); |
| 1578 | drbd_flush(connection); |
| 1579 | |
| 1580 | if (atomic_read(&connection->current_epoch->epoch_size)) { |
| 1581 | epoch = kmalloc(sizeof(struct drbd_epoch), GFP_NOIO); |
| 1582 | if (epoch) |
| 1583 | break; |
| 1584 | } |
| 1585 | |
| 1586 | return 0; |
| 1587 | default: |
| 1588 | drbd_err(connection, "Strangeness in connection->write_ordering %d\n", |
| 1589 | connection->resource->write_ordering); |
| 1590 | return -EIO; |
| 1591 | } |
| 1592 | |
| 1593 | epoch->flags = 0; |
| 1594 | atomic_set(&epoch->epoch_size, 0); |
| 1595 | atomic_set(&epoch->active, 0); |
| 1596 | |
| 1597 | spin_lock(&connection->epoch_lock); |
| 1598 | if (atomic_read(&connection->current_epoch->epoch_size)) { |
| 1599 | list_add(&epoch->list, &connection->current_epoch->list); |
| 1600 | connection->current_epoch = epoch; |
| 1601 | connection->epochs++; |
| 1602 | } else { |
| 1603 | /* The current_epoch got recycled while we allocated this one... */ |
| 1604 | kfree(epoch); |
| 1605 | } |
| 1606 | spin_unlock(&connection->epoch_lock); |
| 1607 | |
| 1608 | return 0; |
| 1609 | } |
| 1610 | |
| 1611 | /* used from receive_RSDataReply (recv_resync_read) |
| 1612 | * and from receive_Data */ |
| 1613 | static struct drbd_peer_request * |
| 1614 | read_in_block(struct drbd_peer_device *peer_device, u64 id, sector_t sector, |
| 1615 | struct packet_info *pi) __must_hold(local) |
| 1616 | { |
| 1617 | struct drbd_device *device = peer_device->device; |
| 1618 | const sector_t capacity = drbd_get_capacity(device->this_bdev); |
| 1619 | struct drbd_peer_request *peer_req; |
| 1620 | struct page *page; |
| 1621 | int digest_size, err; |
| 1622 | unsigned int data_size = pi->size, ds; |
| 1623 | void *dig_in = peer_device->connection->int_dig_in; |
| 1624 | void *dig_vv = peer_device->connection->int_dig_vv; |
| 1625 | unsigned long *data; |
| 1626 | struct p_trim *trim = (pi->cmd == P_TRIM) ? pi->data : NULL; |
| 1627 | |
| 1628 | digest_size = 0; |
| 1629 | if (!trim && peer_device->connection->peer_integrity_tfm) { |
| 1630 | digest_size = crypto_ahash_digestsize(peer_device->connection->peer_integrity_tfm); |
| 1631 | /* |
| 1632 | * FIXME: Receive the incoming digest into the receive buffer |
| 1633 | * here, together with its struct p_data? |
| 1634 | */ |
| 1635 | err = drbd_recv_all_warn(peer_device->connection, dig_in, digest_size); |
| 1636 | if (err) |
| 1637 | return NULL; |
| 1638 | data_size -= digest_size; |
| 1639 | } |
| 1640 | |
| 1641 | if (trim) { |
| 1642 | D_ASSERT(peer_device, data_size == 0); |
| 1643 | data_size = be32_to_cpu(trim->size); |
| 1644 | } |
| 1645 | |
| 1646 | if (!expect(IS_ALIGNED(data_size, 512))) |
| 1647 | return NULL; |
| 1648 | /* prepare for larger trim requests. */ |
| 1649 | if (!trim && !expect(data_size <= DRBD_MAX_BIO_SIZE)) |
| 1650 | return NULL; |
| 1651 | |
| 1652 | /* even though we trust out peer, |
| 1653 | * we sometimes have to double check. */ |
| 1654 | if (sector + (data_size>>9) > capacity) { |
| 1655 | drbd_err(device, "request from peer beyond end of local disk: " |
| 1656 | "capacity: %llus < sector: %llus + size: %u\n", |
| 1657 | (unsigned long long)capacity, |
| 1658 | (unsigned long long)sector, data_size); |
| 1659 | return NULL; |
| 1660 | } |
| 1661 | |
| 1662 | /* GFP_NOIO, because we must not cause arbitrary write-out: in a DRBD |
| 1663 | * "criss-cross" setup, that might cause write-out on some other DRBD, |
| 1664 | * which in turn might block on the other node at this very place. */ |
| 1665 | peer_req = drbd_alloc_peer_req(peer_device, id, sector, data_size, trim == NULL, GFP_NOIO); |
| 1666 | if (!peer_req) |
| 1667 | return NULL; |
| 1668 | |
| 1669 | peer_req->flags |= EE_WRITE; |
| 1670 | if (trim) |
| 1671 | return peer_req; |
| 1672 | |
| 1673 | ds = data_size; |
| 1674 | page = peer_req->pages; |
| 1675 | page_chain_for_each(page) { |
| 1676 | unsigned len = min_t(int, ds, PAGE_SIZE); |
| 1677 | data = kmap(page); |
| 1678 | err = drbd_recv_all_warn(peer_device->connection, data, len); |
| 1679 | if (drbd_insert_fault(device, DRBD_FAULT_RECEIVE)) { |
| 1680 | drbd_err(device, "Fault injection: Corrupting data on receive\n"); |
| 1681 | data[0] = data[0] ^ (unsigned long)-1; |
| 1682 | } |
| 1683 | kunmap(page); |
| 1684 | if (err) { |
| 1685 | drbd_free_peer_req(device, peer_req); |
| 1686 | return NULL; |
| 1687 | } |
| 1688 | ds -= len; |
| 1689 | } |
| 1690 | |
| 1691 | if (digest_size) { |
| 1692 | drbd_csum_ee(peer_device->connection->peer_integrity_tfm, peer_req, dig_vv); |
| 1693 | if (memcmp(dig_in, dig_vv, digest_size)) { |
| 1694 | drbd_err(device, "Digest integrity check FAILED: %llus +%u\n", |
| 1695 | (unsigned long long)sector, data_size); |
| 1696 | drbd_free_peer_req(device, peer_req); |
| 1697 | return NULL; |
| 1698 | } |
| 1699 | } |
| 1700 | device->recv_cnt += data_size >> 9; |
| 1701 | return peer_req; |
| 1702 | } |
| 1703 | |
| 1704 | /* drbd_drain_block() just takes a data block |
| 1705 | * out of the socket input buffer, and discards it. |
| 1706 | */ |
| 1707 | static int drbd_drain_block(struct drbd_peer_device *peer_device, int data_size) |
| 1708 | { |
| 1709 | struct page *page; |
| 1710 | int err = 0; |
| 1711 | void *data; |
| 1712 | |
| 1713 | if (!data_size) |
| 1714 | return 0; |
| 1715 | |
| 1716 | page = drbd_alloc_pages(peer_device, 1, 1); |
| 1717 | |
| 1718 | data = kmap(page); |
| 1719 | while (data_size) { |
| 1720 | unsigned int len = min_t(int, data_size, PAGE_SIZE); |
| 1721 | |
| 1722 | err = drbd_recv_all_warn(peer_device->connection, data, len); |
| 1723 | if (err) |
| 1724 | break; |
| 1725 | data_size -= len; |
| 1726 | } |
| 1727 | kunmap(page); |
| 1728 | drbd_free_pages(peer_device->device, page, 0); |
| 1729 | return err; |
| 1730 | } |
| 1731 | |
| 1732 | static int recv_dless_read(struct drbd_peer_device *peer_device, struct drbd_request *req, |
| 1733 | sector_t sector, int data_size) |
| 1734 | { |
| 1735 | struct bio_vec bvec; |
| 1736 | struct bvec_iter iter; |
| 1737 | struct bio *bio; |
| 1738 | int digest_size, err, expect; |
| 1739 | void *dig_in = peer_device->connection->int_dig_in; |
| 1740 | void *dig_vv = peer_device->connection->int_dig_vv; |
| 1741 | |
| 1742 | digest_size = 0; |
| 1743 | if (peer_device->connection->peer_integrity_tfm) { |
| 1744 | digest_size = crypto_ahash_digestsize(peer_device->connection->peer_integrity_tfm); |
| 1745 | err = drbd_recv_all_warn(peer_device->connection, dig_in, digest_size); |
| 1746 | if (err) |
| 1747 | return err; |
| 1748 | data_size -= digest_size; |
| 1749 | } |
| 1750 | |
| 1751 | /* optimistically update recv_cnt. if receiving fails below, |
| 1752 | * we disconnect anyways, and counters will be reset. */ |
| 1753 | peer_device->device->recv_cnt += data_size>>9; |
| 1754 | |
| 1755 | bio = req->master_bio; |
| 1756 | D_ASSERT(peer_device->device, sector == bio->bi_iter.bi_sector); |
| 1757 | |
| 1758 | bio_for_each_segment(bvec, bio, iter) { |
| 1759 | void *mapped = kmap(bvec.bv_page) + bvec.bv_offset; |
| 1760 | expect = min_t(int, data_size, bvec.bv_len); |
| 1761 | err = drbd_recv_all_warn(peer_device->connection, mapped, expect); |
| 1762 | kunmap(bvec.bv_page); |
| 1763 | if (err) |
| 1764 | return err; |
| 1765 | data_size -= expect; |
| 1766 | } |
| 1767 | |
| 1768 | if (digest_size) { |
| 1769 | drbd_csum_bio(peer_device->connection->peer_integrity_tfm, bio, dig_vv); |
| 1770 | if (memcmp(dig_in, dig_vv, digest_size)) { |
| 1771 | drbd_err(peer_device, "Digest integrity check FAILED. Broken NICs?\n"); |
| 1772 | return -EINVAL; |
| 1773 | } |
| 1774 | } |
| 1775 | |
| 1776 | D_ASSERT(peer_device->device, data_size == 0); |
| 1777 | return 0; |
| 1778 | } |
| 1779 | |
| 1780 | /* |
| 1781 | * e_end_resync_block() is called in ack_sender context via |
| 1782 | * drbd_finish_peer_reqs(). |
| 1783 | */ |
| 1784 | static int e_end_resync_block(struct drbd_work *w, int unused) |
| 1785 | { |
| 1786 | struct drbd_peer_request *peer_req = |
| 1787 | container_of(w, struct drbd_peer_request, w); |
| 1788 | struct drbd_peer_device *peer_device = peer_req->peer_device; |
| 1789 | struct drbd_device *device = peer_device->device; |
| 1790 | sector_t sector = peer_req->i.sector; |
| 1791 | int err; |
| 1792 | |
| 1793 | D_ASSERT(device, drbd_interval_empty(&peer_req->i)); |
| 1794 | |
| 1795 | if (likely((peer_req->flags & EE_WAS_ERROR) == 0)) { |
| 1796 | drbd_set_in_sync(device, sector, peer_req->i.size); |
| 1797 | err = drbd_send_ack(peer_device, P_RS_WRITE_ACK, peer_req); |
| 1798 | } else { |
| 1799 | /* Record failure to sync */ |
| 1800 | drbd_rs_failed_io(device, sector, peer_req->i.size); |
| 1801 | |
| 1802 | err = drbd_send_ack(peer_device, P_NEG_ACK, peer_req); |
| 1803 | } |
| 1804 | dec_unacked(device); |
| 1805 | |
| 1806 | return err; |
| 1807 | } |
| 1808 | |
| 1809 | static int recv_resync_read(struct drbd_peer_device *peer_device, sector_t sector, |
| 1810 | struct packet_info *pi) __releases(local) |
| 1811 | { |
| 1812 | struct drbd_device *device = peer_device->device; |
| 1813 | struct drbd_peer_request *peer_req; |
| 1814 | |
| 1815 | peer_req = read_in_block(peer_device, ID_SYNCER, sector, pi); |
| 1816 | if (!peer_req) |
| 1817 | goto fail; |
| 1818 | |
| 1819 | dec_rs_pending(device); |
| 1820 | |
| 1821 | inc_unacked(device); |
| 1822 | /* corresponding dec_unacked() in e_end_resync_block() |
| 1823 | * respective _drbd_clear_done_ee */ |
| 1824 | |
| 1825 | peer_req->w.cb = e_end_resync_block; |
| 1826 | peer_req->submit_jif = jiffies; |
| 1827 | |
| 1828 | spin_lock_irq(&device->resource->req_lock); |
| 1829 | list_add_tail(&peer_req->w.list, &device->sync_ee); |
| 1830 | spin_unlock_irq(&device->resource->req_lock); |
| 1831 | |
| 1832 | atomic_add(pi->size >> 9, &device->rs_sect_ev); |
| 1833 | if (drbd_submit_peer_request(device, peer_req, WRITE, DRBD_FAULT_RS_WR) == 0) |
| 1834 | return 0; |
| 1835 | |
| 1836 | /* don't care for the reason here */ |
| 1837 | drbd_err(device, "submit failed, triggering re-connect\n"); |
| 1838 | spin_lock_irq(&device->resource->req_lock); |
| 1839 | list_del(&peer_req->w.list); |
| 1840 | spin_unlock_irq(&device->resource->req_lock); |
| 1841 | |
| 1842 | drbd_free_peer_req(device, peer_req); |
| 1843 | fail: |
| 1844 | put_ldev(device); |
| 1845 | return -EIO; |
| 1846 | } |
| 1847 | |
| 1848 | static struct drbd_request * |
| 1849 | find_request(struct drbd_device *device, struct rb_root *root, u64 id, |
| 1850 | sector_t sector, bool missing_ok, const char *func) |
| 1851 | { |
| 1852 | struct drbd_request *req; |
| 1853 | |
| 1854 | /* Request object according to our peer */ |
| 1855 | req = (struct drbd_request *)(unsigned long)id; |
| 1856 | if (drbd_contains_interval(root, sector, &req->i) && req->i.local) |
| 1857 | return req; |
| 1858 | if (!missing_ok) { |
| 1859 | drbd_err(device, "%s: failed to find request 0x%lx, sector %llus\n", func, |
| 1860 | (unsigned long)id, (unsigned long long)sector); |
| 1861 | } |
| 1862 | return NULL; |
| 1863 | } |
| 1864 | |
| 1865 | static int receive_DataReply(struct drbd_connection *connection, struct packet_info *pi) |
| 1866 | { |
| 1867 | struct drbd_peer_device *peer_device; |
| 1868 | struct drbd_device *device; |
| 1869 | struct drbd_request *req; |
| 1870 | sector_t sector; |
| 1871 | int err; |
| 1872 | struct p_data *p = pi->data; |
| 1873 | |
| 1874 | peer_device = conn_peer_device(connection, pi->vnr); |
| 1875 | if (!peer_device) |
| 1876 | return -EIO; |
| 1877 | device = peer_device->device; |
| 1878 | |
| 1879 | sector = be64_to_cpu(p->sector); |
| 1880 | |
| 1881 | spin_lock_irq(&device->resource->req_lock); |
| 1882 | req = find_request(device, &device->read_requests, p->block_id, sector, false, __func__); |
| 1883 | spin_unlock_irq(&device->resource->req_lock); |
| 1884 | if (unlikely(!req)) |
| 1885 | return -EIO; |
| 1886 | |
| 1887 | /* hlist_del(&req->collision) is done in _req_may_be_done, to avoid |
| 1888 | * special casing it there for the various failure cases. |
| 1889 | * still no race with drbd_fail_pending_reads */ |
| 1890 | err = recv_dless_read(peer_device, req, sector, pi->size); |
| 1891 | if (!err) |
| 1892 | req_mod(req, DATA_RECEIVED); |
| 1893 | /* else: nothing. handled from drbd_disconnect... |
| 1894 | * I don't think we may complete this just yet |
| 1895 | * in case we are "on-disconnect: freeze" */ |
| 1896 | |
| 1897 | return err; |
| 1898 | } |
| 1899 | |
| 1900 | static int receive_RSDataReply(struct drbd_connection *connection, struct packet_info *pi) |
| 1901 | { |
| 1902 | struct drbd_peer_device *peer_device; |
| 1903 | struct drbd_device *device; |
| 1904 | sector_t sector; |
| 1905 | int err; |
| 1906 | struct p_data *p = pi->data; |
| 1907 | |
| 1908 | peer_device = conn_peer_device(connection, pi->vnr); |
| 1909 | if (!peer_device) |
| 1910 | return -EIO; |
| 1911 | device = peer_device->device; |
| 1912 | |
| 1913 | sector = be64_to_cpu(p->sector); |
| 1914 | D_ASSERT(device, p->block_id == ID_SYNCER); |
| 1915 | |
| 1916 | if (get_ldev(device)) { |
| 1917 | /* data is submitted to disk within recv_resync_read. |
| 1918 | * corresponding put_ldev done below on error, |
| 1919 | * or in drbd_peer_request_endio. */ |
| 1920 | err = recv_resync_read(peer_device, sector, pi); |
| 1921 | } else { |
| 1922 | if (__ratelimit(&drbd_ratelimit_state)) |
| 1923 | drbd_err(device, "Can not write resync data to local disk.\n"); |
| 1924 | |
| 1925 | err = drbd_drain_block(peer_device, pi->size); |
| 1926 | |
| 1927 | drbd_send_ack_dp(peer_device, P_NEG_ACK, p, pi->size); |
| 1928 | } |
| 1929 | |
| 1930 | atomic_add(pi->size >> 9, &device->rs_sect_in); |
| 1931 | |
| 1932 | return err; |
| 1933 | } |
| 1934 | |
| 1935 | static void restart_conflicting_writes(struct drbd_device *device, |
| 1936 | sector_t sector, int size) |
| 1937 | { |
| 1938 | struct drbd_interval *i; |
| 1939 | struct drbd_request *req; |
| 1940 | |
| 1941 | drbd_for_each_overlap(i, &device->write_requests, sector, size) { |
| 1942 | if (!i->local) |
| 1943 | continue; |
| 1944 | req = container_of(i, struct drbd_request, i); |
| 1945 | if (req->rq_state & RQ_LOCAL_PENDING || |
| 1946 | !(req->rq_state & RQ_POSTPONED)) |
| 1947 | continue; |
| 1948 | /* as it is RQ_POSTPONED, this will cause it to |
| 1949 | * be queued on the retry workqueue. */ |
| 1950 | __req_mod(req, CONFLICT_RESOLVED, NULL); |
| 1951 | } |
| 1952 | } |
| 1953 | |
| 1954 | /* |
| 1955 | * e_end_block() is called in ack_sender context via drbd_finish_peer_reqs(). |
| 1956 | */ |
| 1957 | static int e_end_block(struct drbd_work *w, int cancel) |
| 1958 | { |
| 1959 | struct drbd_peer_request *peer_req = |
| 1960 | container_of(w, struct drbd_peer_request, w); |
| 1961 | struct drbd_peer_device *peer_device = peer_req->peer_device; |
| 1962 | struct drbd_device *device = peer_device->device; |
| 1963 | sector_t sector = peer_req->i.sector; |
| 1964 | int err = 0, pcmd; |
| 1965 | |
| 1966 | if (peer_req->flags & EE_SEND_WRITE_ACK) { |
| 1967 | if (likely((peer_req->flags & EE_WAS_ERROR) == 0)) { |
| 1968 | pcmd = (device->state.conn >= C_SYNC_SOURCE && |
| 1969 | device->state.conn <= C_PAUSED_SYNC_T && |
| 1970 | peer_req->flags & EE_MAY_SET_IN_SYNC) ? |
| 1971 | P_RS_WRITE_ACK : P_WRITE_ACK; |
| 1972 | err = drbd_send_ack(peer_device, pcmd, peer_req); |
| 1973 | if (pcmd == P_RS_WRITE_ACK) |
| 1974 | drbd_set_in_sync(device, sector, peer_req->i.size); |
| 1975 | } else { |
| 1976 | err = drbd_send_ack(peer_device, P_NEG_ACK, peer_req); |
| 1977 | /* we expect it to be marked out of sync anyways... |
| 1978 | * maybe assert this? */ |
| 1979 | } |
| 1980 | dec_unacked(device); |
| 1981 | } |
| 1982 | |
| 1983 | /* we delete from the conflict detection hash _after_ we sent out the |
| 1984 | * P_WRITE_ACK / P_NEG_ACK, to get the sequence number right. */ |
| 1985 | if (peer_req->flags & EE_IN_INTERVAL_TREE) { |
| 1986 | spin_lock_irq(&device->resource->req_lock); |
| 1987 | D_ASSERT(device, !drbd_interval_empty(&peer_req->i)); |
| 1988 | drbd_remove_epoch_entry_interval(device, peer_req); |
| 1989 | if (peer_req->flags & EE_RESTART_REQUESTS) |
| 1990 | restart_conflicting_writes(device, sector, peer_req->i.size); |
| 1991 | spin_unlock_irq(&device->resource->req_lock); |
| 1992 | } else |
| 1993 | D_ASSERT(device, drbd_interval_empty(&peer_req->i)); |
| 1994 | |
| 1995 | drbd_may_finish_epoch(peer_device->connection, peer_req->epoch, EV_PUT + (cancel ? EV_CLEANUP : 0)); |
| 1996 | |
| 1997 | return err; |
| 1998 | } |
| 1999 | |
| 2000 | static int e_send_ack(struct drbd_work *w, enum drbd_packet ack) |
| 2001 | { |
| 2002 | struct drbd_peer_request *peer_req = |
| 2003 | container_of(w, struct drbd_peer_request, w); |
| 2004 | struct drbd_peer_device *peer_device = peer_req->peer_device; |
| 2005 | int err; |
| 2006 | |
| 2007 | err = drbd_send_ack(peer_device, ack, peer_req); |
| 2008 | dec_unacked(peer_device->device); |
| 2009 | |
| 2010 | return err; |
| 2011 | } |
| 2012 | |
| 2013 | static int e_send_superseded(struct drbd_work *w, int unused) |
| 2014 | { |
| 2015 | return e_send_ack(w, P_SUPERSEDED); |
| 2016 | } |
| 2017 | |
| 2018 | static int e_send_retry_write(struct drbd_work *w, int unused) |
| 2019 | { |
| 2020 | struct drbd_peer_request *peer_req = |
| 2021 | container_of(w, struct drbd_peer_request, w); |
| 2022 | struct drbd_connection *connection = peer_req->peer_device->connection; |
| 2023 | |
| 2024 | return e_send_ack(w, connection->agreed_pro_version >= 100 ? |
| 2025 | P_RETRY_WRITE : P_SUPERSEDED); |
| 2026 | } |
| 2027 | |
| 2028 | static bool seq_greater(u32 a, u32 b) |
| 2029 | { |
| 2030 | /* |
| 2031 | * We assume 32-bit wrap-around here. |
| 2032 | * For 24-bit wrap-around, we would have to shift: |
| 2033 | * a <<= 8; b <<= 8; |
| 2034 | */ |
| 2035 | return (s32)a - (s32)b > 0; |
| 2036 | } |
| 2037 | |
| 2038 | static u32 seq_max(u32 a, u32 b) |
| 2039 | { |
| 2040 | return seq_greater(a, b) ? a : b; |
| 2041 | } |
| 2042 | |
| 2043 | static void update_peer_seq(struct drbd_peer_device *peer_device, unsigned int peer_seq) |
| 2044 | { |
| 2045 | struct drbd_device *device = peer_device->device; |
| 2046 | unsigned int newest_peer_seq; |
| 2047 | |
| 2048 | if (test_bit(RESOLVE_CONFLICTS, &peer_device->connection->flags)) { |
| 2049 | spin_lock(&device->peer_seq_lock); |
| 2050 | newest_peer_seq = seq_max(device->peer_seq, peer_seq); |
| 2051 | device->peer_seq = newest_peer_seq; |
| 2052 | spin_unlock(&device->peer_seq_lock); |
| 2053 | /* wake up only if we actually changed device->peer_seq */ |
| 2054 | if (peer_seq == newest_peer_seq) |
| 2055 | wake_up(&device->seq_wait); |
| 2056 | } |
| 2057 | } |
| 2058 | |
| 2059 | static inline int overlaps(sector_t s1, int l1, sector_t s2, int l2) |
| 2060 | { |
| 2061 | return !((s1 + (l1>>9) <= s2) || (s1 >= s2 + (l2>>9))); |
| 2062 | } |
| 2063 | |
| 2064 | /* maybe change sync_ee into interval trees as well? */ |
| 2065 | static bool overlapping_resync_write(struct drbd_device *device, struct drbd_peer_request *peer_req) |
| 2066 | { |
| 2067 | struct drbd_peer_request *rs_req; |
| 2068 | bool rv = 0; |
| 2069 | |
| 2070 | spin_lock_irq(&device->resource->req_lock); |
| 2071 | list_for_each_entry(rs_req, &device->sync_ee, w.list) { |
| 2072 | if (overlaps(peer_req->i.sector, peer_req->i.size, |
| 2073 | rs_req->i.sector, rs_req->i.size)) { |
| 2074 | rv = 1; |
| 2075 | break; |
| 2076 | } |
| 2077 | } |
| 2078 | spin_unlock_irq(&device->resource->req_lock); |
| 2079 | |
| 2080 | return rv; |
| 2081 | } |
| 2082 | |
| 2083 | /* Called from receive_Data. |
| 2084 | * Synchronize packets on sock with packets on msock. |
| 2085 | * |
| 2086 | * This is here so even when a P_DATA packet traveling via sock overtook an Ack |
| 2087 | * packet traveling on msock, they are still processed in the order they have |
| 2088 | * been sent. |
| 2089 | * |
| 2090 | * Note: we don't care for Ack packets overtaking P_DATA packets. |
| 2091 | * |
| 2092 | * In case packet_seq is larger than device->peer_seq number, there are |
