1 /* 2 drbd_int.h 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 #ifndef _DRBD_INT_H 27 #define _DRBD_INT_H 28 29 #include <linux/compiler.h> 30 #include <linux/types.h> 31 #include <linux/list.h> 32 #include <linux/sched.h> 33 #include <linux/bitops.h> 34 #include <linux/slab.h> 35 #include <linux/crypto.h> 36 #include <linux/ratelimit.h> 37 #include <linux/tcp.h> 38 #include <linux/mutex.h> 39 #include <linux/major.h> 40 #include <linux/blkdev.h> 41 #include <linux/genhd.h> 42 #include <linux/idr.h> 43 #include <net/tcp.h> 44 #include <linux/lru_cache.h> 45 #include <linux/prefetch.h> 46 #include <linux/drbd_genl_api.h> 47 #include <linux/drbd.h> 48 #include "drbd_strings.h" 49 #include "drbd_state.h" 50 #include "drbd_protocol.h" 51 52 #ifdef __CHECKER__ 53 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr"))) 54 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read"))) 55 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write"))) 56 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call"))) 57 #else 58 # define __protected_by(x) 59 # define __protected_read_by(x) 60 # define __protected_write_by(x) 61 # define __must_hold(x) 62 #endif 63 64 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0) 65 66 /* module parameter, defined in drbd_main.c */ 67 extern unsigned int minor_count; 68 extern bool disable_sendpage; 69 extern bool allow_oos; 70 void tl_abort_disk_io(struct drbd_device *device); 71 72 #ifdef CONFIG_DRBD_FAULT_INJECTION 73 extern int enable_faults; 74 extern int fault_rate; 75 extern int fault_devs; 76 #endif 77 78 extern char usermode_helper[]; 79 80 81 /* I don't remember why XCPU ... 82 * This is used to wake the asender, 83 * and to interrupt sending the sending task 84 * on disconnect. 85 */ 86 #define DRBD_SIG SIGXCPU 87 88 /* This is used to stop/restart our threads. 89 * Cannot use SIGTERM nor SIGKILL, since these 90 * are sent out by init on runlevel changes 91 * I choose SIGHUP for now. 92 */ 93 #define DRBD_SIGKILL SIGHUP 94 95 #define ID_IN_SYNC (4711ULL) 96 #define ID_OUT_OF_SYNC (4712ULL) 97 #define ID_SYNCER (-1ULL) 98 99 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL) 100 101 struct drbd_device; 102 struct drbd_connection; 103 104 #define __drbd_printk_device(level, device, fmt, args...) \ 105 dev_printk(level, disk_to_dev((device)->vdisk), fmt, ## args) 106 #define __drbd_printk_peer_device(level, peer_device, fmt, args...) \ 107 dev_printk(level, disk_to_dev((peer_device)->device->vdisk), fmt, ## args) 108 #define __drbd_printk_resource(level, resource, fmt, args...) \ 109 printk(level "drbd %s: " fmt, (resource)->name, ## args) 110 #define __drbd_printk_connection(level, connection, fmt, args...) \ 111 printk(level "drbd %s: " fmt, (connection)->resource->name, ## args) 112 113 void drbd_printk_with_wrong_object_type(void); 114 115 #define __drbd_printk_if_same_type(obj, type, func, level, fmt, args...) \ 116 (__builtin_types_compatible_p(typeof(obj), type) || \ 117 __builtin_types_compatible_p(typeof(obj), const type)), \ 118 func(level, (const type)(obj), fmt, ## args) 119 120 #define drbd_printk(level, obj, fmt, args...) \ 121 __builtin_choose_expr( \ 122 __drbd_printk_if_same_type(obj, struct drbd_device *, \ 123 __drbd_printk_device, level, fmt, ## args), \ 124 __builtin_choose_expr( \ 125 __drbd_printk_if_same_type(obj, struct drbd_resource *, \ 126 __drbd_printk_resource, level, fmt, ## args), \ 127 __builtin_choose_expr( \ 128 __drbd_printk_if_same_type(obj, struct drbd_connection *, \ 129 __drbd_printk_connection, level, fmt, ## args), \ 130 __builtin_choose_expr( \ 131 __drbd_printk_if_same_type(obj, struct drbd_peer_device *, \ 132 __drbd_printk_peer_device, level, fmt, ## args), \ 133 drbd_printk_with_wrong_object_type())))) 134 135 #define drbd_dbg(obj, fmt, args...) \ 136 drbd_printk(KERN_DEBUG, obj, fmt, ## args) 137 #define drbd_alert(obj, fmt, args...) \ 138 drbd_printk(KERN_ALERT, obj, fmt, ## args) 139 #define drbd_err(obj, fmt, args...) \ 140 drbd_printk(KERN_ERR, obj, fmt, ## args) 141 #define drbd_warn(obj, fmt, args...) \ 142 drbd_printk(KERN_WARNING, obj, fmt, ## args) 143 #define drbd_info(obj, fmt, args...) \ 144 drbd_printk(KERN_INFO, obj, fmt, ## args) 145 #define drbd_emerg(obj, fmt, args...) \ 146 drbd_printk(KERN_EMERG, obj, fmt, ## args) 147 148 #define dynamic_drbd_dbg(device, fmt, args...) \ 149 dynamic_dev_dbg(disk_to_dev(device->vdisk), fmt, ## args) 150 151 #define D_ASSERT(device, exp) do { \ 152 if (!(exp)) \ 153 drbd_err(device, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__); \ 154 } while (0) 155 156 /** 157 * expect - Make an assertion 158 * 159 * Unlike the assert macro, this macro returns a boolean result. 160 */ 161 #define expect(exp) ({ \ 162 bool _bool = (exp); \ 163 if (!_bool) \ 164 drbd_err(device, "ASSERTION %s FAILED in %s\n", \ 165 #exp, __func__); \ 166 _bool; \ 167 }) 168 169 /* Defines to control fault insertion */ 170 enum { 171 DRBD_FAULT_MD_WR = 0, /* meta data write */ 172 DRBD_FAULT_MD_RD = 1, /* read */ 173 DRBD_FAULT_RS_WR = 2, /* resync */ 174 DRBD_FAULT_RS_RD = 3, 175 DRBD_FAULT_DT_WR = 4, /* data */ 176 DRBD_FAULT_DT_RD = 5, 177 DRBD_FAULT_DT_RA = 6, /* data read ahead */ 178 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */ 179 DRBD_FAULT_AL_EE = 8, /* alloc ee */ 180 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */ 181 182 DRBD_FAULT_MAX, 183 }; 184 185 extern unsigned int 186 _drbd_insert_fault(struct drbd_device *device, unsigned int type); 187 188 static inline int 189 drbd_insert_fault(struct drbd_device *device, unsigned int type) { 190 #ifdef CONFIG_DRBD_FAULT_INJECTION 191 return fault_rate && 192 (enable_faults & (1<<type)) && 193 _drbd_insert_fault(device, type); 194 #else 195 return 0; 196 #endif 197 } 198 199 /* integer division, round _UP_ to the next integer */ 200 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0)) 201 /* usual integer division */ 202 #define div_floor(A, B) ((A)/(B)) 203 204 extern struct ratelimit_state drbd_ratelimit_state; 205 extern struct idr drbd_devices; /* RCU, updates: genl_lock() */ 206 extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */ 207 208 extern const char *cmdname(enum drbd_packet cmd); 209 210 /* for sending/receiving the bitmap, 211 * possibly in some encoding scheme */ 212 struct bm_xfer_ctx { 213 /* "const" 214 * stores total bits and long words 215 * of the bitmap, so we don't need to 216 * call the accessor functions over and again. */ 217 unsigned long bm_bits; 218 unsigned long bm_words; 219 /* during xfer, current position within the bitmap */ 220 unsigned long bit_offset; 221 unsigned long word_offset; 222 223 /* statistics; index: (h->command == P_BITMAP) */ 224 unsigned packets[2]; 225 unsigned bytes[2]; 226 }; 227 228 extern void INFO_bm_xfer_stats(struct drbd_device *device, 229 const char *direction, struct bm_xfer_ctx *c); 230 231 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c) 232 { 233 /* word_offset counts "native long words" (32 or 64 bit), 234 * aligned at 64 bit. 235 * Encoded packet may end at an unaligned bit offset. 236 * In case a fallback clear text packet is transmitted in 237 * between, we adjust this offset back to the last 64bit 238 * aligned "native long word", which makes coding and decoding 239 * the plain text bitmap much more convenient. */ 240 #if BITS_PER_LONG == 64 241 c->word_offset = c->bit_offset >> 6; 242 #elif BITS_PER_LONG == 32 243 c->word_offset = c->bit_offset >> 5; 244 c->word_offset &= ~(1UL); 245 #else 246 # error "unsupported BITS_PER_LONG" 247 #endif 248 } 249 250 extern unsigned int drbd_header_size(struct drbd_connection *connection); 251 252 /**********************************************************************/ 253 enum drbd_thread_state { 254 NONE, 255 RUNNING, 256 EXITING, 257 RESTARTING 258 }; 259 260 struct drbd_thread { 261 spinlock_t t_lock; 262 struct task_struct *task; 263 struct completion stop; 264 enum drbd_thread_state t_state; 265 int (*function) (struct drbd_thread *); 266 struct drbd_resource *resource; 267 struct drbd_connection *connection; 268 int reset_cpu_mask; 269 const char *name; 270 }; 271 272 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi) 273 { 274 /* THINK testing the t_state seems to be uncritical in all cases 275 * (but thread_{start,stop}), so we can read it *without* the lock. 276 * --lge */ 277 278 smp_rmb(); 279 return thi->t_state; 280 } 281 282 struct drbd_work { 283 struct list_head list; 284 int (*cb)(struct drbd_work *, int cancel); 285 }; 286 287 struct drbd_device_work { 288 struct drbd_work w; 289 struct drbd_device *device; 290 }; 291 292 #include "drbd_interval.h" 293 294 extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *); 295 296 struct drbd_request { 297 struct drbd_work w; 298 struct drbd_device *device; 299 300 /* if local IO is not allowed, will be NULL. 301 * if local IO _is_ allowed, holds the locally submitted bio clone, 302 * or, after local IO completion, the ERR_PTR(error). 303 * see drbd_request_endio(). */ 304 struct bio *private_bio; 305 306 struct drbd_interval i; 307 308 /* epoch: used to check on "completion" whether this req was in 309 * the current epoch, and we therefore have to close it, 310 * causing a p_barrier packet to be send, starting a new epoch. 311 * 312 * This corresponds to "barrier" in struct p_barrier[_ack], 313 * and to "barrier_nr" in struct drbd_epoch (and various 314 * comments/function parameters/local variable names). 