FILE
drbd_int.h
File preview with syntax highlighting.
linux-6.19.14/drivers/block/drbd/drbd_int.h
1/* SPDX-License-Identifier: GPL-2.0-only */ 2/* 3 drbd_int.h 4 5 This file is part of DRBD by Philipp Reisner and Lars Ellenberg. 6 7 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. 8 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. 9 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. 10 11 12*/ 13 14#ifndef _DRBD_INT_H 15#define _DRBD_INT_H 16 17#include <crypto/hash.h> 18#include <linux/compiler.h> 19#include <linux/types.h> 20#include <linux/list.h> 21#include <linux/sched/signal.h> 22#include <linux/bitops.h> 23#include <linux/slab.h> 24#include <linux/ratelimit.h> 25#include <linux/tcp.h> 26#include <linux/mutex.h> 27#include <linux/major.h> 28#include <linux/blkdev.h> 29#include <linux/backing-dev.h> 30#include <linux/idr.h> 31#include <linux/dynamic_debug.h> 32#include <net/tcp.h> 33#include <linux/lru_cache.h> 34#include <linux/prefetch.h> 35#include <linux/drbd_genl_api.h> 36#include <linux/drbd.h> 37#include <linux/drbd_config.h> 38#include "drbd_strings.h" 39#include "drbd_state.h" 40#include "drbd_protocol.h" 41#include "drbd_polymorph_printk.h" 42 43/* shared module parameters, defined in drbd_main.c */ 44#ifdef CONFIG_DRBD_FAULT_INJECTION 45extern int drbd_enable_faults; 46extern int drbd_fault_rate; 47#endif 48 49extern unsigned int drbd_minor_count; 50extern char drbd_usermode_helper[]; 51extern int drbd_proc_details; 52 53 54/* This is used to stop/restart our threads. 55 * Cannot use SIGTERM nor SIGKILL, since these 56 * are sent out by init on runlevel changes 57 * I choose SIGHUP for now. 58 */ 59#define DRBD_SIGKILL SIGHUP 60 61#define ID_IN_SYNC (4711ULL) 62#define ID_OUT_OF_SYNC (4712ULL) 63#define ID_SYNCER (-1ULL) 64 65#define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL) 66 67struct drbd_device; 68struct drbd_connection; 69struct drbd_peer_device; 70 71/* Defines to control fault insertion */ 72enum { 73 DRBD_FAULT_MD_WR = 0, /* meta data write */ 74 DRBD_FAULT_MD_RD = 1, /* read */ 75 DRBD_FAULT_RS_WR = 2, /* resync */ 76 DRBD_FAULT_RS_RD = 3, 77 DRBD_FAULT_DT_WR = 4, /* data */ 78 DRBD_FAULT_DT_RD = 5, 79 DRBD_FAULT_DT_RA = 6, /* data read ahead */ 80 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */ 81 DRBD_FAULT_AL_EE = 8, /* alloc ee */ 82 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */ 83 84 DRBD_FAULT_MAX, 85}; 86 87extern unsigned int 88_drbd_insert_fault(struct drbd_device *device, unsigned int type); 89 90static inline int 91drbd_insert_fault(struct drbd_device *device, unsigned int type) { 92#ifdef CONFIG_DRBD_FAULT_INJECTION 93 return drbd_fault_rate && 94 (drbd_enable_faults & (1<<type)) && 95 _drbd_insert_fault(device, type); 96#else 97 return 0; 98#endif 99} 100 101/* integer division, round _UP_ to the next integer */ 102#define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0)) 103/* usual integer division */ 104#define div_floor(A, B) ((A)/(B)) 105 106extern struct ratelimit_state drbd_ratelimit_state; 107extern struct idr drbd_devices; /* RCU, updates: genl_lock() */ 108extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */ 109 110extern const char *cmdname(enum drbd_packet cmd); 111 112/* for sending/receiving the bitmap, 113 * possibly in some encoding scheme */ 114struct bm_xfer_ctx { 115 /* "const" 116 * stores total bits and long words 117 * of the bitmap, so we don't need to 118 * call the accessor functions over and again. */ 119 unsigned long bm_bits; 120 unsigned long bm_words; 121 /* during xfer, current position within the bitmap */ 122 unsigned long bit_offset; 123 unsigned long word_offset; 124 125 /* statistics; index: (h->command == P_BITMAP) */ 126 unsigned packets[2]; 127 unsigned bytes[2]; 128}; 129 130extern void INFO_bm_xfer_stats(struct drbd_peer_device *peer_device, 131 const char *direction, struct bm_xfer_ctx *c); 132 133static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c) 134{ 135 /* word_offset counts "native long words" (32 or 64 bit), 136 * aligned at 64 bit. 137 * Encoded packet may end at an unaligned bit offset. 138 * In case a fallback clear text packet is transmitted in 139 * between, we adjust this offset back to the last 64bit 140 * aligned "native long word", which makes coding and decoding 141 * the plain text bitmap much more convenient. */ 142#if BITS_PER_LONG == 64 143 c->word_offset = c->bit_offset >> 6; 144#elif BITS_PER_LONG == 32 145 c->word_offset = c->bit_offset >> 5; 146 c->word_offset &= ~(1UL); 147#else 148# error "unsupported BITS_PER_LONG" 149#endif 150} 151 152extern unsigned int drbd_header_size(struct drbd_connection *connection); 153 154/**********************************************************************/ 155enum drbd_thread_state { 156 NONE, 157 RUNNING, 158 EXITING, 159 RESTARTING 160}; 161 162struct drbd_thread { 163 spinlock_t t_lock; 164 struct task_struct *task; 165 struct completion stop; 166 enum drbd_thread_state t_state; 167 int (*function) (struct drbd_thread *); 168 struct drbd_resource *resource; 169 struct drbd_connection *connection; 170 int reset_cpu_mask; 171 const char *name; 172}; 173 174static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi) 175{ 176 /* THINK testing the t_state seems to be uncritical in all cases 177 * (but thread_{start,stop}), so we can read it *without* the lock. 178 * --lge */ 179 180 smp_rmb(); 181 return thi->t_state; 182} 183 184struct drbd_work { 185 struct list_head list; 186 int (*cb)(struct drbd_work *, int cancel); 187}; 188 189struct drbd_device_work { 190 struct drbd_work w; 191 struct drbd_device *device; 192}; 193 194#include "drbd_interval.h" 195 196extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *); 197 198extern void lock_all_resources(void); 199extern void unlock_all_resources(void); 200 201struct drbd_request { 202 struct drbd_work w; 203 struct drbd_device *device; 204 205 /* if local IO is not allowed, will be NULL. 206 * if local IO _is_ allowed, holds the locally submitted bio clone, 207 * or, after local IO completion, the ERR_PTR(error). 208 * see drbd_request_endio(). */ 209 struct bio *private_bio; 210 211 struct drbd_interval i; 212 213 /* epoch: used to check on "completion" whether this req was in 214 * the current epoch, and we therefore have to close it, 215 * causing a p_barrier packet to be send, starting a new epoch. 216 * 217 * This corresponds to "barrier" in struct p_barrier[_ack], 218 * and to "barrier_nr" in struct drbd_epoch (and various 219 * comments/function parameters/local variable names). 220 */ 221 unsigned int epoch; 222 223 struct list_head tl_requests; /* ring list in the transfer log */ 224 struct bio *master_bio; /* master bio pointer */ 225 226 /* see struct drbd_device */ 227 struct list_head req_pending_master_completion; 228 struct list_head req_pending_local; 229 230 /* for generic IO accounting */ 231 unsigned long start_jif; 232 233 /* for DRBD internal statistics */ 234 235 /* Minimal set of time stamps to determine if we wait for activity log 236 * transactions, local disk or peer. 32 bit "jiffies" are good enough, 237 * we don't expect a DRBD request to be stalled for several month. 238 */ 239 240 /* before actual request processing */ 241 unsigned long in_actlog_jif; 242 243 /* local disk */ 244 unsigned long pre_submit_jif; 245 246 /* per connection */ 247 unsigned long pre_send_jif; 248 unsigned long acked_jif; 249 unsigned long net_done_jif; 250 251 /* Possibly even more detail to track each phase: 252 * master_completion_jif 253 * how long did it take to complete the master bio 254 * (application visible latency) 255 * allocated_jif 256 * how long the master bio was blocked until we finally allocated 257 * a tracking struct 258 * in_actlog_jif 259 * how long did we wait for activity log transactions 260 * 261 * net_queued_jif 262 * when did we finally queue it for sending 263 * pre_send_jif 264 * when did we start sending it 265 * post_send_jif 266 * how long did we block in the network stack trying to send it 267 * acked_jif 268 * when did we receive (or fake, in protocol A) a remote ACK 269 * net_done_jif 270 * when did we receive final acknowledgement (P_BARRIER_ACK), 271 * or decide, e.g. on connection loss, that we do no longer expect 272 * anything from this peer for this request. 273 * 274 * pre_submit_jif 275 * post_sub_jif 276 * when did we start submiting to the lower level device, 277 * and how long did we block in that submit function 278 * local_completion_jif 279 * how long did it take the lower level device to complete this request 280 */ 281 282 283 /* once it hits 0, we may complete the master_bio */ 284 atomic_t completion_ref; 285 /* once it hits 0, we may destroy this drbd_request object */ 286 struct kref kref; 287 288 unsigned rq_state; /* see comments above _req_mod() */ 289}; 290 291struct drbd_epoch { 292 struct drbd_connection *connection; 293 struct list_head list; 294 unsigned int barrier_nr; 295 atomic_t epoch_size; /* increased on every request added. */ 296 atomic_t active; /* increased on every req. added, and dec on every finished. */ 297 unsigned long flags; 298}; 299 300/* drbd_epoch flag bits */ 301enum { 302 DE_HAVE_BARRIER_NUMBER, 303}; 304 305enum epoch_event { 306 EV_PUT, 307 EV_GOT_BARRIER_NR, 308 EV_BECAME_LAST, 309 EV_CLEANUP = 32, /* used as flag */ 310}; 311 312struct digest_info { 313 int digest_size; 314 void *digest; 315}; 316 317struct drbd_peer_request { 318 struct drbd_work w; 319 struct drbd_peer_device *peer_device; 320 struct drbd_epoch *epoch; /* for writes */ 321 struct page *pages; 322 blk_opf_t opf; 323 atomic_t pending_bios; 324 struct drbd_interval i; 325 /* see comments on ee flag bits below */ 326 unsigned long flags; 327 unsigned long submit_jif; 328 union { 329 u64 block_id; 330 struct digest_info *digest; 331 }; 332}; 333 334/* Equivalent to bio_op and req_op. */ 335#define peer_req_op(peer_req) \ 336 ((peer_req)->opf & REQ_OP_MASK) 337 338/* ee flag bits. 339 * While corresponding bios are in flight, the only modification will be 340 * set_bit WAS_ERROR, which has to be atomic. 