| 2093 | * outstanding packets on the msock. We wait for them to arrive. |
| 2094 | * In case we are the logically next packet, we update device->peer_seq |
| 2095 | * ourselves. Correctly handles 32bit wrap around. |
| 2096 | * |
| 2097 | * Assume we have a 10 GBit connection, that is about 1<<30 byte per second, |
| 2098 | * about 1<<21 sectors per second. So "worst" case, we have 1<<3 == 8 seconds |
| 2099 | * for the 24bit wrap (historical atomic_t guarantee on some archs), and we have |
| 2100 | * 1<<9 == 512 seconds aka ages for the 32bit wrap around... |
| 2101 | * |
| 2102 | * returns 0 if we may process the packet, |
| 2103 | * -ERESTARTSYS if we were interrupted (by disconnect signal). */ |
| 2104 | static int wait_for_and_update_peer_seq(struct drbd_peer_device *peer_device, const u32 peer_seq) |
| 2105 | { |
| 2106 | struct drbd_device *device = peer_device->device; |
| 2107 | DEFINE_WAIT(wait); |
| 2108 | long timeout; |
| 2109 | int ret = 0, tp; |
| 2110 | |
| 2111 | if (!test_bit(RESOLVE_CONFLICTS, &peer_device->connection->flags)) |
| 2112 | return 0; |
| 2113 | |
| 2114 | spin_lock(&device->peer_seq_lock); |
| 2115 | for (;;) { |
| 2116 | if (!seq_greater(peer_seq - 1, device->peer_seq)) { |
| 2117 | device->peer_seq = seq_max(device->peer_seq, peer_seq); |
| 2118 | break; |
| 2119 | } |
| 2120 | |
| 2121 | if (signal_pending(current)) { |
| 2122 | ret = -ERESTARTSYS; |
| 2123 | break; |
| 2124 | } |
| 2125 | |
| 2126 | rcu_read_lock(); |
| 2127 | tp = rcu_dereference(peer_device->connection->net_conf)->two_primaries; |
| 2128 | rcu_read_unlock(); |
| 2129 | |
| 2130 | if (!tp) |
| 2131 | break; |
| 2132 | |
| 2133 | /* Only need to wait if two_primaries is enabled */ |
| 2134 | prepare_to_wait(&device->seq_wait, &wait, TASK_INTERRUPTIBLE); |
| 2135 | spin_unlock(&device->peer_seq_lock); |
| 2136 | rcu_read_lock(); |
| 2137 | timeout = rcu_dereference(peer_device->connection->net_conf)->ping_timeo*HZ/10; |
| 2138 | rcu_read_unlock(); |
| 2139 | timeout = schedule_timeout(timeout); |
| 2140 | spin_lock(&device->peer_seq_lock); |
| 2141 | if (!timeout) { |
| 2142 | ret = -ETIMEDOUT; |
| 2143 | drbd_err(device, "Timed out waiting for missing ack packets; disconnecting\n"); |
| 2144 | break; |
| 2145 | } |
| 2146 | } |
| 2147 | spin_unlock(&device->peer_seq_lock); |
| 2148 | finish_wait(&device->seq_wait, &wait); |
| 2149 | return ret; |
| 2150 | } |
| 2151 | |
| 2152 | /* see also bio_flags_to_wire() |
| 2153 | * DRBD_REQ_*, because we need to semantically map the flags to data packet |
| 2154 | * flags and back. We may replicate to other kernel versions. */ |
| 2155 | static unsigned long wire_flags_to_bio(u32 dpf) |
| 2156 | { |
| 2157 | return (dpf & DP_RW_SYNC ? REQ_SYNC : 0) | |
| 2158 | (dpf & DP_FUA ? REQ_FUA : 0) | |
| 2159 | (dpf & DP_FLUSH ? REQ_FLUSH : 0) | |
| 2160 | (dpf & DP_DISCARD ? REQ_DISCARD : 0); |
| 2161 | } |
| 2162 | |
| 2163 | static void fail_postponed_requests(struct drbd_device *device, sector_t sector, |
| 2164 | unsigned int size) |
| 2165 | { |
| 2166 | struct drbd_interval *i; |
| 2167 | |
| 2168 | repeat: |
| 2169 | drbd_for_each_overlap(i, &device->write_requests, sector, size) { |
| 2170 | struct drbd_request *req; |
| 2171 | struct bio_and_error m; |
| 2172 | |
| 2173 | if (!i->local) |
| 2174 | continue; |
| 2175 | req = container_of(i, struct drbd_request, i); |
| 2176 | if (!(req->rq_state & RQ_POSTPONED)) |
| 2177 | continue; |
| 2178 | req->rq_state &= ~RQ_POSTPONED; |
| 2179 | __req_mod(req, NEG_ACKED, &m); |
| 2180 | spin_unlock_irq(&device->resource->req_lock); |
| 2181 | if (m.bio) |
| 2182 | complete_master_bio(device, &m); |
| 2183 | spin_lock_irq(&device->resource->req_lock); |
| 2184 | goto repeat; |
| 2185 | } |
| 2186 | } |
| 2187 | |
| 2188 | static int handle_write_conflicts(struct drbd_device *device, |
| 2189 | struct drbd_peer_request *peer_req) |
| 2190 | { |
| 2191 | struct drbd_connection *connection = peer_req->peer_device->connection; |
| 2192 | bool resolve_conflicts = test_bit(RESOLVE_CONFLICTS, &connection->flags); |
| 2193 | sector_t sector = peer_req->i.sector; |
| 2194 | const unsigned int size = peer_req->i.size; |
| 2195 | struct drbd_interval *i; |
| 2196 | bool equal; |
| 2197 | int err; |
| 2198 | |
| 2199 | /* |
| 2200 | * Inserting the peer request into the write_requests tree will prevent |
| 2201 | * new conflicting local requests from being added. |
| 2202 | */ |
| 2203 | drbd_insert_interval(&device->write_requests, &peer_req->i); |
| 2204 | |
| 2205 | repeat: |
| 2206 | drbd_for_each_overlap(i, &device->write_requests, sector, size) { |
| 2207 | if (i == &peer_req->i) |
| 2208 | continue; |
| 2209 | if (i->completed) |
| 2210 | continue; |
| 2211 | |
| 2212 | if (!i->local) { |
| 2213 | /* |
| 2214 | * Our peer has sent a conflicting remote request; this |
| 2215 | * should not happen in a two-node setup. Wait for the |
| 2216 | * earlier peer request to complete. |
| 2217 | */ |
| 2218 | err = drbd_wait_misc(device, i); |
| 2219 | if (err) |
| 2220 | goto out; |
| 2221 | goto repeat; |
| 2222 | } |
| 2223 | |
| 2224 | equal = i->sector == sector && i->size == size; |
| 2225 | if (resolve_conflicts) { |
| 2226 | /* |
| 2227 | * If the peer request is fully contained within the |
| 2228 | * overlapping request, it can be considered overwritten |
| 2229 | * and thus superseded; otherwise, it will be retried |
| 2230 | * once all overlapping requests have completed. |
| 2231 | */ |
| 2232 | bool superseded = i->sector <= sector && i->sector + |
| 2233 | (i->size >> 9) >= sector + (size >> 9); |
| 2234 | |
| 2235 | if (!equal) |
| 2236 | drbd_alert(device, "Concurrent writes detected: " |
| 2237 | "local=%llus +%u, remote=%llus +%u, " |
| 2238 | "assuming %s came first\n", |
| 2239 | (unsigned long long)i->sector, i->size, |
| 2240 | (unsigned long long)sector, size, |
| 2241 | superseded ? "local" : "remote"); |
| 2242 | |
| 2243 | peer_req->w.cb = superseded ? e_send_superseded : |
| 2244 | e_send_retry_write; |
| 2245 | list_add_tail(&peer_req->w.list, &device->done_ee); |
| 2246 | queue_work(connection->ack_sender, &peer_req->peer_device->send_acks_work); |
| 2247 | |
| 2248 | err = -ENOENT; |
| 2249 | goto out; |
| 2250 | } else { |
| 2251 | struct drbd_request *req = |
| 2252 | container_of(i, struct drbd_request, i); |
| 2253 | |
| 2254 | if (!equal) |
| 2255 | drbd_alert(device, "Concurrent writes detected: " |
| 2256 | "local=%llus +%u, remote=%llus +%u\n", |
| 2257 | (unsigned long long)i->sector, i->size, |
| 2258 | (unsigned long long)sector, size); |
| 2259 | |
| 2260 | if (req->rq_state & RQ_LOCAL_PENDING || |
| 2261 | !(req->rq_state & RQ_POSTPONED)) { |
| 2262 | /* |
| 2263 | * Wait for the node with the discard flag to |
| 2264 | * decide if this request has been superseded |
| 2265 | * or needs to be retried. |
| 2266 | * Requests that have been superseded will |
| 2267 | * disappear from the write_requests tree. |
| 2268 | * |
| 2269 | * In addition, wait for the conflicting |
| 2270 | * request to finish locally before submitting |
| 2271 | * the conflicting peer request. |
| 2272 | */ |
| 2273 | err = drbd_wait_misc(device, &req->i); |
| 2274 | if (err) { |
| 2275 | _conn_request_state(connection, NS(conn, C_TIMEOUT), CS_HARD); |
| 2276 | fail_postponed_requests(device, sector, size); |
| 2277 | goto out; |
| 2278 | } |
| 2279 | goto repeat; |
| 2280 | } |
| 2281 | /* |
| 2282 | * Remember to restart the conflicting requests after |
| 2283 | * the new peer request has completed. |
| 2284 | */ |
| 2285 | peer_req->flags |= EE_RESTART_REQUESTS; |
| 2286 | } |
| 2287 | } |
| 2288 | err = 0; |
| 2289 | |
| 2290 | out: |
| 2291 | if (err) |
| 2292 | drbd_remove_epoch_entry_interval(device, peer_req); |
| 2293 | return err; |
| 2294 | } |
| 2295 | |
| 2296 | /* mirrored write */ |
| 2297 | static int receive_Data(struct drbd_connection *connection, struct packet_info *pi) |
| 2298 | { |
| 2299 | struct drbd_peer_device *peer_device; |
| 2300 | struct drbd_device *device; |
| 2301 | struct net_conf *nc; |
| 2302 | sector_t sector; |
| 2303 | struct drbd_peer_request *peer_req; |
| 2304 | struct p_data *p = pi->data; |
| 2305 | u32 peer_seq = be32_to_cpu(p->seq_num); |
| 2306 | int rw = WRITE; |
| 2307 | u32 dp_flags; |
| 2308 | int err, tp; |
| 2309 | |
| 2310 | peer_device = conn_peer_device(connection, pi->vnr); |
| 2311 | if (!peer_device) |
| 2312 | return -EIO; |
| 2313 | device = peer_device->device; |
| 2314 | |
| 2315 | if (!get_ldev(device)) { |
| 2316 | int err2; |
| 2317 | |
| 2318 | err = wait_for_and_update_peer_seq(peer_device, peer_seq); |
| 2319 | drbd_send_ack_dp(peer_device, P_NEG_ACK, p, pi->size); |
| 2320 | atomic_inc(&connection->current_epoch->epoch_size); |
| 2321 | err2 = drbd_drain_block(peer_device, pi->size); |
| 2322 | if (!err) |
| 2323 | err = err2; |
| 2324 | return err; |
| 2325 | } |
| 2326 | |
| 2327 | /* |
| 2328 | * Corresponding put_ldev done either below (on various errors), or in |
| 2329 | * drbd_peer_request_endio, if we successfully submit the data at the |
| 2330 | * end of this function. |
| 2331 | */ |
| 2332 | |
| 2333 | sector = be64_to_cpu(p->sector); |
| 2334 | peer_req = read_in_block(peer_device, p->block_id, sector, pi); |
| 2335 | if (!peer_req) { |
| 2336 | put_ldev(device); |
| 2337 | return -EIO; |
| 2338 | } |
| 2339 | |
| 2340 | peer_req->w.cb = e_end_block; |
| 2341 | peer_req->submit_jif = jiffies; |
| 2342 | peer_req->flags |= EE_APPLICATION; |
| 2343 | |
| 2344 | dp_flags = be32_to_cpu(p->dp_flags); |
| 2345 | rw |= wire_flags_to_bio(dp_flags); |
| 2346 | if (pi->cmd == P_TRIM) { |
| 2347 | struct request_queue *q = bdev_get_queue(device->ldev->backing_bdev); |
| 2348 | peer_req->flags |= EE_IS_TRIM; |
| 2349 | if (!blk_queue_discard(q)) |
| 2350 | peer_req->flags |= EE_IS_TRIM_USE_ZEROOUT; |
| 2351 | D_ASSERT(peer_device, peer_req->i.size > 0); |
| 2352 | D_ASSERT(peer_device, rw & REQ_DISCARD); |
| 2353 | D_ASSERT(peer_device, peer_req->pages == NULL); |
| 2354 | } else if (peer_req->pages == NULL) { |
| 2355 | D_ASSERT(device, peer_req->i.size == 0); |
| 2356 | D_ASSERT(device, dp_flags & DP_FLUSH); |
| 2357 | } |
| 2358 | |
| 2359 | if (dp_flags & DP_MAY_SET_IN_SYNC) |
| 2360 | peer_req->flags |= EE_MAY_SET_IN_SYNC; |
| 2361 | |
| 2362 | spin_lock(&connection->epoch_lock); |
| 2363 | peer_req->epoch = connection->current_epoch; |
| 2364 | atomic_inc(&peer_req->epoch->epoch_size); |
| 2365 | atomic_inc(&peer_req->epoch->active); |
| 2366 | spin_unlock(&connection->epoch_lock); |
| 2367 | |
| 2368 | rcu_read_lock(); |
| 2369 | nc = rcu_dereference(peer_device->connection->net_conf); |
| 2370 | tp = nc->two_primaries; |
| 2371 | if (peer_device->connection->agreed_pro_version < 100) { |
| 2372 | switch (nc->wire_protocol) { |
| 2373 | case DRBD_PROT_C: |
| 2374 | dp_flags |= DP_SEND_WRITE_ACK; |
| 2375 | break; |
| 2376 | case DRBD_PROT_B: |
| 2377 | dp_flags |= DP_SEND_RECEIVE_ACK; |
| 2378 | break; |
| 2379 | } |
| 2380 | } |
| 2381 | rcu_read_unlock(); |
| 2382 | |
| 2383 | if (dp_flags & DP_SEND_WRITE_ACK) { |
| 2384 | peer_req->flags |= EE_SEND_WRITE_ACK; |
| 2385 | inc_unacked(device); |
| 2386 | /* corresponding dec_unacked() in e_end_block() |
| 2387 | * respective _drbd_clear_done_ee */ |
| 2388 | } |
| 2389 | |
| 2390 | if (dp_flags & DP_SEND_RECEIVE_ACK) { |
| 2391 | /* I really don't like it that the receiver thread |
| 2392 | * sends on the msock, but anyways */ |
| 2393 | drbd_send_ack(peer_device, P_RECV_ACK, peer_req); |
| 2394 | } |
| 2395 | |
| 2396 | if (tp) { |
| 2397 | /* two primaries implies protocol C */ |
| 2398 | D_ASSERT(device, dp_flags & DP_SEND_WRITE_ACK); |
| 2399 | peer_req->flags |= EE_IN_INTERVAL_TREE; |
| 2400 | err = wait_for_and_update_peer_seq(peer_device, peer_seq); |
| 2401 | if (err) |
| 2402 | goto out_interrupted; |
| 2403 | spin_lock_irq(&device->resource->req_lock); |
| 2404 | err = handle_write_conflicts(device, peer_req); |
| 2405 | if (err) { |
| 2406 | spin_unlock_irq(&device->resource->req_lock); |
| 2407 | if (err == -ENOENT) { |
| 2408 | put_ldev(device); |
| 2409 | return 0; |
| 2410 | } |
| 2411 | goto out_interrupted; |
| 2412 | } |
| 2413 | } else { |
| 2414 | update_peer_seq(peer_device, peer_seq); |
| 2415 | spin_lock_irq(&device->resource->req_lock); |
| 2416 | } |
| 2417 | /* if we use the zeroout fallback code, we process synchronously |
| 2418 | * and we wait for all pending requests, respectively wait for |
| 2419 | * active_ee to become empty in drbd_submit_peer_request(); |
| 2420 | * better not add ourselves here. */ |
| 2421 | if ((peer_req->flags & EE_IS_TRIM_USE_ZEROOUT) == 0) |
| 2422 | list_add_tail(&peer_req->w.list, &device->active_ee); |
| 2423 | spin_unlock_irq(&device->resource->req_lock); |
| 2424 | |
| 2425 | if (device->state.conn == C_SYNC_TARGET) |
| 2426 | wait_event(device->ee_wait, !overlapping_resync_write(device, peer_req)); |
| 2427 | |
| 2428 | if (device->state.pdsk < D_INCONSISTENT) { |
| 2429 | /* In case we have the only disk of the cluster, */ |
| 2430 | drbd_set_out_of_sync(device, peer_req->i.sector, peer_req->i.size); |
| 2431 | peer_req->flags &= ~EE_MAY_SET_IN_SYNC; |
| 2432 | drbd_al_begin_io(device, &peer_req->i); |
| 2433 | peer_req->flags |= EE_CALL_AL_COMPLETE_IO; |
| 2434 | } |
| 2435 | |
| 2436 | err = drbd_submit_peer_request(device, peer_req, rw, DRBD_FAULT_DT_WR); |
| 2437 | if (!err) |
| 2438 | return 0; |
| 2439 | |
| 2440 | /* don't care for the reason here */ |
| 2441 | drbd_err(device, "submit failed, triggering re-connect\n"); |
| 2442 | spin_lock_irq(&device->resource->req_lock); |
| 2443 | list_del(&peer_req->w.list); |
| 2444 | drbd_remove_epoch_entry_interval(device, peer_req); |
| 2445 | spin_unlock_irq(&device->resource->req_lock); |
| 2446 | if (peer_req->flags & EE_CALL_AL_COMPLETE_IO) { |
| 2447 | peer_req->flags &= ~EE_CALL_AL_COMPLETE_IO; |
| 2448 | drbd_al_complete_io(device, &peer_req->i); |
| 2449 | } |
| 2450 | |
| 2451 | out_interrupted: |
| 2452 | drbd_may_finish_epoch(connection, peer_req->epoch, EV_PUT + EV_CLEANUP); |
| 2453 | put_ldev(device); |
| 2454 | drbd_free_peer_req(device, peer_req); |
| 2455 | return err; |
| 2456 | } |
| 2457 | |
| 2458 | /* We may throttle resync, if the lower device seems to be busy, |
| 2459 | * and current sync rate is above c_min_rate. |
| 2460 | * |
| 2461 | * To decide whether or not the lower device is busy, we use a scheme similar |
| 2462 | * to MD RAID is_mddev_idle(): if the partition stats reveal "significant" |
| 2463 | * (more than 64 sectors) of activity we cannot account for with our own resync |
| 2464 | * activity, it obviously is "busy". |
| 2465 | * |
| 2466 | * The current sync rate used here uses only the most recent two step marks, |
| 2467 | * to have a short time average so we can react faster. |
| 2468 | */ |
| 2469 | bool drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector, |
| 2470 | bool throttle_if_app_is_waiting) |
| 2471 | { |
| 2472 | struct lc_element *tmp; |
| 2473 | bool throttle = drbd_rs_c_min_rate_throttle(device); |
| 2474 | |
| 2475 | if (!throttle || throttle_if_app_is_waiting) |
| 2476 | return throttle; |
| 2477 | |
| 2478 | spin_lock_irq(&device->al_lock); |
| 2479 | tmp = lc_find(device->resync, BM_SECT_TO_EXT(sector)); |
| 2480 | if (tmp) { |
| 2481 | struct bm_extent *bm_ext = lc_entry(tmp, struct bm_extent, lce); |
| 2482 | if (test_bit(BME_PRIORITY, &bm_ext->flags)) |
| 2483 | throttle = false; |
| 2484 | /* Do not slow down if app IO is already waiting for this extent, |
| 2485 | * and our progress is necessary for application IO to complete. */ |
| 2486 | } |
| 2487 | spin_unlock_irq(&device->al_lock); |
| 2488 | |
| 2489 | return throttle; |
| 2490 | } |
| 2491 | |
| 2492 | bool drbd_rs_c_min_rate_throttle(struct drbd_device *device) |
| 2493 | { |
| 2494 | struct gendisk *disk = device->ldev->backing_bdev->bd_contains->bd_disk; |
| 2495 | unsigned long db, dt, dbdt; |
| 2496 | unsigned int c_min_rate; |
| 2497 | int curr_events; |
| 2498 | |
| 2499 | rcu_read_lock(); |
| 2500 | c_min_rate = rcu_dereference(device->ldev->disk_conf)->c_min_rate; |
| 2501 | rcu_read_unlock(); |
| 2502 | |
| 2503 | /* feature disabled? */ |
| 2504 | if (c_min_rate == 0) |
| 2505 | return false; |
| 2506 | |
| 2507 | curr_events = (int)part_stat_read(&disk->part0, sectors[0]) + |
| 2508 | (int)part_stat_read(&disk->part0, sectors[1]) - |
| 2509 | atomic_read(&device->rs_sect_ev); |
| 2510 | |
| 2511 | if (atomic_read(&device->ap_actlog_cnt) |
| 2512 | || curr_events - device->rs_last_events > 64) { |
| 2513 | unsigned long rs_left; |
| 2514 | int i; |
| 2515 | |
| 2516 | device->rs_last_events = curr_events; |
| 2517 | |
| 2518 | /* sync speed average over the last 2*DRBD_SYNC_MARK_STEP, |
| 2519 | * approx. */ |
| 2520 | i = (device->rs_last_mark + DRBD_SYNC_MARKS-1) % DRBD_SYNC_MARKS; |
| 2521 | |
| 2522 | if (device->state.conn == C_VERIFY_S || device->state.conn == C_VERIFY_T) |
| 2523 | rs_left = device->ov_left; |
| 2524 | else |
| 2525 | rs_left = drbd_bm_total_weight(device) - device->rs_failed; |
| 2526 | |
| 2527 | dt = ((long)jiffies - (long)device->rs_mark_time[i]) / HZ; |
| 2528 | if (!dt) |
| 2529 | dt++; |
| 2530 | db = device->rs_mark_left[i] - rs_left; |
| 2531 | dbdt = Bit2KB(db/dt); |
| 2532 | |
| 2533 | if (dbdt > c_min_rate) |
| 2534 | return true; |
| 2535 | } |
| 2536 | return false; |
| 2537 | } |
| 2538 | |
| 2539 | static int receive_DataRequest(struct drbd_connection *connection, struct packet_info *pi) |
| 2540 | { |
| 2541 | struct drbd_peer_device *peer_device; |
| 2542 | struct drbd_device *device; |
| 2543 | sector_t sector; |
| 2544 | sector_t capacity; |
| 2545 | struct drbd_peer_request *peer_req; |
| 2546 | struct digest_info *di = NULL; |
| 2547 | int size, verb; |
| 2548 | unsigned int fault_type; |
| 2549 | struct p_block_req *p = pi->data; |
| 2550 | |
| 2551 | peer_device = conn_peer_device(connection, pi->vnr); |
| 2552 | if (!peer_device) |
| 2553 | return -EIO; |
| 2554 | device = peer_device->device; |
| 2555 | capacity = drbd_get_capacity(device->this_bdev); |
| 2556 | |
| 2557 | sector = be64_to_cpu(p->sector); |
| 2558 | size = be32_to_cpu(p->blksize); |
| 2559 | |
| 2560 | if (size <= 0 || !IS_ALIGNED(size, 512) || size > DRBD_MAX_BIO_SIZE) { |
| 2561 | drbd_err(device, "%s:%d: sector: %llus, size: %u\n", __FILE__, __LINE__, |
| 2562 | (unsigned long long)sector, size); |
| 2563 | return -EINVAL; |
| 2564 | } |
| 2565 | if (sector + (size>>9) > capacity) { |
| 2566 | drbd_err(device, "%s:%d: sector: %llus, size: %u\n", __FILE__, __LINE__, |
| 2567 | (unsigned long long)sector, size); |
| 2568 | return -EINVAL; |
| 2569 | } |
| 2570 | |
| 2571 | if (!get_ldev_if_state(device, D_UP_TO_DATE)) { |
| 2572 | verb = 1; |
| 2573 | switch (pi->cmd) { |
| 2574 | case P_DATA_REQUEST: |
| 2575 | drbd_send_ack_rp(peer_device, P_NEG_DREPLY, p); |
| 2576 | break; |
| 2577 | case P_RS_DATA_REQUEST: |
| 2578 | case P_CSUM_RS_REQUEST: |
| 2579 | case P_OV_REQUEST: |
| 2580 | drbd_send_ack_rp(peer_device, P_NEG_RS_DREPLY , p); |
| 2581 | break; |
| 2582 | case P_OV_REPLY: |
| 2583 | verb = 0; |
| 2584 | dec_rs_pending(device); |
| 2585 | drbd_send_ack_ex(peer_device, P_OV_RESULT, sector, size, ID_IN_SYNC); |
| 2586 | break; |
| 2587 | default: |
| 2588 | BUG(); |
| 2589 | } |
| 2590 | if (verb && __ratelimit(&drbd_ratelimit_state)) |
| 2591 | drbd_err(device, "Can not satisfy peer's read request, " |
| 2592 | "no local data.\n"); |
| 2593 | |
| 2594 | /* drain possibly payload */ |
| 2595 | return drbd_drain_block(peer_device, pi->size); |
| 2596 | } |
| 2597 | |
| 2598 | /* GFP_NOIO, because we must not cause arbitrary write-out: in a DRBD |
| 2599 | * "criss-cross" setup, that might cause write-out on some other DRBD, |
| 2600 | * which in turn might block on the other node at this very place. */ |
| 2601 | peer_req = drbd_alloc_peer_req(peer_device, p->block_id, sector, size, |
| 2602 | true /* has real payload */, GFP_NOIO); |
| 2603 | if (!peer_req) { |
| 2604 | put_ldev(device); |
| 2605 | return -ENOMEM; |
| 2606 | } |