315 */ 316 unsigned int epoch; 317 318 struct list_head tl_requests; /* ring list in the transfer log */ 319 struct bio *master_bio; /* master bio pointer */ 320 unsigned long start_time; 321 322 /* once it hits 0, we may complete the master_bio */ 323 atomic_t completion_ref; 324 /* once it hits 0, we may destroy this drbd_request object */ 325 struct kref kref; 326 327 unsigned rq_state; /* see comments above _req_mod() */ 328 }; 329 330 struct drbd_epoch { 331 struct drbd_connection *connection; 332 struct list_head list; 333 unsigned int barrier_nr; 334 atomic_t epoch_size; /* increased on every request added. */ 335 atomic_t active; /* increased on every req. added, and dec on every finished. */ 336 unsigned long flags; 337 }; 338 339 /* Prototype declaration of function defined in drbd_receiver.c */ 340 int drbdd_init(struct drbd_thread *); 341 int drbd_asender(struct drbd_thread *); 342 343 /* drbd_epoch flag bits */ 344 enum { 345 DE_HAVE_BARRIER_NUMBER, 346 }; 347 348 enum epoch_event { 349 EV_PUT, 350 EV_GOT_BARRIER_NR, 351 EV_BECAME_LAST, 352 EV_CLEANUP = 32, /* used as flag */ 353 }; 354 355 struct digest_info { 356 int digest_size; 357 void *digest; 358 }; 359 360 struct drbd_peer_request { 361 struct drbd_work w; 362 struct drbd_peer_device *peer_device; 363 struct drbd_epoch *epoch; /* for writes */ 364 struct page *pages; 365 atomic_t pending_bios; 366 struct drbd_interval i; 367 /* see comments on ee flag bits below */ 368 unsigned long flags; 369 union { 370 u64 block_id; 371 struct digest_info *digest; 372 }; 373 }; 374 375 /* ee flag bits. 376 * While corresponding bios are in flight, the only modification will be 377 * set_bit WAS_ERROR, which has to be atomic. 378 * If no bios are in flight yet, or all have been completed, 379 * non-atomic modification to ee->flags is ok. 380 */ 381 enum { 382 __EE_CALL_AL_COMPLETE_IO, 383 __EE_MAY_SET_IN_SYNC, 384 385 /* is this a TRIM aka REQ_DISCARD? */ 386 __EE_IS_TRIM, 387 /* our lower level cannot handle trim, 388 * and we want to fall back to zeroout instead */ 389 __EE_IS_TRIM_USE_ZEROOUT, 390 391 /* In case a barrier failed, 392 * we need to resubmit without the barrier flag. */ 393 __EE_RESUBMITTED, 394 395 /* we may have several bios per peer request. 396 * if any of those fail, we set this flag atomically 397 * from the endio callback */ 398 __EE_WAS_ERROR, 399 400 /* This ee has a pointer to a digest instead of a block id */ 401 __EE_HAS_DIGEST, 402 403 /* Conflicting local requests need to be restarted after this request */ 404 __EE_RESTART_REQUESTS, 405 406 /* The peer wants a write ACK for this (wire proto C) */ 407 __EE_SEND_WRITE_ACK, 408 409 /* Is set when net_conf had two_primaries set while creating this peer_req */ 410 __EE_IN_INTERVAL_TREE, 411 }; 412 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO) 413 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC) 414 #define EE_IS_TRIM (1<<__EE_IS_TRIM) 415 #define EE_IS_TRIM_USE_ZEROOUT (1<<__EE_IS_TRIM_USE_ZEROOUT) 416 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED) 417 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR) 418 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST) 419 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS) 420 #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK) 421 #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE) 422 423 /* flag bits per device */ 424 enum { 425 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */ 426 MD_DIRTY, /* current uuids and flags not yet on disk */ 427 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */ 428 CL_ST_CHG_SUCCESS, 429 CL_ST_CHG_FAIL, 430 CRASHED_PRIMARY, /* This node was a crashed primary. 431 * Gets cleared when the state.conn 432 * goes into C_CONNECTED state. */ 433 CONSIDER_RESYNC, 434 435 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */ 436 SUSPEND_IO, /* suspend application io */ 437 BITMAP_IO, /* suspend application io; 438 once no more io in flight, start bitmap io */ 439 BITMAP_IO_QUEUED, /* Started bitmap IO */ 440 GO_DISKLESS, /* Disk is being detached, on io-error or admin request. */ 441 WAS_IO_ERROR, /* Local disk failed, returned IO error */ 442 WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */ 443 FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */ 444 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */ 445 RESIZE_PENDING, /* Size change detected locally, waiting for the response from 446 * the peer, if it changed there as well. */ 447 NEW_CUR_UUID, /* Create new current UUID when thawing IO */ 448 AL_SUSPENDED, /* Activity logging is currently suspended. */ 449 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */ 450 B_RS_H_DONE, /* Before resync handler done (already executed) */ 451 DISCARD_MY_DATA, /* discard_my_data flag per volume */ 452 READ_BALANCE_RR, 453 }; 454 455 struct drbd_bitmap; /* opaque for drbd_device */ 456 457 /* definition of bits in bm_flags to be used in drbd_bm_lock 458 * and drbd_bitmap_io and friends. */ 459 enum bm_flag { 460 /* do we need to kfree, or vfree bm_pages? */ 461 BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */ 462 463 /* currently locked for bulk operation */ 464 BM_LOCKED_MASK = 0xf, 465 466 /* in detail, that is: */ 467 BM_DONT_CLEAR = 0x1, 468 BM_DONT_SET = 0x2, 469 BM_DONT_TEST = 0x4, 470 471 /* so we can mark it locked for bulk operation, 472 * and still allow all non-bulk operations */ 473 BM_IS_LOCKED = 0x8, 474 475 /* (test bit, count bit) allowed (common case) */ 476 BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED, 477 478 /* testing bits, as well as setting new bits allowed, but clearing bits 479 * would be unexpected. Used during bitmap receive. Setting new bits 480 * requires sending of "out-of-sync" information, though. */ 481 BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED, 482 483 /* for drbd_bm_write_copy_pages, everything is allowed, 484 * only concurrent bulk operations are locked out. */ 485 BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED, 486 }; 487 488 struct drbd_work_queue { 489 struct list_head q; 490 spinlock_t q_lock; /* to protect the list. */ 491 wait_queue_head_t q_wait; 492 }; 493 494 struct drbd_socket { 495 struct mutex mutex; 496 struct socket *socket; 497 /* this way we get our 498 * send/receive buffers off the stack */ 499 void *sbuf; 500 void *rbuf; 501 }; 502 503 struct drbd_md { 504 u64 md_offset; /* sector offset to 'super' block */ 505 506 u64 la_size_sect; /* last agreed size, unit sectors */ 507 spinlock_t uuid_lock; 508 u64 uuid[UI_SIZE]; 509 u64 device_uuid; 510 u32 flags; 511 u32 md_size_sect; 512 513 s32 al_offset; /* signed relative sector offset to activity log */ 514 s32 bm_offset; /* signed relative sector offset to bitmap */ 515 516 /* cached value of bdev->disk_conf->meta_dev_idx (see below) */ 517 s32 meta_dev_idx; 518 519 /* see al_tr_number_to_on_disk_sector() */ 520 u32 al_stripes; 521 u32 al_stripe_size_4k; 522 u32 al_size_4k; /* cached product of the above */ 523 }; 524 525 struct drbd_backing_dev { 526 struct block_device *backing_bdev; 527 struct block_device *md_bdev; 528 struct drbd_md md; 529 struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */ 530 sector_t known_size; /* last known size of that backing device */ 531 }; 532 533 struct drbd_md_io { 534 unsigned int done; 535 int error; 536 }; 537 538 struct bm_io_work { 539 struct drbd_work w; 540 char *why; 541 enum bm_flag flags; 542 int (*io_fn)(struct drbd_device *device); 543 void (*done)(struct drbd_device *device, int rv); 544 }; 545 546 enum write_ordering_e { 547 WO_none, 548 WO_drain_io, 549 WO_bdev_flush, 550 }; 551 552 struct fifo_buffer { 553 unsigned int head_index; 554 unsigned int size; 555 int total; /* sum of all values */ 556 int values[0]; 557 }; 558 extern struct fifo_buffer *fifo_alloc(int fifo_size); 559 560 /* flag bits per connection */ 561 enum { 562 NET_CONGESTED, /* The data socket is congested */ 563 RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */ 564 SEND_PING, /* whether asender should send a ping asap */ 565 SIGNAL_ASENDER, /* whether asender wants to be interrupted */ 566 GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */ 567 CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */ 568 CONN_WD_ST_CHG_OKAY, 569 CONN_WD_ST_CHG_FAIL, 570 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */ 571 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */ 572 STATE_SENT, /* Do not change state/UUIDs while this is set */ 573 CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC) 574 * pending, from drbd worker context. 575 * If set, bdi_write_congested() returns true, 576 * so shrink_page_list() would not recurse into, 577 * and potentially deadlock on, this drbd worker. 578 */ 579 DISCONNECT_SENT, 580 }; 581 582 struct drbd_resource { 583 char *name; 584 struct kref kref; 585 struct idr devices; /* volume number to device mapping */ 586 struct list_head connections; 587 struct list_head resources; 588 struct res_opts res_opts; 589 struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */ 590 struct mutex adm_mutex; /* mutex to serialize administrative requests */ 591 spinlock_t req_lock; 592 593 unsigned susp:1; /* IO suspended by user */ 594 unsigned susp_nod:1; /* IO suspended because no data */ 595 unsigned susp_fen:1; /* IO suspended because fence peer handler runs */ 596 597 cpumask_var_t cpu_mask; 598 }; 599 600 struct drbd_connection { 601 struct list_head connections; 602 struct drbd_resource *resource; 603 struct kref kref; 604 struct idr peer_devices; /* volume number to peer device mapping */ 605 enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */ 606 struct mutex cstate_mutex; /* Protects graceful disconnects */ 607 unsigned int connect_cnt; /* Inc each time a connection is established */ 608 609 unsigned long flags; 610 struct net_conf *net_conf; /* content protected by rcu */ 611 wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */ 612 613 struct sockaddr_storage my_addr; 614 int my_addr_len; 615 struct sockaddr_storage peer_addr; 616 int peer_addr_len; 617 618 struct drbd_socket data; /* data/barrier/cstate/parameter packets */ 619 struct drbd_socket meta; /* ping/ack (metadata) packets */ 620 int agreed_pro_version; /* actually used protocol version */ 621 u32 agreed_features; 622 unsigned long last_received; /* in jiffies, either socket */ 623 unsigned int ko_count; 624 625 struct list_head transfer_log; /* all requests not yet fully processed */ 626 627 struct crypto_hash *cram_hmac_tfm; 628 struct crypto_hash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */ 629 struct crypto_hash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */ 630 struct crypto_hash *csums_tfm; 631 struct crypto_hash *verify_tfm; 632 void *int_dig_in; 633 void *int_dig_vv; 634 635 /* receiver side */ 636 struct drbd_epoch *current_epoch; 637 spinlock_t epoch_lock; 638 unsigned int epochs; 639 enum write_ordering_e write_ordering; 640 atomic_t current_tle_nr; /* transfer log epoch number */ 641 unsigned current_tle_writes; /* writes seen within this tl epoch */ 642 643 unsigned long last_reconnect_jif; 644 struct drbd_thread receiver; 645 struct drbd_thread worker; 646 struct drbd_thread asender; 647 648 /* sender side */ 649 struct drbd_work_queue sender_work; 650 651 struct { 652 /* whether this sender thread 653 * has processed a single write yet. */ 654 bool seen_any_write_yet; 655 656 /* Which barrier number to send with the next P_BARRIER */ 657 int current_epoch_nr; 658 659 /* how many write requests have been sent 660 * with req->epoch == current_epoch_nr. 661 * If none, no P_BARRIER will be sent. */ 662 unsigned current_epoch_writes; 663 } send; 664 }; 665 666 struct submit_worker { 667 struct workqueue_struct *wq; 668 struct work_struct worker; 669 670 spinlock_t lock; 671 struct list_head writes; 672 }; 673 674 struct drbd_peer_device { 675 struct list_head peer_devices; 676 struct drbd_device *device; 677 struct drbd_connection *connection; 678 }; 679 680 struct drbd_device { 681 struct drbd_resource *resource; 682 struct list_head peer_devices; 683 int vnr; /* volume number within the connection */ 684 struct kref kref; 685 686 /* things that are stored as / read from meta data on disk */ 687 unsigned long flags; 688 689 /* configured by drbdsetup */ 690 struct drbd_backing_dev *ldev __protected_by(local); 691 692 sector_t p_size; /* partner's disk size */ 693 struct request_queue *rq_queue; 694 struct block_device *this_bdev; 695 struct gendisk *vdisk; 696 697 unsigned long last_reattach_jif; 698 struct drbd_work resync_work; 699 struct drbd_work unplug_work; 700 struct drbd_work go_diskless; 701 struct drbd_work md_sync_work; 702 struct drbd_work start_resync_work; 703 struct timer_list resync_timer; 704 struct timer_list md_sync_timer; 705 struct timer_list start_resync_timer; 706 struct timer_list request_timer; 707 #ifdef DRBD_DEBUG_MD_SYNC 708 struct { 709 unsigned int line; 710 const char* func; 711 } last_md_mark_dirty; 712 #endif 713 714 /* Used after attach while negotiating new disk state. */ 715 union drbd_state new_state_tmp; 716 717 union drbd_dev_state state; 718 wait_queue_head_t misc_wait; 719 wait_queue_head_t state_wait; /* upon each state change. */ 720 unsigned int send_cnt; 721 unsigned int recv_cnt; 722 unsigned int read_cnt; 723 unsigned int writ_cnt; 724 unsigned int al_writ_cnt; 725 unsigned int bm_writ_cnt; 726 atomic_t ap_bio_cnt; /* Requests we need to complete */ 727 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */ 728 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */ 729 atomic_t unacked_cnt; /* Need to send replies for */ 730 atomic_t local_cnt; /* Waiting for local completion */ 731 732 /* Interval tree of pending local requests */ 733 struct rb_root read_requests; 734 struct rb_root write_requests; 735 736 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */ 737 unsigned long rs_total; 738 /* number of resync blocks that failed in this run */ 739 unsigned long rs_failed; 740 /* Syncer's start time [unit jiffies] */ 741 unsigned long rs_start; 742 /* cumulated time in PausedSyncX state [unit jiffies] */ 743 unsigned long rs_paused; 744 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */ 745 unsigned long rs_same_csum; 746 #define DRBD_SYNC_MARKS 8 747 #define DRBD_SYNC_MARK_STEP (3*HZ) 748 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */ 749 unsigned long rs_mark_left[DRBD_SYNC_MARKS]; 750 /* marks's time [unit jiffies] */ 751 unsigned long rs_mark_time[DRBD_SYNC_MARKS]; 752 /* current index into rs_mark_{left,time} */ 753 int rs_last_mark; 754 unsigned long rs_last_bcast; /* [unit jiffies] */ 755 756 /* where does the admin want us to start? (sector) */ 757 sector_t ov_start_sector; 758 sector_t ov_stop_sector; 759 /* where are we now? (sector) */ 760 sector_t ov_position; 761 /* Start sector of out of sync range (to merge printk reporting). */ 762 sector_t ov_last_oos_start; 763 /* size of out-of-sync range in sectors. */ 764 sector_t ov_last_oos_size; 765 unsigned long ov_left; /* in bits */ 766 767 struct drbd_bitmap *bitmap; 768 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */ 769 770 /* Used to track operations of resync... */ 771 struct lru_cache *resync; 772 /* Number of locked elements in resync LRU */ 773 unsigned int resync_locked; 774 /* resync extent number waiting for application requests */ 775 unsigned int resync_wenr; 776 777 int open_cnt; 778 u64 *p_uuid; 779 780 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */ 781 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */ 782 struct list_head done_ee; /* need to send P_WRITE_ACK */ 783 struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */ 784 struct list_head net_ee; /* zero-copy network send in progress */ 785 786 int next_barrier_nr; 787 struct list_head resync_reads; 788 atomic_t pp_in_use; /* allocated from page pool */ 789 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */ 790 wait_queue_head_t ee_wait; 791 struct page *md_io_page; /* one page buffer for md_io */ 792 struct drbd_md_io md_io; 793 atomic_t md_io_in_use; /* protects the md_io, md_io_page and md_io_tmpp */ 794 spinlock_t al_lock; 795 wait_queue_head_t al_wait; 796 struct lru_cache *act_log; /* activity log */ 797 unsigned int al_tr_number; 798 int al_tr_cycle; 799 wait_queue_head_t seq_wait; 800 atomic_t packet_seq; 801 unsigned int peer_seq; 802 spinlock_t peer_seq_lock; 803 unsigned int minor; 804 unsigned long comm_bm_set; /* communicated number of set bits. */ 805 struct bm_io_work bm_io_work; 806 u64 ed_uuid; /* UUID of the exposed data */ 807 struct mutex own_state_mutex; 808 struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */ 809 char congestion_reason; /* Why we where congested... */ 810 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */ 811 atomic_t rs_sect_ev; /* for submitted resync data rate, both */ 812 int rs_last_sect_ev; /* counter to compare with */ 813 int rs_last_events; /* counter of read or write "events" (unit sectors) 814 * on the lower level device when we last looked. */ 815 int c_sync_rate; /* current resync rate after syncer throttle magic */ 816 struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */ 817 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */ 818 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */ 819 unsigned int peer_max_bio_size; 820 unsigned int local_max_bio_size; 821 822 /* any requests that would block in drbd_make_request() 823 * are deferred to this single-threaded work queue */ 824 struct submit_worker submit; 825 }; 826 827 struct drbd_config_context { 828 /* assigned from drbd_genlmsghdr */ 829 unsigned int minor; 830 /* assigned from request attributes, if present */ 831 unsigned int volume; 832 #define VOLUME_UNSPECIFIED (-1U) 833 /* pointer into the request skb, 834 * limited lifetime! */ 835 char *resource_name; 836 struct nlattr *my_addr; 837 struct nlattr *peer_addr; 838 839 /* reply buffer */ 840 struct sk_buff *reply_skb; 841 /* pointer into reply buffer */ 842 struct drbd_genlmsghdr *reply_dh; 843 /* resolved from attributes, if possible */ 844 struct drbd_device *device; 845 struct drbd_resource *resource; 846 struct drbd_connection *connection; 847 }; 848 849 static inline struct drbd_device *minor_to_device(unsigned int minor) 850 { 851 return (struct drbd_device *)idr_find(&drbd_devices, minor); 852 } 853 854 static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device) 855 { 856 return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices); 857 } 858 859 #define for_each_resource(resource, _resources) \ 860 list_for_each_entry(resource, _resources, resources) 861 862 #define for_each_resource_rcu(resource, _resources) \ 863 list_for_each_entry_rcu(resource, _resources, resources) 864 865 #define for_each_resource_safe(resource, tmp, _resources) \ 866 list_for_each_entry_safe(resource, tmp, _resources, resources) 867 868 #define for_each_connection(connection, resource) \ 869 list_for_each_entry(connection, &resource->connections, connections) 870 871 #define for_each_connection_rcu(connection, resource) \ 872 list_for_each_entry_rcu(connection, &resource->connections, connections) 873 874 #define for_each_connection_safe(connection, tmp, resource) \ 875 list_for_each_entry_safe(connection, tmp, &resource->connections, connections) 876 877 #define for_each_peer_device(peer_device, device) \ 878 list_for_each_entry(peer_device, &device->peer_devices, peer_devices) 879 880 #define for_each_peer_device_rcu(peer_device, device) \ 881 list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices) 882 883 #define for_each_peer_device_safe(peer_device, tmp, device) \ 884 list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices) 885 886 static inline unsigned int device_to_minor(struct drbd_device *device) 887 { 888 return device->minor; 889 } 890 891 /* 892 * function declarations 893 *************************/ 894 895 /* drbd_main.c */ 896 897 enum dds_flags { 898 DDSF_FORCED = 1, 899 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */ 900 }; 901 902 extern void drbd_init_set_defaults(struct drbd_device *device); 903 extern int drbd_thread_start(struct drbd_thread *thi); 904 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait); 905 #ifdef CONFIG_SMP 906 extern void drbd_thread_current_set_cpu(struct drbd_thread *thi); 907 #else 908 #define drbd_thread_current_set_cpu(A) ({}) 909 #endif 910 extern void tl_release(struct drbd_connection *, unsigned int barrier_nr, 911 unsigned int set_size); 912 extern void tl_clear(struct drbd_connection *); 913 extern void drbd_free_sock(struct drbd_connection *connection); 914 extern int drbd_send(struct drbd_connection *connection, struct socket *sock, 915 void *buf, size_t size, unsigned msg_flags); 916 extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t, 917 unsigned); 918 919 extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd); 920 extern int drbd_send_protocol(struct