341 * If no bios are in flight yet, or all have been completed, 342 * non-atomic modification to ee->flags is ok. 343 */ 344enum { 345 __EE_CALL_AL_COMPLETE_IO, 346 __EE_MAY_SET_IN_SYNC, 347 348 /* is this a TRIM aka REQ_OP_DISCARD? */ 349 __EE_TRIM, 350 /* explicit zero-out requested, or 351 * our lower level cannot handle trim, 352 * and we want to fall back to zeroout instead */ 353 __EE_ZEROOUT, 354 355 /* In case a barrier failed, 356 * we need to resubmit without the barrier flag. */ 357 __EE_RESUBMITTED, 358 359 /* we may have several bios per peer request. 360 * if any of those fail, we set this flag atomically 361 * from the endio callback */ 362 __EE_WAS_ERROR, 363 364 /* This ee has a pointer to a digest instead of a block id */ 365 __EE_HAS_DIGEST, 366 367 /* Conflicting local requests need to be restarted after this request */ 368 __EE_RESTART_REQUESTS, 369 370 /* The peer wants a write ACK for this (wire proto C) */ 371 __EE_SEND_WRITE_ACK, 372 373 /* Is set when net_conf had two_primaries set while creating this peer_req */ 374 __EE_IN_INTERVAL_TREE, 375 376 /* for debugfs: */ 377 /* has this been submitted, or does it still wait for something else? */ 378 __EE_SUBMITTED, 379 380 /* this is/was a write request */ 381 __EE_WRITE, 382 383 /* hand back using mempool_free(e, drbd_buffer_page_pool) */ 384 __EE_RELEASE_TO_MEMPOOL, 385 386 /* this is/was a write same request */ 387 __EE_WRITE_SAME, 388 389 /* this originates from application on peer 390 * (not some resync or verify or other DRBD internal request) */ 391 __EE_APPLICATION, 392 393 /* If it contains only 0 bytes, send back P_RS_DEALLOCATED */ 394 __EE_RS_THIN_REQ, 395}; 396#define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO) 397#define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC) 398#define EE_TRIM (1<<__EE_TRIM) 399#define EE_ZEROOUT (1<<__EE_ZEROOUT) 400#define EE_RESUBMITTED (1<<__EE_RESUBMITTED) 401#define EE_WAS_ERROR (1<<__EE_WAS_ERROR) 402#define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST) 403#define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS) 404#define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK) 405#define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE) 406#define EE_SUBMITTED (1<<__EE_SUBMITTED) 407#define EE_WRITE (1<<__EE_WRITE) 408#define EE_RELEASE_TO_MEMPOOL (1<<__EE_RELEASE_TO_MEMPOOL) 409#define EE_WRITE_SAME (1<<__EE_WRITE_SAME) 410#define EE_APPLICATION (1<<__EE_APPLICATION) 411#define EE_RS_THIN_REQ (1<<__EE_RS_THIN_REQ) 412 413/* flag bits per device */ 414enum { 415 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */ 416 MD_DIRTY, /* current uuids and flags not yet on disk */ 417 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */ 418 CL_ST_CHG_SUCCESS, 419 CL_ST_CHG_FAIL, 420 CRASHED_PRIMARY, /* This node was a crashed primary. 421 * Gets cleared when the state.conn 422 * goes into C_CONNECTED state. */ 423 CONSIDER_RESYNC, 424 425 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */ 426 427 BITMAP_IO, /* suspend application io; 428 once no more io in flight, start bitmap io */ 429 BITMAP_IO_QUEUED, /* Started bitmap IO */ 430 WAS_IO_ERROR, /* Local disk failed, returned IO error */ 431 WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */ 432 FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */ 433 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */ 434 RESIZE_PENDING, /* Size change detected locally, waiting for the response from 435 * the peer, if it changed there as well. */ 436 NEW_CUR_UUID, /* Create new current UUID when thawing IO */ 437 AL_SUSPENDED, /* Activity logging is currently suspended. */ 438 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */ 439 B_RS_H_DONE, /* Before resync handler done (already executed) */ 440 DISCARD_MY_DATA, /* discard_my_data flag per volume */ 441 READ_BALANCE_RR, 442 443 FLUSH_PENDING, /* if set, device->flush_jif is when we submitted that flush 444 * from drbd_flush_after_epoch() */ 445 446 /* cleared only after backing device related structures have been destroyed. */ 447 GOING_DISKLESS, /* Disk is being detached, because of io-error, or admin request. */ 448 449 /* to be used in drbd_device_post_work() */ 450 GO_DISKLESS, /* tell worker to schedule cleanup before detach */ 451 DESTROY_DISK, /* tell worker to close backing devices and destroy related structures. */ 452 MD_SYNC, /* tell worker to call drbd_md_sync() */ 453 RS_START, /* tell worker to start resync/OV */ 454 RS_PROGRESS, /* tell worker that resync made significant progress */ 455 RS_DONE, /* tell worker that resync is done */ 456}; 457 458struct drbd_bitmap; /* opaque for drbd_device */ 459 460/* definition of bits in bm_flags to be used in drbd_bm_lock 461 * and drbd_bitmap_io and friends. */ 462enum bm_flag { 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 488struct 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 494struct 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 503struct 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 525struct drbd_backing_dev { 526 struct block_device *backing_bdev; 527 struct file *backing_bdev_file; 528 struct block_device *md_bdev; 529 struct file *f_md_bdev; 530 struct drbd_md md; 531 struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */ 532 sector_t known_size; /* last known size of that backing device */ 533}; 534 535struct drbd_md_io { 536 struct page *page; 537 unsigned long start_jif; /* last call to drbd_md_get_buffer */ 538 unsigned long submit_jif; /* last _drbd_md_sync_page_io() submit */ 539 const char *current_use; 540 atomic_t in_use; 541 unsigned int done; 542 int error; 543}; 544 545struct bm_io_work { 546 struct drbd_work w; 547 struct drbd_peer_device *peer_device; 548 char *why; 549 enum bm_flag flags; 550 int (*io_fn)(struct drbd_device *device, struct drbd_peer_device *peer_device); 551 void (*done)(struct drbd_device *device, int rv); 552}; 553 554struct fifo_buffer { 555 unsigned int head_index; 556 unsigned int size; 557 int total; /* sum of all values */ 558 int values[] __counted_by(size); 559}; 560extern struct fifo_buffer *fifo_alloc(unsigned int fifo_size); 561 562/* flag bits per connection */ 563enum { 564 NET_CONGESTED, /* The data socket is congested */ 565 RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */ 566 SEND_PING, 567 GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */ 568 CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */ 569 CONN_WD_ST_CHG_OKAY, 570 CONN_WD_ST_CHG_FAIL, 571 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */ 572 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */ 573 STATE_SENT, /* Do not change state/UUIDs while this is set */ 574 CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC) 575 * pending, from drbd worker context. 576 */ 577 DISCONNECT_SENT, 578 579 DEVICE_WORK_PENDING, /* tell worker that some device has pending work */ 580}; 581 582enum which_state { NOW, OLD = NOW, NEW }; 583 584struct drbd_resource { 585 char *name; 586#ifdef CONFIG_DEBUG_FS 587 struct dentry *debugfs_res; 588 struct dentry *debugfs_res_volumes; 589 struct dentry *debugfs_res_connections; 590 struct dentry *debugfs_res_in_flight_summary; 591#endif 592 struct kref kref; 593 struct idr devices; /* volume number to device mapping */ 594 struct list_head connections; 595 struct list_head resources; 596 struct res_opts res_opts; 597 struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */ 598 struct mutex adm_mutex; /* mutex to serialize administrative requests */ 599 spinlock_t req_lock; 600 601 unsigned susp:1; /* IO suspended by user */ 602 unsigned susp_nod:1; /* IO suspended because no data */ 603 unsigned susp_fen:1; /* IO suspended because fence peer handler runs */ 604 605 enum write_ordering_e write_ordering; 606 607 cpumask_var_t cpu_mask; 608}; 609 610struct drbd_thread_timing_details 611{ 612 unsigned long start_jif; 613 void *cb_addr; 614 const char *caller_fn; 615 unsigned int line; 616 unsigned int cb_nr; 617}; 618 619struct drbd_connection { 620 struct list_head connections; 621 struct drbd_resource *resource; 622#ifdef CONFIG_DEBUG_FS 623 struct dentry *debugfs_conn; 624 struct dentry *debugfs_conn_callback_history; 625 struct dentry *debugfs_conn_oldest_requests; 626#endif 627 struct kref kref; 628 struct idr peer_devices; /* volume number to peer device mapping */ 629 enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */ 630 struct mutex cstate_mutex; /* Protects graceful disconnects */ 631 unsigned int connect_cnt; /* Inc each time a connection is established */ 632 633 unsigned long flags; 634 struct net_conf *net_conf; /* content protected by rcu */ 635 wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */ 636 637 struct sockaddr_storage my_addr; 638 int my_addr_len; 639 struct sockaddr_storage peer_addr; 640 int peer_addr_len; 641 642 struct drbd_socket data; /* data/barrier/cstate/parameter packets */ 643 struct drbd_socket meta; /* ping/ack (metadata) packets */ 644 int agreed_pro_version; /* actually used protocol version */ 645 u32 agreed_features; 646 unsigned long last_received; /* in jiffies, either socket */ 647 unsigned int ko_count; 648 649 struct list_head transfer_log; /* all requests not yet fully processed */ 650 651 struct crypto_shash *cram_hmac_tfm; 652 struct crypto_shash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */ 653 struct crypto_shash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */ 654 struct crypto_shash *csums_tfm; 655 struct crypto_shash *verify_tfm; 656 void *int_dig_in; 657 void *int_dig_vv; 658 659 /* receiver side */ 660 struct drbd_epoch *current_epoch; 661 spinlock_t epoch_lock; 662 unsigned int epochs; 663 atomic_t current_tle_nr; /* transfer log epoch number */ 664 unsigned current_tle_writes; /* writes seen within this tl epoch */ 665 666 unsigned long last_reconnect_jif; 667 /* empty member on older kernels without blk_start_plug() */ 668 struct blk_plug receiver_plug; 669 struct drbd_thread receiver; 670 struct drbd_thread worker; 671 struct drbd_thread ack_receiver; 672 struct workqueue_struct *ack_sender; 673 674 /* cached pointers, 675 * so we can look up the oldest pending requests more quickly. 