| 2607 | |
| 2608 | switch (pi->cmd) { |
| 2609 | case P_DATA_REQUEST: |
| 2610 | peer_req->w.cb = w_e_end_data_req; |
| 2611 | fault_type = DRBD_FAULT_DT_RD; |
| 2612 | /* application IO, don't drbd_rs_begin_io */ |
| 2613 | peer_req->flags |= EE_APPLICATION; |
| 2614 | goto submit; |
| 2615 | |
| 2616 | case P_RS_DATA_REQUEST: |
| 2617 | peer_req->w.cb = w_e_end_rsdata_req; |
| 2618 | fault_type = DRBD_FAULT_RS_RD; |
| 2619 | /* used in the sector offset progress display */ |
| 2620 | device->bm_resync_fo = BM_SECT_TO_BIT(sector); |
| 2621 | break; |
| 2622 | |
| 2623 | case P_OV_REPLY: |
| 2624 | case P_CSUM_RS_REQUEST: |
| 2625 | fault_type = DRBD_FAULT_RS_RD; |
| 2626 | di = kmalloc(sizeof(*di) + pi->size, GFP_NOIO); |
| 2627 | if (!di) |
| 2628 | goto out_free_e; |
| 2629 | |
| 2630 | di->digest_size = pi->size; |
| 2631 | di->digest = (((char *)di)+sizeof(struct digest_info)); |
| 2632 | |
| 2633 | peer_req->digest = di; |
| 2634 | peer_req->flags |= EE_HAS_DIGEST; |
| 2635 | |
| 2636 | if (drbd_recv_all(peer_device->connection, di->digest, pi->size)) |
| 2637 | goto out_free_e; |
| 2638 | |
| 2639 | if (pi->cmd == P_CSUM_RS_REQUEST) { |
| 2640 | D_ASSERT(device, peer_device->connection->agreed_pro_version >= 89); |
| 2641 | peer_req->w.cb = w_e_end_csum_rs_req; |
| 2642 | /* used in the sector offset progress display */ |
| 2643 | device->bm_resync_fo = BM_SECT_TO_BIT(sector); |
| 2644 | /* remember to report stats in drbd_resync_finished */ |
| 2645 | device->use_csums = true; |
| 2646 | } else if (pi->cmd == P_OV_REPLY) { |
| 2647 | /* track progress, we may need to throttle */ |
| 2648 | atomic_add(size >> 9, &device->rs_sect_in); |
| 2649 | peer_req->w.cb = w_e_end_ov_reply; |
| 2650 | dec_rs_pending(device); |
| 2651 | /* drbd_rs_begin_io done when we sent this request, |
| 2652 | * but accounting still needs to be done. */ |
| 2653 | goto submit_for_resync; |
| 2654 | } |
| 2655 | break; |
| 2656 | |
| 2657 | case P_OV_REQUEST: |
| 2658 | if (device->ov_start_sector == ~(sector_t)0 && |
| 2659 | peer_device->connection->agreed_pro_version >= 90) { |
| 2660 | unsigned long now = jiffies; |
| 2661 | int i; |
| 2662 | device->ov_start_sector = sector; |
| 2663 | device->ov_position = sector; |
| 2664 | device->ov_left = drbd_bm_bits(device) - BM_SECT_TO_BIT(sector); |
| 2665 | device->rs_total = device->ov_left; |
| 2666 | for (i = 0; i < DRBD_SYNC_MARKS; i++) { |
| 2667 | device->rs_mark_left[i] = device->ov_left; |
| 2668 | device->rs_mark_time[i] = now; |
| 2669 | } |
| 2670 | drbd_info(device, "Online Verify start sector: %llu\n", |
| 2671 | (unsigned long long)sector); |
| 2672 | } |
| 2673 | peer_req->w.cb = w_e_end_ov_req; |
| 2674 | fault_type = DRBD_FAULT_RS_RD; |
| 2675 | break; |
| 2676 | |
| 2677 | default: |
| 2678 | BUG(); |
| 2679 | } |
| 2680 | |
| 2681 | /* Throttle, drbd_rs_begin_io and submit should become asynchronous |
| 2682 | * wrt the receiver, but it is not as straightforward as it may seem. |
| 2683 | * Various places in the resync start and stop logic assume resync |
| 2684 | * requests are processed in order, requeuing this on the worker thread |
| 2685 | * introduces a bunch of new code for synchronization between threads. |
| 2686 | * |
| 2687 | * Unlimited throttling before drbd_rs_begin_io may stall the resync |
| 2688 | * "forever", throttling after drbd_rs_begin_io will lock that extent |
| 2689 | * for application writes for the same time. For now, just throttle |
| 2690 | * here, where the rest of the code expects the receiver to sleep for |
| 2691 | * a while, anyways. |
| 2692 | */ |
| 2693 | |
| 2694 | /* Throttle before drbd_rs_begin_io, as that locks out application IO; |
| 2695 | * this defers syncer requests for some time, before letting at least |
| 2696 | * on request through. The resync controller on the receiving side |
| 2697 | * will adapt to the incoming rate accordingly. |
| 2698 | * |
| 2699 | * We cannot throttle here if remote is Primary/SyncTarget: |
| 2700 | * we would also throttle its application reads. |
| 2701 | * In that case, throttling is done on the SyncTarget only. |
| 2702 | */ |
| 2703 | |
| 2704 | /* Even though this may be a resync request, we do add to "read_ee"; |
| 2705 | * "sync_ee" is only used for resync WRITEs. |
| 2706 | * Add to list early, so debugfs can find this request |
| 2707 | * even if we have to sleep below. */ |
| 2708 | spin_lock_irq(&device->resource->req_lock); |
| 2709 | list_add_tail(&peer_req->w.list, &device->read_ee); |
| 2710 | spin_unlock_irq(&device->resource->req_lock); |
| 2711 | |
| 2712 | update_receiver_timing_details(connection, drbd_rs_should_slow_down); |
| 2713 | if (device->state.peer != R_PRIMARY |
| 2714 | && drbd_rs_should_slow_down(device, sector, false)) |
| 2715 | schedule_timeout_uninterruptible(HZ/10); |
| 2716 | update_receiver_timing_details(connection, drbd_rs_begin_io); |
| 2717 | if (drbd_rs_begin_io(device, sector)) |
| 2718 | goto out_free_e; |
| 2719 | |
| 2720 | submit_for_resync: |
| 2721 | atomic_add(size >> 9, &device->rs_sect_ev); |
| 2722 | |
| 2723 | submit: |
| 2724 | update_receiver_timing_details(connection, drbd_submit_peer_request); |
| 2725 | inc_unacked(device); |
| 2726 | if (drbd_submit_peer_request(device, peer_req, READ, fault_type) == 0) |
| 2727 | return 0; |
| 2728 | |
| 2729 | /* don't care for the reason here */ |
| 2730 | drbd_err(device, "submit failed, triggering re-connect\n"); |
| 2731 | |
| 2732 | out_free_e: |
| 2733 | spin_lock_irq(&device->resource->req_lock); |
| 2734 | list_del(&peer_req->w.list); |
| 2735 | spin_unlock_irq(&device->resource->req_lock); |
| 2736 | /* no drbd_rs_complete_io(), we are dropping the connection anyways */ |
| 2737 | |
| 2738 | put_ldev(device); |
| 2739 | drbd_free_peer_req(device, peer_req); |
| 2740 | return -EIO; |
| 2741 | } |
| 2742 | |
| 2743 | /** |
| 2744 | * drbd_asb_recover_0p - Recover after split-brain with no remaining primaries |
| 2745 | */ |
| 2746 | static int drbd_asb_recover_0p(struct drbd_peer_device *peer_device) __must_hold(local) |
| 2747 | { |
| 2748 | struct drbd_device *device = peer_device->device; |
| 2749 | int self, peer, rv = -100; |
| 2750 | unsigned long ch_self, ch_peer; |
| 2751 | enum drbd_after_sb_p after_sb_0p; |
| 2752 | |
| 2753 | self = device->ldev->md.uuid[UI_BITMAP] & 1; |
| 2754 | peer = device->p_uuid[UI_BITMAP] & 1; |
| 2755 | |
| 2756 | ch_peer = device->p_uuid[UI_SIZE]; |
| 2757 | ch_self = device->comm_bm_set; |
| 2758 | |
| 2759 | rcu_read_lock(); |
| 2760 | after_sb_0p = rcu_dereference(peer_device->connection->net_conf)->after_sb_0p; |
| 2761 | rcu_read_unlock(); |
| 2762 | switch (after_sb_0p) { |
| 2763 | case ASB_CONSENSUS: |
| 2764 | case ASB_DISCARD_SECONDARY: |
| 2765 | case ASB_CALL_HELPER: |
| 2766 | case ASB_VIOLENTLY: |
| 2767 | drbd_err(device, "Configuration error.\n"); |
| 2768 | break; |
| 2769 | case ASB_DISCONNECT: |
| 2770 | break; |
| 2771 | case ASB_DISCARD_YOUNGER_PRI: |
| 2772 | if (self == 0 && peer == 1) { |
| 2773 | rv = -1; |
| 2774 | break; |
| 2775 | } |
| 2776 | if (self == 1 && peer == 0) { |
| 2777 | rv = 1; |
| 2778 | break; |
| 2779 | } |
| 2780 | /* Else fall through to one of the other strategies... */ |
| 2781 | case ASB_DISCARD_OLDER_PRI: |
| 2782 | if (self == 0 && peer == 1) { |
| 2783 | rv = 1; |
| 2784 | break; |
| 2785 | } |
| 2786 | if (self == 1 && peer == 0) { |
| 2787 | rv = -1; |
| 2788 | break; |
| 2789 | } |
| 2790 | /* Else fall through to one of the other strategies... */ |
| 2791 | drbd_warn(device, "Discard younger/older primary did not find a decision\n" |
| 2792 | "Using discard-least-changes instead\n"); |
| 2793 | case ASB_DISCARD_ZERO_CHG: |
| 2794 | if (ch_peer == 0 && ch_self == 0) { |
| 2795 | rv = test_bit(RESOLVE_CONFLICTS, &peer_device->connection->flags) |
| 2796 | ? -1 : 1; |
| 2797 | break; |
| 2798 | } else { |
| 2799 | if (ch_peer == 0) { rv = 1; break; } |
| 2800 | if (ch_self == 0) { rv = -1; break; } |
| 2801 | } |
| 2802 | if (after_sb_0p == ASB_DISCARD_ZERO_CHG) |
| 2803 | break; |
| 2804 | case ASB_DISCARD_LEAST_CHG: |
| 2805 | if (ch_self < ch_peer) |
| 2806 | rv = -1; |
| 2807 | else if (ch_self > ch_peer) |
| 2808 | rv = 1; |
| 2809 | else /* ( ch_self == ch_peer ) */ |
| 2810 | /* Well, then use something else. */ |
| 2811 | rv = test_bit(RESOLVE_CONFLICTS, &peer_device->connection->flags) |
| 2812 | ? -1 : 1; |
| 2813 | break; |
| 2814 | case ASB_DISCARD_LOCAL: |
| 2815 | rv = -1; |
| 2816 | break; |
| 2817 | case ASB_DISCARD_REMOTE: |
| 2818 | rv = 1; |
| 2819 | } |
| 2820 | |
| 2821 | return rv; |
| 2822 | } |
| 2823 | |
| 2824 | /** |
| 2825 | * drbd_asb_recover_1p - Recover after split-brain with one remaining primary |
| 2826 | */ |
| 2827 | static int drbd_asb_recover_1p(struct drbd_peer_device *peer_device) __must_hold(local) |
| 2828 | { |
| 2829 | struct drbd_device *device = peer_device->device; |
| 2830 | int hg, rv = -100; |
| 2831 | enum drbd_after_sb_p after_sb_1p; |
| 2832 | |
| 2833 | rcu_read_lock(); |
| 2834 | after_sb_1p = rcu_dereference(peer_device->connection->net_conf)->after_sb_1p; |
| 2835 | rcu_read_unlock(); |
| 2836 | switch (after_sb_1p) { |
| 2837 | case ASB_DISCARD_YOUNGER_PRI: |
| 2838 | case ASB_DISCARD_OLDER_PRI: |
| 2839 | case ASB_DISCARD_LEAST_CHG: |
| 2840 | case ASB_DISCARD_LOCAL: |
| 2841 | case ASB_DISCARD_REMOTE: |
| 2842 | case ASB_DISCARD_ZERO_CHG: |
| 2843 | drbd_err(device, "Configuration error.\n"); |
| 2844 | break; |
| 2845 | case ASB_DISCONNECT: |
| 2846 | break; |
| 2847 | case ASB_CONSENSUS: |
| 2848 | hg = drbd_asb_recover_0p(peer_device); |
| 2849 | if (hg == -1 && device->state.role == R_SECONDARY) |
| 2850 | rv = hg; |
| 2851 | if (hg == 1 && device->state.role == R_PRIMARY) |
| 2852 | rv = hg; |
| 2853 | break; |
| 2854 | case ASB_VIOLENTLY: |
| 2855 | rv = drbd_asb_recover_0p(peer_device); |
| 2856 | break; |
| 2857 | case ASB_DISCARD_SECONDARY: |
| 2858 | return device->state.role == R_PRIMARY ? 1 : -1; |
| 2859 | case ASB_CALL_HELPER: |
| 2860 | hg = drbd_asb_recover_0p(peer_device); |
| 2861 | if (hg == -1 && device->state.role == R_PRIMARY) { |
| 2862 | enum drbd_state_rv rv2; |
| 2863 | |
| 2864 | /* drbd_change_state() does not sleep while in SS_IN_TRANSIENT_STATE, |
| 2865 | * we might be here in C_WF_REPORT_PARAMS which is transient. |
| 2866 | * we do not need to wait for the after state change work either. */ |
| 2867 | rv2 = drbd_change_state(device, CS_VERBOSE, NS(role, R_SECONDARY)); |
| 2868 | if (rv2 != SS_SUCCESS) { |
| 2869 | drbd_khelper(device, "pri-lost-after-sb"); |
| 2870 | } else { |
| 2871 | drbd_warn(device, "Successfully gave up primary role.\n"); |
| 2872 | rv = hg; |
| 2873 | } |
| 2874 | } else |
| 2875 | rv = hg; |
| 2876 | } |
| 2877 | |
| 2878 | return rv; |
| 2879 | } |
| 2880 | |
| 2881 | /** |
| 2882 | * drbd_asb_recover_2p - Recover after split-brain with two remaining primaries |
| 2883 | */ |
| 2884 | static int drbd_asb_recover_2p(struct drbd_peer_device *peer_device) __must_hold(local) |
| 2885 | { |
| 2886 | struct drbd_device *device = peer_device->device; |
| 2887 | int hg, rv = -100; |
| 2888 | enum drbd_after_sb_p after_sb_2p; |
| 2889 | |
| 2890 | rcu_read_lock(); |
| 2891 | after_sb_2p = rcu_dereference(peer_device->connection->net_conf)->after_sb_2p; |
| 2892 | rcu_read_unlock(); |
| 2893 | switch (after_sb_2p) { |
| 2894 | case ASB_DISCARD_YOUNGER_PRI: |
| 2895 | case ASB_DISCARD_OLDER_PRI: |
| 2896 | case ASB_DISCARD_LEAST_CHG: |
| 2897 | case ASB_DISCARD_LOCAL: |
| 2898 | case ASB_DISCARD_REMOTE: |
| 2899 | case ASB_CONSENSUS: |
| 2900 | case ASB_DISCARD_SECONDARY: |
| 2901 | case ASB_DISCARD_ZERO_CHG: |
| 2902 | drbd_err(device, "Configuration error.\n"); |
| 2903 | break; |
| 2904 | case ASB_VIOLENTLY: |
| 2905 | rv = drbd_asb_recover_0p(peer_device); |
| 2906 | break; |
| 2907 | case ASB_DISCONNECT: |
| 2908 | break; |
| 2909 | case ASB_CALL_HELPER: |
| 2910 | hg = drbd_asb_recover_0p(peer_device); |
| 2911 | if (hg == -1) { |
| 2912 | enum drbd_state_rv rv2; |
| 2913 | |
| 2914 | /* drbd_change_state() does not sleep while in SS_IN_TRANSIENT_STATE, |
| 2915 | * we might be here in C_WF_REPORT_PARAMS which is transient. |
| 2916 | * we do not need to wait for the after state change work either. */ |
| 2917 | rv2 = drbd_change_state(device, CS_VERBOSE, NS(role, R_SECONDARY)); |
| 2918 | if (rv2 != SS_SUCCESS) { |
| 2919 | drbd_khelper(device, "pri-lost-after-sb"); |
| 2920 | } else { |
| 2921 | drbd_warn(device, "Successfully gave up primary role.\n"); |
| 2922 | rv = hg; |
| 2923 | } |
| 2924 | } else |
| 2925 | rv = hg; |
| 2926 | } |
| 2927 | |
| 2928 | return rv; |
| 2929 | } |
| 2930 | |
| 2931 | static void drbd_uuid_dump(struct drbd_device *device, char *text, u64 *uuid, |
| 2932 | u64 bits, u64 flags) |
| 2933 | { |
| 2934 | if (!uuid) { |
| 2935 | drbd_info(device, "%s uuid info vanished while I was looking!\n", text); |
| 2936 | return; |
| 2937 | } |
| 2938 | drbd_info(device, "%s %016llX:%016llX:%016llX:%016llX bits:%llu flags:%llX\n", |
| 2939 | text, |
| 2940 | (unsigned long long)uuid[UI_CURRENT], |
| 2941 | (unsigned long long)uuid[UI_BITMAP], |
| 2942 | (unsigned long long)uuid[UI_HISTORY_START], |
| 2943 | (unsigned long long)uuid[UI_HISTORY_END], |
| 2944 | (unsigned long long)bits, |
| 2945 | (unsigned long long)flags); |
| 2946 | } |
| 2947 | |
| 2948 | /* |
| 2949 | 100 after split brain try auto recover |
| 2950 | 2 C_SYNC_SOURCE set BitMap |
| 2951 | 1 C_SYNC_SOURCE use BitMap |
| 2952 | 0 no Sync |
| 2953 | -1 C_SYNC_TARGET use BitMap |
| 2954 | -2 C_SYNC_TARGET set BitMap |
| 2955 | -100 after split brain, disconnect |
| 2956 | -1000 unrelated data |
| 2957 | -1091 requires proto 91 |
| 2958 | -1096 requires proto 96 |
| 2959 | */ |
| 2960 | static int drbd_uuid_compare(struct drbd_device *const device, int *rule_nr) __must_hold(local) |
| 2961 | { |
| 2962 | struct drbd_peer_device *const peer_device = first_peer_device(device); |
| 2963 | struct drbd_connection *const connection = peer_device ? peer_device->connection : NULL; |
| 2964 | u64 self, peer; |
| 2965 | int i, j; |
| 2966 | |
| 2967 | self = device->ldev->md.uuid[UI_CURRENT] & ~((u64)1); |
| 2968 | peer = device->p_uuid[UI_CURRENT] & ~((u64)1); |
| 2969 | |
| 2970 | *rule_nr = 10; |
| 2971 | if (self == UUID_JUST_CREATED && peer == UUID_JUST_CREATED) |
| 2972 | return 0; |
| 2973 | |
| 2974 | *rule_nr = 20; |
| 2975 | if ((self == UUID_JUST_CREATED || self == (u64)0) && |
| 2976 | peer != UUID_JUST_CREATED) |
| 2977 | return -2; |
| 2978 | |
| 2979 | *rule_nr = 30; |
| 2980 | if (self != UUID_JUST_CREATED && |
| 2981 | (peer == UUID_JUST_CREATED || peer == (u64)0)) |
| 2982 | return 2; |
| 2983 | |
| 2984 | if (self == peer) { |
| 2985 | int rct, dc; /* roles at crash time */ |
| 2986 | |
| 2987 | if (device->p_uuid[UI_BITMAP] == (u64)0 && device->ldev->md.uuid[UI_BITMAP] != (u64)0) { |
| 2988 | |
| 2989 | if (connection->agreed_pro_version < 91) |
| 2990 | return -1091; |
| 2991 | |
| 2992 | if ((device->ldev->md.uuid[UI_BITMAP] & ~((u64)1)) == (device->p_uuid[UI_HISTORY_START] & ~((u64)1)) && |
| 2993 | (device->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1)) == (device->p_uuid[UI_HISTORY_START + 1] & ~((u64)1))) { |
| 2994 | drbd_info(device, "was SyncSource, missed the resync finished event, corrected myself:\n"); |
| 2995 | drbd_uuid_move_history(device); |
| 2996 | device->ldev->md.uuid[UI_HISTORY_START] = device->ldev->md.uuid[UI_BITMAP]; |
| 2997 | device->ldev->md.uuid[UI_BITMAP] = 0; |
| 2998 | |
| 2999 | drbd_uuid_dump(device, "self", device->ldev->md.uuid, |
| 3000 | device->state.disk >= D_NEGOTIATING ? drbd_bm_total_weight(device) : 0, 0); |
| 3001 | *rule_nr = 34; |
| 3002 | } else { |
| 3003 | drbd_info(device, "was SyncSource (peer failed to write sync_uuid)\n"); |
| 3004 | *rule_nr = 36; |
| 3005 | } |
| 3006 | |
| 3007 | return 1; |
| 3008 | } |
| 3009 | |
| 3010 | if (device->ldev->md.uuid[UI_BITMAP] == (u64)0 && device->p_uuid[UI_BITMAP] != (u64)0) { |
| 3011 | |
| 3012 | if (connection->agreed_pro_version < 91) |
| 3013 | return -1091; |
| 3014 | |
| 3015 | if ((device->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1)) == (device->p_uuid[UI_BITMAP] & ~((u64)1)) && |
| 3016 | (device->ldev->md.uuid[UI_HISTORY_START + 1] & ~((u64)1)) == (device->p_uuid[UI_HISTORY_START] & ~((u64)1))) { |
| 3017 | drbd_info(device, "was SyncTarget, peer missed the resync finished event, corrected peer:\n"); |
| 3018 | |
| 3019 | device->p_uuid[UI_HISTORY_START + 1] = device->p_uuid[UI_HISTORY_START]; |
| 3020 | device->p_uuid[UI_HISTORY_START] = device->p_uuid[UI_BITMAP]; |
| 3021 | device->p_uuid[UI_BITMAP] = 0UL; |
| 3022 | |
| 3023 | drbd_uuid_dump(device, "peer", device->p_uuid, device->p_uuid[UI_SIZE], device->p_uuid[UI_FLAGS]); |
| 3024 | *rule_nr = 35; |
| 3025 | } else { |
| 3026 | drbd_info(device, "was SyncTarget (failed to write sync_uuid)\n"); |
| 3027 | *rule_nr = 37; |
| 3028 | } |
| 3029 | |
| 3030 | return -1; |
| 3031 | } |
| 3032 | |
| 3033 | /* Common power [off|failure] */ |
| 3034 | rct = (test_bit(CRASHED_PRIMARY, &device->flags) ? 1 : 0) + |
| 3035 | (device->p_uuid[UI_FLAGS] & 2); |
| 3036 | /* lowest bit is set when we were primary, |
| 3037 | * next bit (weight 2) is set when peer was primary */ |
| 3038 | *rule_nr = 40; |
| 3039 | |
| 3040 | switch (rct) { |
| 3041 | case 0: /* !self_pri && !peer_pri */ return 0; |
| 3042 | case 1: /* self_pri && !peer_pri */ return 1; |
| 3043 | case 2: /* !self_pri && peer_pri */ return -1; |
| 3044 | case 3: /* self_pri && peer_pri */ |
| 3045 | dc = test_bit(RESOLVE_CONFLICTS, &connection->flags); |
| 3046 | return dc ? -1 : 1; |
| 3047 | } |
| 3048 | } |
| 3049 | |
| 3050 | *rule_nr = 50; |
| 3051 | peer = device->p_uuid[UI_BITMAP] & ~((u64)1); |
| 3052 | if (self == peer) |
| 3053 | return -1; |
| 3054 | |
| 3055 | *rule_nr = 51; |
| 3056 | peer = device->p_uuid[UI_HISTORY_START] & ~((u64)1); |
| 3057 | if (self == peer) { |
| 3058 | if (connection->agreed_pro_version < 96 ? |
| 3059 | (device->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1)) == |
| 3060 | (device->p_uuid[UI_HISTORY_START + 1] & ~((u64)1)) : |
| 3061 | peer + UUID_NEW_BM_OFFSET == (device->p_uuid[UI_BITMAP] & ~((u64)1))) { |
| 3062 | /* The last P_SYNC_UUID did not get though. Undo the last start of |
| 3063 | resync as sync source modifications of the peer's UUIDs. */ |
| 3064 | |
| 3065 | if (connection->agreed_pro_version < 91) |
| 3066 | return -1091; |
| 3067 | |
| 3068 | device->p_uuid[UI_BITMAP] = device->p_uuid[UI_HISTORY_START]; |
| 3069 | device->p_uuid[UI_HISTORY_START] = device->p_uuid[UI_HISTORY_START + 1]; |
| 3070 | |
| 3071 | drbd_info(device, "Lost last syncUUID packet, corrected:\n"); |
| 3072 | drbd_uuid_dump(device, "peer", device->p_uuid, device->p_uuid[UI_SIZE], device->p_uuid[UI_FLAGS]); |
| 3073 | |
| 3074 | return -1; |
| 3075 | } |
| 3076 | } |
| 3077 | |
| 3078 | *rule_nr = 60; |
| 3079 | self = device->ldev->md.uuid[UI_CURRENT] & ~((u64)1); |
| 3080 | for (i = UI_HISTORY_START; i <= UI_HISTORY_END; i++) { |
| 3081 | peer = device->p_uuid[i] & ~((u64)1); |
| 3082 | if (self == peer) |
| 3083 | return -2; |
| 3084 | } |
| 3085 | |
| 3086 | *rule_nr = 70; |
| 3087 | self = device->ldev->md.uuid[UI_BITMAP] & ~((u64)1); |
| 3088 | peer = device->p_uuid[UI_CURRENT] & ~((u64)1); |
| 3089 | if (self == peer) |
| 3090 | return 1; |
| 3091 | |
| 3092 | *rule_nr = 71; |
| 3093 | self = device->ldev->md.uuid[UI_HISTORY_START] & ~((u64)1); |
| 3094 | if (self == peer) { |
| 3095 | if (connection->agreed_pro_version < 96 ? |
| 3096 | (device->ldev->md.uuid[UI_HISTORY_START + 1] & ~((u64)1)) == |
| 3097 | (device->p_uuid[UI_HISTORY_START] & ~((u64)1)) : |
| 3098 | self + UUID_NEW_BM_OFFSET == (device->ldev->md.uuid[UI_BITMAP] & ~((u64)1))) { |
| 3099 | /* The last P_SYNC_UUID did not get though. Undo the last start of |
| 3100 | resync as sync source modifications of our UUIDs. */ |
| 3101 | |
| 3102 | if (connection->agreed_pro_version < 91) |
| 3103 | return -1091; |
| 3104 | |