drbd_connection *connection); 921 extern int drbd_send_uuids(struct drbd_peer_device *); 922 extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *); 923 extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *); 924 extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags); 925 extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s); 926 extern int drbd_send_current_state(struct drbd_peer_device *); 927 extern int drbd_send_sync_param(struct drbd_peer_device *); 928 extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr, 929 u32 set_size); 930 extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet, 931 struct drbd_peer_request *); 932 extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet, 933 struct p_block_req *rp); 934 extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet, 935 struct p_data *dp, int data_size); 936 extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet, 937 sector_t sector, int blksize, u64 block_id); 938 extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *); 939 extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet, 940 struct drbd_peer_request *); 941 extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req); 942 extern int drbd_send_drequest(struct drbd_peer_device *, int cmd, 943 sector_t sector, int size, u64 block_id); 944 extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector, 945 int size, void *digest, int digest_size, 946 enum drbd_packet cmd); 947 extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size); 948 949 extern int drbd_send_bitmap(struct drbd_device *device); 950 extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode); 951 extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode); 952 extern void drbd_free_bc(struct drbd_backing_dev *ldev); 953 extern void drbd_device_cleanup(struct drbd_device *device); 954 void drbd_print_uuids(struct drbd_device *device, const char *text); 955 956 extern void conn_md_sync(struct drbd_connection *connection); 957 extern void drbd_md_write(struct drbd_device *device, void *buffer); 958 extern void drbd_md_sync(struct drbd_device *device); 959 extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev); 960 extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); 961 extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); 962 extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local); 963 extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local); 964 extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local); 965 extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); 966 extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local); 967 extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local); 968 extern int drbd_md_test_flag(struct drbd_backing_dev *, int); 969 #ifndef DRBD_DEBUG_MD_SYNC 970 extern void drbd_md_mark_dirty(struct drbd_device *device); 971 #else 972 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ ) 973 extern void drbd_md_mark_dirty_(struct drbd_device *device, 974 unsigned int line, const char *func); 975 #endif 976 extern void drbd_queue_bitmap_io(struct drbd_device *device, 977 int (*io_fn)(struct drbd_device *), 978 void (*done)(struct drbd_device *, int), 979 char *why, enum bm_flag flags); 980 extern int drbd_bitmap_io(struct drbd_device *device, 981 int (*io_fn)(struct drbd_device *), 982 char *why, enum bm_flag flags); 983 extern int drbd_bitmap_io_from_worker(struct drbd_device *device, 984 int (*io_fn)(struct drbd_device *), 985 char *why, enum bm_flag flags); 986 extern int drbd_bmio_set_n_write(struct drbd_device *device); 987 extern int drbd_bmio_clear_n_write(struct drbd_device *device); 988 extern void drbd_ldev_destroy(struct drbd_device *device); 989 990 /* Meta data layout 991 * 992 * We currently have two possible layouts. 993 * Offsets in (512 byte) sectors. 994 * external: 995 * |----------- md_size_sect ------------------| 996 * [ 4k superblock ][ activity log ][ Bitmap ] 997 * | al_offset == 8 | 998 * | bm_offset = al_offset + X | 999 * ==> bitmap sectors = md_size_sect - bm_offset 1000 * 1001 * Variants: 1002 * old, indexed fixed size meta data: 1003 * 1004 * internal: 1005 * |----------- md_size_sect ------------------| 1006 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*] 1007 * | al_offset < 0 | 1008 * | bm_offset = al_offset - Y | 1009 * ==> bitmap sectors = Y = al_offset - bm_offset 1010 * 1011 * [padding*] are zero or up to 7 unused 512 Byte sectors to the 1012 * end of the device, so that the [4k superblock] will be 4k aligned. 1013 * 1014 * The activity log consists of 4k transaction blocks, 1015 * which are written in a ring-buffer, or striped ring-buffer like fashion, 1016 * which are writtensize used to be fixed 32kB, 1017 * but is about to become configurable. 1018 */ 1019 1020 /* Our old fixed size meta data layout 1021 * allows up to about 3.8TB, so if you want more, 1022 * you need to use the "flexible" meta data format. */ 1023 #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */ 1024 #define MD_4kB_SECT 8 1025 #define MD_32kB_SECT 64 1026 1027 /* One activity log extent represents 4M of storage */ 1028 #define AL_EXTENT_SHIFT 22 1029 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT) 1030 1031 /* We could make these currently hardcoded constants configurable 1032 * variables at create-md time (or even re-configurable at runtime?). 1033 * Which will require some more changes to the DRBD "super block" 1034 * and attach code. 1035 * 1036 * updates per transaction: 1037 * This many changes to the active set can be logged with one transaction. 1038 * This number is arbitrary. 1039 * context per transaction: 1040 * This many context extent numbers are logged with each transaction. 1041 * This number is resulting from the transaction block size (4k), the layout 1042 * of the transaction header, and the number of updates per transaction. 1043 * See drbd_actlog.c:struct al_transaction_on_disk 1044 * */ 1045 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary 1046 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4 1047 1048 #if BITS_PER_LONG == 32 1049 #define LN2_BPL 5 1050 #define cpu_to_lel(A) cpu_to_le32(A) 1051 #define lel_to_cpu(A) le32_to_cpu(A) 1052 #elif BITS_PER_LONG == 64 1053 #define LN2_BPL 6 1054 #define cpu_to_lel(A) cpu_to_le64(A) 1055 #define lel_to_cpu(A) le64_to_cpu(A) 1056 #else 1057 #error "LN2 of BITS_PER_LONG unknown!" 1058 #endif 1059 1060 /* resync bitmap */ 1061 /* 16MB sized 'bitmap extent' to track syncer usage */ 1062 struct bm_extent { 1063 int rs_left; /* number of bits set (out of sync) in this extent. */ 1064 int rs_failed; /* number of failed resync requests in this extent. */ 1065 unsigned long flags; 1066 struct lc_element lce; 1067 }; 1068 1069 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */ 1070 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */ 1071 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */ 1072 1073 /* drbd_bitmap.c */ 1074 /* 1075 * We need to store one bit for a block. 1076 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap. 1077 * Bit 0 ==> local node thinks this block is binary identical on both nodes 1078 * Bit 1 ==> local node thinks this block needs to be synced. 1079 */ 1080 1081 #define SLEEP_TIME (HZ/10) 1082 1083 /* We do bitmap IO in units of 4k blocks. 1084 * We also still have a hardcoded 4k per bit relation. */ 1085 #define BM_BLOCK_SHIFT 12 /* 4k per bit */ 1086 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT) 1087 /* mostly arbitrarily set the represented size of one bitmap extent, 1088 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap 1089 * at 4k per bit resolution) */ 1090 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */ 1091 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT) 1092 1093 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12) 1094 #error "HAVE YOU FIXED drbdmeta AS WELL??" 1095 #endif 1096 1097 /* thus many _storage_ sectors are described by one bit */ 1098 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9)) 1099 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9)) 1100 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1) 1101 1102 /* bit to represented kilo byte conversion */ 1103 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10)) 1104 1105 /* in which _bitmap_ extent (resp. sector) the bit for a certain 1106 * _storage_ sector is located in */ 1107 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9)) 1108 1109 /* how much _storage_ sectors we have per bitmap sector */ 1110 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9)) 1111 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1) 1112 1113 /* in one sector of the bitmap, we have this many activity_log extents. */ 1114 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT)) 1115 1116 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT) 1117 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1) 1118 1119 /* the extent in "PER_EXTENT" below is an activity log extent 1120 * we need that many (long words/bytes) to store the bitmap 1121 * of one AL_EXTENT_SIZE chunk of storage. 1122 * we can store the bitmap for that many AL_EXTENTS within 1123 * one sector of the _on_disk_ bitmap: 1124 * bit 0 bit 37 bit 38 bit (512*8)-1 1125 * ...|........|........|.. // ..|........| 1126 * sect. 0 `296 `304 ^(512*8*8)-1 1127 * 1128 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG ) 1129 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128 1130 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4 1131 */ 1132 1133 #define DRBD_MAX_SECTORS_32 (0xffffffffLU) 1134 /* we have a certain meta data variant that has a fixed on-disk size of 128 1135 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity 1136 * log, leaving this many sectors for the bitmap. 1137 */ 1138 1139 #define DRBD_MAX_SECTORS_FIXED_BM \ 1140 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9))) 1141 #if !