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 <crypto/hash.h> 30 #include <linux/compiler.h> 31 #include <linux/types.h> 32 #include <linux/list.h> 33 #include <linux/sched/signal.h> 34 #include <linux/bitops.h> 35 #include <linux/slab.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/backing-dev.h> 42 #include <linux/genhd.h> 43 #include <linux/idr.h> 44 #include <linux/dynamic_debug.h> 45 #include <net/tcp.h> 46 #include <linux/lru_cache.h> 47 #include <linux/prefetch.h> 48 #include <linux/drbd_genl_api.h> 49 #include <linux/drbd.h> 50 #include "drbd_strings.h" 51 #include "drbd_state.h" 52 #include "drbd_protocol.h" 53 54 #ifdef __CHECKER__ 55 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr"))) 56 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read"))) 57 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write"))) 58 #else 59 # define __protected_by(x) 60 # define __protected_read_by(x) 61 # define __protected_write_by(x) 62 #endif 63 64 /* shared module parameters, defined in drbd_main.c */ 65 #ifdef CONFIG_DRBD_FAULT_INJECTION 66 extern int drbd_enable_faults; 67 extern int drbd_fault_rate; 68 #endif 69 70 extern unsigned int drbd_minor_count; 71 extern char drbd_usermode_helper[]; 72 extern int drbd_proc_details; 73 74 75 /* This is used to stop/restart our threads. 76 * Cannot use SIGTERM nor SIGKILL, since these 77 * are sent out by init on runlevel changes 78 * I choose SIGHUP for now. 79 */ 80 #define DRBD_SIGKILL SIGHUP 81 82 #define ID_IN_SYNC (4711ULL) 83 #define ID_OUT_OF_SYNC (4712ULL) 84 #define ID_SYNCER (-1ULL) 85 86 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL) 87 88 struct drbd_device; 89 struct drbd_connection; 90 91 #define __drbd_printk_device(level, device, fmt, args...) \ 92 dev_printk(level, disk_to_dev((device)->vdisk), fmt, ## args) 93 #define __drbd_printk_peer_device(level, peer_device, fmt, args...) \ 94 dev_printk(level, disk_to_dev((peer_device)->device->vdisk), fmt, ## args) 95 #define __drbd_printk_resource(level, resource, fmt, args...) \ 96 printk(level "drbd %s: " fmt, (resource)->name, ## args) 97 #define __drbd_printk_connection(level, connection, fmt, args...) \ 98 printk(level "drbd %s: " fmt, (connection)->resource->name, ## args) 99 100 void drbd_printk_with_wrong_object_type(void); 101 102 #define __drbd_printk_if_same_type(obj, type, func, level, fmt, args...) \ 103 (__builtin_types_compatible_p(typeof(obj), type) || \ 104 __builtin_types_compatible_p(typeof(obj), const type)), \ 105 func(level, (const type)(obj), fmt, ## args) 106 107 #define drbd_printk(level, obj, fmt, args...) \ 108 __builtin_choose_expr( \ 109 __drbd_printk_if_same_type(obj, struct drbd_device *, \ 110 __drbd_printk_device, level, fmt, ## args), \ 111 __builtin_choose_expr( \ 112 __drbd_printk_if_same_type(obj, struct drbd_resource *, \ 113 __drbd_printk_resource, level, fmt, ## args), \ 114 __builtin_choose_expr( \ 115 __drbd_printk_if_same_type(obj, struct drbd_connection *, \ 116 __drbd_printk_connection, level, fmt, ## args), \ 117 __builtin_choose_expr( \ 118 __drbd_printk_if_same_type(obj, struct drbd_peer_device *, \ 119 __drbd_printk_peer_device, level, fmt, ## args), \ 120 drbd_printk_with_wrong_object_type())))) 121 122 #define drbd_dbg(obj, fmt, args...) \ 123 drbd_printk(KERN_DEBUG, obj, fmt, ## args) 124 #define drbd_alert(obj, fmt, args...) \ 125 drbd_printk(KERN_ALERT, obj, fmt, ## args) 126 #define drbd_err(obj, fmt, args...) \ 127 drbd_printk(KERN_ERR, obj, fmt, ## args) 128 #define drbd_warn(obj, fmt, args...) \ 129 drbd_printk(KERN_WARNING, obj, fmt, ## args) 130 #define drbd_info(obj, fmt, args...) \ 131 drbd_printk(KERN_INFO, obj, fmt, ## args) 132 #define drbd_emerg(obj, fmt, args...) \ 133 drbd_printk(KERN_EMERG, obj, fmt, ## args) 134 135 #define dynamic_drbd_dbg(device, fmt, args...) \ 136 dynamic_dev_dbg(disk_to_dev(device->vdisk), fmt, ## args) 137 138 #define D_ASSERT(device, exp) do { \ 139 if (!(exp)) \ 140 drbd_err(device, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__); \ 141 } while (0) 142 143 /** 144 * expect - Make an assertion 145 * 146 * Unlike the assert macro, this macro returns a boolean result. 147 */ 148 #define expect(exp) ({ \ 149 bool _bool = (exp); \ 150 if (!_bool) \ 151 drbd_err(device, "ASSERTION %s FAILED in %s\n", \ 152 #exp, __func__); \ 153 _bool; \ 154 }) 155 156 /* Defines to control fault insertion */ 157 enum { 158 DRBD_FAULT_MD_WR = 0, /* meta data write */ 159 DRBD_FAULT_MD_RD = 1, /* read */ 160 DRBD_FAULT_RS_WR = 2, /* resync */ 161 DRBD_FAULT_RS_RD = 3, 162 DRBD_FAULT_DT_WR = 4, /* data */ 163 DRBD_FAULT_DT_RD = 5, 164 DRBD_FAULT_DT_RA = 6, /* data read ahead */ 165 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */ 166 DRBD_FAULT_AL_EE = 8, /* alloc ee */ 167 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */ 168 169 DRBD_FAULT_MAX, 170 }; 171 172 extern unsigned int 173 _drbd_insert_fault(struct drbd_device *device, unsigned int type); 174 175 static inline int 176 drbd_insert_fault(struct drbd_device *device, unsigned int type) { 177 #ifdef CONFIG_DRBD_FAULT_INJECTION 178 return drbd_fault_rate && 179 (drbd_enable_faults & (1<<type)) && 180 _drbd_insert_fault(device, type); 181 #else 182 return 0; 183 #endif 184 } 185 186 /* integer division, round _UP_ to the next integer */ 187 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0)) 188 /* usual integer division */ 189 #define div_floor(A, B) ((A)/(B)) 190 191 extern struct ratelimit_state drbd_ratelimit_state; 192 extern struct idr drbd_devices; /* RCU, updates: genl_lock() */ 193 extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */ 194 195 extern const char *cmdname(enum drbd_packet cmd); 196 197 /* for sending/receiving the bitmap, 198 * possibly in some encoding scheme */ 199 struct bm_xfer_ctx { 200 /* "const" 201 * stores total bits and long words 202 * of the bitmap, so we don't need to 203 * call the accessor functions over and again. */ 204 unsigned long bm_bits; 205 unsigned long bm_words; 206 /* during xfer, current position within the bitmap */ 207 unsigned long bit_offset; 208 unsigned long word_offset; 209 210 /* statistics; index: (h->command == P_BITMAP) */ 211 unsigned packets[2]; 212 unsigned bytes[2]; 213 }; 214 215 extern void INFO_bm_xfer_stats(struct drbd_device *device, 216 const char *direction, struct bm_xfer_ctx *c); 217 218 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c) 219 { 220 /* word_offset counts "native long words" (32 or 64 bit), 221 * aligned at 64 bit. 222 * Encoded packet may end at an unaligned bit offset. 