| 3105 | __drbd_uuid_set(device, UI_BITMAP, device->ldev->md.uuid[UI_HISTORY_START]); |
| 3106 | __drbd_uuid_set(device, UI_HISTORY_START, device->ldev->md.uuid[UI_HISTORY_START + 1]); |
| 3107 | |
| 3108 | drbd_info(device, "Last syncUUID did not get through, corrected:\n"); |
| 3109 | drbd_uuid_dump(device, "self", device->ldev->md.uuid, |
| 3110 | device->state.disk >= D_NEGOTIATING ? drbd_bm_total_weight(device) : 0, 0); |
| 3111 | |
| 3112 | return 1; |
| 3113 | } |
| 3114 | } |
| 3115 | |
| 3116 | |
| 3117 | *rule_nr = 80; |
| 3118 | peer = device->p_uuid[UI_CURRENT] & ~((u64)1); |
| 3119 | for (i = UI_HISTORY_START; i <= UI_HISTORY_END; i++) { |
| 3120 | self = device->ldev->md.uuid[i] & ~((u64)1); |
| 3121 | if (self == peer) |
| 3122 | return 2; |
| 3123 | } |
| 3124 | |
| 3125 | *rule_nr = 90; |
| 3126 | self = device->ldev->md.uuid[UI_BITMAP] & ~((u64)1); |
| 3127 | peer = device->p_uuid[UI_BITMAP] & ~((u64)1); |
| 3128 | if (self == peer && self != ((u64)0)) |
| 3129 | return 100; |
| 3130 | |
| 3131 | *rule_nr = 100; |
| 3132 | for (i = UI_HISTORY_START; i <= UI_HISTORY_END; i++) { |
| 3133 | self = device->ldev->md.uuid[i] & ~((u64)1); |
| 3134 | for (j = UI_HISTORY_START; j <= UI_HISTORY_END; j++) { |
| 3135 | peer = device->p_uuid[j] & ~((u64)1); |
| 3136 | if (self == peer) |
| 3137 | return -100; |
| 3138 | } |
| 3139 | } |
| 3140 | |
| 3141 | return -1000; |
| 3142 | } |
| 3143 | |
| 3144 | /* drbd_sync_handshake() returns the new conn state on success, or |
| 3145 | CONN_MASK (-1) on failure. |
| 3146 | */ |
| 3147 | static enum drbd_conns drbd_sync_handshake(struct drbd_peer_device *peer_device, |
| 3148 | enum drbd_role peer_role, |
| 3149 | enum drbd_disk_state peer_disk) __must_hold(local) |
| 3150 | { |
| 3151 | struct drbd_device *device = peer_device->device; |
| 3152 | enum drbd_conns rv = C_MASK; |
| 3153 | enum drbd_disk_state mydisk; |
| 3154 | struct net_conf *nc; |
| 3155 | int hg, rule_nr, rr_conflict, tentative; |
| 3156 | |
| 3157 | mydisk = device->state.disk; |
| 3158 | if (mydisk == D_NEGOTIATING) |
| 3159 | mydisk = device->new_state_tmp.disk; |
| 3160 | |
| 3161 | drbd_info(device, "drbd_sync_handshake:\n"); |
| 3162 | |
| 3163 | spin_lock_irq(&device->ldev->md.uuid_lock); |
| 3164 | drbd_uuid_dump(device, "self", device->ldev->md.uuid, device->comm_bm_set, 0); |
| 3165 | drbd_uuid_dump(device, "peer", device->p_uuid, |
| 3166 | device->p_uuid[UI_SIZE], device->p_uuid[UI_FLAGS]); |
| 3167 | |
| 3168 | hg = drbd_uuid_compare(device, &rule_nr); |
| 3169 | spin_unlock_irq(&device->ldev->md.uuid_lock); |
| 3170 | |
| 3171 | drbd_info(device, "uuid_compare()=%d by rule %d\n", hg, rule_nr); |
| 3172 | |
| 3173 | if (hg == -1000) { |
| 3174 | drbd_alert(device, "Unrelated data, aborting!\n"); |
| 3175 | return C_MASK; |
| 3176 | } |
| 3177 | if (hg < -1000) { |
| 3178 | drbd_alert(device, "To resolve this both sides have to support at least protocol %d\n", -hg - 1000); |
| 3179 | return C_MASK; |
| 3180 | } |
| 3181 | |
| 3182 | if ((mydisk == D_INCONSISTENT && peer_disk > D_INCONSISTENT) || |
| 3183 | (peer_disk == D_INCONSISTENT && mydisk > D_INCONSISTENT)) { |
| 3184 | int f = (hg == -100) || abs(hg) == 2; |
| 3185 | hg = mydisk > D_INCONSISTENT ? 1 : -1; |
| 3186 | if (f) |
| 3187 | hg = hg*2; |
| 3188 | drbd_info(device, "Becoming sync %s due to disk states.\n", |
| 3189 | hg > 0 ? "source" : "target"); |
| 3190 | } |
| 3191 | |
| 3192 | if (abs(hg) == 100) |
| 3193 | drbd_khelper(device, "initial-split-brain"); |
| 3194 | |
| 3195 | rcu_read_lock(); |
| 3196 | nc = rcu_dereference(peer_device->connection->net_conf); |
| 3197 | |
| 3198 | if (hg == 100 || (hg == -100 && nc->always_asbp)) { |
| 3199 | int pcount = (device->state.role == R_PRIMARY) |
| 3200 | + (peer_role == R_PRIMARY); |
| 3201 | int forced = (hg == -100); |
| 3202 | |
| 3203 | switch (pcount) { |
| 3204 | case 0: |
| 3205 | hg = drbd_asb_recover_0p(peer_device); |
| 3206 | break; |
| 3207 | case 1: |
| 3208 | hg = drbd_asb_recover_1p(peer_device); |
| 3209 | break; |
| 3210 | case 2: |
| 3211 | hg = drbd_asb_recover_2p(peer_device); |
| 3212 | break; |
| 3213 | } |
| 3214 | if (abs(hg) < 100) { |
| 3215 | drbd_warn(device, "Split-Brain detected, %d primaries, " |
| 3216 | "automatically solved. Sync from %s node\n", |
| 3217 | pcount, (hg < 0) ? "peer" : "this"); |
| 3218 | if (forced) { |
| 3219 | drbd_warn(device, "Doing a full sync, since" |
| 3220 | " UUIDs where ambiguous.\n"); |
| 3221 | hg = hg*2; |
| 3222 | } |
| 3223 | } |
| 3224 | } |
| 3225 | |
| 3226 | if (hg == -100) { |
| 3227 | if (test_bit(DISCARD_MY_DATA, &device->flags) && !(device->p_uuid[UI_FLAGS]&1)) |
| 3228 | hg = -1; |
| 3229 | if (!test_bit(DISCARD_MY_DATA, &device->flags) && (device->p_uuid[UI_FLAGS]&1)) |
| 3230 | hg = 1; |
| 3231 | |
| 3232 | if (abs(hg) < 100) |
| 3233 | drbd_warn(device, "Split-Brain detected, manually solved. " |
| 3234 | "Sync from %s node\n", |
| 3235 | (hg < 0) ? "peer" : "this"); |
| 3236 | } |
| 3237 | rr_conflict = nc->rr_conflict; |
| 3238 | tentative = nc->tentative; |
| 3239 | rcu_read_unlock(); |
| 3240 | |
| 3241 | if (hg == -100) { |
| 3242 | /* FIXME this log message is not correct if we end up here |
| 3243 | * after an attempted attach on a diskless node. |
| 3244 | * We just refuse to attach -- well, we drop the "connection" |
| 3245 | * to that disk, in a way... */ |
| 3246 | drbd_alert(device, "Split-Brain detected but unresolved, dropping connection!\n"); |
| 3247 | drbd_khelper(device, "split-brain"); |
| 3248 | return C_MASK; |
| 3249 | } |
| 3250 | |
| 3251 | if (hg > 0 && mydisk <= D_INCONSISTENT) { |
| 3252 | drbd_err(device, "I shall become SyncSource, but I am inconsistent!\n"); |
| 3253 | return C_MASK; |
| 3254 | } |
| 3255 | |
| 3256 | if (hg < 0 && /* by intention we do not use mydisk here. */ |
| 3257 | device->state.role == R_PRIMARY && device->state.disk >= D_CONSISTENT) { |
| 3258 | switch (rr_conflict) { |
| 3259 | case ASB_CALL_HELPER: |
| 3260 | drbd_khelper(device, "pri-lost"); |
| 3261 | /* fall through */ |
| 3262 | case ASB_DISCONNECT: |
| 3263 | drbd_err(device, "I shall become SyncTarget, but I am primary!\n"); |
| 3264 | return C_MASK; |
| 3265 | case ASB_VIOLENTLY: |
| 3266 | drbd_warn(device, "Becoming SyncTarget, violating the stable-data" |
| 3267 | "assumption\n"); |
| 3268 | } |
| 3269 | } |
| 3270 | |
| 3271 | if (tentative || test_bit(CONN_DRY_RUN, &peer_device->connection->flags)) { |
| 3272 | if (hg == 0) |
| 3273 | drbd_info(device, "dry-run connect: No resync, would become Connected immediately.\n"); |
| 3274 | else |
| 3275 | drbd_info(device, "dry-run connect: Would become %s, doing a %s resync.", |
| 3276 | drbd_conn_str(hg > 0 ? C_SYNC_SOURCE : C_SYNC_TARGET), |
| 3277 | abs(hg) >= 2 ? "full" : "bit-map based"); |
| 3278 | return C_MASK; |
| 3279 | } |
| 3280 | |
| 3281 | if (abs(hg) >= 2) { |
| 3282 | drbd_info(device, "Writing the whole bitmap, full sync required after drbd_sync_handshake.\n"); |
| 3283 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, "set_n_write from sync_handshake", |
| 3284 | BM_LOCKED_SET_ALLOWED)) |
| 3285 | return C_MASK; |
| 3286 | } |
| 3287 | |
| 3288 | if (hg > 0) { /* become sync source. */ |
| 3289 | rv = C_WF_BITMAP_S; |
| 3290 | } else if (hg < 0) { /* become sync target */ |
| 3291 | rv = C_WF_BITMAP_T; |
| 3292 | } else { |
| 3293 | rv = C_CONNECTED; |
| 3294 | if (drbd_bm_total_weight(device)) { |
| 3295 | drbd_info(device, "No resync, but %lu bits in bitmap!\n", |
| 3296 | drbd_bm_total_weight(device)); |
| 3297 | } |
| 3298 | } |
| 3299 | |
| 3300 | return rv; |
| 3301 | } |
| 3302 | |
| 3303 | static enum drbd_after_sb_p convert_after_sb(enum drbd_after_sb_p peer) |
| 3304 | { |
| 3305 | /* ASB_DISCARD_REMOTE - ASB_DISCARD_LOCAL is valid */ |
| 3306 | if (peer == ASB_DISCARD_REMOTE) |
| 3307 | return ASB_DISCARD_LOCAL; |
| 3308 | |
| 3309 | /* any other things with ASB_DISCARD_REMOTE or ASB_DISCARD_LOCAL are invalid */ |
| 3310 | if (peer == ASB_DISCARD_LOCAL) |
| 3311 | return ASB_DISCARD_REMOTE; |
| 3312 | |
| 3313 | /* everything else is valid if they are equal on both sides. */ |
| 3314 | return peer; |
| 3315 | } |
| 3316 | |
| 3317 | static int receive_protocol(struct drbd_connection *connection, struct packet_info *pi) |
| 3318 | { |
| 3319 | struct p_protocol *p = pi->data; |
| 3320 | enum drbd_after_sb_p p_after_sb_0p, p_after_sb_1p, p_after_sb_2p; |
| 3321 | int p_proto, p_discard_my_data, p_two_primaries, cf; |
| 3322 | struct net_conf *nc, *old_net_conf, *new_net_conf = NULL; |
| 3323 | char integrity_alg[SHARED_SECRET_MAX] = ""; |
| 3324 | struct crypto_ahash *peer_integrity_tfm = NULL; |
| 3325 | void *int_dig_in = NULL, *int_dig_vv = NULL; |
| 3326 | |
| 3327 | p_proto = be32_to_cpu(p->protocol); |
| 3328 | p_after_sb_0p = be32_to_cpu(p->after_sb_0p); |
| 3329 | p_after_sb_1p = be32_to_cpu(p->after_sb_1p); |
| 3330 | p_after_sb_2p = be32_to_cpu(p->after_sb_2p); |
| 3331 | p_two_primaries = be32_to_cpu(p->two_primaries); |
| 3332 | cf = be32_to_cpu(p->conn_flags); |
| 3333 | p_discard_my_data = cf & CF_DISCARD_MY_DATA; |
| 3334 | |
| 3335 | if (connection->agreed_pro_version >= 87) { |
| 3336 | int err; |
| 3337 | |
| 3338 | if (pi->size > sizeof(integrity_alg)) |
| 3339 | return -EIO; |
| 3340 | err = drbd_recv_all(connection, integrity_alg, pi->size); |
| 3341 | if (err) |
| 3342 | return err; |
| 3343 | integrity_alg[SHARED_SECRET_MAX - 1] = 0; |
| 3344 | } |
| 3345 | |
| 3346 | if (pi->cmd != P_PROTOCOL_UPDATE) { |
| 3347 | clear_bit(CONN_DRY_RUN, &connection->flags); |
| 3348 | |
| 3349 | if (cf & CF_DRY_RUN) |
| 3350 | set_bit(CONN_DRY_RUN, &connection->flags); |
| 3351 | |
| 3352 | rcu_read_lock(); |
| 3353 | nc = rcu_dereference(connection->net_conf); |
| 3354 | |
| 3355 | if (p_proto != nc->wire_protocol) { |
| 3356 | drbd_err(connection, "incompatible %s settings\n", "protocol"); |
| 3357 | goto disconnect_rcu_unlock; |
| 3358 | } |
| 3359 | |
| 3360 | if (convert_after_sb(p_after_sb_0p) != nc->after_sb_0p) { |
| 3361 | drbd_err(connection, "incompatible %s settings\n", "after-sb-0pri"); |
| 3362 | goto disconnect_rcu_unlock; |
| 3363 | } |
| 3364 | |
| 3365 | if (convert_after_sb(p_after_sb_1p) != nc->after_sb_1p) { |
| 3366 | drbd_err(connection, "incompatible %s settings\n", "after-sb-1pri"); |
| 3367 | goto disconnect_rcu_unlock; |
| 3368 | } |
| 3369 | |
| 3370 | if (convert_after_sb(p_after_sb_2p) != nc->after_sb_2p) { |
| 3371 | drbd_err(connection, "incompatible %s settings\n", "after-sb-2pri"); |
| 3372 | goto disconnect_rcu_unlock; |
| 3373 | } |
| 3374 | |
| 3375 | if (p_discard_my_data && nc->discard_my_data) { |
| 3376 | drbd_err(connection, "incompatible %s settings\n", "discard-my-data"); |
| 3377 | goto disconnect_rcu_unlock; |
| 3378 | } |
| 3379 | |
| 3380 | if (p_two_primaries != nc->two_primaries) { |
| 3381 | drbd_err(connection, "incompatible %s settings\n", "allow-two-primaries"); |
| 3382 | goto disconnect_rcu_unlock; |
| 3383 | } |
| 3384 | |
| 3385 | if (strcmp(integrity_alg, nc->integrity_alg)) { |
| 3386 | drbd_err(connection, "incompatible %s settings\n", "data-integrity-alg"); |
| 3387 | goto disconnect_rcu_unlock; |
| 3388 | } |
| 3389 | |
| 3390 | rcu_read_unlock(); |
| 3391 | } |
| 3392 | |
| 3393 | if (integrity_alg[0]) { |
| 3394 | int hash_size; |
| 3395 | |
| 3396 | /* |
| 3397 | * We can only change the peer data integrity algorithm |
| 3398 | * here. Changing our own data integrity algorithm |
| 3399 | * requires that we send a P_PROTOCOL_UPDATE packet at |
| 3400 | * the same time; otherwise, the peer has no way to |
| 3401 | * tell between which packets the algorithm should |
| 3402 | * change. |
| 3403 | */ |
| 3404 | |
| 3405 | peer_integrity_tfm = crypto_alloc_ahash(integrity_alg, 0, CRYPTO_ALG_ASYNC); |
| 3406 | if (!peer_integrity_tfm) { |
| 3407 | drbd_err(connection, "peer data-integrity-alg %s not supported\n", |
| 3408 | integrity_alg); |
| 3409 | goto disconnect; |
| 3410 | } |
| 3411 | |
| 3412 | hash_size = crypto_ahash_digestsize(peer_integrity_tfm); |
| 3413 | int_dig_in = kmalloc(hash_size, GFP_KERNEL); |
| 3414 | int_dig_vv = kmalloc(hash_size, GFP_KERNEL); |
| 3415 | if (!(int_dig_in && int_dig_vv)) { |
| 3416 | drbd_err(connection, "Allocation of buffers for data integrity checking failed\n"); |
| 3417 | goto disconnect; |
| 3418 | } |
| 3419 | } |
| 3420 | |
| 3421 | new_net_conf = kmalloc(sizeof(struct net_conf), GFP_KERNEL); |
| 3422 | if (!new_net_conf) { |
| 3423 | drbd_err(connection, "Allocation of new net_conf failed\n"); |
| 3424 | goto disconnect; |
| 3425 | } |
| 3426 | |
| 3427 | mutex_lock(&connection->data.mutex); |
| 3428 | mutex_lock(&connection->resource->conf_update); |
| 3429 | old_net_conf = connection->net_conf; |
| 3430 | *new_net_conf = *old_net_conf; |
| 3431 | |
| 3432 | new_net_conf->wire_protocol = p_proto; |
| 3433 | new_net_conf->after_sb_0p = convert_after_sb(p_after_sb_0p); |
| 3434 | new_net_conf->after_sb_1p = convert_after_sb(p_after_sb_1p); |
| 3435 | new_net_conf->after_sb_2p = convert_after_sb(p_after_sb_2p); |
| 3436 | new_net_conf->two_primaries = p_two_primaries; |
| 3437 | |
| 3438 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
| 3439 | mutex_unlock(&connection->resource->conf_update); |
| 3440 | mutex_unlock(&connection->data.mutex); |
| 3441 | |
| 3442 | crypto_free_ahash(connection->peer_integrity_tfm); |
| 3443 | kfree(connection->int_dig_in); |
| 3444 | kfree(connection->int_dig_vv); |
| 3445 | connection->peer_integrity_tfm = peer_integrity_tfm; |
| 3446 | connection->int_dig_in = int_dig_in; |
| 3447 | connection->int_dig_vv = int_dig_vv; |
| 3448 | |
| 3449 | if (strcmp(old_net_conf->integrity_alg, integrity_alg)) |
| 3450 | drbd_info(connection, "peer data-integrity-alg: %s\n", |
| 3451 | integrity_alg[0] ? integrity_alg : "(none)"); |
| 3452 | |
| 3453 | synchronize_rcu(); |
| 3454 | kfree(old_net_conf); |
| 3455 | return 0; |
| 3456 | |
| 3457 | disconnect_rcu_unlock: |
| 3458 | rcu_read_unlock(); |
| 3459 | disconnect: |
| 3460 | crypto_free_ahash(peer_integrity_tfm); |
| 3461 | kfree(int_dig_in); |
| 3462 | kfree(int_dig_vv); |
| 3463 | conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_HARD); |
| 3464 | return -EIO; |
| 3465 | } |
| 3466 | |
| 3467 | /* helper function |
| 3468 | * input: alg name, feature name |
| 3469 | * return: NULL (alg name was "") |
| 3470 | * ERR_PTR(error) if something goes wrong |
| 3471 | * or the crypto hash ptr, if it worked out ok. */ |
| 3472 | static struct crypto_ahash *drbd_crypto_alloc_digest_safe(const struct drbd_device *device, |
| 3473 | const char *alg, const char *name) |
| 3474 | { |
| 3475 | struct crypto_ahash *tfm; |
| 3476 | |
| 3477 | if (!alg[0]) |
| 3478 | return NULL; |
| 3479 | |
| 3480 | tfm = crypto_alloc_ahash(alg, 0, CRYPTO_ALG_ASYNC); |
| 3481 | if (IS_ERR(tfm)) { |
| 3482 | drbd_err(device, "Can not allocate \"%s\" as %s (reason: %ld)\n", |
| 3483 | alg, name, PTR_ERR(tfm)); |
| 3484 | return tfm; |
| 3485 | } |
| 3486 | return tfm; |
| 3487 | } |
| 3488 | |
| 3489 | static int ignore_remaining_packet(struct drbd_connection *connection, struct packet_info *pi) |
| 3490 | { |
| 3491 | void *buffer = connection->data.rbuf; |
| 3492 | int size = pi->size; |
| 3493 | |
| 3494 | while (size) { |
| 3495 | int s = min_t(int, size, DRBD_SOCKET_BUFFER_SIZE); |
| 3496 | s = drbd_recv(connection, buffer, s); |
| 3497 | if (s <= 0) { |
| 3498 | if (s < 0) |
| 3499 | return s; |
| 3500 | break; |
| 3501 | } |
| 3502 | size -= s; |
| 3503 | } |
| 3504 | if (size) |
| 3505 | return -EIO; |
| 3506 | return 0; |
| 3507 | } |
| 3508 | |
| 3509 | /* |
| 3510 | * config_unknown_volume - device configuration command for unknown volume |
| 3511 | * |
| 3512 | * When a device is added to an existing connection, the node on which the |
| 3513 | * device is added first will send configuration commands to its peer but the |
| 3514 | * peer will not know about the device yet. It will warn and ignore these |
| 3515 | * commands. Once the device is added on the second node, the second node will |
| 3516 | * send the same device configuration commands, but in the other direction. |
| 3517 | * |
| 3518 | * (We can also end up here if drbd is misconfigured.) |
| 3519 | */ |
| 3520 | static int config_unknown_volume(struct drbd_connection *connection, struct packet_info *pi) |
| 3521 | { |
| 3522 | drbd_warn(connection, "%s packet received for volume %u, which is not configured locally\n", |
| 3523 | cmdname(pi->cmd), pi->vnr); |
| 3524 | return ignore_remaining_packet(connection, pi); |
| 3525 | } |
| 3526 | |
| 3527 | static int receive_SyncParam(struct drbd_connection *connection, struct packet_info *pi) |
| 3528 | { |
| 3529 | struct drbd_peer_device *peer_device; |
| 3530 | struct drbd_device *device; |
| 3531 | struct p_rs_param_95 *p; |
| 3532 | unsigned int header_size, data_size, exp_max_sz; |
| 3533 | struct crypto_ahash *verify_tfm = NULL; |
| 3534 | struct crypto_ahash *csums_tfm = NULL; |
| 3535 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
| 3536 | struct disk_conf *old_disk_conf = NULL, *new_disk_conf = NULL; |
| 3537 | const int apv = connection->agreed_pro_version; |
| 3538 | struct fifo_buffer *old_plan = NULL, *new_plan = NULL; |
| 3539 | int fifo_size = 0; |
| 3540 | int err; |
| 3541 | |
| 3542 | peer_device = conn_peer_device(connection, pi->vnr); |
| 3543 | if (!peer_device) |
| 3544 | return config_unknown_volume(connection, pi); |
| 3545 | device = peer_device->device; |
| 3546 | |
| 3547 | exp_max_sz = apv <= 87 ? sizeof(struct p_rs_param) |
| 3548 | : apv == 88 ? sizeof(struct p_rs_param) |
| 3549 | + SHARED_SECRET_MAX |
| 3550 | : apv <= 94 ? sizeof(struct p_rs_param_89) |
| 3551 | : /* apv >= 95 */ sizeof(struct p_rs_param_95); |
| 3552 | |
| 3553 | if (pi->size > exp_max_sz) { |
| 3554 | drbd_err(device, "SyncParam packet too long: received %u, expected <= %u bytes\n", |
| 3555 | pi->size, exp_max_sz); |
| 3556 | return -EIO; |
| 3557 | } |
| 3558 | |
| 3559 | if (apv <= 88) { |
| 3560 | header_size = sizeof(struct p_rs_param); |
| 3561 | data_size = pi->size - header_size; |
| 3562 | } else if (apv <= 94) { |
| 3563 | header_size = sizeof(struct p_rs_param_89); |
| 3564 | data_size = pi->size - header_size; |
| 3565 | D_ASSERT(device, data_size == 0); |
| 3566 | } else { |
| 3567 | header_size = sizeof(struct p_rs_param_95); |
| 3568 | data_size = pi->size - header_size; |
| 3569 | D_ASSERT(device, data_size == 0); |
| 3570 | } |
| 3571 | |
| 3572 | /* initialize verify_alg and csums_alg */ |
| 3573 | p = pi->data; |
| 3574 | memset(p->verify_alg, 0, 2 * SHARED_SECRET_MAX); |
| 3575 | |
| 3576 | err = drbd_recv_all(peer_device->connection, p, header_size); |
| 3577 | if (err) |
| 3578 | return err; |
| 3579 | |
| 3580 | mutex_lock(&connection->resource->conf_update); |
| 3581 | old_net_conf = peer_device->connection->net_conf; |
| 3582 | if (get_ldev(device)) { |
| 3583 | new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL); |
| 3584 | if (!new_disk_conf) { |
| 3585 | put_ldev(device); |
| 3586 | mutex_unlock(&connection->resource->conf_update); |
| 3587 | drbd_err(device, "Allocation of new disk_conf failed\n"); |
| 3588 | return -ENOMEM; |
| 3589 | } |
| 3590 | |
| 3591 | old_disk_conf = device->ldev->disk_conf; |
| 3592 | *new_disk_conf = *old_disk_conf; |
| 3593 | |
| 3594 | new_disk_conf->resync_rate = be32_to_cpu(p->resync_rate); |
| 3595 | } |
| 3596 | |
| 3597 | if (apv >= 88) { |
| 3598 | if (apv == 88) { |
| 3599 | if (data_size > SHARED_SECRET_MAX || data_size == 0) { |
| 3600 | drbd_err(device, "verify-alg of wrong size, " |
| 3601 | "peer wants %u, accepting only up to %u byte\n", |
| 3602 | data_size, SHARED_SECRET_MAX); |
| 3603 | err = -EIO; |
| 3604 | goto reconnect; |
| 3605 | } |
| 3606 | |
| 3607 | err = drbd_recv_all(peer_device->connection, p->verify_alg, data_size); |
| 3608 | if (err) |
| 3609 | goto reconnect; |
| 3610 | /* we expect NUL terminated string */ |
| 3611 | /* but just in case someone tries to be evil */ |
| 3612 | D_ASSERT(device, p->verify_alg[data_size-1] == 0); |