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32 1142 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32 1143 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32 1144 #else 1145 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM 1146 /* 16 TB in units of sectors */ 1147 #if BITS_PER_LONG == 32 1148 /* adjust by one page worth of bitmap, 1149 * so we won't wrap around in drbd_bm_find_next_bit. 1150 * you should use 64bit OS for that much storage, anyways. */ 1151 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff) 1152 #else 1153 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */ 1154 #define DRBD_MAX_SECTORS_FLEX (1UL << 51) 1155 /* corresponds to (1UL << 38) bits right now. */ 1156 #endif 1157 #endif 1158 1159 /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE, 1160 * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte. 1161 * Since we may live in a mixed-platform cluster, 1162 * we limit us to a platform agnostic constant here for now. 1163 * A followup commit may allow even bigger BIO sizes, 1164 * once we thought that through. */ 1165 #define DRBD_MAX_BIO_SIZE (1U << 20) 1166 #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE 1167 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE 1168 #endif 1169 #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */ 1170 1171 #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */ 1172 #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */ 1173 1174 /* For now, don't allow more than one activity log extent worth of data 1175 * to be discarded in one go. We may need to rework drbd_al_begin_io() 1176 * to allow for even larger discard ranges */ 1177 #define DRBD_MAX_DISCARD_SIZE AL_EXTENT_SIZE 1178 #define DRBD_MAX_DISCARD_SECTORS (DRBD_MAX_DISCARD_SIZE >> 9) 1179 1180 extern int drbd_bm_init(struct drbd_device *device); 1181 extern int drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits); 1182 extern void drbd_bm_cleanup(struct drbd_device *device); 1183 extern void drbd_bm_set_all(struct drbd_device *device); 1184 extern void drbd_bm_clear_all(struct drbd_device *device); 1185 /* set/clear/test only a few bits at a time */ 1186 extern int drbd_bm_set_bits( 1187 struct drbd_device *device, unsigned long s, unsigned long e); 1188 extern int drbd_bm_clear_bits( 1189 struct drbd_device *device, unsigned long s, unsigned long e); 1190 extern int drbd_bm_count_bits( 1191 struct drbd_device *device, const unsigned long s, const unsigned long e); 1192 /* bm_set_bits variant for use while holding drbd_bm_lock, 1193 * may process the whole bitmap in one go */ 1194 extern void _drbd_bm_set_bits(struct drbd_device *device, 1195 const unsigned long s, const unsigned long e); 1196 extern int drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr); 1197 extern int drbd_bm_e_weight(struct drbd_device *device, unsigned long enr); 1198 extern int drbd_bm_write_page(struct drbd_device *device, unsigned int idx) __must_hold(local); 1199 extern int drbd_bm_read(struct drbd_device *device) __must_hold(local); 1200 extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr); 1201 extern int drbd_bm_write(struct drbd_device *device) __must_hold(local); 1202 extern int drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local); 1203 extern int drbd_bm_write_all(struct drbd_device *device) __must_hold(local); 1204 extern int drbd_bm_write_copy_pages(struct drbd_device *device) __must_hold(local); 1205 extern size_t drbd_bm_words(struct drbd_device *device); 1206 extern unsigned long drbd_bm_bits(struct drbd_device *device); 1207 extern sector_t drbd_bm_capacity(struct drbd_device *device); 1208 1209 #define DRBD_END_OF_BITMAP (~(unsigned long)0) 1210 extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); 1211 /* bm_find_next variants for use while you hold drbd_bm_lock() */ 1212 extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); 1213 extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo); 1214 extern unsigned long _drbd_bm_total_weight(struct drbd_device *device); 1215 extern unsigned long drbd_bm_total_weight(struct drbd_device *device); 1216 extern int drbd_bm_rs_done(struct drbd_device *device); 1217 /* for receive_bitmap */ 1218 extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset, 1219 size_t number, unsigned long *buffer); 1220 /* for _drbd_send_bitmap */ 1221 extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset, 1222 size_t number, unsigned long *buffer); 1223 1224 extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags); 1225 extern void drbd_bm_unlock(struct drbd_device *device); 1226 /* drbd_main.c */ 1227 1228 extern struct kmem_cache *drbd_request_cache; 1229 extern struct kmem_cache *drbd_ee_cache; /* peer requests */ 1230 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */ 1231 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */ 1232 extern mempool_t *drbd_request_mempool; 1233 extern mempool_t *drbd_ee_mempool; 1234 1235 /* drbd's page pool, used to buffer data received from the peer, 1236 * or data requested by the peer. 1237 * 1238 * This does not have an emergency reserve. 1239 * 1240 * When allocating from this pool, it first takes pages from the pool. 1241 * Only if the pool is depleted will try to allocate from the system. 1242 * 1243 * The assumption is that pages taken from this pool will be processed, 1244 * and given back, "quickly", and then can be recycled, so we can avoid 1245 * frequent calls to alloc_page(), and still will be able to make progress even 1246 * under memory pressure. 1247 */ 1248 extern struct page *drbd_pp_pool; 1249 extern spinlock_t drbd_pp_lock; 1250 extern int drbd_pp_vacant; 1251 extern wait_queue_head_t drbd_pp_wait; 1252 1253 /* We also need a standard (emergency-reserve backed) page pool 1254 * for meta data IO (activity log, bitmap). 1255 * We can keep it global, as long as it is used as "N pages at a time". 1256 * 128 should be plenty, currently we probably can get away with as few as 1. 1257 */ 1258 #define DRBD_MIN_POOL_PAGES 128 1259 extern mempool_t *drbd_md_io_page_pool; 1260 1261 /* We also need to make sure we get a bio 1262 * when we need it for housekeeping purposes */ 1263 extern struct bio_set *drbd_md_io_bio_set; 1264 /* to allocate from that set */ 1265 extern struct bio *bio_alloc_drbd(gfp_t gfp_mask); 1266 1267 extern rwlock_t global_state_lock; 1268 1269 extern int conn_lowest_minor(struct drbd_connection *connection); 1270 extern enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor); 1271 extern void drbd_destroy_device(struct kref *kref); 1272 extern void drbd_delete_device(struct drbd_device *device); 1273 1274 extern struct drbd_resource *drbd_create_resource(const char *name); 1275 extern void drbd_free_resource(struct drbd_resource *resource); 1276 1277 extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts); 1278 extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts); 1279 extern void drbd_destroy_connection(struct kref *kref); 1280 extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len, 1281 void *peer_addr, int peer_addr_len); 1282 extern struct drbd_resource *drbd_find_resource(const char *name); 1283 extern void drbd_destroy_resource(struct kref *kref); 1284 extern void conn_free_crypto(struct drbd_connection *connection); 1285 1286 extern int proc_details; 1287 1288 /* drbd_req */ 1289 extern void do_submit(struct work_struct *ws); 1290 extern void __drbd_make_request(struct drbd_device *, struct bio *, unsigned long); 1291 extern void drbd_make_request(struct request_queue *q, struct bio *bio); 1292 extern int drbd_read_remote(struct drbd_device *device, struct drbd_request *req); 1293 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec); 1294 extern int is_valid_ar_handle(struct drbd_request *, sector_t); 1295 1296 1297 /* drbd_nl.c */ 1298 extern int drbd_msg_put_info(struct sk_buff *skb, const char *info); 1299 extern void drbd_suspend_io(struct drbd_device *device); 1300 extern void drbd_resume_io(struct drbd_device *device); 1301 extern char *ppsize(char *buf, unsigned long long size); 1302 extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int); 1303 enum determine_dev_size { 1304 DS_ERROR_SHRINK = -3, 1305 DS_ERROR_SPACE_MD = -2, 1306 DS_ERROR = -1, 1307 DS_UNCHANGED = 0, 1308 DS_SHRUNK = 1, 1309 DS_GREW = 2, 1310 DS_GREW_FROM_ZERO = 3, 1311 }; 1312 extern enum determine_dev_size 1313 drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local); 1314 extern void resync_after_online_grow(struct drbd_device *); 1315 extern void drbd_reconsider_max_bio_size(struct drbd_device *device); 1316 extern enum drbd_state_rv drbd_set_role(struct drbd_device *device, 1317 enum drbd_role new_role, 1318 int force); 1319 extern bool conn_try_outdate_peer(struct drbd_connection *connection); 1320 extern void conn_try_outdate_peer_async(struct drbd_connection *connection); 1321 extern int drbd_khelper(struct drbd_device *device, char *cmd); 1322 1323 /* drbd_worker.c */ 1324 /* bi_end_io handlers */ 1325 extern void drbd_md_io_complete(struct bio *bio, int error); 1326 extern void drbd_peer_request_endio(struct bio *bio, int error); 1327 extern void drbd_request_endio(struct bio *bio, int error); 1328 extern int drbd_worker(struct drbd_thread *thi); 1329 enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor); 1330 void drbd_resync_after_changed(struct drbd_device *device); 1331 extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side); 1332 extern void resume_next_sg(struct drbd_device *device); 1333 extern void suspend_other_sg(struct drbd_device *device); 1334 extern int drbd_resync_finished(struct drbd_device *device); 1335 /* maybe rather drbd_main.c ? */ 1336 extern void *drbd_md_get_buffer(struct drbd_device *device); 1337 extern void drbd_md_put_buffer(struct drbd_device *device); 1338 extern int drbd_md_sync_page_io(struct drbd_device *device, 1339 struct drbd_backing_dev *bdev, sector_t sector, int rw); 1340 extern void drbd_ov_out_of_sync_found(struct drbd_device *, sector_t, int); 1341 extern void wait_until_done_or_force_detached(struct drbd_device *device, 1342 struct drbd_backing_dev *bdev, unsigned int *done); 1343 extern void drbd_rs_controller_reset(struct drbd_device *device); 1344 1345 static inline void ov_out_of_sync_print(struct drbd_device *device) 1346 { 1347 if (device->ov_last_oos_size) { 1348 drbd_err(device, "Out of sync: start=%llu, size=%lu (sectors)\n", 1349 (unsigned long long)device->ov_last_oos_start, 1350 (unsigned long)device->ov_last_oos_size); 1351 } 1352 device->ov_last_oos_size = 0; 1353 } 1354 1355 1356 extern void drbd_csum_bio(struct crypto_hash *, struct bio *, void *); 1357 extern void drbd_csum_ee(struct crypto_hash *, struct drbd_peer_request *, void *); 1358 /* worker callbacks */ 1359 extern int w_e_end_data_req(struct drbd_work *, int); 1360 extern int w_e_end_rsdata_req(struct drbd_work *, int); 1361 extern int w_e_end_csum_rs_req(struct drbd_work *, int); 1362 extern int w_e_end_ov_reply(struct drbd_work *, int); 1363 extern int w_e_end_ov_req(struct drbd_work *, int); 1364 extern int w_ov_finished(struct drbd_work *, int); 1365 extern int w_resync_timer(struct drbd_work *, int); 1366 extern int w_send_write_hint(struct drbd_work *, int); 1367 extern int w_send_dblock(struct drbd_work *, int); 1368 extern int w_send_read_req(struct drbd_work *, int); 1369 extern int w_e_reissue(struct drbd_work *, int); 1370 extern int w_restart_disk_io(struct drbd_work *, int); 1371 extern int w_send_out_of_sync(struct drbd_work *, int); 1372 extern int w_start_resync(struct drbd_work *, int); 1373 1374 extern void resync_timer_fn(unsigned long data); 1375 extern void start_resync_timer_fn(unsigned long data); 1376 1377 extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req); 1378 1379 /* drbd_receiver.c */ 1380 extern int drbd_receiver(struct drbd_thread *thi); 1381 extern int drbd_asender(struct drbd_thread *thi); 1382 extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device); 1383 extern bool drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector); 1384 extern int drbd_submit_peer_request(struct drbd_device *, 1385 struct drbd_peer_request *, const unsigned, 1386 const int); 1387 extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *); 1388 extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64, 1389 sector_t, unsigned int, 1390 bool, 1391 gfp_t) __must_hold(local); 1392 extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *, 1393 int); 1394 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0) 1395 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1) 1396 extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool); 1397 extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled); 1398 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed); 1399 extern int drbd_connected(struct drbd_peer_device *); 1400 1401 /* Yes, there is kernel_setsockopt, but only since 2.6.18. 1402 * So we have our own copy of it here. */ 1403 static inline int drbd_setsockopt(struct socket *sock, int level, int optname, 1404 char *optval, int optlen) 1405 { 1406 mm_segment_t oldfs = get_fs(); 1407 char __user *uoptval; 1408 int err; 1409 1410 uoptval = (char __user __force *)optval; 1411 1412 set_fs(KERNEL_DS); 1413 if (level == SOL_SOCKET) 1414 err = sock_setsockopt(sock, level, optname, uoptval, optlen); 1415 else 1416 err = sock->ops->setsockopt(sock, level, optname, uoptval, 1417 optlen); 1418 set_fs(oldfs); 1419 return err; 1420 } 1421 1422 static inline void drbd_tcp_cork(struct socket *sock) 1423 { 1424 int val = 1; 1425 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK, 1426 (char*)&val, sizeof(val)); 1427 } 1428 1429 static inline void drbd_tcp_uncork(struct socket *sock) 1430 { 1431 int val = 0; 1432 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK, 1433 (char*)&val, sizeof(val)); 1434 } 1435 1436 static inline void drbd_tcp_nodelay(struct socket *sock) 1437 { 1438 int val = 1; 1439 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY, 1440 (char*)&val, sizeof(val)); 1441 } 1442 1443 static inline void drbd_tcp_quickack(struct socket *sock) 1444 { 1445 int val = 2; 1446 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK, 1447 (char*)&val, sizeof(val)); 1448 } 1449 1450 /* sets the number of 512 byte sectors of our virtual device */ 1451 static inline void drbd_set_my_capacity(struct drbd_device *device, 1452 sector_t size) 1453 { 1454 /* set_capacity(device->this_bdev->bd_disk, size); */ 1455 set_capacity(device->vdisk, size); 1456 device->this_bdev->bd_inode->i_size = (loff_t)size << 9; 1457 } 1458 1459 /* 1460 * used to submit our private bio 1461 */ 1462 static inline void drbd_generic_make_request(struct drbd_device *device, 1463 int fault_type, struct bio *bio) 1464 { 1465 __release(local); 1466 if (!bio->bi_bdev) { 1467 printk(KERN_ERR "drbd%d: drbd_generic_make_request: " 1468 "bio->bi_bdev == NULL\n", 1469 device_to_minor(device)); 1470 dump_stack(); 1471 bio_endio(bio, -ENODEV); 1472 return; 1473 } 1474 1475 if (drbd_insert_fault(device, fault_type)) 1476 bio_endio(bio, -EIO); 1477 else 1478 generic_make_request(bio); 1479 } 1480 1481 void drbd_bump_write_ordering(struct drbd_connection *connection, enum write_ordering_e wo); 1482 1483 /* drbd_proc.c */ 1484 extern struct proc_dir_entry *drbd_proc; 1485 extern const struct file_operations drbd_proc_fops; 1486 extern const char *drbd_conn_str(enum drbd_conns s); 1487 extern const char *drbd_role_str(enum drbd_role s); 1488 1489 /* drbd_actlog.c */ 1490 extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i); 1491 extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i); 1492 extern void drbd_al_begin_io_commit(struct drbd_device *device, bool delegate); 1493 extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i); 1494 extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i, bool delegate); 1495 extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i); 1496 extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector); 1497 extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector); 1498 extern int drbd_try_rs_begin_io(struct drbd_device *device, sector_t sector); 1499 extern void drbd_rs_cancel_all(struct drbd_device *device); 1500 extern int drbd_rs_del_all(struct drbd_device *device); 1501 extern void drbd_rs_failed_io(struct drbd_device *device, 1502 sector_t sector, int size); 1503 extern void drbd_advance_rs_marks(struct drbd_device *device, unsigned long still_to_go); 1504 extern void __drbd_set_in_sync(struct drbd_device *device, sector_t sector, 1505 int size, const char *file, const unsigned int line); 1506 #define drbd_set_in_sync(device, sector, size) \ 1507 __drbd_set_in_sync(device, sector, size, __FILE__, __LINE__) 1508 extern int __drbd_set_out_of_sync(struct drbd_device *device, sector_t sector, 1509 int size, const char *file, const unsigned int line); 1510 #define drbd_set_out_of_sync(device, sector, size) \ 1511 __drbd_set_out_of_sync(device, sector, size, __FILE__, __LINE__) 1512 extern void drbd_al_shrink(struct drbd_device *device); 1513 extern int drbd_initialize_al(struct drbd_device *, void *); 1514 1515 /* drbd_nl.c */ 1516 /* state info broadcast */ 1517 struct sib_info { 1518 enum drbd_state_info_bcast_reason sib_reason; 1519 union { 1520 struct { 1521 char *helper_name; 1522 unsigned helper_exit_code; 1523 }; 1524 struct { 1525 union drbd_state os; 1526 union drbd_state ns; 1527 }; 1528 }; 1529 }; 1530 void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib); 1531 1532 /* 1533 * inline helper functions 1534 *************************/ 1535 1536 /* see also page_chain_add and friends in drbd_receiver.c */ 1537 static inline struct page *page_chain_next(struct page *page) 1538 { 1539 return (struct page *)page_private(page); 1540 } 1541 #define page_chain_for_each(page) \ 1542 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \ 1543 page = page_chain_next(page)) 1544 #define page_chain_for_each_safe(page, n) \ 1545 for (; page && ({ n = page_chain_next(page); 1; }); page = n) 1546 1547 1548 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req) 1549 { 1550 struct page *page = peer_req->pages; 1551 page_chain_for_each(page) { 1552 if (page_count(page) > 1) 1553 return 1; 1554 } 1555 return 0; 1556 } 1557 1558 static inline enum drbd_state_rv 1559 _drbd_set_state(struct drbd_device *device, union drbd_state ns, 1560 enum chg_state_flags flags, struct completion *done) 1561 { 1562 enum drbd_state_rv rv; 1563 1564 read_lock(&global_state_lock); 1565 rv = __drbd_set_state(device, ns, flags, done); 1566 read_unlock(&global_state_lock); 1567 1568 return rv; 1569 } 1570 1571 static inline union drbd_state drbd_read_state(struct drbd_device *device) 1572 { 1573 struct drbd_resource *resource = device->resource; 1574 union drbd_state rv; 1575 1576 rv.i = device->state.i; 1577 rv.susp = resource->susp; 1578 rv.susp_nod = resource->susp_nod; 1579 rv.susp_fen = resource->susp_fen; 1580 1581 return rv; 1582 } 1583 1584 enum drbd_force_detach_flags { 1585 DRBD_READ_ERROR, 1586 DRBD_WRITE_ERROR, 1587 DRBD_META_IO_ERROR, 1588 DRBD_FORCE_DETACH, 1589 }; 1590 1591 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__) 1592 static inline void __drbd_chk_io_error_(struct drbd_device *device, 1593 enum drbd_force_detach_flags df, 1594 const char *where) 1595 { 1596 enum drbd_io_error_p ep; 1597 1598 rcu_read_lock(); 1599 ep = rcu_dereference(device->ldev->disk_conf)->on_io_error; 1600 rcu_read_unlock(); 1601 switch (ep) { 1602 case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */ 1603 if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) { 1604 if (__ratelimit(&drbd_ratelimit_state)) 1605 drbd_err(device, "Local IO failed in %s.\n", where); 1606 if (device->state.disk > D_INCONSISTENT) 1607 _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL); 1608 break; 1609 } 1610 /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */ 1611 case EP_DETACH: 1612 case EP_CALL_HELPER: 1613 /* Remember whether we saw a READ or WRITE error. 1614 * 1615 * Recovery of the affected area for WRITE failure is covered 1616 * by the activity log. 1617 * READ errors may fall outside that area though. Certain READ 1618 * errors can be "healed" by writing good data to the affected 1619 * blocks, which triggers block re-allocation in lower layers. 1620 * 1621 * If we can not write the bitmap after a READ error, 1622 * we may need to trigger a full sync (see w_go_diskless()). 1623 * 1624 * Force-detach is not really an IO error, but rather a 1625 * desperate measure to try to deal with a completely 1626 * unresponsive lower level IO stack. 