676 * protected by resource->req_lock */ 677 struct drbd_request *req_next; /* DRBD 9: todo.req_next */ 678 struct drbd_request *req_ack_pending; 679 struct drbd_request *req_not_net_done; 680 681 /* sender side */ 682 struct drbd_work_queue sender_work; 683 684#define DRBD_THREAD_DETAILS_HIST 16 685 unsigned int w_cb_nr; /* keeps counting up */ 686 unsigned int r_cb_nr; /* keeps counting up */ 687 struct drbd_thread_timing_details w_timing_details[DRBD_THREAD_DETAILS_HIST]; 688 struct drbd_thread_timing_details r_timing_details[DRBD_THREAD_DETAILS_HIST]; 689 690 struct { 691 unsigned long last_sent_barrier_jif; 692 693 /* whether this sender thread 694 * has processed a single write yet. */ 695 bool seen_any_write_yet; 696 697 /* Which barrier number to send with the next P_BARRIER */ 698 int current_epoch_nr; 699 700 /* how many write requests have been sent 701 * with req->epoch == current_epoch_nr. 702 * If none, no P_BARRIER will be sent. */ 703 unsigned current_epoch_writes; 704 } send; 705}; 706 707static inline bool has_net_conf(struct drbd_connection *connection) 708{ 709 bool has_net_conf; 710 711 rcu_read_lock(); 712 has_net_conf = rcu_dereference(connection->net_conf); 713 rcu_read_unlock(); 714 715 return has_net_conf; 716} 717 718void __update_timing_details( 719 struct drbd_thread_timing_details *tdp, 720 unsigned int *cb_nr, 721 void *cb, 722 const char *fn, const unsigned int line); 723 724#define update_worker_timing_details(c, cb) \ 725 __update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ ) 726#define update_receiver_timing_details(c, cb) \ 727 __update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ ) 728 729struct submit_worker { 730 struct workqueue_struct *wq; 731 struct work_struct worker; 732 733 /* protected by ..->resource->req_lock */ 734 struct list_head writes; 735}; 736 737struct drbd_peer_device { 738 struct list_head peer_devices; 739 struct drbd_device *device; 740 struct drbd_connection *connection; 741 struct work_struct send_acks_work; 742#ifdef CONFIG_DEBUG_FS 743 struct dentry *debugfs_peer_dev; 744#endif 745}; 746 747struct drbd_device { 748 struct drbd_resource *resource; 749 struct list_head peer_devices; 750 struct list_head pending_bitmap_io; 751 752 unsigned long flush_jif; 753#ifdef CONFIG_DEBUG_FS 754 struct dentry *debugfs_minor; 755 struct dentry *debugfs_vol; 756 struct dentry *debugfs_vol_oldest_requests; 757 struct dentry *debugfs_vol_act_log_extents; 758 struct dentry *debugfs_vol_resync_extents; 759 struct dentry *debugfs_vol_data_gen_id; 760 struct dentry *debugfs_vol_ed_gen_id; 761#endif 762 763 unsigned int vnr; /* volume number within the connection */ 764 unsigned int minor; /* device minor number */ 765 766 struct kref kref; 767 768 /* things that are stored as / read from meta data on disk */ 769 unsigned long flags; 770 771 /* configured by drbdsetup */ 772 struct drbd_backing_dev *ldev; 773 774 sector_t p_size; /* partner's disk size */ 775 struct request_queue *rq_queue; 776 struct gendisk *vdisk; 777 778 unsigned long last_reattach_jif; 779 struct drbd_work resync_work; 780 struct drbd_work unplug_work; 781 struct timer_list resync_timer; 782 struct timer_list md_sync_timer; 783 struct timer_list start_resync_timer; 784 struct timer_list request_timer; 785 786 /* Used after attach while negotiating new disk state. */ 787 union drbd_state new_state_tmp; 788 789 union drbd_dev_state state; 790 wait_queue_head_t misc_wait; 791 wait_queue_head_t state_wait; /* upon each state change. */ 792 unsigned int send_cnt; 793 unsigned int recv_cnt; 794 unsigned int read_cnt; 795 unsigned int writ_cnt; 796 unsigned int al_writ_cnt; 797 unsigned int bm_writ_cnt; 798 atomic_t ap_bio_cnt; /* Requests we need to complete */ 799 atomic_t ap_actlog_cnt; /* Requests waiting for activity log */ 800 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */ 801 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */ 802 atomic_t unacked_cnt; /* Need to send replies for */ 803 atomic_t local_cnt; /* Waiting for local completion */ 804 atomic_t suspend_cnt; 805 806 /* Interval tree of pending local requests */ 807 struct rb_root read_requests; 808 struct rb_root write_requests; 809 810 /* for statistics and timeouts */ 811 /* [0] read, [1] write */ 812 struct list_head pending_master_completion[2]; 813 struct list_head pending_completion[2]; 814 815 /* use checksums for *this* resync */ 816 bool use_csums; 817 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */ 818 unsigned long rs_total; 819 /* number of resync blocks that failed in this run */ 820 unsigned long rs_failed; 821 /* Syncer's start time [unit jiffies] */ 822 unsigned long rs_start; 823 /* cumulated time in PausedSyncX state [unit jiffies] */ 824 unsigned long rs_paused; 825 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */ 826 unsigned long rs_same_csum; 827#define DRBD_SYNC_MARKS 8 828#define DRBD_SYNC_MARK_STEP (3*HZ) 829 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */ 830 unsigned long rs_mark_left[DRBD_SYNC_MARKS]; 831 /* marks's time [unit jiffies] */ 832 unsigned long rs_mark_time[DRBD_SYNC_MARKS]; 833 /* current index into rs_mark_{left,time} */ 834 int rs_last_mark; 835 unsigned long rs_last_bcast; /* [unit jiffies] */ 836 837 /* where does the admin want us to start? (sector) */ 838 sector_t ov_start_sector; 839 sector_t ov_stop_sector; 840 /* where are we now? (sector) */ 841 sector_t ov_position; 842 /* Start sector of out of sync range (to merge printk reporting). */ 843 sector_t ov_last_oos_start; 844 /* size of out-of-sync range in sectors. */ 845 sector_t ov_last_oos_size; 846 unsigned long ov_left; /* in bits */ 847 848 struct drbd_bitmap *bitmap; 849 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */ 850 851 /* Used to track operations of resync... */ 852 struct lru_cache *resync; 853 /* Number of locked elements in resync LRU */ 854 unsigned int resync_locked; 855 /* resync extent number waiting for application requests */ 856 unsigned int resync_wenr; 857 858 int open_cnt; 859 u64 *p_uuid; 860 861 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */ 862 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */ 863 struct list_head done_ee; /* need to send P_WRITE_ACK */ 864 struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */ 865 866 struct list_head resync_reads; 867 atomic_t pp_in_use; /* allocated from page pool */ 868 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */ 869 wait_queue_head_t ee_wait; 870 struct drbd_md_io md_io; 871 spinlock_t al_lock; 872 wait_queue_head_t al_wait; 873 struct lru_cache *act_log; /* activity log */ 874 unsigned int al_tr_number; 875 int al_tr_cycle; 876 wait_queue_head_t seq_wait; 877 atomic_t packet_seq; 878 unsigned int peer_seq; 879 spinlock_t peer_seq_lock; 880 unsigned long comm_bm_set; /* communicated number of set bits. */ 881 struct bm_io_work bm_io_work; 882 u64 ed_uuid; /* UUID of the exposed data */ 883 struct mutex own_state_mutex; 884 struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */ 885 char congestion_reason; /* Why we where congested... */ 886 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */ 887 atomic_t rs_sect_ev; /* for submitted resync data rate, both */ 888 int rs_last_sect_ev; /* counter to compare with */ 889 int rs_last_events; /* counter of read or write "events" (unit sectors) 890 * on the lower level device when we last looked. */ 891 int c_sync_rate; /* current resync rate after syncer throttle magic */ 892 struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */ 893 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */ 894 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */ 895 unsigned int peer_max_bio_size; 896 unsigned int local_max_bio_size; 897 898 /* any requests that would block in drbd_make_request() 899 * are deferred to this single-threaded work queue */ 900 struct submit_worker submit; 901}; 902 903struct drbd_bm_aio_ctx { 904 struct drbd_device *device; 905 struct list_head list; /* on device->pending_bitmap_io */; 906 unsigned long start_jif; 907 atomic_t in_flight; 908 unsigned int done; 909 unsigned flags; 910#define BM_AIO_COPY_PAGES 1 911#define BM_AIO_WRITE_HINTED 2 912#define BM_AIO_WRITE_ALL_PAGES 4 913#define BM_AIO_READ 8 914 int error; 915 struct kref kref; 916}; 917 918struct drbd_config_context { 919 /* assigned from drbd_genlmsghdr */ 920 unsigned int minor; 921 /* assigned from request attributes, if present */ 922 unsigned int volume; 923#define VOLUME_UNSPECIFIED (-1U) 924 /* pointer into the request skb, 925 * limited lifetime! */ 926 char *resource_name; 927 struct nlattr *my_addr; 928 struct nlattr *peer_addr; 929 930 /* reply buffer */ 931 struct sk_buff *reply_skb; 932 /* pointer into reply buffer */ 933 struct drbd_genlmsghdr *reply_dh; 934 /* resolved from attributes, if possible */ 935 struct drbd_device *device; 936 struct drbd_resource *resource; 937 struct drbd_connection *connection; 938}; 939 940static inline struct drbd_device *minor_to_device(unsigned int minor) 941{ 942 return (struct drbd_device *)idr_find(&drbd_devices, minor); 943} 944 945static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device) 946{ 947 return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices); 948} 949 950static inline struct drbd_peer_device * 951conn_peer_device(struct drbd_connection *connection, int volume_number) 952{ 953 return idr_find(&connection->peer_devices, volume_number); 954} 955 956#define for_each_resource(resource, _resources) \ 957 list_for_each_entry(resource, _resources, resources) 958 959#define