223 * In case a fallback clear text packet is transmitted in 224 * between, we adjust this offset back to the last 64bit 225 * aligned "native long word", which makes coding and decoding 226 * the plain text bitmap much more convenient. */ 227 #if BITS_PER_LONG == 64 228 c->word_offset = c->bit_offset >> 6; 229 #elif BITS_PER_LONG == 32 230 c->word_offset = c->bit_offset >> 5; 231 c->word_offset &= ~(1UL); 232 #else 233 # error "unsupported BITS_PER_LONG" 234 #endif 235 } 236 237 extern unsigned int drbd_header_size(struct drbd_connection *connection); 238 239 /**********************************************************************/ 240 enum drbd_thread_state { 241 NONE, 242 RUNNING, 243 EXITING, 244 RESTARTING 245 }; 246 247 struct drbd_thread { 248 spinlock_t t_lock; 249 struct task_struct *task; 250 struct completion stop; 251 enum drbd_thread_state t_state; 252 int (*function) (struct drbd_thread *); 253 struct drbd_resource *resource; 254 struct drbd_connection *connection; 255 int reset_cpu_mask; 256 const char *name; 257 }; 258 259 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi) 260 { 261 /* THINK testing the t_state seems to be uncritical in all cases 262 * (but thread_{start,stop}), so we can read it *without* the lock. 263 * --lge */ 264 265 smp_rmb(); 266 return thi->t_state; 267 } 268 269 struct drbd_work { 270 struct list_head list; 271 int (*cb)(struct drbd_work *, int cancel); 272 }; 273 274 struct drbd_device_work { 275 struct drbd_work w; 276 struct drbd_device *device; 277 }; 278 279 #include "drbd_interval.h" 280 281 extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *); 282 283 extern void lock_all_resources(void); 284 extern void unlock_all_resources(void); 285 286 struct drbd_request { 287 struct drbd_work w; 288 struct drbd_device *device; 289 290 /* if local IO is not allowed, will be NULL. 291 * if local IO _is_ allowed, holds the locally submitted bio clone, 292 * or, after local IO completion, the ERR_PTR(error). 293 * see drbd_request_endio(). */ 294 struct bio *private_bio; 295 296 struct drbd_interval i; 297 298 /* epoch: used to check on "completion" whether this req was in 299 * the current epoch, and we therefore have to close it, 300 * causing a p_barrier packet to be send, starting a new epoch. 301 * 302 * This corresponds to "barrier" in struct p_barrier[_ack], 303 * and to "barrier_nr" in struct drbd_epoch (and various 304 * comments/function parameters/local variable names). 305 */ 306 unsigned int epoch; 307 308 struct list_head tl_requests; /* ring list in the transfer log */ 309 struct bio *master_bio; /* master bio pointer */ 310 311 /* see struct drbd_device */ 312 struct list_head req_pending_master_completion; 313 struct list_head req_pending_local; 314 315 /* for generic IO accounting */ 316 unsigned long start_jif; 317 318 /* for DRBD internal statistics */ 319 320 /* Minimal set of time stamps to determine if we wait for activity log 321 * transactions, local disk or peer. 32 bit "jiffies" are good enough, 322 * we don't expect a DRBD request to be stalled for several month. 323 */ 324 325 /* before actual request processing */ 326 unsigned long in_actlog_jif; 327 328 /* local disk */ 329 unsigned long pre_submit_jif; 330 331 /* per connection */ 332 unsigned long pre_send_jif; 333 unsigned long acked_jif; 334 unsigned long net_done_jif; 335 336 /* Possibly even more detail to track each phase: 337 * master_completion_jif 338 * how long did it take to complete the master bio 339 * (application visible latency) 340 * allocated_jif 341 * how long the master bio was blocked until we finally allocated 342 * a tracking struct 343 * in_actlog_jif 344 * how long did we wait for activity log transactions 345 * 346 * net_queued_jif 347 * when did we finally queue it for sending 348 * pre_send_jif 349 * when did we start sending it 350 * post_send_jif 351 * how long did we block in the network stack trying to send it 352 * acked_jif 353 * when did we receive (or fake, in protocol A) a remote ACK 354 * net_done_jif 355 * when did we receive final acknowledgement (P_BARRIER_ACK), 356 * or decide, e.g. on connection loss, that we do no longer expect 357 * anything from this peer for this request. 358 * 359 * pre_submit_jif 360 * post_sub_jif 361 * when did we start submiting to the lower level device, 362 * and how long did we block in that submit function 363 * local_completion_jif 364 * how long did it take the lower level device to complete this request 365 */ 366 367 368 /* once it hits 0, we may complete the master_bio */ 369 atomic_t completion_ref; 370 /* once it hits 0, we may destroy this drbd_request object */ 371 struct kref kref; 372 373 unsigned rq_state; /* see comments above _req_mod() */ 374 }; 375 376 struct drbd_epoch { 377 struct drbd_connection *connection; 378 struct list_head list; 379 unsigned int barrier_nr; 380 atomic_t epoch_size; /* increased on every request added. */ 381 atomic_t active; /* increased on every req. added, and dec on every finished. */ 382 unsigned long flags; 383 }; 384 385 /* Prototype declaration of function defined in drbd_receiver.c */ 386 int drbdd_init(struct drbd_thread *); 387 int drbd_asender(struct drbd_thread *); 388 389 /* drbd_epoch flag bits */ 390 enum { 391 DE_HAVE_BARRIER_NUMBER, 392 }; 393 394 enum epoch_event { 395 EV_PUT, 396 EV_GOT_BARRIER_NR, 397 EV_BECAME_LAST, 398 EV_CLEANUP = 32, /* used as flag */ 399 }; 400 401 struct digest_info { 402 int digest_size; 403 void *digest; 404 }; 405 406 struct drbd_peer_request { 407 struct drbd_work w; 408 struct drbd_peer_device *peer_device; 409 struct drbd_epoch *epoch; /* for writes */ 410 struct page *pages; 411 atomic_t pending_bios; 412 struct drbd_interval i; 413 /* see comments on ee flag bits below */ 414 unsigned long flags; 415 unsigned long submit_jif; 416 union { 417 u64 block_id; 418 struct digest_info *digest; 419 }; 420 }; 421 422 /* ee flag bits. 423 * While corresponding bios are in flight, the only modification will be 424 * set_bit WAS_ERROR, which has to be atomic. 425 * If no bios are in flight yet, or all have been completed, 426 * non-atomic modification to ee->flags is ok. 427 */ 428 enum { 429 __EE_CALL_AL_COMPLETE_IO, 430 __EE_MAY_SET_IN_SYNC, 431 432 /* is this a TRIM aka REQ_DISCARD? */ 433 __EE_IS_TRIM, 434 435 /* In case a barrier failed, 436 * we need to resubmit without the barrier flag. */ 437 __EE_RESUBMITTED, 438 439 /* we may have several bios per peer request. 