| 3613 | p->verify_alg[data_size-1] = 0; |
| 3614 | |
| 3615 | } else /* apv >= 89 */ { |
| 3616 | /* we still expect NUL terminated strings */ |
| 3617 | /* but just in case someone tries to be evil */ |
| 3618 | D_ASSERT(device, p->verify_alg[SHARED_SECRET_MAX-1] == 0); |
| 3619 | D_ASSERT(device, p->csums_alg[SHARED_SECRET_MAX-1] == 0); |
| 3620 | p->verify_alg[SHARED_SECRET_MAX-1] = 0; |
| 3621 | p->csums_alg[SHARED_SECRET_MAX-1] = 0; |
| 3622 | } |
| 3623 | |
| 3624 | if (strcmp(old_net_conf->verify_alg, p->verify_alg)) { |
| 3625 | if (device->state.conn == C_WF_REPORT_PARAMS) { |
| 3626 | drbd_err(device, "Different verify-alg settings. me=\"%s\" peer=\"%s\"\n", |
| 3627 | old_net_conf->verify_alg, p->verify_alg); |
| 3628 | goto disconnect; |
| 3629 | } |
| 3630 | verify_tfm = drbd_crypto_alloc_digest_safe(device, |
| 3631 | p->verify_alg, "verify-alg"); |
| 3632 | if (IS_ERR(verify_tfm)) { |
| 3633 | verify_tfm = NULL; |
| 3634 | goto disconnect; |
| 3635 | } |
| 3636 | } |
| 3637 | |
| 3638 | if (apv >= 89 && strcmp(old_net_conf->csums_alg, p->csums_alg)) { |
| 3639 | if (device->state.conn == C_WF_REPORT_PARAMS) { |
| 3640 | drbd_err(device, "Different csums-alg settings. me=\"%s\" peer=\"%s\"\n", |
| 3641 | old_net_conf->csums_alg, p->csums_alg); |
| 3642 | goto disconnect; |
| 3643 | } |
| 3644 | csums_tfm = drbd_crypto_alloc_digest_safe(device, |
| 3645 | p->csums_alg, "csums-alg"); |
| 3646 | if (IS_ERR(csums_tfm)) { |
| 3647 | csums_tfm = NULL; |
| 3648 | goto disconnect; |
| 3649 | } |
| 3650 | } |
| 3651 | |
| 3652 | if (apv > 94 && new_disk_conf) { |
| 3653 | new_disk_conf->c_plan_ahead = be32_to_cpu(p->c_plan_ahead); |
| 3654 | new_disk_conf->c_delay_target = be32_to_cpu(p->c_delay_target); |
| 3655 | new_disk_conf->c_fill_target = be32_to_cpu(p->c_fill_target); |
| 3656 | new_disk_conf->c_max_rate = be32_to_cpu(p->c_max_rate); |
| 3657 | |
| 3658 | fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
| 3659 | if (fifo_size != device->rs_plan_s->size) { |
| 3660 | new_plan = fifo_alloc(fifo_size); |
| 3661 | if (!new_plan) { |
| 3662 | drbd_err(device, "kmalloc of fifo_buffer failed"); |
| 3663 | put_ldev(device); |
| 3664 | goto disconnect; |
| 3665 | } |
| 3666 | } |
| 3667 | } |
| 3668 | |
| 3669 | if (verify_tfm || csums_tfm) { |
| 3670 | new_net_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); |
| 3671 | if (!new_net_conf) { |
| 3672 | drbd_err(device, "Allocation of new net_conf failed\n"); |
| 3673 | goto disconnect; |
| 3674 | } |
| 3675 | |
| 3676 | *new_net_conf = *old_net_conf; |
| 3677 | |
| 3678 | if (verify_tfm) { |
| 3679 | strcpy(new_net_conf->verify_alg, p->verify_alg); |
| 3680 | new_net_conf->verify_alg_len = strlen(p->verify_alg) + 1; |
| 3681 | crypto_free_ahash(peer_device->connection->verify_tfm); |
| 3682 | peer_device->connection->verify_tfm = verify_tfm; |
| 3683 | drbd_info(device, "using verify-alg: \"%s\"\n", p->verify_alg); |
| 3684 | } |
| 3685 | if (csums_tfm) { |
| 3686 | strcpy(new_net_conf->csums_alg, p->csums_alg); |
| 3687 | new_net_conf->csums_alg_len = strlen(p->csums_alg) + 1; |
| 3688 | crypto_free_ahash(peer_device->connection->csums_tfm); |
| 3689 | peer_device->connection->csums_tfm = csums_tfm; |
| 3690 | drbd_info(device, "using csums-alg: \"%s\"\n", p->csums_alg); |
| 3691 | } |
| 3692 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
| 3693 | } |
| 3694 | } |
| 3695 | |
| 3696 | if (new_disk_conf) { |
| 3697 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
| 3698 | put_ldev(device); |
| 3699 | } |
| 3700 | |
| 3701 | if (new_plan) { |
| 3702 | old_plan = device->rs_plan_s; |
| 3703 | rcu_assign_pointer(device->rs_plan_s, new_plan); |
| 3704 | } |
| 3705 | |
| 3706 | mutex_unlock(&connection->resource->conf_update); |
| 3707 | synchronize_rcu(); |
| 3708 | if (new_net_conf) |
| 3709 | kfree(old_net_conf); |
| 3710 | kfree(old_disk_conf); |
| 3711 | kfree(old_plan); |
| 3712 | |
| 3713 | return 0; |
| 3714 | |
| 3715 | reconnect: |
| 3716 | if (new_disk_conf) { |
| 3717 | put_ldev(device); |
| 3718 | kfree(new_disk_conf); |
| 3719 | } |
| 3720 | mutex_unlock(&connection->resource->conf_update); |
| 3721 | return -EIO; |
| 3722 | |
| 3723 | disconnect: |
| 3724 | kfree(new_plan); |
| 3725 | if (new_disk_conf) { |
| 3726 | put_ldev(device); |
| 3727 | kfree(new_disk_conf); |
| 3728 | } |
| 3729 | mutex_unlock(&connection->resource->conf_update); |
| 3730 | /* just for completeness: actually not needed, |
| 3731 | * as this is not reached if csums_tfm was ok. */ |
| 3732 | crypto_free_ahash(csums_tfm); |
| 3733 | /* but free the verify_tfm again, if csums_tfm did not work out */ |
| 3734 | crypto_free_ahash(verify_tfm); |
| 3735 | conn_request_state(peer_device->connection, NS(conn, C_DISCONNECTING), CS_HARD); |
| 3736 | return -EIO; |
| 3737 | } |
| 3738 | |
| 3739 | /* warn if the arguments differ by more than 12.5% */ |
| 3740 | static void warn_if_differ_considerably(struct drbd_device *device, |
| 3741 | const char *s, sector_t a, sector_t b) |
| 3742 | { |
| 3743 | sector_t d; |
| 3744 | if (a == 0 || b == 0) |
| 3745 | return; |
| 3746 | d = (a > b) ? (a - b) : (b - a); |
| 3747 | if (d > (a>>3) || d > (b>>3)) |
| 3748 | drbd_warn(device, "Considerable difference in %s: %llus vs. %llus\n", s, |
| 3749 | (unsigned long long)a, (unsigned long long)b); |
| 3750 | } |
| 3751 | |
| 3752 | static int receive_sizes(struct drbd_connection *connection, struct packet_info *pi) |
| 3753 | { |
| 3754 | struct drbd_peer_device *peer_device; |
| 3755 | struct drbd_device *device; |
| 3756 | struct p_sizes *p = pi->data; |
| 3757 | enum determine_dev_size dd = DS_UNCHANGED; |
| 3758 | sector_t p_size, p_usize, p_csize, my_usize; |
| 3759 | int ldsc = 0; /* local disk size changed */ |
| 3760 | enum dds_flags ddsf; |
| 3761 | |
| 3762 | peer_device = conn_peer_device(connection, pi->vnr); |
| 3763 | if (!peer_device) |
| 3764 | return config_unknown_volume(connection, pi); |
| 3765 | device = peer_device->device; |
| 3766 | |
| 3767 | p_size = be64_to_cpu(p->d_size); |
| 3768 | p_usize = be64_to_cpu(p->u_size); |
| 3769 | p_csize = be64_to_cpu(p->c_size); |
| 3770 | |
| 3771 | /* just store the peer's disk size for now. |
| 3772 | * we still need to figure out whether we accept that. */ |
| 3773 | device->p_size = p_size; |
| 3774 | |
| 3775 | if (get_ldev(device)) { |
| 3776 | rcu_read_lock(); |
| 3777 | my_usize = rcu_dereference(device->ldev->disk_conf)->disk_size; |
| 3778 | rcu_read_unlock(); |
| 3779 | |
| 3780 | warn_if_differ_considerably(device, "lower level device sizes", |
| 3781 | p_size, drbd_get_max_capacity(device->ldev)); |
| 3782 | warn_if_differ_considerably(device, "user requested size", |
| 3783 | p_usize, my_usize); |
| 3784 | |
| 3785 | /* if this is the first connect, or an otherwise expected |
| 3786 | * param exchange, choose the minimum */ |
| 3787 | if (device->state.conn == C_WF_REPORT_PARAMS) |
| 3788 | p_usize = min_not_zero(my_usize, p_usize); |
| 3789 | |
| 3790 | /* Never shrink a device with usable data during connect. |
| 3791 | But allow online shrinking if we are connected. */ |
| 3792 | if (drbd_new_dev_size(device, device->ldev, p_usize, 0) < |
| 3793 | drbd_get_capacity(device->this_bdev) && |
| 3794 | device->state.disk >= D_OUTDATED && |
| 3795 | device->state.conn < C_CONNECTED) { |
| 3796 | drbd_err(device, "The peer's disk size is too small!\n"); |
| 3797 | conn_request_state(peer_device->connection, NS(conn, C_DISCONNECTING), CS_HARD); |
| 3798 | put_ldev(device); |
| 3799 | return -EIO; |
| 3800 | } |
| 3801 | |
| 3802 | if (my_usize != p_usize) { |
| 3803 | struct disk_conf *old_disk_conf, *new_disk_conf = NULL; |
| 3804 | |
| 3805 | new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL); |
| 3806 | if (!new_disk_conf) { |
| 3807 | drbd_err(device, "Allocation of new disk_conf failed\n"); |
| 3808 | put_ldev(device); |
| 3809 | return -ENOMEM; |
| 3810 | } |
| 3811 | |
| 3812 | mutex_lock(&connection->resource->conf_update); |
| 3813 | old_disk_conf = device->ldev->disk_conf; |
| 3814 | *new_disk_conf = *old_disk_conf; |
| 3815 | new_disk_conf->disk_size = p_usize; |
| 3816 | |
| 3817 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
| 3818 | mutex_unlock(&connection->resource->conf_update); |
| 3819 | synchronize_rcu(); |
| 3820 | kfree(old_disk_conf); |
| 3821 | |
| 3822 | drbd_info(device, "Peer sets u_size to %lu sectors\n", |
| 3823 | (unsigned long)my_usize); |
| 3824 | } |
| 3825 | |
| 3826 | put_ldev(device); |
| 3827 | } |
| 3828 | |
| 3829 | device->peer_max_bio_size = be32_to_cpu(p->max_bio_size); |
| 3830 | /* Leave drbd_reconsider_max_bio_size() before drbd_determine_dev_size(). |
| 3831 | In case we cleared the QUEUE_FLAG_DISCARD from our queue in |
| 3832 | drbd_reconsider_max_bio_size(), we can be sure that after |
| 3833 | drbd_determine_dev_size() no REQ_DISCARDs are in the queue. */ |
| 3834 | |
| 3835 | ddsf = be16_to_cpu(p->dds_flags); |
| 3836 | if (get_ldev(device)) { |
| 3837 | drbd_reconsider_max_bio_size(device, device->ldev); |
| 3838 | dd = drbd_determine_dev_size(device, ddsf, NULL); |
| 3839 | put_ldev(device); |
| 3840 | if (dd == DS_ERROR) |
| 3841 | return -EIO; |
| 3842 | drbd_md_sync(device); |
| 3843 | } else { |
| 3844 | /* |
| 3845 | * I am diskless, need to accept the peer's *current* size. |
| 3846 | * I must NOT accept the peers backing disk size, |
| 3847 | * it may have been larger than mine all along... |
| 3848 | * |
| 3849 | * At this point, the peer knows more about my disk, or at |
| 3850 | * least about what we last agreed upon, than myself. |
| 3851 | * So if his c_size is less than his d_size, the most likely |
| 3852 | * reason is that *my* d_size was smaller last time we checked. |
| 3853 | * |
| 3854 | * However, if he sends a zero current size, |
| 3855 | * take his (user-capped or) backing disk size anyways. |
| 3856 | */ |
| 3857 | drbd_reconsider_max_bio_size(device, NULL); |
| 3858 | drbd_set_my_capacity(device, p_csize ?: p_usize ?: p_size); |
| 3859 | } |
| 3860 | |
| 3861 | if (get_ldev(device)) { |
| 3862 | if (device->ldev->known_size != drbd_get_capacity(device->ldev->backing_bdev)) { |
| 3863 | device->ldev->known_size = drbd_get_capacity(device->ldev->backing_bdev); |
| 3864 | ldsc = 1; |
| 3865 | } |
| 3866 | |
| 3867 | put_ldev(device); |
| 3868 | } |
| 3869 | |
| 3870 | if (device->state.conn > C_WF_REPORT_PARAMS) { |
| 3871 | if (be64_to_cpu(p->c_size) != |
| 3872 | drbd_get_capacity(device->this_bdev) || ldsc) { |
| 3873 | /* we have different sizes, probably peer |
| 3874 | * needs to know my new size... */ |
| 3875 | drbd_send_sizes(peer_device, 0, ddsf); |
| 3876 | } |
| 3877 | if (test_and_clear_bit(RESIZE_PENDING, &device->flags) || |
| 3878 | (dd == DS_GREW && device->state.conn == C_CONNECTED)) { |
| 3879 | if (device->state.pdsk >= D_INCONSISTENT && |
| 3880 | device->state.disk >= D_INCONSISTENT) { |
| 3881 | if (ddsf & DDSF_NO_RESYNC) |
| 3882 | drbd_info(device, "Resync of new storage suppressed with --assume-clean\n"); |
| 3883 | else |
| 3884 | resync_after_online_grow(device); |
| 3885 | } else |
| 3886 | set_bit(RESYNC_AFTER_NEG, &device->flags); |
| 3887 | } |
| 3888 | } |
| 3889 | |
| 3890 | return 0; |
| 3891 | } |
| 3892 | |
| 3893 | static int receive_uuids(struct drbd_connection *connection, struct packet_info *pi) |
| 3894 | { |
| 3895 | struct drbd_peer_device *peer_device; |
| 3896 | struct drbd_device *device; |
| 3897 | struct p_uuids *p = pi->data; |
| 3898 | u64 *p_uuid; |
| 3899 | int i, updated_uuids = 0; |
| 3900 | |
| 3901 | peer_device = conn_peer_device(connection, pi->vnr); |
| 3902 | if (!peer_device) |
| 3903 | return config_unknown_volume(connection, pi); |
| 3904 | device = peer_device->device; |
| 3905 | |
| 3906 | p_uuid = kmalloc(sizeof(u64)*UI_EXTENDED_SIZE, GFP_NOIO); |
| 3907 | if (!p_uuid) { |
| 3908 | drbd_err(device, "kmalloc of p_uuid failed\n"); |
| 3909 | return false; |
| 3910 | } |
| 3911 | |
| 3912 | for (i = UI_CURRENT; i < UI_EXTENDED_SIZE; i++) |
| 3913 | p_uuid[i] = be64_to_cpu(p->uuid[i]); |
| 3914 | |
| 3915 | kfree(device->p_uuid); |
| 3916 | device->p_uuid = p_uuid; |
| 3917 | |
| 3918 | if (device->state.conn < C_CONNECTED && |
| 3919 | device->state.disk < D_INCONSISTENT && |
| 3920 | device->state.role == R_PRIMARY && |
| 3921 | (device->ed_uuid & ~((u64)1)) != (p_uuid[UI_CURRENT] & ~((u64)1))) { |
| 3922 | drbd_err(device, "Can only connect to data with current UUID=%016llX\n", |
| 3923 | (unsigned long long)device->ed_uuid); |
| 3924 | conn_request_state(peer_device->connection, NS(conn, C_DISCONNECTING), CS_HARD); |
| 3925 | return -EIO; |
| 3926 | } |
| 3927 | |
| 3928 | if (get_ldev(device)) { |
| 3929 | int skip_initial_sync = |
| 3930 | device->state.conn == C_CONNECTED && |
| 3931 | peer_device->connection->agreed_pro_version >= 90 && |
| 3932 | device->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && |
| 3933 | (p_uuid[UI_FLAGS] & 8); |
| 3934 | if (skip_initial_sync) { |
| 3935 | drbd_info(device, "Accepted new current UUID, preparing to skip initial sync\n"); |
| 3936 | drbd_bitmap_io(device, &drbd_bmio_clear_n_write, |
| 3937 | "clear_n_write from receive_uuids", |
| 3938 | BM_LOCKED_TEST_ALLOWED); |
| 3939 | _drbd_uuid_set(device, UI_CURRENT, p_uuid[UI_CURRENT]); |
| 3940 | _drbd_uuid_set(device, UI_BITMAP, 0); |
| 3941 | _drbd_set_state(_NS2(device, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), |
| 3942 | CS_VERBOSE, NULL); |
| 3943 | drbd_md_sync(device); |
| 3944 | updated_uuids = 1; |
| 3945 | } |
| 3946 | put_ldev(device); |
| 3947 | } else if (device->state.disk < D_INCONSISTENT && |
| 3948 | device->state.role == R_PRIMARY) { |
| 3949 | /* I am a diskless primary, the peer just created a new current UUID |
| 3950 | for me. */ |
| 3951 | updated_uuids = drbd_set_ed_uuid(device, p_uuid[UI_CURRENT]); |
| 3952 | } |
| 3953 | |
| 3954 | /* Before we test for the disk state, we should wait until an eventually |
| 3955 | ongoing cluster wide state change is finished. That is important if |
| 3956 | we are primary and are detaching from our disk. We need to see the |
| 3957 | new disk state... */ |
| 3958 | mutex_lock(device->state_mutex); |
| 3959 | mutex_unlock(device->state_mutex); |
| 3960 | if (device->state.conn >= C_CONNECTED && device->state.disk < D_INCONSISTENT) |
| 3961 | updated_uuids |= drbd_set_ed_uuid(device, p_uuid[UI_CURRENT]); |
| 3962 | |
| 3963 | if (updated_uuids) |
| 3964 | drbd_print_uuids(device, "receiver updated UUIDs to"); |
| 3965 | |
| 3966 | return 0; |
| 3967 | } |
| 3968 | |
| 3969 | /** |
| 3970 | * convert_state() - Converts the peer's view of the cluster state to our point of view |
| 3971 | * @ps: The state as seen by the peer. |
| 3972 | */ |
| 3973 | static union drbd_state convert_state(union drbd_state ps) |
| 3974 | { |
| 3975 | union drbd_state ms; |
| 3976 | |
| 3977 | static enum drbd_conns c_tab[] = { |
| 3978 | [C_WF_REPORT_PARAMS] = C_WF_REPORT_PARAMS, |
| 3979 | [C_CONNECTED] = C_CONNECTED, |
| 3980 | |
| 3981 | [C_STARTING_SYNC_S] = C_STARTING_SYNC_T, |
| 3982 | [C_STARTING_SYNC_T] = C_STARTING_SYNC_S, |
| 3983 | [C_DISCONNECTING] = C_TEAR_DOWN, /* C_NETWORK_FAILURE, */ |
| 3984 | [C_VERIFY_S] = C_VERIFY_T, |
| 3985 | [C_MASK] = C_MASK, |
| 3986 | }; |
| 3987 | |
| 3988 | ms.i = ps.i; |
| 3989 | |
| 3990 | ms.conn = c_tab[ps.conn]; |
| 3991 | ms.peer = ps.role; |
| 3992 | ms.role = ps.peer; |
| 3993 | ms.pdsk = ps.disk; |
| 3994 | ms.disk = ps.pdsk; |
| 3995 | ms.peer_isp = (ps.aftr_isp | ps.user_isp); |
| 3996 | |
| 3997 | return ms; |
| 3998 | } |
| 3999 | |
| 4000 | static int receive_req_state(struct drbd_connection *connection, struct packet_info *pi) |
| 4001 | { |
| 4002 | struct drbd_peer_device *peer_device; |
| 4003 | struct drbd_device *device; |
| 4004 | struct p_req_state *p = pi->data; |
| 4005 | union drbd_state mask, val; |
| 4006 | enum drbd_state_rv rv; |
| 4007 | |
| 4008 | peer_device = conn_peer_device(connection, pi->vnr); |
| 4009 | if (!peer_device) |
| 4010 | return -EIO; |
| 4011 | device = peer_device->device; |
| 4012 | |
| 4013 | mask.i = be32_to_cpu(p->mask); |
| 4014 | val.i = be32_to_cpu(p->val); |
| 4015 | |
| 4016 | if (test_bit(RESOLVE_CONFLICTS, &peer_device->connection->flags) && |
| 4017 | mutex_is_locked(device->state_mutex)) { |
| 4018 | drbd_send_sr_reply(peer_device, SS_CONCURRENT_ST_CHG); |
| 4019 | return 0; |
| 4020 | } |
| 4021 | |
| 4022 | mask = convert_state(mask); |
| 4023 | val = convert_state(val); |
| 4024 | |
| 4025 | rv = drbd_change_state(device, CS_VERBOSE, mask, val); |
| 4026 | drbd_send_sr_reply(peer_device, rv); |
| 4027 | |
| 4028 | drbd_md_sync(device); |
| 4029 | |
| 4030 | return 0; |
| 4031 | } |
| 4032 | |
| 4033 | static int receive_req_conn_state(struct drbd_connection *connection, struct packet_info *pi) |
| 4034 | { |
| 4035 | struct p_req_state *p = pi->data; |
| 4036 | union drbd_state mask, val; |
| 4037 | enum drbd_state_rv rv; |
| 4038 | |
| 4039 | mask.i = be32_to_cpu(p->mask); |
| 4040 | val.i = be32_to_cpu(p->val); |
| 4041 | |
| 4042 | if (test_bit(RESOLVE_CONFLICTS, &connection->flags) && |
| 4043 | mutex_is_locked(&connection->cstate_mutex)) { |
| 4044 | conn_send_sr_reply(connection, SS_CONCURRENT_ST_CHG); |
| 4045 | return 0; |
| 4046 | } |
| 4047 | |
| 4048 | mask = convert_state(mask); |
| 4049 | val = convert_state(val); |
| 4050 | |
| 4051 | rv = conn_request_state(connection, mask, val, CS_VERBOSE | CS_LOCAL_ONLY | CS_IGN_OUTD_FAIL); |
| 4052 | conn_send_sr_reply(connection, rv); |
| 4053 | |
| 4054 | return 0; |
| 4055 | } |
| 4056 | |
| 4057 | static int receive_state(struct drbd_connection *connection, struct packet_info *pi) |
| 4058 | { |
| 4059 | struct drbd_peer_device *peer_device; |
| 4060 | struct drbd_device *device; |
| 4061 | struct p_state *p = pi->data; |
| 4062 | union drbd_state os, ns, peer_state; |
| 4063 | enum drbd_disk_state real_peer_disk; |
| 4064 | enum chg_state_flags cs_flags; |
| 4065 | int rv; |
| 4066 | |
| 4067 | peer_device = conn_peer_device(connection, pi->vnr); |
| 4068 | if (!peer_device) |
| 4069 | return config_unknown_volume(connection, pi); |
| 4070 | device = peer_device->device; |
| 4071 | |
| 4072 | peer_state.i = be32_to_cpu(p->state); |
| 4073 | |
| 4074 | real_peer_disk = peer_state.disk; |
| 4075 | if (peer_state.disk == D_NEGOTIATING) { |
| 4076 | real_peer_disk = device->p_uuid[UI_FLAGS] & 4 ? D_INCONSISTENT : D_CONSISTENT; |
| 4077 | drbd_info(device, "real peer disk state = %s\n", drbd_disk_str(real_peer_disk)); |
| 4078 | } |
| 4079 | |
| 4080 | spin_lock_irq(&device->resource->req_lock); |
| 4081 | retry: |
| 4082 | os = ns = drbd_read_state(device); |
| 4083 | spin_unlock_irq(&device->resource->req_lock); |
| 4084 | |
| 4085 | /* If some other part of the code (ack_receiver thread, timeout) |
| 4086 | * already decided to close the connection again, |
| 4087 | * we must not "re-establish" it here. */ |
| 4088 | if (os.conn <= C_TEAR_DOWN) |
| 4089 | return -ECONNRESET; |
| 4090 | |
| 4091 | /* If this is the "end of sync" confirmation, usually the peer disk |
| 4092 | * transitions from D_INCONSISTENT to D_UP_TO_DATE. For empty (0 bits |
| 4093 | * set) resync started in PausedSyncT, or if the timing of pause-/ |
| 4094 | * unpause-sync events has been "just right", the peer disk may |
| 4095 | * transition from D_CONSISTENT to D_UP_TO_DATE as well. |
| 4096 | */ |
| 4097 | if ((os.pdsk == D_INCONSISTENT || os.pdsk == D_CONSISTENT) && |
| 4098 | real_peer_disk == D_UP_TO_DATE && |
| 4099 | os.conn > C_CONNECTED && os.disk == D_UP_TO_DATE) { |
| 4100 | /* If we are (becoming) SyncSource, but peer is still in sync |
| 4101 | * preparation, ignore its uptodate-ness to avoid flapping, it |
| 4102 | * will change to inconsistent once the peer reaches active |
| 4103 | * syncing states. |
| 4104 | * It may have changed syncer-paused flags, however, so we |
| 4105 | * cannot ignore this completely. */ |
| 4106 | if (peer_state.conn > C_CONNECTED && |
| 4107 | peer_state.conn < C_SYNC_SOURCE) |