for_each_resource_rcu(resource, _resources) \ 960 list_for_each_entry_rcu(resource, _resources, resources) 961 962#define for_each_resource_safe(resource, tmp, _resources) \ 963 list_for_each_entry_safe(resource, tmp, _resources, resources) 964 965#define for_each_connection(connection, resource) \ 966 list_for_each_entry(connection, &resource->connections, connections) 967 968#define for_each_connection_rcu(connection, resource) \ 969 list_for_each_entry_rcu(connection, &resource->connections, connections) 970 971#define for_each_connection_safe(connection, tmp, resource) \ 972 list_for_each_entry_safe(connection, tmp, &resource->connections, connections) 973 974#define for_each_peer_device(peer_device, device) \ 975 list_for_each_entry(peer_device, &device->peer_devices, peer_devices) 976 977#define for_each_peer_device_rcu(peer_device, device) \ 978 list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices) 979 980#define for_each_peer_device_safe(peer_device, tmp, device) \ 981 list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices) 982 983static inline unsigned int device_to_minor(struct drbd_device *device) 984{ 985 return device->minor; 986} 987 988/* 989 * function declarations 990 *************************/ 991 992/* drbd_main.c */ 993 994enum dds_flags { 995 DDSF_FORCED = 1, 996 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */ 997}; 998 999extern void drbd_init_set_defaults(struct drbd_device *device); 1000extern int drbd_thread_start(struct drbd_thread *thi); 1001extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait); 1002#ifdef CONFIG_SMP 1003extern void drbd_thread_current_set_cpu(struct drbd_thread *thi); 1004#else 1005#define drbd_thread_current_set_cpu(A) ({}) 1006#endif 1007extern void tl_release(struct drbd_connection *, unsigned int barrier_nr, 1008 unsigned int set_size); 1009extern void tl_clear(struct drbd_connection *); 1010extern void drbd_free_sock(struct drbd_connection *connection); 1011extern int drbd_send(struct drbd_connection *connection, struct socket *sock, 1012 void *buf, size_t size, unsigned msg_flags); 1013extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t, 1014 unsigned); 1015 1016extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd); 1017extern int drbd_send_protocol(struct drbd_connection *connection); 1018extern int drbd_send_uuids(struct drbd_peer_device *); 1019extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *); 1020extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *); 1021extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags); 1022extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s); 1023extern int drbd_send_current_state(struct drbd_peer_device *); 1024extern int drbd_send_sync_param(struct drbd_peer_device *); 1025extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr, 1026 u32 set_size); 1027extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet, 1028 struct drbd_peer_request *); 1029extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet, 1030 struct p_block_req *rp); 1031extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet, 1032 struct p_data *dp, int data_size); 1033extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet, 1034 sector_t sector, int blksize, u64 block_id); 1035extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *); 1036extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet, 1037 struct drbd_peer_request *); 1038extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req); 1039extern int drbd_send_drequest(struct drbd_peer_device *, int cmd, 1040 sector_t sector, int size, u64 block_id); 1041extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector, 1042 int size, void *digest, int digest_size, 1043 enum drbd_packet cmd); 1044extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size); 1045 1046extern int drbd_send_bitmap(struct drbd_device *device, struct drbd_peer_device *peer_device); 1047extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode); 1048extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode); 1049extern int drbd_send_rs_deallocated(struct drbd_peer_device *, struct drbd_peer_request *); 1050extern void drbd_backing_dev_free(struct drbd_device *device, struct drbd_backing_dev *ldev); 1051extern void drbd_device_cleanup(struct drbd_device *device); 1052extern void drbd_print_uuids(struct drbd_device *device, const char *text); 1053extern void drbd_queue_unplug(struct drbd_device *device); 1054 1055extern void conn_md_sync(struct drbd_connection *connection); 1056extern void drbd_md_write(struct drbd_device *device, void *buffer); 1057extern void drbd_md_sync(struct drbd_device *device); 1058extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev); 1059extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); 1060extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); 1061extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local); 1062extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local); 1063extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local); 1064extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); 1065extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local); 1066extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local); 1067extern int drbd_md_test_flag(struct drbd_backing_dev *, int); 1068extern void drbd_md_mark_dirty(struct drbd_device *device); 1069extern void drbd_queue_bitmap_io(struct drbd_device *device, 1070 int (*io_fn)(struct drbd_device *, struct drbd_peer_device *), 1071 void (*done)(struct drbd_device *, int), 1072 char *why, enum bm_flag flags, 1073 struct drbd_peer_device *peer_device); 1074extern int drbd_bitmap_io(struct drbd_device *device, 1075 int (*io_fn)(struct drbd_device *, struct drbd_peer_device *), 1076 char *why, enum bm_flag flags, 1077 struct drbd_peer_device *peer_device); 1078extern int drbd_bitmap_io_from_worker(struct drbd_device *device, 1079 int (*io_fn)(struct drbd_device *, struct drbd_peer_device *), 1080 char *why, enum bm_flag flags, 1081 struct drbd_peer_device *peer_device); 1082extern int drbd_bmio_set_n_write(struct drbd_device *device, 1083 struct drbd_peer_device *peer_device) __must_hold(local); 1084extern int drbd_bmio_clear_n_write(struct drbd_device *device, 1085 struct drbd_peer_device *peer_device) __must_hold(local); 1086 1087/* Meta data layout 1088 * 1089 * We currently have two possible layouts. 1090 * Offsets in (512 byte) sectors. 1091 * external: 1092 * |----------- md_size_sect ------------------| 1093 * [ 4k superblock ][ activity log ][ Bitmap ] 1094 * | al_offset == 8 | 1095 * | bm_offset = al_offset + X | 1096 * ==> bitmap sectors = md_size_sect - bm_offset 1097 * 1098 * Variants: 1099 * old, indexed fixed size meta data: 1100 * 1101 * internal: 1102 * |----------- md_size_sect ------------------| 1103 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*] 1104 * | al_offset < 0 | 1105 * | bm_offset = al_offset - Y | 1106 * ==> bitmap sectors = Y = al_offset - bm_offset 1107 * 1108 * [padding*] are zero or up to 7 unused 512 Byte sectors to the 1109 * end of the device, so that the [4k superblock] will be 4k aligned. 1110 * 1111 * The activity log consists of 4k transaction blocks, 1112 * which are written in a ring-buffer, or striped ring-buffer like fashion, 1113 * which are writtensize used to be fixed 32kB, 1114 * but is about to become configurable. 1115 */ 1116 1117/* Our old fixed size meta data layout 1118 * allows up to about 3.8TB, so if you want more, 1119 * you need to use the "flexible" meta data format. */ 1120#define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */ 1121#define MD_4kB_SECT 8 1122#define MD_32kB_SECT 64 1123 1124/* One activity log extent represents 4M of storage */ 1125#define AL_EXTENT_SHIFT 22 1126#define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT) 1127 1128/* We could make these currently hardcoded constants configurable 1129 * variables at create-md time (or even re-configurable at runtime?). 1130 * Which will require some more changes to the DRBD "super block" 1131 * and attach code. 1132 * 1133 * updates per transaction: 1134 * This many changes to the active set can be logged with one transaction. 1135 * This number is arbitrary. 1136 * context per transaction: 1137 * This many context extent numbers are logged with each transaction. 1138 * This number is resulting from the transaction block size (4k), the layout 1139 * of the transaction header, and the number of updates per transaction. 1140 * See drbd_actlog.c:struct al_transaction_on_disk 1141 * */ 1142#define AL_UPDATES_PER_TRANSACTION 64 // arbitrary 1143#define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4 1144 1145#if BITS_PER_LONG == 32 1146#define LN2_BPL 5 1147#define cpu_to_lel(A) cpu_to_le32(A) 1148#define lel_to_cpu(A) le32_to_cpu(A) 1149#elif BITS_PER_LONG == 64 1150#define LN2_BPL 6 1151#define cpu_to_lel(A) cpu_to_le64(A) 1152#define lel_to_cpu(A) le64_to_cpu(A) 1153#else 1154#error "LN2 of BITS_PER_LONG unknown!" 1155#endif 1156 1157/* resync bitmap */ 1158/* 16MB sized 'bitmap extent' to track syncer usage */ 1159struct bm_extent { 1160 int rs_left; /* number of bits set (out of sync) in this extent. */ 1161 int rs_failed; /* number of failed resync requests in this extent. */ 1162 unsigned long flags; 1163 struct lc_element lce; 1164}; 1165 1166#define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */ 1167#define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */ 1168#define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */ 1169 1170/* drbd_bitmap.c */ 1171/* 1172 * We need to store one bit for a block. 1173 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap. 1174 * Bit 0 ==> local node thinks this block is binary identical on both nodes 1175 * Bit 1 ==> local node thinks this block needs to be synced. 1176 */ 1177 1178#define SLEEP_TIME (HZ/10) 1179 1180/* We do bitmap IO in units of 4k blocks. 