440 * if any of those fail, we set this flag atomically 441 * from the endio callback */ 442 __EE_WAS_ERROR, 443 444 /* This ee has a pointer to a digest instead of a block id */ 445 __EE_HAS_DIGEST, 446 447 /* Conflicting local requests need to be restarted after this request */ 448 __EE_RESTART_REQUESTS, 449 450 /* The peer wants a write ACK for this (wire proto C) */ 451 __EE_SEND_WRITE_ACK, 452 453 /* Is set when net_conf had two_primaries set while creating this peer_req */ 454 __EE_IN_INTERVAL_TREE, 455 456 /* for debugfs: */ 457 /* has this been submitted, or does it still wait for something else? */ 458 __EE_SUBMITTED, 459 460 /* this is/was a write request */ 461 __EE_WRITE, 462 463 /* this is/was a write same request */ 464 __EE_WRITE_SAME, 465 466 /* this originates from application on peer 467 * (not some resync or verify or other DRBD internal request) */ 468 __EE_APPLICATION, 469 470 /* If it contains only 0 bytes, send back P_RS_DEALLOCATED */ 471 __EE_RS_THIN_REQ, 472 }; 473 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO) 474 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC) 475 #define EE_IS_TRIM (1<<__EE_IS_TRIM) 476 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED) 477 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR) 478 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST) 479 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS) 480 #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK) 481 #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE) 482 #define EE_SUBMITTED (1<<__EE_SUBMITTED) 483 #define EE_WRITE (1<<__EE_WRITE) 484 #define EE_WRITE_SAME (1<<__EE_WRITE_SAME) 485 #define EE_APPLICATION (1<<__EE_APPLICATION) 486 #define EE_RS_THIN_REQ (1<<__EE_RS_THIN_REQ) 487 488 /* flag bits per device */ 489 enum { 490 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */ 491 MD_DIRTY, /* current uuids and flags not yet on disk */ 492 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */ 493 CL_ST_CHG_SUCCESS, 494 CL_ST_CHG_FAIL, 495 CRASHED_PRIMARY, /* This node was a crashed primary. 496 * Gets cleared when the state.conn 497 * goes into C_CONNECTED state. */ 498 CONSIDER_RESYNC, 499 500 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */ 501 502 BITMAP_IO, /* suspend application io; 503 once no more io in flight, start bitmap io */ 504 BITMAP_IO_QUEUED, /* Started bitmap IO */ 505 WAS_IO_ERROR, /* Local disk failed, returned IO error */ 506 WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */ 507 FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */ 508 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */ 509 RESIZE_PENDING, /* Size change detected locally, waiting for the response from 510 * the peer, if it changed there as well. */ 511 NEW_CUR_UUID, /* Create new current UUID when thawing IO */ 512 AL_SUSPENDED, /* Activity logging is currently suspended. */ 513 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */ 514 B_RS_H_DONE, /* Before resync handler done (already executed) */ 515 DISCARD_MY_DATA, /* discard_my_data flag per volume */ 516 READ_BALANCE_RR, 517 518 FLUSH_PENDING, /* if set, device->flush_jif is when we submitted that flush 519 * from drbd_flush_after_epoch() */ 520 521 /* cleared only after backing device related structures have been destroyed. */ 522 GOING_DISKLESS, /* Disk is being detached, because of io-error, or admin request. */ 523 524 /* to be used in drbd_device_post_work() */ 525 GO_DISKLESS, /* tell worker to schedule cleanup before detach */ 526 DESTROY_DISK, /* tell worker to close backing devices and destroy related structures. */ 527 MD_SYNC, /* tell worker to call drbd_md_sync() */ 528 RS_START, /* tell worker to start resync/OV */ 529 RS_PROGRESS, /* tell worker that resync made significant progress */ 530 RS_DONE, /* tell worker that resync is done */ 531 }; 532 533 struct drbd_bitmap; /* opaque for drbd_device */ 534 535 /* definition of bits in bm_flags to be used in drbd_bm_lock 536 * and drbd_bitmap_io and friends. */ 537 enum bm_flag { 538 /* currently locked for bulk operation */ 539 BM_LOCKED_MASK = 0xf, 540 541 /* in detail, that is: */ 542 BM_DONT_CLEAR = 0x1, 543 BM_DONT_SET = 0x2, 544 BM_DONT_TEST = 0x4, 545 546 /* so we can mark it locked for bulk operation, 547 * and still allow all non-bulk operations */ 548 BM_IS_LOCKED = 0x8, 549 550 /* (test bit, count bit) allowed (common case) */ 551 BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED, 552 553 /* testing bits, as well as setting new bits allowed, but clearing bits 554 * would be unexpected. Used during bitmap receive. Setting new bits 555 * requires sending of "out-of-sync" information, though. */ 556 BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED, 557 558 /* for drbd_bm_write_copy_pages, everything is allowed, 559 * only concurrent bulk operations are locked out. */ 560 BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED, 561 }; 562 563 struct drbd_work_queue { 564 struct list_head q; 565 spinlock_t q_lock; /* to protect the list. */ 566 wait_queue_head_t q_wait; 567 }; 568 569 struct drbd_socket { 570 struct mutex mutex; 571 struct socket *socket; 572 /* this way we get our 573 * send/receive buffers off the stack */ 574 void *sbuf; 575 void *rbuf; 576 }; 577 578 struct drbd_md { 579 u64 md_offset; /* sector offset to 'super' block */ 580 581 u64 la_size_sect; /* last agreed size, unit sectors */ 582 spinlock_t uuid_lock; 583 u64 uuid[UI_SIZE]; 584 u64 device_uuid; 585 u32 flags; 586 u32 md_size_sect; 587 588 s32 al_offset; /* signed relative sector offset to activity log */ 589 s32 bm_offset; /* signed relative sector offset to bitmap */ 590 591 /* cached value of bdev->disk_conf->meta_dev_idx (see below) */ 592 s32 meta_dev_idx; 593 594 /* see al_tr_number_to_on_disk_sector() */ 595 u32 al_stripes; 596 u32 al_stripe_size_4k; 597 u32 al_size_4k; /* cached product of the above */ 598 }; 599 600 struct drbd_backing_dev { 601 struct block_device *backing_bdev; 602 struct block_device *md_bdev; 603 struct drbd_md md; 604 struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */ 605 sector_t known_size; /* last known size of that backing device */ 606 }; 607 608 struct drbd_md_io { 609 struct page *page; 610 unsigned long start_jif; /* last call to drbd_md_get_buffer */ 611 unsigned long submit_jif; /* last _drbd_md_sync_page_io() submit */ 612 const char *current_use; 613 atomic_t in_use; 614 unsigned int done; 615 int error; 616 }; 617 618 struct bm_io_work { 619 struct drbd_work w; 620 char *why; 621 enum bm_flag flags; 622 int (*io_fn)(struct drbd_device *device); 623 void (*done)(struct drbd_device *device, int rv); 624 }; 625 626 struct fifo_buffer { 627 unsigned int head_index; 628 unsigned int size; 629 int total; /* sum of all values */ 630 int values[0]; 631 }; 632 extern struct fifo_buffer *fifo_alloc(int fifo_size); 633 634 /* flag bits per connection */ 635 enum { 636 NET_CONGESTED, /* The data socket is congested */ 637 RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */ 638 SEND_PING, 639 GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */ 640 CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */ 641 CONN_WD_ST_CHG_OKAY, 642 CONN_WD_ST_CHG_FAIL, 643 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */ 644 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */ 645 STATE_SENT, /* Do not change state/UUIDs while this is set */ 646 CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC) 647 * pending, from drbd worker context. 648 * If set, bdi_write_congested() returns true, 649 * so shrink_page_list() would not recurse into, 650 * and potentially deadlock on, this drbd worker. 