| 4108 | real_peer_disk = D_INCONSISTENT; |
| 4109 | |
| 4110 | /* if peer_state changes to connected at the same time, |
| 4111 | * it explicitly notifies us that it finished resync. |
| 4112 | * Maybe we should finish it up, too? */ |
| 4113 | else if (os.conn >= C_SYNC_SOURCE && |
| 4114 | peer_state.conn == C_CONNECTED) { |
| 4115 | if (drbd_bm_total_weight(device) <= device->rs_failed) |
| 4116 | drbd_resync_finished(device); |
| 4117 | return 0; |
| 4118 | } |
| 4119 | } |
| 4120 | |
| 4121 | /* explicit verify finished notification, stop sector reached. */ |
| 4122 | if (os.conn == C_VERIFY_T && os.disk == D_UP_TO_DATE && |
| 4123 | peer_state.conn == C_CONNECTED && real_peer_disk == D_UP_TO_DATE) { |
| 4124 | ov_out_of_sync_print(device); |
| 4125 | drbd_resync_finished(device); |
| 4126 | return 0; |
| 4127 | } |
| 4128 | |
| 4129 | /* peer says his disk is inconsistent, while we think it is uptodate, |
| 4130 | * and this happens while the peer still thinks we have a sync going on, |
| 4131 | * but we think we are already done with the sync. |
| 4132 | * We ignore this to avoid flapping pdsk. |
| 4133 | * This should not happen, if the peer is a recent version of drbd. */ |
| 4134 | if (os.pdsk == D_UP_TO_DATE && real_peer_disk == D_INCONSISTENT && |
| 4135 | os.conn == C_CONNECTED && peer_state.conn > C_SYNC_SOURCE) |
| 4136 | real_peer_disk = D_UP_TO_DATE; |
| 4137 | |
| 4138 | if (ns.conn == C_WF_REPORT_PARAMS) |
| 4139 | ns.conn = C_CONNECTED; |
| 4140 | |
| 4141 | if (peer_state.conn == C_AHEAD) |
| 4142 | ns.conn = C_BEHIND; |
| 4143 | |
| 4144 | if (device->p_uuid && peer_state.disk >= D_NEGOTIATING && |
| 4145 | get_ldev_if_state(device, D_NEGOTIATING)) { |
| 4146 | int cr; /* consider resync */ |
| 4147 | |
| 4148 | /* if we established a new connection */ |
| 4149 | cr = (os.conn < C_CONNECTED); |
| 4150 | /* if we had an established connection |
| 4151 | * and one of the nodes newly attaches a disk */ |
| 4152 | cr |= (os.conn == C_CONNECTED && |
| 4153 | (peer_state.disk == D_NEGOTIATING || |
| 4154 | os.disk == D_NEGOTIATING)); |
| 4155 | /* if we have both been inconsistent, and the peer has been |
| 4156 | * forced to be UpToDate with --overwrite-data */ |
| 4157 | cr |= test_bit(CONSIDER_RESYNC, &device->flags); |
| 4158 | /* if we had been plain connected, and the admin requested to |
| 4159 | * start a sync by "invalidate" or "invalidate-remote" */ |
| 4160 | cr |= (os.conn == C_CONNECTED && |
| 4161 | (peer_state.conn >= C_STARTING_SYNC_S && |
| 4162 | peer_state.conn <= C_WF_BITMAP_T)); |
| 4163 | |
| 4164 | if (cr) |
| 4165 | ns.conn = drbd_sync_handshake(peer_device, peer_state.role, real_peer_disk); |
| 4166 | |
| 4167 | put_ldev(device); |
| 4168 | if (ns.conn == C_MASK) { |
| 4169 | ns.conn = C_CONNECTED; |
| 4170 | if (device->state.disk == D_NEGOTIATING) { |
| 4171 | drbd_force_state(device, NS(disk, D_FAILED)); |
| 4172 | } else if (peer_state.disk == D_NEGOTIATING) { |
| 4173 | drbd_err(device, "Disk attach process on the peer node was aborted.\n"); |
| 4174 | peer_state.disk = D_DISKLESS; |
| 4175 | real_peer_disk = D_DISKLESS; |
| 4176 | } else { |
| 4177 | if (test_and_clear_bit(CONN_DRY_RUN, &peer_device->connection->flags)) |
| 4178 | return -EIO; |
| 4179 | D_ASSERT(device, os.conn == C_WF_REPORT_PARAMS); |
| 4180 | conn_request_state(peer_device->connection, NS(conn, C_DISCONNECTING), CS_HARD); |
| 4181 | return -EIO; |
| 4182 | } |
| 4183 | } |
| 4184 | } |
| 4185 | |
| 4186 | spin_lock_irq(&device->resource->req_lock); |
| 4187 | if (os.i != drbd_read_state(device).i) |
| 4188 | goto retry; |
| 4189 | clear_bit(CONSIDER_RESYNC, &device->flags); |
| 4190 | ns.peer = peer_state.role; |
| 4191 | ns.pdsk = real_peer_disk; |
| 4192 | ns.peer_isp = (peer_state.aftr_isp | peer_state.user_isp); |
| 4193 | if ((ns.conn == C_CONNECTED || ns.conn == C_WF_BITMAP_S) && ns.disk == D_NEGOTIATING) |
| 4194 | ns.disk = device->new_state_tmp.disk; |
| 4195 | cs_flags = CS_VERBOSE + (os.conn < C_CONNECTED && ns.conn >= C_CONNECTED ? 0 : CS_HARD); |
| 4196 | if (ns.pdsk == D_CONSISTENT && drbd_suspended(device) && ns.conn == C_CONNECTED && os.conn < C_CONNECTED && |
| 4197 | test_bit(NEW_CUR_UUID, &device->flags)) { |
| 4198 | /* Do not allow tl_restart(RESEND) for a rebooted peer. We can only allow this |
| 4199 | for temporal network outages! */ |
| 4200 | spin_unlock_irq(&device->resource->req_lock); |
| 4201 | drbd_err(device, "Aborting Connect, can not thaw IO with an only Consistent peer\n"); |
| 4202 | tl_clear(peer_device->connection); |
| 4203 | drbd_uuid_new_current(device); |
| 4204 | clear_bit(NEW_CUR_UUID, &device->flags); |
| 4205 | conn_request_state(peer_device->connection, NS2(conn, C_PROTOCOL_ERROR, susp, 0), CS_HARD); |
| 4206 | return -EIO; |
| 4207 | } |
| 4208 | rv = _drbd_set_state(device, ns, cs_flags, NULL); |
| 4209 | ns = drbd_read_state(device); |
| 4210 | spin_unlock_irq(&device->resource->req_lock); |
| 4211 | |
| 4212 | if (rv < SS_SUCCESS) { |
| 4213 | conn_request_state(peer_device->connection, NS(conn, C_DISCONNECTING), CS_HARD); |
| 4214 | return -EIO; |
| 4215 | } |
| 4216 | |
| 4217 | if (os.conn > C_WF_REPORT_PARAMS) { |
| 4218 | if (ns.conn > C_CONNECTED && peer_state.conn <= C_CONNECTED && |
| 4219 | peer_state.disk != D_NEGOTIATING ) { |
| 4220 | /* we want resync, peer has not yet decided to sync... */ |
| 4221 | /* Nowadays only used when forcing a node into primary role and |
| 4222 | setting its disk to UpToDate with that */ |
| 4223 | drbd_send_uuids(peer_device); |
| 4224 | drbd_send_current_state(peer_device); |
| 4225 | } |
| 4226 | } |
| 4227 | |
| 4228 | clear_bit(DISCARD_MY_DATA, &device->flags); |
| 4229 | |
| 4230 | drbd_md_sync(device); /* update connected indicator, la_size_sect, ... */ |
| 4231 | |
| 4232 | return 0; |
| 4233 | } |
| 4234 | |
| 4235 | static int receive_sync_uuid(struct drbd_connection *connection, struct packet_info *pi) |
| 4236 | { |
| 4237 | struct drbd_peer_device *peer_device; |
| 4238 | struct drbd_device *device; |
| 4239 | struct p_rs_uuid *p = pi->data; |
| 4240 | |
| 4241 | peer_device = conn_peer_device(connection, pi->vnr); |
| 4242 | if (!peer_device) |
| 4243 | return -EIO; |
| 4244 | device = peer_device->device; |
| 4245 | |
| 4246 | wait_event(device->misc_wait, |
| 4247 | device->state.conn == C_WF_SYNC_UUID || |
| 4248 | device->state.conn == C_BEHIND || |
| 4249 | device->state.conn < C_CONNECTED || |
| 4250 | device->state.disk < D_NEGOTIATING); |
| 4251 | |
| 4252 | /* D_ASSERT(device, device->state.conn == C_WF_SYNC_UUID ); */ |
| 4253 | |
| 4254 | /* Here the _drbd_uuid_ functions are right, current should |
| 4255 | _not_ be rotated into the history */ |
| 4256 | if (get_ldev_if_state(device, D_NEGOTIATING)) { |
| 4257 | _drbd_uuid_set(device, UI_CURRENT, be64_to_cpu(p->uuid)); |
| 4258 | _drbd_uuid_set(device, UI_BITMAP, 0UL); |
| 4259 | |
| 4260 | drbd_print_uuids(device, "updated sync uuid"); |
| 4261 | drbd_start_resync(device, C_SYNC_TARGET); |
| 4262 | |
| 4263 | put_ldev(device); |
| 4264 | } else |
| 4265 | drbd_err(device, "Ignoring SyncUUID packet!\n"); |
| 4266 | |
| 4267 | return 0; |
| 4268 | } |
| 4269 | |
| 4270 | /** |
| 4271 | * receive_bitmap_plain |
| 4272 | * |
| 4273 | * Return 0 when done, 1 when another iteration is needed, and a negative error |
| 4274 | * code upon failure. |
| 4275 | */ |
| 4276 | static int |
| 4277 | receive_bitmap_plain(struct drbd_peer_device *peer_device, unsigned int size, |
| 4278 | unsigned long *p, struct bm_xfer_ctx *c) |
| 4279 | { |
| 4280 | unsigned int data_size = DRBD_SOCKET_BUFFER_SIZE - |
| 4281 | drbd_header_size(peer_device->connection); |
| 4282 | unsigned int num_words = min_t(size_t, data_size / sizeof(*p), |
| 4283 | c->bm_words - c->word_offset); |
| 4284 | unsigned int want = num_words * sizeof(*p); |
| 4285 | int err; |
| 4286 | |
| 4287 | if (want != size) { |
| 4288 | drbd_err(peer_device, "%s:want (%u) != size (%u)\n", __func__, want, size); |
| 4289 | return -EIO; |
| 4290 | } |
| 4291 | if (want == 0) |
| 4292 | return 0; |
| 4293 | err = drbd_recv_all(peer_device->connection, p, want); |
| 4294 | if (err) |
| 4295 | return err; |
| 4296 | |
| 4297 | drbd_bm_merge_lel(peer_device->device, c->word_offset, num_words, p); |
| 4298 | |
| 4299 | c->word_offset += num_words; |
| 4300 | c->bit_offset = c->word_offset * BITS_PER_LONG; |
| 4301 | if (c->bit_offset > c->bm_bits) |
| 4302 | c->bit_offset = c->bm_bits; |
| 4303 | |
| 4304 | return 1; |
| 4305 | } |
| 4306 | |
| 4307 | static enum drbd_bitmap_code dcbp_get_code(struct p_compressed_bm *p) |
| 4308 | { |
| 4309 | return (enum drbd_bitmap_code)(p->encoding & 0x0f); |
| 4310 | } |
| 4311 | |
| 4312 | static int dcbp_get_start(struct p_compressed_bm *p) |
| 4313 | { |
| 4314 | return (p->encoding & 0x80) != 0; |
| 4315 | } |
| 4316 | |
| 4317 | static int dcbp_get_pad_bits(struct p_compressed_bm *p) |
| 4318 | { |
| 4319 | return (p->encoding >> 4) & 0x7; |
| 4320 | } |
| 4321 | |
| 4322 | /** |
| 4323 | * recv_bm_rle_bits |
| 4324 | * |
| 4325 | * Return 0 when done, 1 when another iteration is needed, and a negative error |
| 4326 | * code upon failure. |
| 4327 | */ |
| 4328 | static int |
| 4329 | recv_bm_rle_bits(struct drbd_peer_device *peer_device, |
| 4330 | struct p_compressed_bm *p, |
| 4331 | struct bm_xfer_ctx *c, |
| 4332 | unsigned int len) |
| 4333 | { |
| 4334 | struct bitstream bs; |
| 4335 | u64 look_ahead; |
| 4336 | u64 rl; |
| 4337 | u64 tmp; |
| 4338 | unsigned long s = c->bit_offset; |
| 4339 | unsigned long e; |
| 4340 | int toggle = dcbp_get_start(p); |
| 4341 | int have; |
| 4342 | int bits; |
| 4343 | |
| 4344 | bitstream_init(&bs, p->code, len, dcbp_get_pad_bits(p)); |
| 4345 | |
| 4346 | bits = bitstream_get_bits(&bs, &look_ahead, 64); |
| 4347 | if (bits < 0) |
| 4348 | return -EIO; |
| 4349 | |
| 4350 | for (have = bits; have > 0; s += rl, toggle = !toggle) { |
| 4351 | bits = vli_decode_bits(&rl, look_ahead); |
| 4352 | if (bits <= 0) |
| 4353 | return -EIO; |
| 4354 | |
| 4355 | if (toggle) { |
| 4356 | e = s + rl -1; |
| 4357 | if (e >= c->bm_bits) { |
| 4358 | drbd_err(peer_device, "bitmap overflow (e:%lu) while decoding bm RLE packet\n", e); |
| 4359 | return -EIO; |
| 4360 | } |
| 4361 | _drbd_bm_set_bits(peer_device->device, s, e); |
| 4362 | } |
| 4363 | |
| 4364 | if (have < bits) { |
| 4365 | drbd_err(peer_device, "bitmap decoding error: h:%d b:%d la:0x%08llx l:%u/%u\n", |
| 4366 | have, bits, look_ahead, |
| 4367 | (unsigned int)(bs.cur.b - p->code), |
| 4368 | (unsigned int)bs.buf_len); |
| 4369 | return -EIO; |
| 4370 | } |
| 4371 | /* if we consumed all 64 bits, assign 0; >> 64 is "undefined"; */ |
| 4372 | if (likely(bits < 64)) |
| 4373 | look_ahead >>= bits; |
| 4374 | else |
| 4375 | look_ahead = 0; |
| 4376 | have -= bits; |
| 4377 | |
| 4378 | bits = bitstream_get_bits(&bs, &tmp, 64 - have); |
| 4379 | if (bits < 0) |
| 4380 | return -EIO; |
| 4381 | look_ahead |= tmp << have; |
| 4382 | have += bits; |
| 4383 | } |
| 4384 | |
| 4385 | c->bit_offset = s; |
| 4386 | bm_xfer_ctx_bit_to_word_offset(c); |
| 4387 | |
| 4388 | return (s != c->bm_bits); |
| 4389 | } |
| 4390 | |
| 4391 | /** |
| 4392 | * decode_bitmap_c |
| 4393 | * |
| 4394 | * Return 0 when done, 1 when another iteration is needed, and a negative error |
| 4395 | * code upon failure. |
| 4396 | */ |
| 4397 | static int |
| 4398 | decode_bitmap_c(struct drbd_peer_device *peer_device, |
| 4399 | struct p_compressed_bm *p, |
| 4400 | struct bm_xfer_ctx *c, |
| 4401 | unsigned int len) |
| 4402 | { |
| 4403 | if (dcbp_get_code(p) == RLE_VLI_Bits) |
| 4404 | return recv_bm_rle_bits(peer_device, p, c, len - sizeof(*p)); |
| 4405 | |
| 4406 | /* other variants had been implemented for evaluation, |
| 4407 | * but have been dropped as this one turned out to be "best" |
| 4408 | * during all our tests. */ |
| 4409 | |
| 4410 | drbd_err(peer_device, "receive_bitmap_c: unknown encoding %u\n", p->encoding); |
| 4411 | conn_request_state(peer_device->connection, NS(conn, C_PROTOCOL_ERROR), CS_HARD); |
| 4412 | return -EIO; |
| 4413 | } |
| 4414 | |
| 4415 | void INFO_bm_xfer_stats(struct drbd_device *device, |
| 4416 | const char *direction, struct bm_xfer_ctx *c) |
| 4417 | { |
| 4418 | /* what would it take to transfer it "plaintext" */ |
| 4419 | unsigned int header_size = drbd_header_size(first_peer_device(device)->connection); |
| 4420 | unsigned int data_size = DRBD_SOCKET_BUFFER_SIZE - header_size; |
| 4421 | unsigned int plain = |
| 4422 | header_size * (DIV_ROUND_UP(c->bm_words, data_size) + 1) + |
| 4423 | c->bm_words * sizeof(unsigned long); |
| 4424 | unsigned int total = c->bytes[0] + c->bytes[1]; |
| 4425 | unsigned int r; |
| 4426 | |
| 4427 | /* total can not be zero. but just in case: */ |
| 4428 | if (total == 0) |
| 4429 | return; |
| 4430 | |
| 4431 | /* don't report if not compressed */ |
| 4432 | if (total >= plain) |
| 4433 | return; |
| 4434 | |
| 4435 | /* total < plain. check for overflow, still */ |
| 4436 | r = (total > UINT_MAX/1000) ? (total / (plain/1000)) |
| 4437 | : (1000 * total / plain); |
| 4438 | |
| 4439 | if (r > 1000) |
| 4440 | r = 1000; |
| 4441 | |
| 4442 | r = 1000 - r; |
| 4443 | drbd_info(device, "%s bitmap stats [Bytes(packets)]: plain %u(%u), RLE %u(%u), " |
| 4444 | "total %u; compression: %u.%u%%\n", |
| 4445 | direction, |
| 4446 | c->bytes[1], c->packets[1], |
| 4447 | c->bytes[0], c->packets[0], |
| 4448 | total, r/10, r % 10); |
| 4449 | } |
| 4450 | |
| 4451 | /* Since we are processing the bitfield from lower addresses to higher, |
| 4452 | it does not matter if the process it in 32 bit chunks or 64 bit |
| 4453 | chunks as long as it is little endian. (Understand it as byte stream, |
| 4454 | beginning with the lowest byte...) If we would use big endian |
| 4455 | we would need to process it from the highest address to the lowest, |
| 4456 | in order to be agnostic to the 32 vs 64 bits issue. |
| 4457 | |
| 4458 | returns 0 on failure, 1 if we successfully received it. */ |
| 4459 | static int receive_bitmap(struct drbd_connection *connection, struct packet_info *pi) |
| 4460 | { |
| 4461 | struct drbd_peer_device *peer_device; |
| 4462 | struct drbd_device *device; |
| 4463 | struct bm_xfer_ctx c; |
| 4464 | int err; |
| 4465 | |
| 4466 | peer_device = conn_peer_device(connection, pi->vnr); |
| 4467 | if (!peer_device) |
| 4468 | return -EIO; |
| 4469 | device = peer_device->device; |
| 4470 | |
| 4471 | drbd_bm_lock(device, "receive bitmap", BM_LOCKED_SET_ALLOWED); |
| 4472 | /* you are supposed to send additional out-of-sync information |
| 4473 | * if you actually set bits during this phase */ |
| 4474 | |
| 4475 | c = (struct bm_xfer_ctx) { |
| 4476 | .bm_bits = drbd_bm_bits(device), |
| 4477 | .bm_words = drbd_bm_words(device), |
| 4478 | }; |
| 4479 | |
| 4480 | for(;;) { |
| 4481 | if (pi->cmd == P_BITMAP) |
| 4482 | err = receive_bitmap_plain(peer_device, pi->size, pi->data, &c); |
| 4483 | else if (pi->cmd == P_COMPRESSED_BITMAP) { |
| 4484 | /* MAYBE: sanity check that we speak proto >= 90, |
| 4485 | * and the feature is enabled! */ |
| 4486 | struct p_compressed_bm *p = pi->data; |
| 4487 | |
| 4488 | if (pi->size > DRBD_SOCKET_BUFFER_SIZE - drbd_header_size(connection)) { |
| 4489 | drbd_err(device, "ReportCBitmap packet too large\n"); |
| 4490 | err = -EIO; |
| 4491 | goto out; |
| 4492 | } |
| 4493 | if (pi->size <= sizeof(*p)) { |
| 4494 | drbd_err(device, "ReportCBitmap packet too small (l:%u)\n", pi->size); |
| 4495 | err = -EIO; |
| 4496 | goto out; |
| 4497 | } |
| 4498 | err = drbd_recv_all(peer_device->connection, p, pi->size); |
| 4499 | if (err) |
| 4500 | goto out; |
| 4501 | err = decode_bitmap_c(peer_device, p, &c, pi->size); |
| 4502 | } else { |
| 4503 | drbd_warn(device, "receive_bitmap: cmd neither ReportBitMap nor ReportCBitMap (is 0x%x)", pi->cmd); |
| 4504 | err = -EIO; |
| 4505 | goto out; |
| 4506 | } |
| 4507 | |
| 4508 | c.packets[pi->cmd == P_BITMAP]++; |
| 4509 | c.bytes[pi->cmd == P_BITMAP] += drbd_header_size(connection) + pi->size; |
| 4510 | |
| 4511 | if (err <= 0) { |
| 4512 | if (err < 0) |
| 4513 | goto out; |
| 4514 | break; |
| 4515 | } |
| 4516 | err = drbd_recv_header(peer_device->connection, pi); |
| 4517 | if (err) |
| 4518 | goto out; |
| 4519 | } |
| 4520 | |
| 4521 | INFO_bm_xfer_stats(device, "receive", &c); |
| 4522 | |
| 4523 | if (device->state.conn == C_WF_BITMAP_T) { |
| 4524 | enum drbd_state_rv rv; |
| 4525 | |
| 4526 | err = drbd_send_bitmap(device); |
| 4527 | if (err) |
| 4528 | goto out; |
| 4529 | /* Omit CS_ORDERED with this state transition to avoid deadlocks. */ |
| 4530 | rv = _drbd_request_state(device, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE); |
| 4531 | D_ASSERT(device, rv == SS_SUCCESS); |
| 4532 | } else if (device->state.conn != C_WF_BITMAP_S) { |
| 4533 | /* admin may have requested C_DISCONNECTING, |
| 4534 | * other threads may have noticed network errors */ |
| 4535 | drbd_info(device, "unexpected cstate (%s) in receive_bitmap\n", |
| 4536 | drbd_conn_str(device->state.conn)); |
| 4537 | } |
| 4538 | err = 0; |
| 4539 | |
| 4540 | out: |
| 4541 | drbd_bm_unlock(device); |
| 4542 | if (!err && device->state.conn == C_WF_BITMAP_S) |
| 4543 | drbd_start_resync(device, C_SYNC_SOURCE); |
| 4544 | return err; |
| 4545 | } |
| 4546 | |
| 4547 | static int receive_skip(struct drbd_connection *connection, struct packet_info *pi) |
| 4548 | { |
| 4549 | drbd_warn(connection, "skipping unknown optional packet type %d, l: %d!\n", |
| 4550 | pi->cmd, pi->size); |
| 4551 | |
| 4552 | return ignore_remaining_packet(connection, pi); |
| 4553 | } |
| 4554 | |
| 4555 | static int receive_UnplugRemote(struct drbd_connection *connection, struct packet_info *pi) |
| 4556 | { |
| 4557 | /* Make sure we've acked all the TCP data associated |
| 4558 | * with the data requests being unplugged */ |
| 4559 | drbd_tcp_quickack(connection->data.socket); |
| 4560 | |
| 4561 | return 0; |
| 4562 | } |
| 4563 | |
| 4564 | static int receive_out_of_sync(struct drbd_connection *connection, struct packet_info *pi) |
| 4565 | { |
| 4566 | struct drbd_peer_device *peer_device; |
| 4567 | struct drbd_device *device; |
| 4568 | struct p_block_desc *p = pi->data; |
| 4569 | |
| 4570 | peer_device = conn_peer_device(connection, pi->vnr); |
| 4571 | if (!peer_device) |
| 4572 | return -EIO; |
| 4573 | device = peer_device->device; |
| 4574 | |
| 4575 | switch (device->state.conn) { |
| 4576 | case C_WF_SYNC_UUID: |
| 4577 | case C_WF_BITMAP_T: |
| 4578 | case C_BEHIND: |
| 4579 | break; |
| 4580 | default: |
| 4581 | drbd_err(device, "ASSERT FAILED cstate = %s, expected: WFSyncUUID|WFBitMapT|Behind\n", |
| 4582 | drbd_conn_str(device->state.conn)); |
| 4583 | } |
| 4584 | |
| 4585 | drbd_set_out_of_sync(device, be64_to_cpu(p->sector), be32_to_cpu(p->blksize)); |
| 4586 | |
| 4587 | return 0; |
| 4588 | } |
| 4589 | |
| 4590 | struct data_cmd { |
| 4591 | int expect_payload; |
| 4592 | size_t pkt_size; |
| 4593 | int (*fn)(struct drbd_connection *, struct packet_info *); |
| 4594 | }; |
| 4595 | |
| 4596 | static struct data_cmd drbd_cmd_handler[] = { |
| 4597 | [P_DATA] = { 1, sizeof(struct p_data), receive_Data }, |
| 4598 | [P_DATA_REPLY] = { 1, sizeof(struct p_data), receive_DataReply }, |
| 4599 | [P_RS_DATA_REPLY] = { 1, sizeof(struct p_data), receive_RSDataReply } , |
| 4600 | [P_BARRIER] = { 0, sizeof(struct p_barrier), receive_Barrier } , |
| 4601 | [P_BITMAP] = { 1, 0, receive_bitmap } , |
| 4602 | [P_COMPRESSED_BITMAP] = { 1, 0, receive_bitmap } , |
| 4603 | [P_UNPLUG_REMOTE] = { 0, 0, receive_UnplugRemote }, |
| 4604 | [P_DATA_REQUEST] = { 0, sizeof(struct p_block_req), receive_DataRequest }, |
| 4605 | [P_RS_DATA_REQUEST] = { 0, sizeof(struct p_block_req), receive_DataRequest }, |
| 4606 | [P_SYNC_PARAM] = { 1, 0, receive_SyncParam }, |
| 4607 | [P_SYNC_PARAM89] = { 1, 0, receive_SyncParam }, |
| 4608 | [P_PROTOCOL] = { 1, sizeof(struct p_protocol), receive_protocol }, |
| 4609 | [P_UUIDS] = { 0, sizeof(struct p_uuids), receive_uuids }, |