1181 * We also still have a hardcoded 4k per bit relation. */ 1182#define BM_BLOCK_SHIFT 12 /* 4k per bit */ 1183#define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT) 1184/* mostly arbitrarily set the represented size of one bitmap extent, 1185 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap 1186 * at 4k per bit resolution) */ 1187#define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */ 1188#define BM_EXT_SIZE (1<<BM_EXT_SHIFT) 1189 1190#if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12) 1191#error "HAVE YOU FIXED drbdmeta AS WELL??" 1192#endif 1193 1194/* thus many _storage_ sectors are described by one bit */ 1195#define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9)) 1196#define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9)) 1197#define BM_SECT_PER_BIT BM_BIT_TO_SECT(1) 1198 1199/* bit to represented kilo byte conversion */ 1200#define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10)) 1201 1202/* in which _bitmap_ extent (resp. sector) the bit for a certain 1203 * _storage_ sector is located in */ 1204#define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9)) 1205#define BM_BIT_TO_EXT(x) ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT)) 1206 1207/* first storage sector a bitmap extent corresponds to */ 1208#define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9)) 1209/* how much _storage_ sectors we have per bitmap extent */ 1210#define BM_SECT_PER_EXT BM_EXT_TO_SECT(1) 1211/* how many bits are covered by one bitmap extent (resync extent) */ 1212#define BM_BITS_PER_EXT (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT)) 1213 1214#define BM_BLOCKS_PER_BM_EXT_MASK (BM_BITS_PER_EXT - 1) 1215 1216 1217/* in one sector of the bitmap, we have this many activity_log extents. */ 1218#define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT)) 1219 1220/* the extent in "PER_EXTENT" below is an activity log extent 1221 * we need that many (long words/bytes) to store the bitmap 1222 * of one AL_EXTENT_SIZE chunk of storage. 1223 * we can store the bitmap for that many AL_EXTENTS within 1224 * one sector of the _on_disk_ bitmap: 1225 * bit 0 bit 37 bit 38 bit (512*8)-1 1226 * ...|........|........|.. // ..|........| 1227 * sect. 0 `296 `304 ^(512*8*8)-1 1228 * 1229#define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG ) 1230#define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128 1231#define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4 1232 */ 1233 1234#define DRBD_MAX_SECTORS_32 (0xffffffffLU) 1235/* we have a certain meta data variant that has a fixed on-disk size of 128 1236 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity 1237 * log, leaving this many sectors for the bitmap. 1238 */ 1239 1240#define DRBD_MAX_SECTORS_FIXED_BM \ 1241 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9))) 1242#define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM 1243/* 16 TB in units of sectors */ 1244#if BITS_PER_LONG == 32 1245/* adjust by one page worth of bitmap, 1246 * so we won't wrap around in drbd_bm_find_next_bit. 1247 * you should use 64bit OS for that much storage, anyways. */ 1248#define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff) 1249#else 1250/* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */ 1251#define DRBD_MAX_SECTORS_FLEX (1UL << 51) 1252/* corresponds to (1UL << 38) bits right now. */ 1253#endif 1254 1255/* Estimate max bio size as 256 * PAGE_SIZE, 1256 * so for typical PAGE_SIZE of 4k, that is (1<<20) Byte. 1257 * Since we may live in a mixed-platform cluster, 1258 * we limit us to a platform agnostic constant here for now. 1259 * A followup commit may allow even bigger BIO sizes, 1260 * once we thought that through. */ 1261#define DRBD_MAX_BIO_SIZE (1U << 20) 1262#if DRBD_MAX_BIO_SIZE > (BIO_MAX_VECS << PAGE_SHIFT) 1263#error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE 1264#endif 1265#define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */ 1266 1267#define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */ 1268#define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */ 1269 1270/* For now, don't allow more than half of what we can "activate" in one 1271 * activity log transaction to be discarded in one go. We may need to rework 1272 * drbd_al_begin_io() to allow for even larger discard ranges */ 1273#define DRBD_MAX_BATCH_BIO_SIZE (AL_UPDATES_PER_TRANSACTION/2*AL_EXTENT_SIZE) 1274#define DRBD_MAX_BBIO_SECTORS (DRBD_MAX_BATCH_BIO_SIZE >> 9) 1275 1276extern int drbd_bm_init(struct drbd_device *device); 1277extern int drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits); 1278extern void drbd_bm_cleanup(struct drbd_device *device); 1279extern void drbd_bm_set_all(struct drbd_device *device); 1280extern void drbd_bm_clear_all(struct drbd_device *device); 1281/* set/clear/test only a few bits at a time */ 1282extern int drbd_bm_set_bits( 1283 struct drbd_device *device, unsigned long s, unsigned long e); 1284extern int drbd_bm_clear_bits( 1285 struct drbd_device *device, unsigned long s, unsigned long e); 1286extern int drbd_bm_count_bits( 1287 struct drbd_device *device, const unsigned long s, const unsigned long e); 1288/* bm_set_bits variant for use while holding drbd_bm_lock, 1289 * may process the whole bitmap in one go */ 1290extern void _drbd_bm_set_bits(struct drbd_device *device, 1291 const unsigned long s, const unsigned long e); 1292extern int drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr); 1293extern int drbd_bm_e_weight(struct drbd_device *device, unsigned long enr); 1294extern int drbd_bm_read(struct drbd_device *device, 1295 struct drbd_peer_device *peer_device) __must_hold(local); 1296extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr); 1297extern int drbd_bm_write(struct drbd_device *device, 1298 struct drbd_peer_device *peer_device) __must_hold(local); 1299extern void drbd_bm_reset_al_hints(struct drbd_device *device) __must_hold(local); 1300extern int drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local); 1301extern int drbd_bm_write_lazy(struct drbd_device *device, unsigned upper_idx) __must_hold(local); 1302extern int drbd_bm_write_all(struct drbd_device *device, 1303 struct drbd_peer_device *peer_device) __must_hold(local); 1304extern int drbd_bm_write_copy_pages(struct drbd_device *device, 1305 struct drbd_peer_device *peer_device) __must_hold(local); 1306extern size_t drbd_bm_words(struct drbd_device *device); 1307extern unsigned long drbd_bm_bits(struct drbd_device *device); 1308extern sector_t drbd_bm_capacity(struct drbd_device *device); 1309 1310#define DRBD_END_OF_BITMAP (~(unsigned long)0) 1311extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); 1312/* bm_find_next variants for use while you hold drbd_bm_lock() */ 1313extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); 1314extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo); 1315extern unsigned long _drbd_bm_total_weight(struct drbd_device *device); 1316extern unsigned long drbd_bm_total_weight(struct drbd_device *device); 1317/* for receive_bitmap */ 1318extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset, 1319 size_t number, unsigned long *buffer); 1320/* for _drbd_send_bitmap */ 1321extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset, 1322 size_t number, unsigned long *buffer); 1323 1324extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags); 1325extern void drbd_bm_unlock(struct drbd_device *device); 1326/* drbd_main.c */ 1327 1328extern struct kmem_cache *drbd_request_cache; 1329extern struct kmem_cache *drbd_ee_cache; /* peer requests */ 1330extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */ 1331extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */ 1332extern mempool_t drbd_request_mempool; 1333extern mempool_t drbd_ee_mempool; 1334 1335/* We also need a standard (emergency-reserve backed) page pool 1336 * for meta data IO (activity log, bitmap). 1337 * We can keep it global, as long as it is used as "N pages at a time". 1338 * 128 should be plenty, currently we probably can get away with as few as 1. 1339 */ 1340#define DRBD_MIN_POOL_PAGES 128 1341extern mempool_t drbd_md_io_page_pool; 1342extern mempool_t drbd_buffer_page_pool; 1343 1344/* We also need to make sure we get a bio 1345 * when we need it for housekeeping purposes */ 1346extern struct bio_set drbd_md_io_bio_set; 1347 1348/* And a bio_set for cloning */ 1349extern struct bio_set drbd_io_bio_set; 1350 1351extern struct mutex resources_mutex; 1352 1353extern enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor); 1354extern void drbd_destroy_device(struct kref *kref); 1355extern void drbd_delete_device(struct drbd_device *device); 1356 1357extern struct drbd_resource *drbd_create_resource(const char *name); 1358extern void drbd_free_resource(struct drbd_resource *resource); 1359 1360extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts); 1361extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts); 1362extern void drbd_destroy_connection(struct kref *kref); 1363extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len, 1364 void *peer_addr, int peer_addr_len); 1365extern struct drbd_resource *drbd_find_resource(const char *name); 1366extern void drbd_destroy_resource(struct kref *kref); 1367extern void conn_free_crypto(struct drbd_connection *connection); 1368 1369/* drbd_req */ 1370extern void do_submit(struct work_struct *ws); 1371extern void __drbd_make_request(struct drbd_device *, struct bio *); 1372void drbd_submit_bio(struct bio *bio); 1373 1374/* drbd_nl.c */ 1375 1376extern struct mutex notification_mutex; 1377 1378extern void drbd_suspend_io(struct drbd_device *device); 1379extern void drbd_resume_io(struct drbd_device *device); 1380extern char *ppsize(char *buf, unsigned long long size); 1381extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int); 1382enum determine_dev_size { 1383 DS_ERROR_SHRINK = -3, 1384 DS_ERROR_SPACE_MD = -2, 1385 DS_ERROR = -1, 1386 DS_UNCHANGED = 0, 1387 DS_SHRUNK = 1, 1388 DS_GREW = 2, 1389 