651 */ 652 DISCONNECT_SENT, 653 654 DEVICE_WORK_PENDING, /* tell worker that some device has pending work */ 655 }; 656 657 enum which_state { NOW, OLD = NOW, NEW }; 658 659 struct drbd_resource { 660 char *name; 661 #ifdef CONFIG_DEBUG_FS 662 struct dentry *debugfs_res; 663 struct dentry *debugfs_res_volumes; 664 struct dentry *debugfs_res_connections; 665 struct dentry *debugfs_res_in_flight_summary; 666 #endif 667 struct kref kref; 668 struct idr devices; /* volume number to device mapping */ 669 struct list_head connections; 670 struct list_head resources; 671 struct res_opts res_opts; 672 struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */ 673 struct mutex adm_mutex; /* mutex to serialize administrative requests */ 674 spinlock_t req_lock; 675 676 unsigned susp:1; /* IO suspended by user */ 677 unsigned susp_nod:1; /* IO suspended because no data */ 678 unsigned susp_fen:1; /* IO suspended because fence peer handler runs */ 679 680 enum write_ordering_e write_ordering; 681 682 cpumask_var_t cpu_mask; 683 }; 684 685 struct drbd_thread_timing_details 686 { 687 unsigned long start_jif; 688 void *cb_addr; 689 const char *caller_fn; 690 unsigned int line; 691 unsigned int cb_nr; 692 }; 693 694 struct drbd_connection { 695 struct list_head connections; 696 struct drbd_resource *resource; 697 #ifdef CONFIG_DEBUG_FS 698 struct dentry *debugfs_conn; 699 struct dentry *debugfs_conn_callback_history; 700 struct dentry *debugfs_conn_oldest_requests; 701 #endif 702 struct kref kref; 703 struct idr peer_devices; /* volume number to peer device mapping */ 704 enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */ 705 struct mutex cstate_mutex; /* Protects graceful disconnects */ 706 unsigned int connect_cnt; /* Inc each time a connection is established */ 707 708 unsigned long flags; 709 struct net_conf *net_conf; /* content protected by rcu */ 710 wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */ 711 712 struct sockaddr_storage my_addr; 713 int my_addr_len; 714 struct sockaddr_storage peer_addr; 715 int peer_addr_len; 716 717 struct drbd_socket data; /* data/barrier/cstate/parameter packets */ 718 struct drbd_socket meta; /* ping/ack (metadata) packets */ 719 int agreed_pro_version; /* actually used protocol version */ 720 u32 agreed_features; 721 unsigned long last_received; /* in jiffies, either socket */ 722 unsigned int ko_count; 723 724 struct list_head transfer_log; /* all requests not yet fully processed */ 725 726 struct crypto_shash *cram_hmac_tfm; 727 struct crypto_ahash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */ 728 struct crypto_ahash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */ 729 struct crypto_ahash *csums_tfm; 730 struct crypto_ahash *verify_tfm; 731 void *int_dig_in; 732 void *int_dig_vv; 733 734 /* receiver side */ 735 struct drbd_epoch *current_epoch; 736 spinlock_t epoch_lock; 737 unsigned int epochs; 738 atomic_t current_tle_nr; /* transfer log epoch number */ 739 unsigned current_tle_writes; /* writes seen within this tl epoch */ 740 741 unsigned long last_reconnect_jif; 742 /* empty member on older kernels without blk_start_plug() */ 743 struct blk_plug receiver_plug; 744 struct drbd_thread receiver; 745 struct drbd_thread worker; 746 struct drbd_thread ack_receiver; 747 struct workqueue_struct *ack_sender; 748 749 /* cached pointers, 750 * so we can look up the oldest pending requests more quickly. 751 * protected by resource->req_lock */ 752 struct drbd_request *req_next; /* DRBD 9: todo.req_next */ 753 struct drbd_request *req_ack_pending; 754 struct drbd_request *req_not_net_done; 755 756 /* sender side */ 757 struct drbd_work_queue sender_work; 758 759 #define DRBD_THREAD_DETAILS_HIST 16 760 unsigned int w_cb_nr; /* keeps counting up */ 761 unsigned int r_cb_nr; /* keeps counting up */ 762 struct drbd_thread_timing_details w_timing_details[DRBD_THREAD_DETAILS_HIST]; 763 struct drbd_thread_timing_details r_timing_details[DRBD_THREAD_DETAILS_HIST]; 764 765 struct { 766 unsigned long last_sent_barrier_jif; 767 768 /* whether this sender thread 769 * has processed a single write yet. */ 770 bool seen_any_write_yet; 771 772 /* Which barrier number to send with the next P_BARRIER */ 773 int current_epoch_nr; 774 775 /* how many write requests have been sent 776 * with req->epoch == current_epoch_nr. 777 * If none, no P_BARRIER will be sent. */ 778 unsigned current_epoch_writes; 779 } send; 780 }; 781 782 static inline bool has_net_conf(struct drbd_connection *connection) 783 { 784 bool has_net_conf; 785 786 rcu_read_lock(); 787 has_net_conf = rcu_dereference(connection->net_conf); 788 rcu_read_unlock(); 789 790 return has_net_conf; 791 } 792 793 void __update_timing_details( 794 struct drbd_thread_timing_details *tdp, 795 unsigned int *cb_nr, 796 void *cb, 797 const char *fn, const unsigned int line); 798 799 #define update_worker_timing_details(c, cb) \ 800 __update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ ) 801 #define update_receiver_timing_details(c, cb) \ 802 __update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ ) 803 804 struct submit_worker { 805 struct workqueue_struct *wq; 806 struct work_struct worker; 807 808 /* protected by ..->resource->req_lock */ 809 struct list_head writes; 810 }; 811 812 struct drbd_peer_device { 813 struct list_head peer_devices; 814 struct drbd_device *device; 815 struct drbd_connection *connection; 816 struct work_struct send_acks_work; 817 #ifdef CONFIG_DEBUG_FS 818 struct dentry *debugfs_peer_dev; 819 #endif 820 }; 821 822 struct drbd_device { 823 struct drbd_resource *resource; 824 struct list_head peer_devices; 825 struct list_head pending_bitmap_io; 826 827 unsigned long flush_jif; 828 #ifdef CONFIG_DEBUG_FS 829 struct dentry *debugfs_minor; 830 struct dentry *debugfs_vol; 831 struct dentry *debugfs_vol_oldest_requests; 832 struct dentry *debugfs_vol_act_log_extents; 833 struct dentry *debugfs_vol_resync_extents; 834 struct dentry *debugfs_vol_data_gen_id; 835 struct dentry *debugfs_vol_ed_gen_id; 836 #endif 837 838 unsigned int vnr; /* volume number within the connection */ 839 unsigned int minor; /* device minor number */ 840 841 struct kref kref; 842 843 /* things that are stored as / read from meta data on disk */ 844 unsigned long flags; 845 846 /* configured by drbdsetup */ 847 struct drbd_backing_dev *ldev __protected_by(local); 848 849 sector_t p_size; /* partner's disk size */ 850 struct request_queue *rq_queue; 851 struct block_device *this_bdev; 852 struct gendisk *vdisk; 853 854 unsigned long last_reattach_jif; 855 struct drbd_work resync_work; 856 struct drbd_work unplug_work; 857 struct timer_list resync_timer; 858 struct timer_list md_sync_timer; 859 struct timer_list start_resync_timer; 860 struct timer_list request_timer; 861 862 /* Used after attach while negotiating new disk state. */ 863 union drbd_state new_state_tmp; 864 865 union drbd_dev_state state; 866 wait_queue_head_t misc_wait; 867 wait_queue_head_t state_wait; /* upon each state change. */ 868 unsigned int send_cnt; 869 unsigned int recv_cnt; 870 unsigned int read_cnt; 871 unsigned int writ_cnt; 872 unsigned int al_writ_cnt; 873 unsigned int bm_writ_cnt; 874 atomic_t ap_bio_cnt; /* Requests we need to complete */ 875 atomic_t ap_actlog_cnt; /* Requests waiting for activity log */ 876 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */ 877 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */ 878 atomic_t unacked_cnt; /* Need to send replies for */ 879 atomic_t local_cnt; /* Waiting for local completion */ 880 atomic_t suspend_cnt; 881 882 /* Interval tree of pending local requests */ 883 struct rb_root read_requests; 884 struct rb_root write_requests; 885 886 /* for statistics and timeouts */ 887 /* [0] read, [1] write */ 888 struct list_head pending_master_completion[2]; 889 struct list_head pending_completion[2]; 890 891 /* use checksums for *this* resync */ 892 bool use_csums; 893 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */ 894 unsigned long rs_total; 895 /* number of resync blocks that failed in this run */ 896 unsigned long rs_failed; 897 /* Syncer's start time [unit jiffies] */ 898 unsigned long rs_start; 899 /* cumulated time in PausedSyncX state [unit jiffies] */ 900 unsigned long rs_paused; 901 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */ 902 unsigned long rs_same_csum; 903 #define DRBD_SYNC_MARKS 8 904 #define DRBD_SYNC_MARK_STEP (3*HZ) 905 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */ 906 unsigned long rs_mark_left[DRBD_SYNC_MARKS]; 907 /* marks's time [unit jiffies] */ 908 unsigned long rs_mark_time[DRBD_SYNC_MARKS]; 909 /* current index into rs_mark_{left,time} */ 910 int rs_last_mark; 911 unsigned long rs_last_bcast; /* [unit jiffies] */ 912 913 /* where does the admin want us to start? (sector) */ 914 sector_t ov_start_sector; 915 sector_t ov_stop_sector; 916 /* where are we now? (sector) */ 917 sector_t ov_position; 918 /* Start sector of out of sync range (to merge printk reporting). */ 919 sector_t ov_last_oos_start; 920 /* size of out-of-sync range in sectors. */ 921 sector_t ov_last_oos_size; 922 unsigned long ov_left; /* in bits */ 923 924 struct drbd_bitmap *bitmap; 925 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */ 926 927 /* Used to track operations of resync... */ 928 struct lru_cache *resync; 929 /* Number of locked elements in resync LRU */ 930 unsigned int resync_locked; 931 /* resync extent number waiting for application requests */ 932 unsigned int resync_wenr; 933 934 int open_cnt; 935 u64 *p_uuid; 936 937 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */ 938 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */ 939 struct list_head done_ee; /* need to send P_WRITE_ACK */ 940 struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */ 941 struct list_head net_ee; /* zero-copy network send in progress */ 942 943 int next_barrier_nr; 944 struct list_head resync_reads; 945 atomic_t pp_in_use; /* allocated from page pool */ 946 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */ 947 wait_queue_head_t ee_wait; 948 struct drbd_md_io md_io; 949 spinlock_t al_lock; 950 wait_queue_head_t al_wait; 951 struct lru_cache *act_log; /* activity log */ 952 unsigned int al_tr_number; 953 int al_tr_cycle; 954 wait_queue_head_t seq_wait; 955 atomic_t packet_seq; 956 unsigned int peer_seq; 957 spinlock_t peer_seq_lock; 958 unsigned long comm_bm_set; /* communicated number of set bits. */ 959 struct bm_io_work bm_io_work; 960 u64 ed_uuid; /* UUID of the exposed data */ 961 struct mutex own_state_mutex; 962 struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */ 963 char congestion_reason; /* Why we where congested... */ 964 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */ 965 atomic_t rs_sect_ev; /* for submitted resync data rate, both */ 966 int rs_last_sect_ev; /* counter to compare with */ 967 int rs_last_events; /* counter of read or write "events" (unit sectors) 968 * on the lower level device when we last looked. */ 969 int c_sync_rate; /* current resync rate after syncer throttle magic */ 970 struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */ 971 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */ 972 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */ 973 unsigned int peer_max_bio_size; 974 unsigned int local_max_bio_size; 975 976 /* any requests that would block in drbd_make_request() 977 * are deferred to this single-threaded work queue */ 978 struct submit_worker submit; 979 }; 980 981 struct drbd_bm_aio_ctx { 982 struct drbd_device *device; 983 struct list_head list; /* on device->pending_bitmap_io */; 984 unsigned long start_jif; 985 atomic_t in_flight; 986 unsigned int done; 987 unsigned flags; 988 #define BM_AIO_COPY_PAGES 1 989 #define BM_AIO_WRITE_HINTED 2 990 #define BM_AIO_WRITE_ALL_PAGES 4 991 #define BM_AIO_READ 8 992 int error; 993 struct kref kref; 994 }; 995 996 struct drbd_config_context { 997 /* assigned from drbd_genlmsghdr */ 998 unsigned int minor; 999 /* assigned from request attributes, if present */ 1000 unsigned int volume; 1001 #define VOLUME_UNSPECIFIED (-1U) 1002 /* pointer into the request skb, 1003 * limited lifetime! */ 1004 char *resource_name; 1005 struct nlattr *my_addr; 1006 struct nlattr *peer_addr; 1007 1008 /* reply buffer */ 1009 struct sk_buff *reply_skb; 1010 /* pointer into reply buffer */ 1011 struct drbd_genlmsghdr *reply_dh; 1012 /* resolved from attributes, if possible */ 1013 struct drbd_device *device; 1014 struct drbd_resource *resource; 1015 struct drbd_connection *connection; 1016 }; 1017 1018 static inline struct drbd_device *minor_to_device(unsigned int minor) 1019 { 1020 return (struct drbd_device *)idr_find(&drbd_devices, minor); 1021 } 1022 1023 static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device) 1024 { 1025 return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices); 1026 } 1027 1028 static inline struct drbd_peer_device * 1029 conn_peer_device(struct drbd_connection *connection, int volume_number) 1030 { 1031 return idr_find(&connection->peer_devices, volume_number); 1032 } 1033 1034 #define for_each_resource(resource, _resources) \ 1035 list_for_each_entry(resource, _resources, resources) 1036 1037 #define for_each_resource_rcu(resource, _resources) \ 1038 list_for_each_entry_rcu(resource, _resources, resources) 1039 1040 #define for_each_resource_safe(resource, tmp, _resources) \ 1041 list_for_each_entry_safe(resource, tmp, _resources, resources) 1042 1043 #define for_each_connection(connection, resource) \ 1044 list_for_each_entry(connection, &resource->connections, connections) 1045 1046 #define for_each_connection_rcu(connection, resource) \ 1047 list_for_each_entry_rcu(connection, &resource->connections, connections) 1048 1049 #define for_each_connection_safe(connection, tmp, resource) \ 1050 list_for_each_entry_safe(connection, tmp, &resource->connections, connections) 1051 1052 #define for_each_peer_device(peer_device, device) \ 1053 list_for_each_entry(peer_device, &device->peer_devices, peer_devices) 1054 1055 #define for_each_peer_device_rcu(peer_device, device) \ 1056 list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices) 1057 1058 #define for_each_peer_device_safe(peer_device, tmp, device) \ 1059 list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices) 1060 1061 static inline unsigned int device_to_minor(struct drbd_device *device) 1062 { 1063 return device->minor; 