| 4610 | [P_SIZES] = { 0, sizeof(struct p_sizes), receive_sizes }, |
| 4611 | [P_STATE] = { 0, sizeof(struct p_state), receive_state }, |
| 4612 | [P_STATE_CHG_REQ] = { 0, sizeof(struct p_req_state), receive_req_state }, |
| 4613 | [P_SYNC_UUID] = { 0, sizeof(struct p_rs_uuid), receive_sync_uuid }, |
| 4614 | [P_OV_REQUEST] = { 0, sizeof(struct p_block_req), receive_DataRequest }, |
| 4615 | [P_OV_REPLY] = { 1, sizeof(struct p_block_req), receive_DataRequest }, |
| 4616 | [P_CSUM_RS_REQUEST] = { 1, sizeof(struct p_block_req), receive_DataRequest }, |
| 4617 | [P_DELAY_PROBE] = { 0, sizeof(struct p_delay_probe93), receive_skip }, |
| 4618 | [P_OUT_OF_SYNC] = { 0, sizeof(struct p_block_desc), receive_out_of_sync }, |
| 4619 | [P_CONN_ST_CHG_REQ] = { 0, sizeof(struct p_req_state), receive_req_conn_state }, |
| 4620 | [P_PROTOCOL_UPDATE] = { 1, sizeof(struct p_protocol), receive_protocol }, |
| 4621 | [P_TRIM] = { 0, sizeof(struct p_trim), receive_Data }, |
| 4622 | }; |
| 4623 | |
| 4624 | static void drbdd(struct drbd_connection *connection) |
| 4625 | { |
| 4626 | struct packet_info pi; |
| 4627 | size_t shs; /* sub header size */ |
| 4628 | int err; |
| 4629 | |
| 4630 | while (get_t_state(&connection->receiver) == RUNNING) { |
| 4631 | struct data_cmd *cmd; |
| 4632 | |
| 4633 | drbd_thread_current_set_cpu(&connection->receiver); |
| 4634 | update_receiver_timing_details(connection, drbd_recv_header); |
| 4635 | if (drbd_recv_header(connection, &pi)) |
| 4636 | goto err_out; |
| 4637 | |
| 4638 | cmd = &drbd_cmd_handler[pi.cmd]; |
| 4639 | if (unlikely(pi.cmd >= ARRAY_SIZE(drbd_cmd_handler) || !cmd->fn)) { |
| 4640 | drbd_err(connection, "Unexpected data packet %s (0x%04x)", |
| 4641 | cmdname(pi.cmd), pi.cmd); |
| 4642 | goto err_out; |
| 4643 | } |
| 4644 | |
| 4645 | shs = cmd->pkt_size; |
| 4646 | if (pi.size > shs && !cmd->expect_payload) { |
| 4647 | drbd_err(connection, "No payload expected %s l:%d\n", |
| 4648 | cmdname(pi.cmd), pi.size); |
| 4649 | goto err_out; |
| 4650 | } |
| 4651 | |
| 4652 | if (shs) { |
| 4653 | update_receiver_timing_details(connection, drbd_recv_all_warn); |
| 4654 | err = drbd_recv_all_warn(connection, pi.data, shs); |
| 4655 | if (err) |
| 4656 | goto err_out; |
| 4657 | pi.size -= shs; |
| 4658 | } |
| 4659 | |
| 4660 | update_receiver_timing_details(connection, cmd->fn); |
| 4661 | err = cmd->fn(connection, &pi); |
| 4662 | if (err) { |
| 4663 | drbd_err(connection, "error receiving %s, e: %d l: %d!\n", |
| 4664 | cmdname(pi.cmd), err, pi.size); |
| 4665 | goto err_out; |
| 4666 | } |
| 4667 | } |
| 4668 | return; |
| 4669 | |
| 4670 | err_out: |
| 4671 | conn_request_state(connection, NS(conn, C_PROTOCOL_ERROR), CS_HARD); |
| 4672 | } |
| 4673 | |
| 4674 | static void conn_disconnect(struct drbd_connection *connection) |
| 4675 | { |
| 4676 | struct drbd_peer_device *peer_device; |
| 4677 | enum drbd_conns oc; |
| 4678 | int vnr; |
| 4679 | |
| 4680 | if (connection->cstate == C_STANDALONE) |
| 4681 | return; |
| 4682 | |
| 4683 | /* We are about to start the cleanup after connection loss. |
| 4684 | * Make sure drbd_make_request knows about that. |
| 4685 | * Usually we should be in some network failure state already, |
| 4686 | * but just in case we are not, we fix it up here. |
| 4687 | */ |
| 4688 | conn_request_state(connection, NS(conn, C_NETWORK_FAILURE), CS_HARD); |
| 4689 | |
| 4690 | /* ack_receiver does not clean up anything. it must not interfere, either */ |
| 4691 | drbd_thread_stop(&connection->ack_receiver); |
| 4692 | if (connection->ack_sender) { |
| 4693 | destroy_workqueue(connection->ack_sender); |
| 4694 | connection->ack_sender = NULL; |
| 4695 | } |
| 4696 | drbd_free_sock(connection); |
| 4697 | |
| 4698 | rcu_read_lock(); |
| 4699 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
| 4700 | struct drbd_device *device = peer_device->device; |
| 4701 | kref_get(&device->kref); |
| 4702 | rcu_read_unlock(); |
| 4703 | drbd_disconnected(peer_device); |
| 4704 | kref_put(&device->kref, drbd_destroy_device); |
| 4705 | rcu_read_lock(); |
| 4706 | } |
| 4707 | rcu_read_unlock(); |
| 4708 | |
| 4709 | if (!list_empty(&connection->current_epoch->list)) |
| 4710 | drbd_err(connection, "ASSERTION FAILED: connection->current_epoch->list not empty\n"); |
| 4711 | /* ok, no more ee's on the fly, it is safe to reset the epoch_size */ |
| 4712 | atomic_set(&connection->current_epoch->epoch_size, 0); |
| 4713 | connection->send.seen_any_write_yet = false; |
| 4714 | |
| 4715 | drbd_info(connection, "Connection closed\n"); |
| 4716 | |
| 4717 | if (conn_highest_role(connection) == R_PRIMARY && conn_highest_pdsk(connection) >= D_UNKNOWN) |
| 4718 | conn_try_outdate_peer_async(connection); |
| 4719 | |
| 4720 | spin_lock_irq(&connection->resource->req_lock); |
| 4721 | oc = connection->cstate; |
| 4722 | if (oc >= C_UNCONNECTED) |
| 4723 | _conn_request_state(connection, NS(conn, C_UNCONNECTED), CS_VERBOSE); |
| 4724 | |
| 4725 | spin_unlock_irq(&connection->resource->req_lock); |
| 4726 | |
| 4727 | if (oc == C_DISCONNECTING) |
| 4728 | conn_request_state(connection, NS(conn, C_STANDALONE), CS_VERBOSE | CS_HARD); |
| 4729 | } |
| 4730 | |
| 4731 | static int drbd_disconnected(struct drbd_peer_device *peer_device) |
| 4732 | { |
| 4733 | struct drbd_device *device = peer_device->device; |
| 4734 | unsigned int i; |
| 4735 | |
| 4736 | /* wait for current activity to cease. */ |
| 4737 | spin_lock_irq(&device->resource->req_lock); |
| 4738 | _drbd_wait_ee_list_empty(device, &device->active_ee); |
| 4739 | _drbd_wait_ee_list_empty(device, &device->sync_ee); |
| 4740 | _drbd_wait_ee_list_empty(device, &device->read_ee); |
| 4741 | spin_unlock_irq(&device->resource->req_lock); |
| 4742 | |
| 4743 | /* We do not have data structures that would allow us to |
| 4744 | * get the rs_pending_cnt down to 0 again. |
| 4745 | * * On C_SYNC_TARGET we do not have any data structures describing |
| 4746 | * the pending RSDataRequest's we have sent. |
| 4747 | * * On C_SYNC_SOURCE there is no data structure that tracks |
| 4748 | * the P_RS_DATA_REPLY blocks that we sent to the SyncTarget. |
| 4749 | * And no, it is not the sum of the reference counts in the |
| 4750 | * resync_LRU. The resync_LRU tracks the whole operation including |
| 4751 | * the disk-IO, while the rs_pending_cnt only tracks the blocks |
| 4752 | * on the fly. */ |
| 4753 | drbd_rs_cancel_all(device); |
| 4754 | device->rs_total = 0; |
| 4755 | device->rs_failed = 0; |
| 4756 | atomic_set(&device->rs_pending_cnt, 0); |
| 4757 | wake_up(&device->misc_wait); |
| 4758 | |
| 4759 | del_timer_sync(&device->resync_timer); |
| 4760 | resync_timer_fn((unsigned long)device); |
| 4761 | |
| 4762 | /* wait for all w_e_end_data_req, w_e_end_rsdata_req, w_send_barrier, |
| 4763 | * w_make_resync_request etc. which may still be on the worker queue |
| 4764 | * to be "canceled" */ |
| 4765 | drbd_flush_workqueue(&peer_device->connection->sender_work); |
| 4766 | |
| 4767 | drbd_finish_peer_reqs(device); |
| 4768 | |
| 4769 | /* This second workqueue flush is necessary, since drbd_finish_peer_reqs() |
| 4770 | might have issued a work again. The one before drbd_finish_peer_reqs() is |
| 4771 | necessary to reclain net_ee in drbd_finish_peer_reqs(). */ |
| 4772 | drbd_flush_workqueue(&peer_device->connection->sender_work); |
| 4773 | |
| 4774 | /* need to do it again, drbd_finish_peer_reqs() may have populated it |
| 4775 | * again via drbd_try_clear_on_disk_bm(). */ |
| 4776 | drbd_rs_cancel_all(device); |
| 4777 | |
| 4778 | kfree(device->p_uuid); |
| 4779 | device->p_uuid = NULL; |
| 4780 | |
| 4781 | if (!drbd_suspended(device)) |
| 4782 | tl_clear(peer_device->connection); |
| 4783 | |
| 4784 | drbd_md_sync(device); |
| 4785 | |
| 4786 | /* serialize with bitmap writeout triggered by the state change, |
| 4787 | * if any. */ |
| 4788 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); |
| 4789 | |
| 4790 | /* tcp_close and release of sendpage pages can be deferred. I don't |
| 4791 | * want to use SO_LINGER, because apparently it can be deferred for |
| 4792 | * more than 20 seconds (longest time I checked). |
| 4793 | * |
| 4794 | * Actually we don't care for exactly when the network stack does its |
| 4795 | * put_page(), but release our reference on these pages right here. |
| 4796 | */ |
| 4797 | i = drbd_free_peer_reqs(device, &device->net_ee); |
| 4798 | if (i) |
| 4799 | drbd_info(device, "net_ee not empty, killed %u entries\n", i); |
| 4800 | i = atomic_read(&device->pp_in_use_by_net); |
| 4801 | if (i) |
| 4802 | drbd_info(device, "pp_in_use_by_net = %d, expected 0\n", i); |
| 4803 | i = atomic_read(&device->pp_in_use); |
| 4804 | if (i) |
| 4805 | drbd_info(device, "pp_in_use = %d, expected 0\n", i); |
| 4806 | |
| 4807 | D_ASSERT(device, list_empty(&device->read_ee)); |
| 4808 | D_ASSERT(device, list_empty(&device->active_ee)); |
| 4809 | D_ASSERT(device, list_empty(&device->sync_ee)); |
| 4810 | D_ASSERT(device, list_empty(&device->done_ee)); |
| 4811 | |
| 4812 | return 0; |
| 4813 | } |
| 4814 | |
| 4815 | /* |
| 4816 | * We support PRO_VERSION_MIN to PRO_VERSION_MAX. The protocol version |
| 4817 | * we can agree on is stored in agreed_pro_version. |
| 4818 | * |
| 4819 | * feature flags and the reserved array should be enough room for future |
| 4820 | * enhancements of the handshake protocol, and possible plugins... |
| 4821 | * |
| 4822 | * for now, they are expected to be zero, but ignored. |
| 4823 | */ |
| 4824 | static int drbd_send_features(struct drbd_connection *connection) |
| 4825 | { |
| 4826 | struct drbd_socket *sock; |
| 4827 | struct p_connection_features *p; |
| 4828 | |
| 4829 | sock = &connection->data; |
| 4830 | p = conn_prepare_command(connection, sock); |
| 4831 | if (!p) |
| 4832 | return -EIO; |
| 4833 | memset(p, 0, sizeof(*p)); |
| 4834 | p->protocol_min = cpu_to_be32(PRO_VERSION_MIN); |
| 4835 | p->protocol_max = cpu_to_be32(PRO_VERSION_MAX); |
| 4836 | p->feature_flags = cpu_to_be32(PRO_FEATURES); |
| 4837 | return conn_send_command(connection, sock, P_CONNECTION_FEATURES, sizeof(*p), NULL, 0); |
| 4838 | } |
| 4839 | |
| 4840 | /* |
| 4841 | * return values: |
| 4842 | * 1 yes, we have a valid connection |
| 4843 | * 0 oops, did not work out, please try again |
| 4844 | * -1 peer talks different language, |
| 4845 | * no point in trying again, please go standalone. |
| 4846 | */ |
| 4847 | static int drbd_do_features(struct drbd_connection *connection) |
| 4848 | { |
| 4849 | /* ASSERT current == connection->receiver ... */ |
| 4850 | struct p_connection_features *p; |
| 4851 | const int expect = sizeof(struct p_connection_features); |
| 4852 | struct packet_info pi; |
| 4853 | int err; |
| 4854 | |
| 4855 | err = drbd_send_features(connection); |
| 4856 | if (err) |
| 4857 | return 0; |
| 4858 | |
| 4859 | err = drbd_recv_header(connection, &pi); |
| 4860 | if (err) |
| 4861 | return 0; |
| 4862 | |
| 4863 | if (pi.cmd != P_CONNECTION_FEATURES) { |
| 4864 | drbd_err(connection, "expected ConnectionFeatures packet, received: %s (0x%04x)\n", |
| 4865 | cmdname(pi.cmd), pi.cmd); |
| 4866 | return -1; |
| 4867 | } |
| 4868 | |
| 4869 | if (pi.size != expect) { |
| 4870 | drbd_err(connection, "expected ConnectionFeatures length: %u, received: %u\n", |
| 4871 | expect, pi.size); |
| 4872 | return -1; |
| 4873 | } |
| 4874 | |
| 4875 | p = pi.data; |
| 4876 | err = drbd_recv_all_warn(connection, p, expect); |
| 4877 | if (err) |
| 4878 | return 0; |
| 4879 | |
| 4880 | p->protocol_min = be32_to_cpu(p->protocol_min); |
| 4881 | p->protocol_max = be32_to_cpu(p->protocol_max); |
| 4882 | if (p->protocol_max == 0) |
| 4883 | p->protocol_max = p->protocol_min; |
| 4884 | |
| 4885 | if (PRO_VERSION_MAX < p->protocol_min || |
| 4886 | PRO_VERSION_MIN > p->protocol_max) |
| 4887 | goto incompat; |
| 4888 | |
| 4889 | connection->agreed_pro_version = min_t(int, PRO_VERSION_MAX, p->protocol_max); |
| 4890 | connection->agreed_features = PRO_FEATURES & be32_to_cpu(p->feature_flags); |
| 4891 | |
| 4892 | drbd_info(connection, "Handshake successful: " |
| 4893 | "Agreed network protocol version %d\n", connection->agreed_pro_version); |
| 4894 | |
| 4895 | drbd_info(connection, "Agreed to%ssupport TRIM on protocol level\n", |
| 4896 | connection->agreed_features & FF_TRIM ? " " : " not "); |
| 4897 | |
| 4898 | return 1; |
| 4899 | |
| 4900 | incompat: |
| 4901 | drbd_err(connection, "incompatible DRBD dialects: " |
| 4902 | "I support %d-%d, peer supports %d-%d\n", |
| 4903 | PRO_VERSION_MIN, PRO_VERSION_MAX, |
| 4904 | p->protocol_min, p->protocol_max); |
| 4905 | return -1; |
| 4906 | } |
| 4907 | |
| 4908 | #if !defined(CONFIG_CRYPTO_HMAC) && !defined(CONFIG_CRYPTO_HMAC_MODULE) |
| 4909 | static int drbd_do_auth(struct drbd_connection *connection) |
| 4910 | { |
| 4911 | drbd_err(connection, "This kernel was build without CONFIG_CRYPTO_HMAC.\n"); |
| 4912 | drbd_err(connection, "You need to disable 'cram-hmac-alg' in drbd.conf.\n"); |
| 4913 | return -1; |
| 4914 | } |
| 4915 | #else |
| 4916 | #define CHALLENGE_LEN 64 |
| 4917 | |
| 4918 | /* Return value: |
| 4919 | 1 - auth succeeded, |
| 4920 | 0 - failed, try again (network error), |
| 4921 | -1 - auth failed, don't try again. |
| 4922 | */ |
| 4923 | |
| 4924 | static int drbd_do_auth(struct drbd_connection *connection) |
| 4925 | { |
| 4926 | struct drbd_socket *sock; |
| 4927 | char my_challenge[CHALLENGE_LEN]; /* 64 Bytes... */ |
| 4928 | char *response = NULL; |
| 4929 | char *right_response = NULL; |
| 4930 | char *peers_ch = NULL; |
| 4931 | unsigned int key_len; |
| 4932 | char secret[SHARED_SECRET_MAX]; /* 64 byte */ |
| 4933 | unsigned int resp_size; |
| 4934 | SHASH_DESC_ON_STACK(desc, connection->cram_hmac_tfm); |
| 4935 | struct packet_info pi; |
| 4936 | struct net_conf *nc; |
| 4937 | int err, rv; |
| 4938 | |
| 4939 | /* FIXME: Put the challenge/response into the preallocated socket buffer. */ |
| 4940 | |
| 4941 | rcu_read_lock(); |
| 4942 | nc = rcu_dereference(connection->net_conf); |
| 4943 | key_len = strlen(nc->shared_secret); |
| 4944 | memcpy(secret, nc->shared_secret, key_len); |
| 4945 | rcu_read_unlock(); |
| 4946 | |
| 4947 | desc->tfm = connection->cram_hmac_tfm; |
| 4948 | desc->flags = 0; |
| 4949 | |
| 4950 | rv = crypto_shash_setkey(connection->cram_hmac_tfm, (u8 *)secret, key_len); |
| 4951 | if (rv) { |
| 4952 | drbd_err(connection, "crypto_shash_setkey() failed with %d\n", rv); |
| 4953 | rv = -1; |
| 4954 | goto fail; |
| 4955 | } |
| 4956 | |
| 4957 | get_random_bytes(my_challenge, CHALLENGE_LEN); |
| 4958 | |
| 4959 | sock = &connection->data; |
| 4960 | if (!conn_prepare_command(connection, sock)) { |
| 4961 | rv = 0; |
| 4962 | goto fail; |
| 4963 | } |
| 4964 | rv = !conn_send_command(connection, sock, P_AUTH_CHALLENGE, 0, |
| 4965 | my_challenge, CHALLENGE_LEN); |
| 4966 | if (!rv) |
| 4967 | goto fail; |
| 4968 | |
| 4969 | err = drbd_recv_header(connection, &pi); |
| 4970 | if (err) { |
| 4971 | rv = 0; |
| 4972 | goto fail; |
| 4973 | } |
| 4974 | |
| 4975 | if (pi.cmd != P_AUTH_CHALLENGE) { |
| 4976 | drbd_err(connection, "expected AuthChallenge packet, received: %s (0x%04x)\n", |
| 4977 | cmdname(pi.cmd), pi.cmd); |
| 4978 | rv = 0; |
| 4979 | goto fail; |
| 4980 | } |
| 4981 | |
| 4982 | if (pi.size > CHALLENGE_LEN * 2) { |
| 4983 | drbd_err(connection, "expected AuthChallenge payload too big.\n"); |
| 4984 | rv = -1; |
| 4985 | goto fail; |
| 4986 | } |
| 4987 | |
| 4988 | if (pi.size < CHALLENGE_LEN) { |
| 4989 | drbd_err(connection, "AuthChallenge payload too small.\n"); |
| 4990 | rv = -1; |
| 4991 | goto fail; |
| 4992 | } |
| 4993 | |
| 4994 | peers_ch = kmalloc(pi.size, GFP_NOIO); |
| 4995 | if (peers_ch == NULL) { |
| 4996 | drbd_err(connection, "kmalloc of peers_ch failed\n"); |
| 4997 | rv = -1; |
| 4998 | goto fail; |
| 4999 | } |
| 5000 | |
| 5001 | err = drbd_recv_all_warn(connection, peers_ch, pi.size); |
| 5002 | if (err) { |
| 5003 | rv = 0; |
| 5004 | goto fail; |
| 5005 | } |
| 5006 | |
| 5007 | if (!memcmp(my_challenge, peers_ch, CHALLENGE_LEN)) { |
| 5008 | drbd_err(connection, "Peer presented the same challenge!\n"); |
| 5009 | rv = -1; |
| 5010 | goto fail; |
| 5011 | } |
| 5012 | |
| 5013 | resp_size = crypto_shash_digestsize(connection->cram_hmac_tfm); |
| 5014 | response = kmalloc(resp_size, GFP_NOIO); |
| 5015 | if (response == NULL) { |
| 5016 | drbd_err(connection, "kmalloc of response failed\n"); |
| 5017 | rv = -1; |
| 5018 | goto fail; |
| 5019 | } |
| 5020 | |
| 5021 | rv = crypto_shash_digest(desc, peers_ch, pi.size, response); |
| 5022 | if (rv) { |
| 5023 | drbd_err(connection, "crypto_hash_digest() failed with %d\n", rv); |
| 5024 | rv = -1; |
| 5025 | goto fail; |
| 5026 | } |
| 5027 | |
| 5028 | if (!conn_prepare_command(connection, sock)) { |
| 5029 | rv = 0; |
| 5030 | goto fail; |
| 5031 | } |
| 5032 | rv = !conn_send_command(connection, sock, P_AUTH_RESPONSE, 0, |
| 5033 | response, resp_size); |
| 5034 | if (!rv) |
| 5035 | goto fail; |
| 5036 | |
| 5037 | err = drbd_recv_header(connection, &pi); |
| 5038 | if (err) { |
| 5039 | rv = 0; |
| 5040 | goto fail; |
| 5041 | } |
| 5042 | |
| 5043 | if (pi.cmd != P_AUTH_RESPONSE) { |
| 5044 | drbd_err(connection, "expected AuthResponse packet, received: %s (0x%04x)\n", |
| 5045 | cmdname(pi.cmd), pi.cmd); |
| 5046 | rv = 0; |
| 5047 | goto fail; |
| 5048 | } |
| 5049 | |
| 5050 | if (pi.size != resp_size) { |
| 5051 | drbd_err(connection, "expected AuthResponse payload of wrong size\n"); |
| 5052 | rv = 0; |
| 5053 | goto fail; |
| 5054 | } |
| 5055 | |
| 5056 | err = drbd_recv_all_warn(connection, response , resp_size); |
| 5057 | if (err) { |
| 5058 | rv = 0; |
| 5059 | goto fail; |
| 5060 | } |
| 5061 | |
| 5062 | right_response = kmalloc(resp_size, GFP_NOIO); |
| 5063 | if (right_response == NULL) { |
| 5064 | drbd_err(connection, "kmalloc of right_response failed\n"); |
| 5065 | rv = -1; |
| 5066 | goto fail; |
| 5067 | } |
| 5068 | |
| 5069 | rv = crypto_shash_digest(desc, my_challenge, CHALLENGE_LEN, |
| 5070 | right_response); |
| 5071 | if (rv) { |
| 5072 | drbd_err(connection, "crypto_hash_digest() failed with %d\n", rv); |
| 5073 | rv = -1; |
| 5074 | goto fail; |
| 5075 | } |
| 5076 | |
| 5077 | rv = !memcmp(response, right_response, resp_size); |
| 5078 | |
| 5079 | if (rv) |
| 5080 | drbd_info(connection, "Peer authenticated using %d bytes HMAC\n", |
| 5081 | resp_size); |
| 5082 | else |
| 5083 | rv = -1; |
| 5084 | |
| 5085 | fail: |
| 5086 | kfree(peers_ch); |
| 5087 | kfree(response); |
| 5088 | kfree(right_response); |
| 5089 | shash_desc_zero(desc); |
| 5090 | |
| 5091 | return rv; |
| 5092 | } |
| 5093 | #endif |
| 5094 | |
| 5095 | int drbd_receiver(struct drbd_thread *thi) |
| 5096 | { |
| 5097 | struct drbd_connection *connection = thi->connection; |
| 5098 | int h; |
| 5099 | |
| 5100 | drbd_info(connection, "receiver (re)started\n"); |
| 5101 | |
| 5102 | do { |
| 5103 | h = conn_connect(connection); |
| 5104 | if (h == 0) { |
| 5105 | conn_disconnect(connection); |
| 5106 | schedule_timeout_interruptible(HZ); |
| 5107 | } |
| 5108 | if (h == -1) { |
| 5109 | drbd_warn(connection, "Discarding network configuration.\n"); |
| 5110 | conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_HARD); |
| 5111 | } |
| 5112 | } while (h == 0); |
| 5113 | |
| 5114 | if (h > 0) |
| 5115 | drbdd(connection); |
| 5116 | |
| 5117 | conn_disconnect(connection); |
| 5118 | |
| 5119 | drbd_info(connection, "receiver terminated\n"); |
| 5120 | return 0; |
| 5121 | } |
| 5122 | |
| 5123 | /* ********* acknowledge sender ******** */ |
| 5124 | |
| 5125 | static int got_conn_RqSReply(struct drbd_connection *connection, struct packet_info *pi) |
| 5126 | { |
| 5127 | struct p_req_state_reply *p = pi->data; |
| 5128 | int retcode = be32_to_cpu(p->retcode); |
| 5129 | |
| 5130 | if (retcode >= SS_SUCCESS) { |
| 5131 | set_bit(CONN_WD_ST_CHG_OKAY, &connection->flags); |
| 5132 | } else { |
| 5133 | set_bit(CONN_WD_ST_CHG_FAIL, &connection->flags); |