DS_GREW_FROM_ZERO = 3, 1390}; 1391extern enum determine_dev_size 1392drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local); 1393extern void resync_after_online_grow(struct drbd_device *); 1394extern void drbd_reconsider_queue_parameters(struct drbd_device *device, 1395 struct drbd_backing_dev *bdev, struct o_qlim *o); 1396extern enum drbd_state_rv drbd_set_role(struct drbd_device *device, 1397 enum drbd_role new_role, 1398 int force); 1399extern bool conn_try_outdate_peer(struct drbd_connection *connection); 1400extern void conn_try_outdate_peer_async(struct drbd_connection *connection); 1401extern enum drbd_peer_state conn_khelper(struct drbd_connection *connection, char *cmd); 1402extern int drbd_khelper(struct drbd_device *device, char *cmd); 1403 1404/* drbd_worker.c */ 1405/* bi_end_io handlers */ 1406extern void drbd_md_endio(struct bio *bio); 1407extern void drbd_peer_request_endio(struct bio *bio); 1408extern void drbd_request_endio(struct bio *bio); 1409extern int drbd_worker(struct drbd_thread *thi); 1410enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor); 1411void drbd_resync_after_changed(struct drbd_device *device); 1412extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side); 1413extern void resume_next_sg(struct drbd_device *device); 1414extern void suspend_other_sg(struct drbd_device *device); 1415extern int drbd_resync_finished(struct drbd_peer_device *peer_device); 1416/* maybe rather drbd_main.c ? */ 1417extern void *drbd_md_get_buffer(struct drbd_device *device, const char *intent); 1418extern void drbd_md_put_buffer(struct drbd_device *device); 1419extern int drbd_md_sync_page_io(struct drbd_device *device, 1420 struct drbd_backing_dev *bdev, sector_t sector, enum req_op op); 1421extern void drbd_ov_out_of_sync_found(struct drbd_peer_device *peer_device, 1422 sector_t sector, int size); 1423extern void wait_until_done_or_force_detached(struct drbd_device *device, 1424 struct drbd_backing_dev *bdev, unsigned int *done); 1425extern void drbd_rs_controller_reset(struct drbd_peer_device *peer_device); 1426 1427static inline void ov_out_of_sync_print(struct drbd_peer_device *peer_device) 1428{ 1429 struct drbd_device *device = peer_device->device; 1430 1431 if (device->ov_last_oos_size) { 1432 drbd_err(peer_device, "Out of sync: start=%llu, size=%lu (sectors)\n", 1433 (unsigned long long)device->ov_last_oos_start, 1434 (unsigned long)device->ov_last_oos_size); 1435 } 1436 device->ov_last_oos_size = 0; 1437} 1438 1439 1440extern void drbd_csum_bio(struct crypto_shash *, struct bio *, void *); 1441extern void drbd_csum_ee(struct crypto_shash *, struct drbd_peer_request *, 1442 void *); 1443/* worker callbacks */ 1444extern int w_e_end_data_req(struct drbd_work *, int); 1445extern int w_e_end_rsdata_req(struct drbd_work *, int); 1446extern int w_e_end_csum_rs_req(struct drbd_work *, int); 1447extern int w_e_end_ov_reply(struct drbd_work *, int); 1448extern int w_e_end_ov_req(struct drbd_work *, int); 1449extern int w_ov_finished(struct drbd_work *, int); 1450extern int w_resync_timer(struct drbd_work *, int); 1451extern int w_send_write_hint(struct drbd_work *, int); 1452extern int w_send_dblock(struct drbd_work *, int); 1453extern int w_send_read_req(struct drbd_work *, int); 1454extern int w_restart_disk_io(struct drbd_work *, int); 1455extern int w_send_out_of_sync(struct drbd_work *, int); 1456 1457extern void resync_timer_fn(struct timer_list *t); 1458extern void start_resync_timer_fn(struct timer_list *t); 1459 1460extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req); 1461 1462/* drbd_receiver.c */ 1463extern int drbd_issue_discard_or_zero_out(struct drbd_device *device, 1464 sector_t start, unsigned int nr_sectors, int flags); 1465extern int drbd_receiver(struct drbd_thread *thi); 1466extern int drbd_ack_receiver(struct drbd_thread *thi); 1467extern void drbd_send_acks_wf(struct work_struct *ws); 1468extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device); 1469extern bool drbd_rs_should_slow_down(struct drbd_peer_device *peer_device, sector_t sector, 1470 bool throttle_if_app_is_waiting); 1471extern int drbd_submit_peer_request(struct drbd_peer_request *peer_req); 1472extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *); 1473extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64, 1474 sector_t, unsigned int, 1475 unsigned int, 1476 gfp_t) __must_hold(local); 1477extern void drbd_free_peer_req(struct drbd_device *device, struct drbd_peer_request *req); 1478extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool); 1479extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed); 1480extern int drbd_connected(struct drbd_peer_device *); 1481 1482/* sets the number of 512 byte sectors of our virtual device */ 1483void drbd_set_my_capacity(struct drbd_device *device, sector_t size); 1484 1485/* 1486 * used to submit our private bio 1487 */ 1488static inline void drbd_submit_bio_noacct(struct drbd_device *device, 1489 int fault_type, struct bio *bio) 1490{ 1491 __release(local); 1492 if (!bio->bi_bdev) { 1493 drbd_err(device, "drbd_submit_bio_noacct: bio->bi_bdev == NULL\n"); 1494 bio->bi_status = BLK_STS_IOERR; 1495 bio_endio(bio); 1496 return; 1497 } 1498 1499 if (drbd_insert_fault(device, fault_type)) 1500 bio_io_error(bio); 1501 else 1502 submit_bio_noacct(bio); 1503} 1504 1505void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev, 1506 enum write_ordering_e wo); 1507 1508/* drbd_proc.c */ 1509extern struct proc_dir_entry *drbd_proc; 1510int drbd_seq_show(struct seq_file *seq, void *v); 1511 1512/* drbd_actlog.c */ 1513extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i); 1514extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i); 1515extern void drbd_al_begin_io_commit(struct drbd_device *device); 1516extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i); 1517extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i); 1518extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i); 1519extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector); 1520extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector); 1521extern int drbd_try_rs_begin_io(struct drbd_peer_device *peer_device, sector_t sector); 1522extern void drbd_rs_cancel_all(struct drbd_device *device); 1523extern int drbd_rs_del_all(struct drbd_device *device); 1524extern void drbd_rs_failed_io(struct drbd_peer_device *peer_device, 1525 sector_t sector, int size); 1526extern void drbd_advance_rs_marks(struct drbd_peer_device *peer_device, unsigned long still_to_go); 1527 1528enum update_sync_bits_mode { RECORD_RS_FAILED, SET_OUT_OF_SYNC, SET_IN_SYNC }; 1529extern int __drbd_change_sync(struct drbd_peer_device *peer_device, sector_t sector, int size, 1530 enum update_sync_bits_mode mode); 1531#define drbd_set_in_sync(peer_device, sector, size) \ 1532 __drbd_change_sync(peer_device, sector, size, SET_IN_SYNC) 1533#define drbd_set_out_of_sync(peer_device, sector, size) \ 1534 __drbd_change_sync(peer_device, sector, size, SET_OUT_OF_SYNC) 1535#define drbd_rs_failed_io(peer_device, sector, size) \ 1536 __drbd_change_sync(peer_device, sector, size, RECORD_RS_FAILED) 1537extern void drbd_al_shrink(struct drbd_device *device); 1538extern int drbd_al_initialize(struct drbd_device *, void *); 1539 1540/* drbd_nl.c */ 1541/* state info broadcast */ 1542struct sib_info { 1543 enum drbd_state_info_bcast_reason sib_reason; 1544 union { 1545 struct { 1546 char *helper_name; 1547 unsigned helper_exit_code; 1548 }; 1549 struct { 1550 union drbd_state os; 1551 union drbd_state ns; 1552 }; 1553 }; 1554}; 1555void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib); 1556 1557extern int notify_resource_state(struct sk_buff *, 1558 unsigned int, 1559 struct drbd_resource *, 1560 struct resource_info *, 1561 enum drbd_notification_type); 1562extern int notify_device_state(struct sk_buff *, 1563 unsigned int, 1564 struct drbd_device *, 1565 struct device_info *, 1566 enum drbd_notification_type); 1567extern int notify_connection_state(struct sk_buff *, 1568 unsigned int, 1569 struct drbd_connection *, 1570 struct connection_info *, 1571 enum drbd_notification_type); 1572extern int notify_peer_device_state(struct sk_buff *, 1573 unsigned int, 1574 struct drbd_peer_device *, 1575 struct peer_device_info *, 1576 enum drbd_notification_type); 1577extern void notify_helper(enum drbd_notification_type, struct drbd_device *, 1578 struct drbd_connection *, const char *, int); 1579 1580/* 1581 * inline helper functions 1582 *************************/ 1583 1584/* see also page_chain_add and friends in drbd_receiver.c */ 1585static inline struct page *page_chain_next(struct page *page) 1586{ 1587 return (struct page *)page_private(page); 1588} 1589#define page_chain_for_each(page) \ 1590 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \ 1591 page = page_chain_next(page)) 1592#define page_chain_for_each_safe(page, n) \ 1593 for (; page && ({ n = page_chain_next(page); 1; }); page = n) 1594 1595 1596static inline union drbd_state drbd_read_state(struct drbd_device *device) 1597{ 1598 struct drbd_resource *resource = device->resource; 1599 union drbd_state rv; 1600 1601 rv.i = device->state.i; 1602 rv.susp = resource->susp; 1603 rv.susp_nod = resource->susp_nod; 1604 rv.susp_fen = resource->susp_fen; 1605 1606 return rv; 1607} 1608 1609enum drbd_force_detach_flags { 1610 DRBD_READ_ERROR, 1611 DRBD_WRITE_ERROR, 1612 DRBD_META_IO_ERROR, 1613 DRBD_FORCE_DETACH, 1614}; 1615 1616#define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__) 1617static inline void __drbd_chk_io_error_(struct drbd_device *device, 1618 enum drbd_force_detach_flags df, 1619 const char *where) 1620{ 1621 enum drbd_io_error_p ep; 1622 1623 rcu_read_lock(); 1624 ep = rcu_dereference(device->ldev->disk_conf)->on_io_error; 1625 rcu_read_unlock(); 1626 switch (ep) { 1627 case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */ 1628 if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) { 1629 if (drbd_ratelimit()) 1630 drbd_err(device, "Local IO failed in %s.