1627 * Still it should be treated as a WRITE error. 1628 * 1629 * Meta IO error is always WRITE error: 1630 * we read meta data only once during attach, 1631 * which will fail in case of errors. 1632 */ 1633 set_bit(WAS_IO_ERROR, &device->flags); 1634 if (df == DRBD_READ_ERROR) 1635 set_bit(WAS_READ_ERROR, &device->flags); 1636 if (df == DRBD_FORCE_DETACH) 1637 set_bit(FORCE_DETACH, &device->flags); 1638 if (device->state.disk > D_FAILED) { 1639 _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL); 1640 drbd_err(device, 1641 "Local IO failed in %s. Detaching...\n", where); 1642 } 1643 break; 1644 } 1645 } 1646 1647 /** 1648 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers 1649 * @device: DRBD device. 1650 * @error: Error code passed to the IO completion callback 1651 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data 1652 * 1653 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED) 1654 */ 1655 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__) 1656 static inline void drbd_chk_io_error_(struct drbd_device *device, 1657 int error, enum drbd_force_detach_flags forcedetach, const char *where) 1658 { 1659 if (error) { 1660 unsigned long flags; 1661 spin_lock_irqsave(&device->resource->req_lock, flags); 1662 __drbd_chk_io_error_(device, forcedetach, where); 1663 spin_unlock_irqrestore(&device->resource->req_lock, flags); 1664 } 1665 } 1666 1667 1668 /** 1669 * drbd_md_first_sector() - Returns the first sector number of the meta data area 1670 * @bdev: Meta data block device. 1671 * 1672 * BTW, for internal meta data, this happens to be the maximum capacity 1673 * we could agree upon with our peer node. 1674 */ 1675 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev) 1676 { 1677 switch (bdev->md.meta_dev_idx) { 1678 case DRBD_MD_INDEX_INTERNAL: 1679 case DRBD_MD_INDEX_FLEX_INT: 1680 return bdev->md.md_offset + bdev->md.bm_offset; 1681 case DRBD_MD_INDEX_FLEX_EXT: 1682 default: 1683 return bdev->md.md_offset; 1684 } 1685 } 1686 1687 /** 1688 * drbd_md_last_sector() - Return the last sector number of the meta data area 1689 * @bdev: Meta data block device. 1690 */ 1691 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev) 1692 { 1693 switch (bdev->md.meta_dev_idx) { 1694 case DRBD_MD_INDEX_INTERNAL: 1695 case DRBD_MD_INDEX_FLEX_INT: 1696 return bdev->md.md_offset + MD_4kB_SECT -1; 1697 case DRBD_MD_INDEX_FLEX_EXT: 1698 default: 1699 return bdev->md.md_offset + bdev->md.md_size_sect -1; 1700 } 1701 } 1702 1703 /* Returns the number of 512 byte sectors of the device */ 1704 static inline sector_t drbd_get_capacity(struct block_device *bdev) 1705 { 1706 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */ 1707 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0; 1708 } 1709 1710 /** 1711 * drbd_get_max_capacity() - Returns the capacity we announce to out peer 1712 * @bdev: Meta data block device. 1713 * 1714 * returns the capacity we announce to out peer. we clip ourselves at the 1715 * various MAX_SECTORS, because if we don't, current implementation will 1716 * oops sooner or later 1717 */ 1718 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev) 1719 { 1720 sector_t s; 1721 1722 switch (bdev->md.meta_dev_idx) { 1723 case DRBD_MD_INDEX_INTERNAL: 1724 case DRBD_MD_INDEX_FLEX_INT: 1725 s = drbd_get_capacity(bdev->backing_bdev) 1726 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX, 1727 drbd_md_first_sector(bdev)) 1728 : 0; 1729 break; 1730 case DRBD_MD_INDEX_FLEX_EXT: 1731 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX, 1732 drbd_get_capacity(bdev->backing_bdev)); 1733 /* clip at maximum size the meta device can support */ 1734 s = min_t(sector_t, s, 1735 BM_EXT_TO_SECT(bdev->md.md_size_sect 1736 - bdev->md.bm_offset)); 1737 break; 1738 default: 1739 s = min_t(sector_t, DRBD_MAX_SECTORS, 1740 drbd_get_capacity(bdev->backing_bdev)); 1741 } 1742 return s; 1743 } 1744 1745 /** 1746 * drbd_md_ss() - Return the sector number of our meta data super block 1747 * @bdev: Meta data block device. 1748 */ 1749 static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev) 1750 { 1751 const int meta_dev_idx = bdev->md.meta_dev_idx; 1752 1753 if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT) 1754 return 0; 1755 1756 /* Since drbd08, internal meta data is always "flexible". 1757 * position: last 4k aligned block of 4k size */ 1758 if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL || 1759 meta_dev_idx == DRBD_MD_INDEX_FLEX_INT) 1760 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8; 1761 1762 /* external, some index; this is the old fixed size layout */ 1763 return MD_128MB_SECT * bdev->md.meta_dev_idx; 1764 } 1765 1766 static inline void 1767 drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w) 1768 { 1769 unsigned long flags; 1770 spin_lock_irqsave(&q->q_lock, flags); 1771 list_add(&w->list, &q->q); 1772 spin_unlock_irqrestore(&q->q_lock, flags); 1773 wake_up(&q->q_wait); 1774 } 1775 1776 static inline void 1777 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w) 1778 { 1779 unsigned long flags; 1780 spin_lock_irqsave(&q->q_lock, flags); 1781 list_add_tail(&w->list, &q->q); 1782 spin_unlock_irqrestore(&q->q_lock, flags); 1783 wake_up(&q->q_wait); 1784 } 1785 1786 extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue); 1787 1788 static inline void wake_asender(struct drbd_connection *connection) 1789 { 1790 if (test_bit(SIGNAL_ASENDER, &connection->flags)) 1791 force_sig(DRBD_SIG, connection->asender.task); 1792 } 1793 1794 static inline void request_ping(struct drbd_connection *connection) 1795 { 1796 set_bit(SEND_PING, &connection->flags); 1797 wake_asender(connection); 1798 } 1799 1800 extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *); 1801 extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *); 1802 extern int conn_send_command(struct drbd_connection *, struct drbd_socket *, 1803 enum drbd_packet, unsigned int, void *, 1804 unsigned int); 1805 extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *, 1806 enum drbd_packet, unsigned int, void *, 1807 unsigned int); 1808 1809 extern int drbd_send_ping(struct drbd_connection *connection); 1810 extern int drbd_send_ping_ack(struct drbd_connection *connection); 1811 extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state); 1812 extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state); 1813 1814 static inline void drbd_thread_stop(struct drbd_thread *thi) 1815 { 1816 _drbd_thread_stop(thi, false, true); 1817 } 1818 1819 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi) 1820 { 1821 _drbd_thread_stop(thi, false, false); 1822 } 1823 1824 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi) 1825 { 1826 _drbd_thread_stop(thi, true, false); 1827 } 1828 1829 /* counts how many answer packets packets we expect from our peer, 1830 * for either explicit application requests, 1831 * or implicit barrier packets as necessary. 1832 * increased: 1833 * w_send_barrier 1834 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ); 1835 * it is much easier and equally valid to count what we queue for the 1836 * worker, even before it actually was queued or send. 1837 * (drbd_make_request_common; recovery path on read io-error) 1838 * decreased: 1839 * got_BarrierAck (respective tl_clear, tl_clear_barrier) 1840 * _req_mod(req, DATA_RECEIVED) 1841 * [from receive_DataReply] 1842 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED) 1843 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)] 1844 * for some reason it is NOT decreased in got_NegAck, 1845 * but in the resulting cleanup code from report_params. 1846 * we should try to remember the reason for that... 1847 * _req_mod(req, SEND_FAILED or SEND_CANCELED) 1848 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING) 1849 * [from tl_clear_barrier] 1850 */ 1851 static inline void inc_ap_pending(struct drbd_device *device) 1852 { 1853 atomic_inc(&device->ap_pending_cnt); 1854 } 1855 1856 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \ 1857 if (atomic_read(&device->which) < 0) \ 1858 drbd_err(device, "in %s:%d: " #which " = %d < 0 !\n", \ 1859 func, line, \ 1860 atomic_read(&device->which)) 1861 1862 #define dec_ap_pending(device) _dec_ap_pending(device, __FUNCTION__, __LINE__) 1863 static inline void _dec_ap_pending(struct drbd_device *device, const char *func, int line) 1864 { 1865 if (atomic_dec_and_test(&device->ap_pending_cnt)) 1866 wake_up(&device->misc_wait); 1867 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line); 1868 } 1869 1870 /* counts how many resync-related answers we still expect from the peer 1871 * increase decrease 1872 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY) 1873 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER) 1874 * (or P_NEG_ACK with ID_SYNCER) 1875 */ 1876 static inline void inc_rs_pending(struct drbd_device *device) 1877 { 1878 atomic_inc(&device->rs_pending_cnt); 1879 } 1880 1881 #define dec_rs_pending(device) _dec_rs_pending(device, __FUNCTION__, __LINE__) 1882 static inline void _dec_rs_pending(struct drbd_device *device, const char *func, int line) 1883 { 1884 atomic_dec(&device->rs_pending_cnt); 1885 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line); 1886 } 1887 1888 /* counts how many answers we still need to send to the peer. 1889 * increased on 1890 * receive_Data unless protocol A; 1891 * we need to send a P_RECV_ACK (proto B) 1892 * or P_WRITE_ACK (proto C) 1893 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK 1894 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA 1895 * receive_Barrier_* we need to send a P_BARRIER_ACK 1896 */ 1897 static inline void inc_unacked(struct drbd_device *device) 1898 { 1899 atomic_inc(&device->unacked_cnt); 1900 } 1901 1902 #define dec_unacked(device) _dec_unacked(device, __FUNCTION__, __LINE__) 1903 static inline void _dec_unacked(struct drbd_device *device, const char *func, int line) 1904 { 1905 atomic_dec(&device->unacked_cnt); 1906 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line); 1907 } 1908 1909 #define sub_unacked(device, n) _sub_unacked(device, n, __FUNCTION__, __LINE__) 1910 static inline void _sub_unacked(struct drbd_device *device, int n, const char *func, int line) 1911 { 1912 atomic_sub(n, &device->unacked_cnt); 1913 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line); 1914 } 1915 1916 /** 1917 * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev 1918 * @M: DRBD device. 