1064 } 1065 1066 /* 1067 * function declarations 1068 *************************/ 1069 1070 /* drbd_main.c */ 1071 1072 enum dds_flags { 1073 DDSF_FORCED = 1, 1074 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */ 1075 }; 1076 1077 extern void drbd_init_set_defaults(struct drbd_device *device); 1078 extern int drbd_thread_start(struct drbd_thread *thi); 1079 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait); 1080 #ifdef CONFIG_SMP 1081 extern void drbd_thread_current_set_cpu(struct drbd_thread *thi); 1082 #else 1083 #define drbd_thread_current_set_cpu(A) ({}) 1084 #endif 1085 extern void tl_release(struct drbd_connection *, unsigned int barrier_nr, 1086 unsigned int set_size); 1087 extern void tl_clear(struct drbd_connection *); 1088 extern void drbd_free_sock(struct drbd_connection *connection); 1089 extern int drbd_send(struct drbd_connection *connection, struct socket *sock, 1090 void *buf, size_t size, unsigned msg_flags); 1091 extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t, 1092 unsigned); 1093 1094 extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd); 1095 extern int drbd_send_protocol(struct drbd_connection *connection); 1096 extern int drbd_send_uuids(struct drbd_peer_device *); 1097 extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *); 1098 extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *); 1099 extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags); 1100 extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s); 1101 extern int drbd_send_current_state(struct drbd_peer_device *); 1102 extern int drbd_send_sync_param(struct drbd_peer_device *); 1103 extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr, 1104 u32 set_size); 1105 extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet, 1106 struct drbd_peer_request *); 1107 extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet, 1108 struct p_block_req *rp); 1109 extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet, 1110 struct p_data *dp, int data_size); 1111 extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet, 1112 sector_t sector, int blksize, u64 block_id); 1113 extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *); 1114 extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet, 1115 struct drbd_peer_request *); 1116 extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req); 1117 extern int drbd_send_drequest(struct drbd_peer_device *, int cmd, 1118 sector_t sector, int size, u64 block_id); 1119 extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector, 1120 int size, void *digest, int digest_size, 1121 enum drbd_packet cmd); 1122 extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size); 1123 1124 extern int drbd_send_bitmap(struct drbd_device *device); 1125 extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode); 1126 extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode); 1127 extern int drbd_send_rs_deallocated(struct drbd_peer_device *, struct drbd_peer_request *); 1128 extern void drbd_backing_dev_free(struct drbd_device *device, struct drbd_backing_dev *ldev); 1129 extern void drbd_device_cleanup(struct drbd_device *device); 1130 extern void drbd_print_uuids(struct drbd_device *device, const char *text); 1131 extern void drbd_queue_unplug(struct drbd_device *device); 1132 1133 extern void conn_md_sync(struct drbd_connection *connection); 1134 extern void drbd_md_write(struct drbd_device *device, void *buffer); 1135 extern void drbd_md_sync(struct drbd_device *device); 1136 extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev); 1137 extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); 1138 extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); 1139 extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local); 1140 extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local); 1141 extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local); 1142 extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); 1143 extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local); 1144 extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local); 1145 extern int drbd_md_test_flag(struct drbd_backing_dev *, int); 1146 extern void drbd_md_mark_dirty(struct drbd_device *device); 1147 extern void drbd_queue_bitmap_io(struct drbd_device *device, 1148 int (*io_fn)(struct drbd_device *), 1149 void (*done)(struct drbd_device *, int), 1150 char *why, enum bm_flag flags); 1151 extern int drbd_bitmap_io(struct drbd_device *device, 1152 int (*io_fn)(struct drbd_device *), 1153 char *why, enum bm_flag flags); 1154 extern int drbd_bitmap_io_from_worker(struct drbd_device *device, 1155 int (*io_fn)(struct drbd_device *), 1156 char *why, enum bm_flag flags); 1157 extern int drbd_bmio_set_n_write(struct drbd_device *device) __must_hold(local); 1158 extern int drbd_bmio_clear_n_write(struct drbd_device *device) __must_hold(local); 1159 1160 /* Meta data layout 1161 * 1162 * We currently have two possible layouts. 1163 * Offsets in (512 byte) sectors. 1164 * external: 1165 * |----------- md_size_sect ------------------| 1166 * [ 4k superblock ][ activity log ][ Bitmap ] 1167 * | al_offset == 8 | 1168 * | bm_offset = al_offset + X | 1169 * ==> bitmap sectors = md_size_sect - bm_offset 1170 * 1171 * Variants: 1172 * old, indexed fixed size meta data: 1173 * 1174 * internal: 1175 * |----------- md_size_sect ------------------| 1176 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*] 1177 * | al_offset < 0 | 1178 * | bm_offset = al_offset - Y | 1179 * ==> bitmap sectors = Y = al_offset - bm_offset 1180 * 1181 * [padding*] are zero or up to 7 unused 512 Byte sectors to the 1182 * end of the device, so that the [4k superblock] will be 4k aligned. 1183 * 1184 * The activity log consists of 4k transaction blocks, 1185 * which are written in a ring-buffer, or striped ring-buffer like fashion, 1186 * which are writtensize used to be fixed 32kB, 1187 * but is about to become configurable. 1188 */ 1189 1190 /* Our old fixed size meta data layout 1191 * allows up to about 3.8TB, so if you want more, 1192 * you need to use the "flexible" meta data format. */ 1193 #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */ 1194 #define MD_4kB_SECT 8 1195 #define MD_32kB_SECT 64 1196 1197 /* One activity log extent represents 4M of storage */ 1198 #define AL_EXTENT_SHIFT 22 1199 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT) 1200 1201 /* We could make these currently hardcoded constants configurable 1202 * variables at create-md time (or even re-configurable at runtime?). 1203 * Which will require some more changes to the DRBD "super block" 1204 * and attach code. 1205 * 1206 * updates per transaction: 1207 * This many changes to the active set can be logged with one transaction. 1208 * This number is arbitrary. 1209 * context per transaction: 1210 * This many context extent numbers are logged with each transaction. 1211 * This number is resulting from the transaction block size (4k), the layout 1212 * of the transaction header, and the number of updates per transaction. 