\n", where); 1631 if (device->state.disk > D_INCONSISTENT) 1632 _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL); 1633 break; 1634 } 1635 fallthrough; /* for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */ 1636 case EP_DETACH: 1637 case EP_CALL_HELPER: 1638 /* Remember whether we saw a READ or WRITE error. 1639 * 1640 * Recovery of the affected area for WRITE failure is covered 1641 * by the activity log. 1642 * READ errors may fall outside that area though. Certain READ 1643 * errors can be "healed" by writing good data to the affected 1644 * blocks, which triggers block re-allocation in lower layers. 1645 * 1646 * If we can not write the bitmap after a READ error, 1647 * we may need to trigger a full sync (see w_go_diskless()). 1648 * 1649 * Force-detach is not really an IO error, but rather a 1650 * desperate measure to try to deal with a completely 1651 * unresponsive lower level IO stack. 1652 * Still it should be treated as a WRITE error. 1653 * 1654 * Meta IO error is always WRITE error: 1655 * we read meta data only once during attach, 1656 * which will fail in case of errors. 1657 */ 1658 set_bit(WAS_IO_ERROR, &device->flags); 1659 if (df == DRBD_READ_ERROR) 1660 set_bit(WAS_READ_ERROR, &device->flags); 1661 if (df == DRBD_FORCE_DETACH) 1662 set_bit(FORCE_DETACH, &device->flags); 1663 if (device->state.disk > D_FAILED) { 1664 _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL); 1665 drbd_err(device, 1666 "Local IO failed in %s. Detaching...\n", where); 1667 } 1668 break; 1669 } 1670} 1671 1672/** 1673 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers 1674 * @device: DRBD device. 1675 * @error: Error code passed to the IO completion callback 1676 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data 1677 * 1678 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED) 1679 */ 1680#define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__) 1681static inline void drbd_chk_io_error_(struct drbd_device *device, 1682 int error, enum drbd_force_detach_flags forcedetach, const char *where) 1683{ 1684 if (error) { 1685 unsigned long flags; 1686 spin_lock_irqsave(&device->resource->req_lock, flags); 1687 __drbd_chk_io_error_(device, forcedetach, where); 1688 spin_unlock_irqrestore(&device->resource->req_lock, flags); 1689 } 1690} 1691 1692 1693/** 1694 * drbd_md_first_sector() - Returns the first sector number of the meta data area 1695 * @bdev: Meta data block device. 1696 * 1697 * BTW, for internal meta data, this happens to be the maximum capacity 1698 * we could agree upon with our peer node. 1699 */ 1700static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev) 1701{ 1702 switch (bdev->md.meta_dev_idx) { 1703 case DRBD_MD_INDEX_INTERNAL: 1704 case DRBD_MD_INDEX_FLEX_INT: 1705 return bdev->md.md_offset + bdev->md.bm_offset; 1706 case DRBD_MD_INDEX_FLEX_EXT: 1707 default: 1708 return bdev->md.md_offset; 1709 } 1710} 1711 1712/** 1713 * drbd_md_last_sector() - Return the last sector number of the meta data area 1714 * @bdev: Meta data block device. 1715 */ 1716static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev) 1717{ 1718 switch (bdev->md.meta_dev_idx) { 1719 case DRBD_MD_INDEX_INTERNAL: 1720 case DRBD_MD_INDEX_FLEX_INT: 1721 return bdev->md.md_offset + MD_4kB_SECT -1; 1722 case DRBD_MD_INDEX_FLEX_EXT: 1723 default: 1724 return bdev->md.md_offset + bdev->md.md_size_sect -1; 1725 } 1726} 1727 1728/* Returns the number of 512 byte sectors of the device */ 1729static inline sector_t drbd_get_capacity(struct block_device *bdev) 1730{ 1731 return bdev ? bdev_nr_sectors(bdev) : 0; 1732} 1733 1734/** 1735 * drbd_get_max_capacity() - Returns the capacity we announce to out peer 1736 * @bdev: Meta data block device. 1737 * 1738 * returns the capacity we announce to out peer. we clip ourselves at the 1739 * various MAX_SECTORS, because if we don't, current implementation will 1740 * oops sooner or later 1741 */ 1742static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev) 1743{ 1744 sector_t s; 1745 1746 switch (bdev->md.meta_dev_idx) { 1747 case DRBD_MD_INDEX_INTERNAL: 1748 case DRBD_MD_INDEX_FLEX_INT: 1749 s = drbd_get_capacity(bdev->backing_bdev) 1750 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX, 1751 drbd_md_first_sector(bdev)) 1752 : 0; 1753 break; 1754 case DRBD_MD_INDEX_FLEX_EXT: 1755 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX, 1756 drbd_get_capacity(bdev->backing_bdev)); 1757 /* clip at maximum size the meta device can support */ 1758 s = min_t(sector_t, s, 1759 BM_EXT_TO_SECT(bdev->md.md_size_sect 1760 - bdev->md.bm_offset)); 1761 break; 1762 default: 1763 s = min_t(sector_t, DRBD_MAX_SECTORS, 1764 drbd_get_capacity(bdev->backing_bdev)); 1765 } 1766 return s; 1767} 1768 1769/** 1770 * drbd_md_ss() - Return the sector number of our meta data super block 1771 * @bdev: Meta data block device. 1772 */ 1773static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev) 1774{ 1775 const int meta_dev_idx = bdev->md.meta_dev_idx; 1776 1777 if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT) 1778 return 0; 1779 1780 /* Since drbd08, internal meta data is always "flexible". 1781 * position: last 4k aligned block of 4k size */ 1782 if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL || 1783 meta_dev_idx == DRBD_MD_INDEX_FLEX_INT) 1784 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8; 1785 1786 /* external, some index; this is the old fixed size layout */ 1787 return MD_128MB_SECT * bdev->md.meta_dev_idx; 1788} 1789 1790static inline void 1791drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w) 1792{ 1793 unsigned long flags; 1794 spin_lock_irqsave(&q->q_lock, flags); 1795 list_add_tail(&w->list, &q->q); 1796 spin_unlock_irqrestore(&q->q_lock, flags); 1797 wake_up(&q->q_wait); 1798} 1799 1800static inline void 1801drbd_queue_work_if_unqueued(struct drbd_work_queue *q, struct drbd_work *w) 1802{ 1803 unsigned long flags; 1804 spin_lock_irqsave(&q->q_lock, flags); 1805 if (list_empty_careful(&w->list)) 1806 list_add_tail(&w->list, &q->q); 1807 spin_unlock_irqrestore(&q->q_lock, flags); 1808 wake_up(&q->q_wait); 1809} 1810 1811static inline void 1812drbd_device_post_work(struct drbd_device *device, int work_bit) 1813{ 1814 if (!test_and_set_bit(work_bit, &device->flags)) { 1815 struct drbd_connection *connection = 1816 first_peer_device(device)->connection; 1817 struct drbd_work_queue *q = &connection->sender_work; 1818 if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags)) 1819 wake_up(&q->q_wait); 1820 } 1821} 1822 1823extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue); 1824 1825/* To get the ack_receiver out of the blocking network stack, 1826 * so it can change its sk_rcvtimeo from idle- to ping-timeout, 1827 * and send a ping, we need to send a signal. 1828 * Which signal we send is irrelevant. */ 1829static inline void wake_ack_receiver(struct drbd_connection *connection) 1830{ 1831 struct task_struct *task = connection->ack_receiver.task; 1832 if (task && get_t_state(&connection->ack_receiver) == RUNNING) 1833 send_sig(SIGXCPU, task, 1); 1834} 1835 1836static inline void request_ping(struct drbd_connection *connection) 1837{ 1838 set_bit(SEND_PING, &connection->flags); 1839 wake_ack_receiver(connection); 1840} 1841 1842extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *); 1843extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *); 1844extern int conn_send_command(struct drbd_connection *, struct drbd_socket *, 1845 enum drbd_packet, unsigned int, void *, 1846 unsigned int); 1847extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *, 1848 enum drbd_packet, unsigned int, void *, 1849 unsigned int); 1850 1851extern int drbd_send_ping(struct drbd_connection *connection); 1852extern int drbd_send_ping_ack(struct drbd_connection *connection); 1853extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state); 1854extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state); 1855 1856static inline void drbd_thread_stop(struct drbd_thread *thi) 1857{ 1858 _drbd_thread_stop(thi, false, true); 1859} 1860 1861static inline void drbd_thread_stop_nowait(struct drbd_thread *thi) 1862{ 1863 _drbd_thread_stop(thi, false, false); 1864} 1865 1866static inline void drbd_thread_restart_nowait(struct drbd_thread *thi) 1867{ 1868 _drbd_thread_stop(thi, true, false); 1869} 1870 1871/* counts how many answer packets packets we expect from our peer, 1872 * for either explicit application requests, 1873 * or implicit barrier packets as necessary. 1874 * increased: 1875 * w_send_barrier 1876 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ); 1877 * it is much easier and equally valid to count what we queue for the 1878 * worker, even before it actually was queued or send. 1879 * (drbd_make_request_common; recovery path on read io-error) 1880 * decreased: 1881 * got_BarrierAck (respective tl_clear, tl_clear_barrier) 1882 * _req_mod(req, DATA_RECEIVED) 1883 * [from receive_DataReply] 1884 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED) 1885 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)] 1886 * for some reason it is NOT decreased in got_NegAck, 1887 * but in the resulting cleanup code from report_params. 