1919 * 1920 * You have to call put_ldev() when finished working with device->ldev. 1921 */ 1922 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT)) 1923 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS)) 1924 1925 static inline void put_ldev(struct drbd_device *device) 1926 { 1927 int i = atomic_dec_return(&device->local_cnt); 1928 1929 /* This may be called from some endio handler, 1930 * so we must not sleep here. */ 1931 1932 __release(local); 1933 D_ASSERT(device, i >= 0); 1934 if (i == 0) { 1935 if (device->state.disk == D_DISKLESS) 1936 /* even internal references gone, safe to destroy */ 1937 drbd_ldev_destroy(device); 1938 if (device->state.disk == D_FAILED) { 1939 /* all application IO references gone. */ 1940 if (!test_and_set_bit(GO_DISKLESS, &device->flags)) 1941 drbd_queue_work(&first_peer_device(device)->connection->sender_work, 1942 &device->go_diskless); 1943 } 1944 wake_up(&device->misc_wait); 1945 } 1946 } 1947 1948 #ifndef __CHECKER__ 1949 static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins) 1950 { 1951 int io_allowed; 1952 1953 /* never get a reference while D_DISKLESS */ 1954 if (device->state.disk == D_DISKLESS) 1955 return 0; 1956 1957 atomic_inc(&device->local_cnt); 1958 io_allowed = (device->state.disk >= mins); 1959 if (!io_allowed) 1960 put_ldev(device); 1961 return io_allowed; 1962 } 1963 #else 1964 extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins); 1965 #endif 1966 1967 /* you must have an "get_ldev" reference */ 1968 static inline void drbd_get_syncer_progress(struct drbd_device *device, 1969 unsigned long *bits_left, unsigned int *per_mil_done) 1970 { 1971 /* this is to break it at compile time when we change that, in case we 1972 * want to support more than (1<<32) bits on a 32bit arch. */ 1973 typecheck(unsigned long, device->rs_total); 1974 1975 /* note: both rs_total and rs_left are in bits, i.e. in 1976 * units of BM_BLOCK_SIZE. 1977 * for the percentage, we don't care. */ 1978 1979 if (device->state.conn == C_VERIFY_S || device->state.conn == C_VERIFY_T) 1980 *bits_left = device->ov_left; 1981 else 1982 *bits_left = drbd_bm_total_weight(device) - device->rs_failed; 1983 /* >> 10 to prevent overflow, 1984 * +1 to prevent division by zero */ 1985 if (*bits_left > device->rs_total) { 1986 /* doh. maybe a logic bug somewhere. 1987 * may also be just a race condition 1988 * between this and a disconnect during sync. 1989 * for now, just prevent in-kernel buffer overflow. 1990 */ 1991 smp_rmb(); 1992 drbd_warn(device, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n", 1993 drbd_conn_str(device->state.conn), 1994 *bits_left, device->rs_total, device->rs_failed); 1995 *per_mil_done = 0; 1996 } else { 1997 /* Make sure the division happens in long context. 1998 * We allow up to one petabyte storage right now, 1999 * at a granularity of 4k per bit that is 2**38 bits. 2000 * After shift right and multiplication by 1000, 2001 * this should still fit easily into a 32bit long, 2002 * so we don't need a 64bit division on 32bit arch. 2003 * Note: currently we don't support such large bitmaps on 32bit 2004 * arch anyways, but no harm done to be prepared for it here. 2005 */ 2006 unsigned int shift = device->rs_total > UINT_MAX ? 16 : 10; 2007 unsigned long left = *bits_left >> shift; 2008 unsigned long total = 1UL + (device->rs_total >> shift); 2009 unsigned long tmp = 1000UL - left * 1000UL/total; 2010 *per_mil_done = tmp; 2011 } 2012 } 2013 2014 2015 /* this throttles on-the-fly application requests 2016 * according to max_buffers settings; 2017 * maybe re-implement using semaphores? */ 2018 static inline int drbd_get_max_buffers(struct drbd_device *device) 2019 { 2020 struct net_conf *nc; 2021 int mxb; 2022 2023 rcu_read_lock(); 2024 nc = rcu_dereference(first_peer_device(device)->connection->net_conf); 2025 mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */ 2026 rcu_read_unlock(); 2027 2028 return mxb; 2029 } 2030 2031 static inline int drbd_state_is_stable(struct drbd_device *device) 2032 { 2033 union drbd_dev_state s = device->state; 2034 2035 /* DO NOT add a default clause, we want the compiler to warn us 2036 * for any newly introduced state we may have forgotten to add here */ 2037 2038 switch ((enum drbd_conns)s.conn) { 2039 /* new io only accepted when there is no connection, ... */ 2040 case C_STANDALONE: 2041 case C_WF_CONNECTION: 2042 /* ... or there is a well established connection. */ 2043 case C_CONNECTED: 2044 case C_SYNC_SOURCE: 2045 case C_SYNC_TARGET: 2046 case C_VERIFY_S: 2047 case C_VERIFY_T: 2048 case C_PAUSED_SYNC_S: 2049 case C_PAUSED_SYNC_T: 2050 case C_AHEAD: 2051 case C_BEHIND: 2052 /* transitional states, IO allowed */ 2053 case C_DISCONNECTING: 2054 case C_UNCONNECTED: 2055 case C_TIMEOUT: 2056 case C_BROKEN_PIPE: 2057 case C_NETWORK_FAILURE: 2058 case C_PROTOCOL_ERROR: 2059 case C_TEAR_DOWN: 2060 case C_WF_REPORT_PARAMS: 2061 case C_STARTING_SYNC_S: 2062 case C_STARTING_SYNC_T: 2063 break; 2064 2065 /* Allow IO in BM exchange states with new protocols */ 2066 case C_WF_BITMAP_S: 2067 if (first_peer_device(device)->connection->agreed_pro_version < 96) 2068 return 0; 2069 break; 2070 2071 /* no new io accepted in these states */ 2072 case C_WF_BITMAP_T: 2073 case C_WF_SYNC_UUID: 2074 case C_MASK: 2075 /* not "stable" */ 2076 return 0; 2077 } 2078 2079 switch ((enum drbd_disk_state)s.disk) { 2080 case D_DISKLESS: 2081 case D_INCONSISTENT: 2082 case D_OUTDATED: 2083 case D_CONSISTENT: 2084 case D_UP_TO_DATE: 2085 case D_FAILED: 2086 /* disk state is stable as well. */ 2087 break; 2088 2089 /* no new io accepted during transitional states */ 2090 case D_ATTACHING: 2091 case D_NEGOTIATING: 2092 case D_UNKNOWN: 2093 case D_MASK: 2094 /* not "stable" */ 2095 return 0; 2096 } 2097 2098 return 1; 2099 } 2100 2101 static inline int drbd_suspended(struct drbd_device *device) 2102 { 2103 struct drbd_resource *resource = device->resource; 2104 2105 return resource->susp || resource->susp_fen || resource->susp_nod; 2106 } 2107 2108 static inline bool may_inc_ap_bio(struct drbd_device *device) 2109 { 2110 int mxb = drbd_get_max_buffers(device); 2111 2112 if (drbd_suspended(device)) 2113 return false; 2114 if (test_bit(SUSPEND_IO, &device->flags)) 2115 return false; 2116 2117 /* to avoid potential deadlock or bitmap corruption, 2118 * in various places, we only allow new application io 2119 * to start during "stable" states. */ 2120 2121 /* no new io accepted when attaching or detaching the disk */ 2122 if (!drbd_state_is_stable(device)) 2123 return false; 2124 2125 /* since some older kernels don't have atomic_add_unless, 2126 * and we are within the spinlock anyways, we have this workaround. */ 2127 if (atomic_read(&device->ap_bio_cnt) > mxb) 2128 return false; 2129 if (test_bit(BITMAP_IO, &device->flags)) 2130 return false; 2131 return true; 2132 } 2133 2134 static inline bool inc_ap_bio_cond(struct drbd_device *device) 2135 { 2136 bool rv = false; 2137 2138 spin_lock_irq(&device->resource->req_lock); 2139 rv = may_inc_ap_bio(device); 2140 if (rv) 2141 atomic_inc(&device->ap_bio_cnt); 2142 spin_unlock_irq(&device->resource->req_lock); 2143 2144 return rv; 2145 } 2146 2147 static inline void inc_ap_bio(struct drbd_device *device) 2148 { 2149 /* we wait here 2150 * as long as the device is suspended 2151 * until the bitmap is no longer on the fly during connection 2152 * handshake as long as we would exceed the max_buffer limit. 2153 * 2154 * to avoid races with the reconnect code, 2155 * we need to atomic_inc within the spinlock. */ 2156 2157 wait_event(device->misc_wait, inc_ap_bio_cond(device)); 2158 } 2159 2160 static inline void dec_ap_bio(struct drbd_device *device) 2161 { 2162 int mxb = drbd_get_max_buffers(device); 2163 int ap_bio = atomic_dec_return(&device->ap_bio_cnt); 2164 2165 D_ASSERT(device, ap_bio >= 0); 2166 2167 if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) { 2168 if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags)) 2169 drbd_queue_work(&first_peer_device(device)-> 2170 connection->sender_work, 2171 &device->bm_io_work.w); 2172 } 2173 2174 /* this currently does wake_up for every dec_ap_bio! 2175 * maybe rather introduce some type of hysteresis? 2176 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */ 2177 if (ap_bio < mxb) 2178 wake_up(&device->misc_wait); 2179 } 2180 2181 static inline bool verify_can_do_stop_sector(struct drbd_device *device) 2182 { 2183 return first_peer_device(device)->connection->agreed_pro_version >= 97 && 2184 first_peer_device(device)->connection->agreed_pro_version != 100; 2185 } 2186 2187 static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val) 2188 { 2189 int changed = device->ed_uuid != val; 2190 device->ed_uuid = val; 2191 return changed; 2192 } 2193 2194 static inline int drbd_queue_order_type(struct drbd_device *device) 2195 { 2196 /* sorry, we currently have no working implementation 2197 * of distributed TCQ stuff */ 2198 #ifndef QUEUE_ORDERED_NONE 2199 #define QUEUE_ORDERED_NONE 0 2200 #endif 2201 return QUEUE_ORDERED_NONE; 2202 } 2203 2204 static inline void drbd_md_flush(struct drbd_device *device) 2205 { 2206 int r; 2207 2208 if (device->ldev == NULL) { 2209 drbd_warn(device, "device->ldev == NULL in drbd_md_flush\n"); 2210 return; 2211 } 2212 2213 if (test_bit(MD_NO_FUA, &device->flags)) 2214 return; 2215 2216 r = blkdev_issue_flush(device->ldev->md_bdev, GFP_NOIO, NULL); 2217 if (r) { 2218 set_bit(MD_NO_FUA, &device->flags); 2219 drbd_err(device, "meta data flush failed with status %d, disabling md-flushes\n", r); 2220 } 2221 } 2222 2223 static inline struct drbd_connection *first_connection(struct drbd_resource *resource) 2224 { 2225 return list_first_entry_or_null(&resource->connections, 2226 struct drbd_connection, connections); 2227 } 2228 2229 #endif 2230