1213 * See drbd_actlog.c:struct al_transaction_on_disk 1214 * */ 1215 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary 1216 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4 1217 1218 #if BITS_PER_LONG == 32 1219 #define LN2_BPL 5 1220 #define cpu_to_lel(A) cpu_to_le32(A) 1221 #define lel_to_cpu(A) le32_to_cpu(A) 1222 #elif BITS_PER_LONG == 64 1223 #define LN2_BPL 6 1224 #define cpu_to_lel(A) cpu_to_le64(A) 1225 #define lel_to_cpu(A) le64_to_cpu(A) 1226 #else 1227 #error "LN2 of BITS_PER_LONG unknown!" 1228 #endif 1229 1230 /* resync bitmap */ 1231 /* 16MB sized 'bitmap extent' to track syncer usage */ 1232 struct bm_extent { 1233 int rs_left; /* number of bits set (out of sync) in this extent. */ 1234 int rs_failed; /* number of failed resync requests in this extent. */ 1235 unsigned long flags; 1236 struct lc_element lce; 1237 }; 1238 1239 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */ 1240 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */ 1241 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */ 1242 1243 /* drbd_bitmap.c */ 1244 /* 1245 * We need to store one bit for a block. 1246 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap. 1247 * Bit 0 ==> local node thinks this block is binary identical on both nodes 1248 * Bit 1 ==> local node thinks this block needs to be synced. 1249 */ 1250 1251 #define SLEEP_TIME (HZ/10) 1252 1253 /* We do bitmap IO in units of 4k blocks. 1254 * We also still have a hardcoded 4k per bit relation. */ 1255 #define BM_BLOCK_SHIFT 12 /* 4k per bit */ 1256 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT) 1257 /* mostly arbitrarily set the represented size of one bitmap extent, 1258 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap 1259 * at 4k per bit resolution) */ 1260 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */ 1261 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT) 1262 1263 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12) 1264 #error "HAVE YOU FIXED drbdmeta AS WELL??" 1265 #endif 1266 1267 /* thus many _storage_ sectors are described by one bit */ 1268 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9)) 1269 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9)) 1270 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1) 1271 1272 /* bit to represented kilo byte conversion */ 1273 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10)) 1274 1275 /* in which _bitmap_ extent (resp. sector) the bit for a certain 1276 * _storage_ sector is located in */ 1277 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9)) 1278 #define BM_BIT_TO_EXT(x) ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT)) 1279 1280 /* first storage sector a bitmap extent corresponds to */ 1281 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9)) 1282 /* how much _storage_ sectors we have per bitmap extent */ 1283 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1) 1284 /* how many bits are covered by one bitmap extent (resync extent) */ 1285 #define BM_BITS_PER_EXT (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT)) 1286 1287 #define BM_BLOCKS_PER_BM_EXT_MASK (BM_BITS_PER_EXT - 1) 1288 1289 1290 /* in one sector of the bitmap, we have this many activity_log extents. */ 1291 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT)) 1292 1293 /* the extent in "PER_EXTENT" below is an activity log extent 1294 * we need that many (long words/bytes) to store the bitmap 1295 * of one AL_EXTENT_SIZE chunk of storage. 1296 * we can store the bitmap for that many AL_EXTENTS within 1297 * one sector of the _on_disk_ bitmap: 1298 * bit 0 bit 37 bit 38 bit (512*8)-1 1299 * ...|........|........|.. // ..|........| 1300 * sect. 0 `296 `304 ^(512*8*8)-1 1301 * 1302 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG ) 1303 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128 1304 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4 1305 */ 1306 1307 #define DRBD_MAX_SECTORS_32 (0xffffffffLU) 1308 /* we have a certain meta data variant that has a fixed on-disk size of 128 1309 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity 1310 * log, leaving this many sectors for the bitmap. 1311 */ 1312 1313 #define DRBD_MAX_SECTORS_FIXED_BM \ 1314 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9))) 1315 #if !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32 1316 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32 1317 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32 1318 #else 1319 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM 1320 /* 16 TB in units of sectors */ 1321 #if BITS_PER_LONG == 32 1322 /* adjust by one page worth of bitmap, 1323 * so we won't wrap around in drbd_bm_find_next_bit. 1324 * you should use 64bit OS for that much storage, anyways. */ 1325 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff) 1326 #else 1327 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */ 1328 #define DRBD_MAX_SECTORS_FLEX (1UL << 51) 1329 /* corresponds to (1UL << 38) bits right now. */ 1330 #endif 1331 #endif 1332 1333 /* Estimate max bio size as 256 * PAGE_SIZE, 1334 * so for typical PAGE_SIZE of 4k, that is (1<<20) Byte. 1335 * Since we may live in a mixed-platform cluster, 1336 * we limit us to a platform agnostic constant here for now. 1337 * A followup commit may allow even bigger BIO sizes, 1338 * once we thought that through. */ 1339 #define DRBD_MAX_BIO_SIZE (1U << 20) 1340 #if DRBD_MAX_BIO_SIZE > (BIO_MAX_PAGES << PAGE_SHIFT) 1341 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE 1342 #endif 1343 #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */ 1344 1345 #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */ 1346 #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */ 1347 1348 /* For now, don't allow more than half of what we can "activate" in one 1349 * activity log transaction to be discarded in one go. We may need to rework 1350 * drbd_al_begin_io() to allow for even larger discard ranges */ 1351 #define DRBD_MAX_BATCH_BIO_SIZE (AL_UPDATES_PER_TRANSACTION/2*AL_EXTENT_SIZE) 1352 #define DRBD_MAX_BBIO_SECTORS (DRBD_MAX_BATCH_BIO_SIZE >> 9) 1353 1354 extern int drbd_bm_init(struct drbd_device *device); 1355 extern int drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits); 1356 extern void drbd_bm_cleanup(struct drbd_device *device); 1357 extern void drbd_bm_set_all(struct drbd_device *device); 1358 extern void drbd_bm_clear_all(struct drbd_device *device); 1359 /* set/clear/test only a few bits at a time */ 1360 extern int drbd_bm_set_bits( 1361 struct drbd_device *device, unsigned long s, unsigned long e); 1362 extern int drbd_bm_clear_bits( 1363 struct drbd_device *device, unsigned long s, unsigned long e); 1364 extern int drbd_bm_count_bits( 1365 struct drbd_device *device, const unsigned long s, const unsigned long e); 1366 /* bm_set_bits variant for use while holding drbd_bm_lock, 1367 * may process the whole bitmap in one go */ 1368 extern void _drbd_bm_set_bits(struct drbd_device *device, 1369 