1888 * we should try to remember the reason for that... 1889 * _req_mod(req, SEND_FAILED or SEND_CANCELED) 1890 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING) 1891 * [from tl_clear_barrier] 1892 */ 1893static inline void inc_ap_pending(struct drbd_device *device) 1894{ 1895 atomic_inc(&device->ap_pending_cnt); 1896} 1897 1898#define dec_ap_pending(device) ((void)expect((device), __dec_ap_pending(device) >= 0)) 1899static inline int __dec_ap_pending(struct drbd_device *device) 1900{ 1901 int ap_pending_cnt = atomic_dec_return(&device->ap_pending_cnt); 1902 1903 if (ap_pending_cnt == 0) 1904 wake_up(&device->misc_wait); 1905 return ap_pending_cnt; 1906} 1907 1908/* counts how many resync-related answers we still expect from the peer 1909 * increase decrease 1910 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY) 1911 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER) 1912 * (or P_NEG_ACK with ID_SYNCER) 1913 */ 1914static inline void inc_rs_pending(struct drbd_peer_device *peer_device) 1915{ 1916 atomic_inc(&peer_device->device->rs_pending_cnt); 1917} 1918 1919#define dec_rs_pending(peer_device) \ 1920 ((void)expect((peer_device), __dec_rs_pending(peer_device) >= 0)) 1921static inline int __dec_rs_pending(struct drbd_peer_device *peer_device) 1922{ 1923 return atomic_dec_return(&peer_device->device->rs_pending_cnt); 1924} 1925 1926/* counts how many answers we still need to send to the peer. 1927 * increased on 1928 * receive_Data unless protocol A; 1929 * we need to send a P_RECV_ACK (proto B) 1930 * or P_WRITE_ACK (proto C) 1931 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK 1932 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA 1933 * receive_Barrier_* we need to send a P_BARRIER_ACK 1934 */ 1935static inline void inc_unacked(struct drbd_device *device) 1936{ 1937 atomic_inc(&device->unacked_cnt); 1938} 1939 1940#define dec_unacked(device) ((void)expect(device, __dec_unacked(device) >= 0)) 1941static inline int __dec_unacked(struct drbd_device *device) 1942{ 1943 return atomic_dec_return(&device->unacked_cnt); 1944} 1945 1946#define sub_unacked(device, n) ((void)expect(device, __sub_unacked(device) >= 0)) 1947static inline int __sub_unacked(struct drbd_device *device, int n) 1948{ 1949 return atomic_sub_return(n, &device->unacked_cnt); 1950} 1951 1952static inline bool is_sync_target_state(enum drbd_conns connection_state) 1953{ 1954 return connection_state == C_SYNC_TARGET || 1955 connection_state == C_PAUSED_SYNC_T; 1956} 1957 1958static inline bool is_sync_source_state(enum drbd_conns connection_state) 1959{ 1960 return connection_state == C_SYNC_SOURCE || 1961 connection_state == C_PAUSED_SYNC_S; 1962} 1963 1964static inline bool is_sync_state(enum drbd_conns connection_state) 1965{ 1966 return is_sync_source_state(connection_state) || 1967 is_sync_target_state(connection_state); 1968} 1969 1970/** 1971 * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev 1972 * @_device: DRBD device. 1973 * @_min_state: Minimum device state required for success. 1974 * 1975 * You have to call put_ldev() when finished working with device->ldev. 1976 */ 1977#define get_ldev_if_state(_device, _min_state) \ 1978 (_get_ldev_if_state((_device), (_min_state)) ? \ 1979 ({ __acquire(x); true; }) : false) 1980#define get_ldev(_device) get_ldev_if_state(_device, D_INCONSISTENT) 1981 1982static inline void put_ldev(struct drbd_device *device) 1983{ 1984 enum drbd_disk_state disk_state = device->state.disk; 1985 /* We must check the state *before* the atomic_dec becomes visible, 1986 * or we have a theoretical race where someone hitting zero, 1987 * while state still D_FAILED, will then see D_DISKLESS in the 1988 * condition below and calling into destroy, where he must not, yet. */ 1989 int i = atomic_dec_return(&device->local_cnt); 1990 1991 /* This may be called from some endio handler, 1992 * so we must not sleep here. */ 1993 1994 __release(local); 1995 D_ASSERT(device, i >= 0); 1996 if (i == 0) { 1997 if (disk_state == D_DISKLESS) 1998 /* even internal references gone, safe to destroy */ 1999 drbd_device_post_work(device, DESTROY_DISK); 2000 if (disk_state == D_FAILED) 2001 /* all application IO references gone. */ 2002 if (!test_and_set_bit(GOING_DISKLESS, &device->flags)) 2003 drbd_device_post_work(device, GO_DISKLESS); 2004 wake_up(&device->misc_wait); 2005 } 2006} 2007 2008#ifndef __CHECKER__ 2009static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins) 2010{ 2011 int io_allowed; 2012 2013 /* never get a reference while D_DISKLESS */ 2014 if (device->state.disk == D_DISKLESS) 2015 return 0; 2016 2017 atomic_inc(&device->local_cnt); 2018 io_allowed = (device->state.disk >= mins); 2019 if (!io_allowed) 2020 put_ldev(device); 2021 return io_allowed; 2022} 2023#else 2024extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins); 2025#endif 2026 2027/* this throttles on-the-fly application requests 2028 * according to max_buffers settings; 2029 * maybe re-implement using semaphores? */ 2030static inline int drbd_get_max_buffers(struct drbd_device *device) 2031{ 2032 struct net_conf *nc; 2033 int mxb; 2034 2035 rcu_read_lock(); 2036 nc = rcu_dereference(first_peer_device(device)->connection->net_conf); 2037 mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */ 2038 rcu_read_unlock(); 2039 2040 return mxb; 2041} 2042 2043static inline int drbd_state_is_stable(struct drbd_device *device) 2044{ 2045 union drbd_dev_state s = device->state; 2046 2047 /* DO NOT add a default clause, we want the compiler to warn us 2048 * for any newly introduced state we may have forgotten to add here */ 2049 2050 switch ((enum drbd_conns)s.conn) { 2051 /* new io only accepted when there is no connection, ... */ 2052 case C_STANDALONE: 2053 case C_WF_CONNECTION: 2054 /* ... or there is a well established connection. */ 2055 case C_CONNECTED: 2056 case C_SYNC_SOURCE: 2057 case C_SYNC_TARGET: 2058 case C_VERIFY_S: 2059 case C_VERIFY_T: 2060 case C_PAUSED_SYNC_S: 2061 case C_PAUSED_SYNC_T: 2062 case C_AHEAD: 2063 case C_BEHIND: 2064 /* transitional states, IO allowed */ 2065 case C_DISCONNECTING: 2066 case C_UNCONNECTED: 2067 case C_TIMEOUT: 2068 case C_BROKEN_PIPE: 2069 case C_NETWORK_FAILURE: 2070 case C_PROTOCOL_ERROR: 2071 case C_TEAR_DOWN: 2072 case C_WF_REPORT_PARAMS: 2073 case C_STARTING_SYNC_S: 2074 case C_STARTING_SYNC_T: 2075 break; 2076 2077 /* Allow IO in BM exchange states with new protocols */ 2078 case C_WF_BITMAP_S: 2079 if (first_peer_device(device)->connection->agreed_pro_version < 96) 2080 return 0; 2081 break; 2082 2083 /* no new io accepted in these states */ 2084 case C_WF_BITMAP_T: 2085 case C_WF_SYNC_UUID: 2086 case C_MASK: 2087 /* not "stable" */ 2088 return 0; 2089 } 2090 2091 switch ((enum drbd_disk_state)s.disk) { 2092 case D_DISKLESS: 2093 case D_INCONSISTENT: 2094 case D_OUTDATED: 2095 case D_CONSISTENT: 2096 case D_UP_TO_DATE: 2097 case D_FAILED: 2098 /* disk state is stable as well. */ 2099 break; 2100 2101 /* no new io accepted during transitional states */ 2102 case D_ATTACHING: 2103 case D_NEGOTIATING: 2104 case D_UNKNOWN: 2105 case D_MASK: 2106 /* not "stable" */ 2107 return 0; 2108 } 2109 2110 return 1; 2111} 2112 2113static inline int drbd_suspended(struct drbd_device *device) 2114{ 2115 struct drbd_resource *resource = device->resource; 2116 2117 return resource->susp || resource->susp_fen || resource->susp_nod; 2118} 2119 2120static inline bool may_inc_ap_bio(struct drbd_device *device) 2121{ 2122 int mxb = drbd_get_max_buffers(device); 2123 2124 if (drbd_suspended(device)) 2125 return false; 2126 if (atomic_read(&device->suspend_cnt)) 2127 return false; 2128 2129 /* to avoid potential deadlock or bitmap corruption, 2130 * in various places, we only allow new application io 2131 * to start during "stable" states. */ 2132 2133 /* no new io accepted when attaching or detaching the disk */ 2134 if (!drbd_state_is_stable(device)) 2135 return false; 2136 2137 /* since some older kernels don't have atomic_add_unless, 2138 * and we are within the spinlock anyways, we have this workaround. */ 2139 if (atomic_read(&device->ap_bio_cnt) > mxb) 2140 return false; 2141 if (test_bit(BITMAP_IO, &device->flags)) 2142 return false; 2143 return true; 2144} 2145 2146static inline bool inc_ap_bio_cond(struct drbd_device *device) 2147{ 2148 bool rv = false; 2149 2150 spin_lock_irq(&device->resource->req_lock); 2151 rv = may_inc_ap_bio(device); 2152 if (rv) 2153 atomic_inc(&device->ap_bio_cnt); 2154 spin_unlock_irq(&device->resource->req_lock); 2155 2156 return rv; 2157} 2158 2159static inline void inc_ap_bio(struct drbd_device *device) 2160{ 2161 /* we wait here 2162 * as long as the device is suspended 2163 * until the bitmap is no longer on the fly during connection 2164 * handshake as long as we would exceed the max_buffer limit. 2165 * 2166 * to avoid races with the reconnect code, 2167 * we need to atomic_inc within the spinlock. */ 2168 2169 wait_event(device->misc_wait, inc_ap_bio_cond(device)); 2170} 2171 2172static inline void dec_ap_bio(struct drbd_device *device) 2173{ 2174 int mxb = drbd_get_max_buffers(device); 2175 int ap_bio = atomic_dec_return(&device->ap_bio_cnt); 2176 2177 D_ASSERT(device, ap_bio >= 0); 2178 2179 if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) { 2180 if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags)) 2181 drbd_queue_work(&first_peer_device(device)-> 2182 connection->sender_work, 2183 &device->bm_io_work.w); 2184 } 2185 2186 /* this currently does wake_up for every dec_ap_bio! 2187 * maybe rather introduce some type of hysteresis? 2188 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */ 2189 if (ap_bio < mxb) 2190 wake_up(&device->misc_wait); 2191} 2192 2193static inline bool verify_can_do_stop_sector(struct drbd_device *device) 2194{ 2195 return first_peer_device(device)->connection->agreed_pro_version >= 97 && 2196 first_peer_device(device)->connection->agreed_pro_version != 100; 2197} 2198 2199static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val) 2200{ 2201 int changed = device->ed_uuid != val; 2202 device->ed_uuid = val; 2203 return changed; 2204} 2205 2206static inline int drbd_queue_order_type(struct drbd_device *device) 2207{ 2208 /* sorry, we currently have no working implementation 2209 * of distributed TCQ stuff */ 2210#ifndef QUEUE_ORDERED_NONE 2211#define QUEUE_ORDERED_NONE 0 2212#endif 2213 return QUEUE_ORDERED_NONE; 2214} 2215 2216static inline struct drbd_connection *first_connection(struct drbd_resource *resource) 2217{ 2218 return list_first_entry_or_null(&resource->connections, 2219 struct drbd_connection, connections); 2220} 2221 2222#endif
~/
Read-only file preview.