| 5134 | drbd_err(connection, "Requested state change failed by peer: %s (%d)\n", |
| 5135 | drbd_set_st_err_str(retcode), retcode); |
| 5136 | } |
| 5137 | wake_up(&connection->ping_wait); |
| 5138 | |
| 5139 | return 0; |
| 5140 | } |
| 5141 | |
| 5142 | static int got_RqSReply(struct drbd_connection *connection, struct packet_info *pi) |
| 5143 | { |
| 5144 | struct drbd_peer_device *peer_device; |
| 5145 | struct drbd_device *device; |
| 5146 | struct p_req_state_reply *p = pi->data; |
| 5147 | int retcode = be32_to_cpu(p->retcode); |
| 5148 | |
| 5149 | peer_device = conn_peer_device(connection, pi->vnr); |
| 5150 | if (!peer_device) |
| 5151 | return -EIO; |
| 5152 | device = peer_device->device; |
| 5153 | |
| 5154 | if (test_bit(CONN_WD_ST_CHG_REQ, &connection->flags)) { |
| 5155 | D_ASSERT(device, connection->agreed_pro_version < 100); |
| 5156 | return got_conn_RqSReply(connection, pi); |
| 5157 | } |
| 5158 | |
| 5159 | if (retcode >= SS_SUCCESS) { |
| 5160 | set_bit(CL_ST_CHG_SUCCESS, &device->flags); |
| 5161 | } else { |
| 5162 | set_bit(CL_ST_CHG_FAIL, &device->flags); |
| 5163 | drbd_err(device, "Requested state change failed by peer: %s (%d)\n", |
| 5164 | drbd_set_st_err_str(retcode), retcode); |
| 5165 | } |
| 5166 | wake_up(&device->state_wait); |
| 5167 | |
| 5168 | return 0; |
| 5169 | } |
| 5170 | |
| 5171 | static int got_Ping(struct drbd_connection *connection, struct packet_info *pi) |
| 5172 | { |
| 5173 | return drbd_send_ping_ack(connection); |
| 5174 | |
| 5175 | } |
| 5176 | |
| 5177 | static int got_PingAck(struct drbd_connection *connection, struct packet_info *pi) |
| 5178 | { |
| 5179 | /* restore idle timeout */ |
| 5180 | connection->meta.socket->sk->sk_rcvtimeo = connection->net_conf->ping_int*HZ; |
| 5181 | if (!test_and_set_bit(GOT_PING_ACK, &connection->flags)) |
| 5182 | wake_up(&connection->ping_wait); |
| 5183 | |
| 5184 | return 0; |
| 5185 | } |
| 5186 | |
| 5187 | static int got_IsInSync(struct drbd_connection *connection, struct packet_info *pi) |
| 5188 | { |
| 5189 | struct drbd_peer_device *peer_device; |
| 5190 | struct drbd_device *device; |
| 5191 | struct p_block_ack *p = pi->data; |
| 5192 | sector_t sector = be64_to_cpu(p->sector); |
| 5193 | int blksize = be32_to_cpu(p->blksize); |
| 5194 | |
| 5195 | peer_device = conn_peer_device(connection, pi->vnr); |
| 5196 | if (!peer_device) |
| 5197 | return -EIO; |
| 5198 | device = peer_device->device; |
| 5199 | |
| 5200 | D_ASSERT(device, peer_device->connection->agreed_pro_version >= 89); |
| 5201 | |
| 5202 | update_peer_seq(peer_device, be32_to_cpu(p->seq_num)); |
| 5203 | |
| 5204 | if (get_ldev(device)) { |
| 5205 | drbd_rs_complete_io(device, sector); |
| 5206 | drbd_set_in_sync(device, sector, blksize); |
| 5207 | /* rs_same_csums is supposed to count in units of BM_BLOCK_SIZE */ |
| 5208 | device->rs_same_csum += (blksize >> BM_BLOCK_SHIFT); |
| 5209 | put_ldev(device); |
| 5210 | } |
| 5211 | dec_rs_pending(device); |
| 5212 | atomic_add(blksize >> 9, &device->rs_sect_in); |
| 5213 | |
| 5214 | return 0; |
| 5215 | } |
| 5216 | |
| 5217 | static int |
| 5218 | validate_req_change_req_state(struct drbd_device *device, u64 id, sector_t sector, |
| 5219 | struct rb_root *root, const char *func, |
| 5220 | enum drbd_req_event what, bool missing_ok) |
| 5221 | { |
| 5222 | struct drbd_request *req; |
| 5223 | struct bio_and_error m; |
| 5224 | |
| 5225 | spin_lock_irq(&device->resource->req_lock); |
| 5226 | req = find_request(device, root, id, sector, missing_ok, func); |
| 5227 | if (unlikely(!req)) { |
| 5228 | spin_unlock_irq(&device->resource->req_lock); |
| 5229 | return -EIO; |
| 5230 | } |
| 5231 | __req_mod(req, what, &m); |
| 5232 | spin_unlock_irq(&device->resource->req_lock); |
| 5233 | |
| 5234 | if (m.bio) |
| 5235 | complete_master_bio(device, &m); |
| 5236 | return 0; |
| 5237 | } |
| 5238 | |
| 5239 | static int got_BlockAck(struct drbd_connection *connection, struct packet_info *pi) |
| 5240 | { |
| 5241 | struct drbd_peer_device *peer_device; |
| 5242 | struct drbd_device *device; |
| 5243 | struct p_block_ack *p = pi->data; |
| 5244 | sector_t sector = be64_to_cpu(p->sector); |
| 5245 | int blksize = be32_to_cpu(p->blksize); |
| 5246 | enum drbd_req_event what; |
| 5247 | |
| 5248 | peer_device = conn_peer_device(connection, pi->vnr); |
| 5249 | if (!peer_device) |
| 5250 | return -EIO; |
| 5251 | device = peer_device->device; |
| 5252 | |
| 5253 | update_peer_seq(peer_device, be32_to_cpu(p->seq_num)); |
| 5254 | |
| 5255 | if (p->block_id == ID_SYNCER) { |
| 5256 | drbd_set_in_sync(device, sector, blksize); |
| 5257 | dec_rs_pending(device); |
| 5258 | return 0; |
| 5259 | } |
| 5260 | switch (pi->cmd) { |
| 5261 | case P_RS_WRITE_ACK: |
| 5262 | what = WRITE_ACKED_BY_PEER_AND_SIS; |
| 5263 | break; |
| 5264 | case P_WRITE_ACK: |
| 5265 | what = WRITE_ACKED_BY_PEER; |
| 5266 | break; |
| 5267 | case P_RECV_ACK: |
| 5268 | what = RECV_ACKED_BY_PEER; |
| 5269 | break; |
| 5270 | case P_SUPERSEDED: |
| 5271 | what = CONFLICT_RESOLVED; |
| 5272 | break; |
| 5273 | case P_RETRY_WRITE: |
| 5274 | what = POSTPONE_WRITE; |
| 5275 | break; |
| 5276 | default: |
| 5277 | BUG(); |
| 5278 | } |
| 5279 | |
| 5280 | return validate_req_change_req_state(device, p->block_id, sector, |
| 5281 | &device->write_requests, __func__, |
| 5282 | what, false); |
| 5283 | } |
| 5284 | |
| 5285 | static int got_NegAck(struct drbd_connection *connection, struct packet_info *pi) |
| 5286 | { |
| 5287 | struct drbd_peer_device *peer_device; |
| 5288 | struct drbd_device *device; |
| 5289 | struct p_block_ack *p = pi->data; |
| 5290 | sector_t sector = be64_to_cpu(p->sector); |
| 5291 | int size = be32_to_cpu(p->blksize); |
| 5292 | int err; |
| 5293 | |
| 5294 | peer_device = conn_peer_device(connection, pi->vnr); |
| 5295 | if (!peer_device) |
| 5296 | return -EIO; |
| 5297 | device = peer_device->device; |
| 5298 | |
| 5299 | update_peer_seq(peer_device, be32_to_cpu(p->seq_num)); |
| 5300 | |
| 5301 | if (p->block_id == ID_SYNCER) { |
| 5302 | dec_rs_pending(device); |
| 5303 | drbd_rs_failed_io(device, sector, size); |
| 5304 | return 0; |
| 5305 | } |
| 5306 | |
| 5307 | err = validate_req_change_req_state(device, p->block_id, sector, |
| 5308 | &device->write_requests, __func__, |
| 5309 | NEG_ACKED, true); |
| 5310 | if (err) { |
| 5311 | /* Protocol A has no P_WRITE_ACKs, but has P_NEG_ACKs. |
| 5312 | The master bio might already be completed, therefore the |
| 5313 | request is no longer in the collision hash. */ |
| 5314 | /* In Protocol B we might already have got a P_RECV_ACK |
| 5315 | but then get a P_NEG_ACK afterwards. */ |
| 5316 | drbd_set_out_of_sync(device, sector, size); |
| 5317 | } |
| 5318 | return 0; |
| 5319 | } |
| 5320 | |
| 5321 | static int got_NegDReply(struct drbd_connection *connection, struct packet_info *pi) |
| 5322 | { |
| 5323 | struct drbd_peer_device *peer_device; |
| 5324 | struct drbd_device *device; |
| 5325 | struct p_block_ack *p = pi->data; |
| 5326 | sector_t sector = be64_to_cpu(p->sector); |
| 5327 | |
| 5328 | peer_device = conn_peer_device(connection, pi->vnr); |
| 5329 | if (!peer_device) |
| 5330 | return -EIO; |
| 5331 | device = peer_device->device; |
| 5332 | |
| 5333 | update_peer_seq(peer_device, be32_to_cpu(p->seq_num)); |
| 5334 | |
| 5335 | drbd_err(device, "Got NegDReply; Sector %llus, len %u.\n", |
| 5336 | (unsigned long long)sector, be32_to_cpu(p->blksize)); |
| 5337 | |
| 5338 | return validate_req_change_req_state(device, p->block_id, sector, |
| 5339 | &device->read_requests, __func__, |
| 5340 | NEG_ACKED, false); |
| 5341 | } |
| 5342 | |
| 5343 | static int got_NegRSDReply(struct drbd_connection *connection, struct packet_info *pi) |
| 5344 | { |
| 5345 | struct drbd_peer_device *peer_device; |
| 5346 | struct drbd_device *device; |
| 5347 | sector_t sector; |
| 5348 | int size; |
| 5349 | struct p_block_ack *p = pi->data; |
| 5350 | |
| 5351 | peer_device = conn_peer_device(connection, pi->vnr); |
| 5352 | if (!peer_device) |
| 5353 | return -EIO; |
| 5354 | device = peer_device->device; |
| 5355 | |
| 5356 | sector = be64_to_cpu(p->sector); |
| 5357 | size = be32_to_cpu(p->blksize); |
| 5358 | |
| 5359 | update_peer_seq(peer_device, be32_to_cpu(p->seq_num)); |
| 5360 | |
| 5361 | dec_rs_pending(device); |
| 5362 | |
| 5363 | if (get_ldev_if_state(device, D_FAILED)) { |
| 5364 | drbd_rs_complete_io(device, sector); |
| 5365 | switch (pi->cmd) { |
| 5366 | case P_NEG_RS_DREPLY: |
| 5367 | drbd_rs_failed_io(device, sector, size); |
| 5368 | case P_RS_CANCEL: |
| 5369 | break; |
| 5370 | default: |
| 5371 | BUG(); |
| 5372 | } |
| 5373 | put_ldev(device); |
| 5374 | } |
| 5375 | |
| 5376 | return 0; |
| 5377 | } |
| 5378 | |
| 5379 | static int got_BarrierAck(struct drbd_connection *connection, struct packet_info *pi) |
| 5380 | { |
| 5381 | struct p_barrier_ack *p = pi->data; |
| 5382 | struct drbd_peer_device *peer_device; |
| 5383 | int vnr; |
| 5384 | |
| 5385 | tl_release(connection, p->barrier, be32_to_cpu(p->set_size)); |
| 5386 | |
| 5387 | rcu_read_lock(); |
| 5388 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
| 5389 | struct drbd_device *device = peer_device->device; |
| 5390 | |
| 5391 | if (device->state.conn == C_AHEAD && |
| 5392 | atomic_read(&device->ap_in_flight) == 0 && |
| 5393 | !test_and_set_bit(AHEAD_TO_SYNC_SOURCE, &device->flags)) { |
| 5394 | device->start_resync_timer.expires = jiffies + HZ; |
| 5395 | add_timer(&device->start_resync_timer); |
| 5396 | } |
| 5397 | } |
| 5398 | rcu_read_unlock(); |
| 5399 | |
| 5400 | return 0; |
| 5401 | } |
| 5402 | |
| 5403 | static int got_OVResult(struct drbd_connection *connection, struct packet_info *pi) |
| 5404 | { |
| 5405 | struct drbd_peer_device *peer_device; |
| 5406 | struct drbd_device *device; |
| 5407 | struct p_block_ack *p = pi->data; |
| 5408 | struct drbd_device_work *dw; |
| 5409 | sector_t sector; |
| 5410 | int size; |
| 5411 | |
| 5412 | peer_device = conn_peer_device(connection, pi->vnr); |
| 5413 | if (!peer_device) |
| 5414 | return -EIO; |
| 5415 | device = peer_device->device; |
| 5416 | |
| 5417 | sector = be64_to_cpu(p->sector); |
| 5418 | size = be32_to_cpu(p->blksize); |
| 5419 | |
| 5420 | update_peer_seq(peer_device, be32_to_cpu(p->seq_num)); |
| 5421 | |
| 5422 | if (be64_to_cpu(p->block_id) == ID_OUT_OF_SYNC) |
| 5423 | drbd_ov_out_of_sync_found(device, sector, size); |
| 5424 | else |
| 5425 | ov_out_of_sync_print(device); |
| 5426 | |
| 5427 | if (!get_ldev(device)) |
| 5428 | return 0; |
| 5429 | |
| 5430 | drbd_rs_complete_io(device, sector); |
| 5431 | dec_rs_pending(device); |
| 5432 | |
| 5433 | --device->ov_left; |
| 5434 | |
| 5435 | /* let's advance progress step marks only for every other megabyte */ |
| 5436 | if ((device->ov_left & 0x200) == 0x200) |
| 5437 | drbd_advance_rs_marks(device, device->ov_left); |
| 5438 | |
| 5439 | if (device->ov_left == 0) { |
| 5440 | dw = kmalloc(sizeof(*dw), GFP_NOIO); |
| 5441 | if (dw) { |
| 5442 | dw->w.cb = w_ov_finished; |
| 5443 | dw->device = device; |
| 5444 | drbd_queue_work(&peer_device->connection->sender_work, &dw->w); |
| 5445 | } else { |
| 5446 | drbd_err(device, "kmalloc(dw) failed."); |
| 5447 | ov_out_of_sync_print(device); |
| 5448 | drbd_resync_finished(device); |
| 5449 | } |
| 5450 | } |
| 5451 | put_ldev(device); |
| 5452 | return 0; |
| 5453 | } |
| 5454 | |
| 5455 | static int got_skip(struct drbd_connection *connection, struct packet_info *pi) |
| 5456 | { |
| 5457 | return 0; |
| 5458 | } |
| 5459 | |
| 5460 | struct meta_sock_cmd { |
| 5461 | size_t pkt_size; |
| 5462 | int (*fn)(struct drbd_connection *connection, struct packet_info *); |
| 5463 | }; |
| 5464 | |
| 5465 | static void set_rcvtimeo(struct drbd_connection *connection, bool ping_timeout) |
| 5466 | { |
| 5467 | long t; |
| 5468 | struct net_conf *nc; |
| 5469 | |
| 5470 | rcu_read_lock(); |
| 5471 | nc = rcu_dereference(connection->net_conf); |
| 5472 | t = ping_timeout ? nc->ping_timeo : nc->ping_int; |
| 5473 | rcu_read_unlock(); |
| 5474 | |
| 5475 | t *= HZ; |
| 5476 | if (ping_timeout) |
| 5477 | t /= 10; |
| 5478 | |
| 5479 | connection->meta.socket->sk->sk_rcvtimeo = t; |
| 5480 | } |
| 5481 | |
| 5482 | static void set_ping_timeout(struct drbd_connection *connection) |
| 5483 | { |
| 5484 | set_rcvtimeo(connection, 1); |
| 5485 | } |
| 5486 | |
| 5487 | static void set_idle_timeout(struct drbd_connection *connection) |
| 5488 | { |
| 5489 | set_rcvtimeo(connection, 0); |
| 5490 | } |
| 5491 | |
| 5492 | static struct meta_sock_cmd ack_receiver_tbl[] = { |
| 5493 | [P_PING] = { 0, got_Ping }, |
| 5494 | [P_PING_ACK] = { 0, got_PingAck }, |
| 5495 | [P_RECV_ACK] = { sizeof(struct p_block_ack), got_BlockAck }, |
| 5496 | [P_WRITE_ACK] = { sizeof(struct p_block_ack), got_BlockAck }, |
| 5497 | [P_RS_WRITE_ACK] = { sizeof(struct p_block_ack), got_BlockAck }, |
| 5498 | [P_SUPERSEDED] = { sizeof(struct p_block_ack), got_BlockAck }, |
| 5499 | [P_NEG_ACK] = { sizeof(struct p_block_ack), got_NegAck }, |
| 5500 | [P_NEG_DREPLY] = { sizeof(struct p_block_ack), got_NegDReply }, |
| 5501 | [P_NEG_RS_DREPLY] = { sizeof(struct p_block_ack), got_NegRSDReply }, |
| 5502 | [P_OV_RESULT] = { sizeof(struct p_block_ack), got_OVResult }, |
| 5503 | [P_BARRIER_ACK] = { sizeof(struct p_barrier_ack), got_BarrierAck }, |
| 5504 | [P_STATE_CHG_REPLY] = { sizeof(struct p_req_state_reply), got_RqSReply }, |
| 5505 | [P_RS_IS_IN_SYNC] = { sizeof(struct p_block_ack), got_IsInSync }, |
| 5506 | [P_DELAY_PROBE] = { sizeof(struct p_delay_probe93), got_skip }, |
| 5507 | [P_RS_CANCEL] = { sizeof(struct p_block_ack), got_NegRSDReply }, |
| 5508 | [P_CONN_ST_CHG_REPLY]={ sizeof(struct p_req_state_reply), got_conn_RqSReply }, |
| 5509 | [P_RETRY_WRITE] = { sizeof(struct p_block_ack), got_BlockAck }, |
| 5510 | }; |
| 5511 | |
| 5512 | int drbd_ack_receiver(struct drbd_thread *thi) |
| 5513 | { |
| 5514 | struct drbd_connection *connection = thi->connection; |
| 5515 | struct meta_sock_cmd *cmd = NULL; |
| 5516 | struct packet_info pi; |
| 5517 | unsigned long pre_recv_jif; |
| 5518 | int rv; |
| 5519 | void *buf = connection->meta.rbuf; |
| 5520 | int received = 0; |
| 5521 | unsigned int header_size = drbd_header_size(connection); |
| 5522 | int expect = header_size; |
| 5523 | bool ping_timeout_active = false; |
| 5524 | struct sched_param param = { .sched_priority = 2 }; |
| 5525 | |
| 5526 | rv = sched_setscheduler(current, SCHED_RR, ¶m); |
| 5527 | if (rv < 0) |
| 5528 | drbd_err(connection, "drbd_ack_receiver: ERROR set priority, ret=%d\n", rv); |
| 5529 | |
| 5530 | while (get_t_state(thi) == RUNNING) { |
| 5531 | drbd_thread_current_set_cpu(thi); |
| 5532 | |
| 5533 | conn_reclaim_net_peer_reqs(connection); |
| 5534 | |
| 5535 | if (test_and_clear_bit(SEND_PING, &connection->flags)) { |
| 5536 | if (drbd_send_ping(connection)) { |
| 5537 | drbd_err(connection, "drbd_send_ping has failed\n"); |
| 5538 | goto reconnect; |
| 5539 | } |
| 5540 | set_ping_timeout(connection); |
| 5541 | ping_timeout_active = true; |
| 5542 | } |
| 5543 | |
| 5544 | pre_recv_jif = jiffies; |
| 5545 | rv = drbd_recv_short(connection->meta.socket, buf, expect-received, 0); |
| 5546 | |
| 5547 | /* Note: |
| 5548 | * -EINTR (on meta) we got a signal |
| 5549 | * -EAGAIN (on meta) rcvtimeo expired |
| 5550 | * -ECONNRESET other side closed the connection |
| 5551 | * -ERESTARTSYS (on data) we got a signal |
| 5552 | * rv < 0 other than above: unexpected error! |
| 5553 | * rv == expected: full header or command |
| 5554 | * rv < expected: "woken" by signal during receive |
| 5555 | * rv == 0 : "connection shut down by peer" |
| 5556 | */ |
| 5557 | if (likely(rv > 0)) { |
| 5558 | received += rv; |
| 5559 | buf += rv; |
| 5560 | } else if (rv == 0) { |
| 5561 | if (test_bit(DISCONNECT_SENT, &connection->flags)) { |
| 5562 | long t; |
| 5563 | rcu_read_lock(); |
| 5564 | t = rcu_dereference(connection->net_conf)->ping_timeo * HZ/10; |
| 5565 | rcu_read_unlock(); |
| 5566 | |
| 5567 | t = wait_event_timeout(connection->ping_wait, |
| 5568 | connection->cstate < C_WF_REPORT_PARAMS, |
| 5569 | t); |
| 5570 | if (t) |
| 5571 | break; |
| 5572 | } |
| 5573 | drbd_err(connection, "meta connection shut down by peer.\n"); |
| 5574 | goto reconnect; |
| 5575 | } else if (rv == -EAGAIN) { |
| 5576 | /* If the data socket received something meanwhile, |
| 5577 | * that is good enough: peer is still alive. */ |
| 5578 | if (time_after(connection->last_received, pre_recv_jif)) |
| 5579 | continue; |
| 5580 | if (ping_timeout_active) { |
| 5581 | drbd_err(connection, "PingAck did not arrive in time.\n"); |
| 5582 | goto reconnect; |
| 5583 | } |
| 5584 | set_bit(SEND_PING, &connection->flags); |
| 5585 | continue; |
| 5586 | } else if (rv == -EINTR) { |
| 5587 | /* maybe drbd_thread_stop(): the while condition will notice. |
| 5588 | * maybe woken for send_ping: we'll send a ping above, |
| 5589 | * and change the rcvtimeo */ |
| 5590 | flush_signals(current); |
| 5591 | continue; |
| 5592 | } else { |
| 5593 | drbd_err(connection, "sock_recvmsg returned %d\n", rv); |
| 5594 | goto reconnect; |
| 5595 | } |
| 5596 | |
| 5597 | if (received == expect && cmd == NULL) { |
| 5598 | if (decode_header(connection, connection->meta.rbuf, &pi)) |
| 5599 | goto reconnect; |
| 5600 | cmd = &ack_receiver_tbl[pi.cmd]; |
| 5601 | if (pi.cmd >= ARRAY_SIZE(ack_receiver_tbl) || !cmd->fn) { |
| 5602 | drbd_err(connection, "Unexpected meta packet %s (0x%04x)\n", |
| 5603 | cmdname(pi.cmd), pi.cmd); |
| 5604 | goto disconnect; |
| 5605 | } |
| 5606 | expect = header_size + cmd->pkt_size; |
| 5607 | if (pi.size != expect - header_size) { |
| 5608 | drbd_err(connection, "Wrong packet size on meta (c: %d, l: %d)\n", |
| 5609 | pi.cmd, pi.size); |
| 5610 | goto reconnect; |
| 5611 | } |
| 5612 | } |
| 5613 | if (received == expect) { |
| 5614 | bool err; |
| 5615 | |
| 5616 | err = cmd->fn(connection, &pi); |
| 5617 | if (err) { |
| 5618 | drbd_err(connection, "%pf failed\n", cmd->fn); |
| 5619 | goto reconnect; |
| 5620 | } |
| 5621 | |
| 5622 | connection->last_received = jiffies; |
| 5623 | |
| 5624 | if (cmd == &ack_receiver_tbl[P_PING_ACK]) { |
| 5625 | set_idle_timeout(connection); |
| 5626 | ping_timeout_active = false; |
| 5627 | } |
| 5628 | |
| 5629 | buf = connection->meta.rbuf; |
| 5630 | received = 0; |
| 5631 | expect = header_size; |
| 5632 | cmd = NULL; |
| 5633 | } |
| 5634 | } |
| 5635 | |
| 5636 | if (0) { |
| 5637 | reconnect: |
| 5638 | conn_request_state(connection, NS(conn, C_NETWORK_FAILURE), CS_HARD); |
| 5639 | conn_md_sync(connection); |
| 5640 | } |
| 5641 | if (0) { |
| 5642 | disconnect: |
| 5643 | conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_HARD); |
| 5644 | } |
| 5645 | |
| 5646 | drbd_info(connection, "ack_receiver terminated\n"); |
| 5647 | |
| 5648 | return 0; |
| 5649 | } |
| 5650 | |
| 5651 | void drbd_send_acks_wf(struct work_struct *ws) |
| 5652 | { |
| 5653 | struct drbd_peer_device *peer_device = |
| 5654 | container_of(ws, struct drbd_peer_device, send_acks_work); |
| 5655 | struct drbd_connection *connection = peer_device->connection; |
| 5656 | struct drbd_device *device = peer_device->device; |
| 5657 | struct net_conf *nc; |
| 5658 | int tcp_cork, err; |
| 5659 | |
| 5660 | rcu_read_lock(); |
| 5661 | nc = rcu_dereference(connection->net_conf); |
| 5662 | tcp_cork = nc->tcp_cork; |
| 5663 | rcu_read_unlock(); |
| 5664 | |
| 5665 | if (tcp_cork) |
| 5666 | drbd_tcp_cork(connection->meta.socket); |
| 5667 | |
| 5668 | err = drbd_finish_peer_reqs(device); |
| 5669 | kref_put(&device->kref, drbd_destroy_device); |
| 5670 | /* get is in drbd_endio_write_sec_final(). That is necessary to keep the |
| 5671 | struct work_struct send_acks_work alive, which is in the peer_device object */ |
| 5672 | |
| 5673 | if (err) { |
| 5674 | conn_request_state(connection, NS(conn, C_NETWORK_FAILURE), CS_HARD); |
| 5675 | return; |
| 5676 | } |
| 5677 | |
| 5678 | if (tcp_cork) |
| 5679 | drbd_tcp_uncork(connection->meta.socket); |
| 5680 | |
| 5681 | return; |
| 5682 | } |