const unsigned long s, const unsigned long e); 1370 extern int drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr); 1371 extern int drbd_bm_e_weight(struct drbd_device *device, unsigned long enr); 1372 extern int drbd_bm_read(struct drbd_device *device) __must_hold(local); 1373 extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr); 1374 extern int drbd_bm_write(struct drbd_device *device) __must_hold(local); 1375 extern void drbd_bm_reset_al_hints(struct drbd_device *device) __must_hold(local); 1376 extern int drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local); 1377 extern int drbd_bm_write_lazy(struct drbd_device *device, unsigned upper_idx) __must_hold(local); 1378 extern int drbd_bm_write_all(struct drbd_device *device) __must_hold(local); 1379 extern int drbd_bm_write_copy_pages(struct drbd_device *device) __must_hold(local); 1380 extern size_t drbd_bm_words(struct drbd_device *device); 1381 extern unsigned long drbd_bm_bits(struct drbd_device *device); 1382 extern sector_t drbd_bm_capacity(struct drbd_device *device); 1383 1384 #define DRBD_END_OF_BITMAP (~(unsigned long)0) 1385 extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); 1386 /* bm_find_next variants for use while you hold drbd_bm_lock() */ 1387 extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); 1388 extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo); 1389 extern unsigned long _drbd_bm_total_weight(struct drbd_device *device); 1390 extern unsigned long drbd_bm_total_weight(struct drbd_device *device); 1391 /* for receive_bitmap */ 1392 extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset, 1393 size_t number, unsigned long *buffer); 1394 /* for _drbd_send_bitmap */ 1395 extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset, 1396 size_t number, unsigned long *buffer); 1397 1398 extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags); 1399 extern void drbd_bm_unlock(struct drbd_device *device); 1400 /* drbd_main.c */ 1401 1402 extern struct kmem_cache *drbd_request_cache; 1403 extern struct kmem_cache *drbd_ee_cache; /* peer requests */ 1404 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */ 1405 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */ 1406 extern mempool_t drbd_request_mempool; 1407 extern mempool_t drbd_ee_mempool; 1408 1409 /* drbd's page pool, used to buffer data received from the peer, 1410 * or data requested by the peer. 1411 * 1412 * This does not have an emergency reserve. 1413 * 1414 * When allocating from this pool, it first takes pages from the pool. 1415 * Only if the pool is depleted will try to allocate from the system. 1416 * 1417 * The assumption is that pages taken from this pool will be processed, 1418 * and given back, "quickly", and then can be recycled, so we can avoid 1419 * frequent calls to alloc_page(), and still will be able to make progress even 1420 * under memory pressure. 1421 */ 1422 extern struct page *drbd_pp_pool; 1423 extern spinlock_t drbd_pp_lock; 1424 extern int drbd_pp_vacant; 1425 extern wait_queue_head_t drbd_pp_wait; 1426 1427 /* We also need a standard (emergency-reserve backed) page pool 1428 * for meta data IO (activity log, bitmap). 1429 * We can keep it global, as long as it is used as "N pages at a time". 1430 * 128 should be plenty, currently we probably can get away with as few as 1. 1431 */ 1432 #define DRBD_MIN_POOL_PAGES 128 1433 extern mempool_t drbd_md_io_page_pool; 1434 1435 /* We also need to make sure we get a bio 1436 * when we need it for housekeeping purposes */ 1437 extern struct bio_set drbd_md_io_bio_set; 1438 /* to allocate from that set */ 1439 extern struct bio *bio_alloc_drbd(gfp_t gfp_mask); 1440 1441 /* And a bio_set for cloning */ 1442 extern struct bio_set drbd_io_bio_set; 1443 1444 extern struct mutex resources_mutex; 1445 1446 extern int conn_lowest_minor(struct drbd_connection *connection); 1447 extern enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor); 1448 extern void drbd_destroy_device(struct kref *kref); 1449 extern void drbd_delete_device(struct drbd_device *device); 1450 1451 extern struct drbd_resource *drbd_create_resource(const char *name); 1452 extern void drbd_free_resource(struct drbd_resource *resource); 1453 1454 extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts); 1455 extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts); 1456 extern void drbd_destroy_connection(struct kref *kref); 1457 extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len, 1458 void *peer_addr, int peer_addr_len); 1459 extern struct drbd_resource *drbd_find_resource(const char *name); 1460 extern void drbd_destroy_resource(struct kref *kref); 1461 extern void conn_free_crypto(struct drbd_connection *connection); 1462 1463 /* drbd_req */ 1464 extern void do_submit(struct work_struct *ws); 1465 extern void __drbd_make_request(struct drbd_device *, struct bio *, unsigned long); 1466 extern blk_qc_t drbd_make_request(struct request_queue *q, struct bio *bio); 1467 extern int drbd_read_remote(struct drbd_device *device, struct drbd_request *req); 1468 extern int is_valid_ar_handle(struct drbd_request *, sector_t); 1469 1470 1471 /* drbd_nl.c */ 1472 1473 extern struct mutex notification_mutex; 1474 1475 extern void drbd_suspend_io(struct drbd_device *device); 1476 extern void drbd_resume_io(struct drbd_device *device); 1477 extern char *ppsize(char *buf, unsigned long long size); 1478 extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int); 1479 enum determine_dev_size { 1480 DS_ERROR_SHRINK = -3, 1481 DS_ERROR_SPACE_MD = -2, 1482 DS_ERROR = -1, 1483 DS_UNCHANGED = 0, 1484 DS_SHRUNK = 1, 1485 DS_GREW = 2, 1486 DS_GREW_FROM_ZERO = 3, 1487 }; 1488 extern enum determine_dev_size 1489 drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local); 1490 extern void resync_after_online_grow(struct drbd_device *); 1491 extern void drbd_reconsider_queue_parameters(struct drbd_device *device, 1492 struct drbd_backing_dev *bdev, struct o_qlim *o); 1493 extern enum drbd_state_rv drbd_set_role(struct drbd_device *device, 1494 enum drbd_role new_role, 1495 int force); 1496 extern bool conn_try_outdate_peer(struct drbd_connection *connection); 1497 extern void conn_try_outdate_peer_async(struct drbd_connection *connection); 1498 extern enum drbd_peer_state conn_khelper(struct drbd_connection *connection, char *cmd); 1499 extern int drbd_khelper(struct drbd_device *device, char *cmd); 1500 1501 /* drbd_worker.c */ 1502 /* bi_end_io handlers */ 1503 extern void drbd_md_endio(struct bio *bio); 1504 extern void drbd_peer_request_endio(struct bio *bio); 1505 extern void drbd_request_endio(struct bio *bio); 1506 extern int drbd_worker(struct drbd_thread *thi); 1507 enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor); 1508 void drbd_resync_after_changed(struct drbd_device *device); 1509 extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side); 1510 extern void resume_next_sg(struct drbd_device *device); 1511 extern void suspend_other_sg(struct drbd_device *device); 1512 extern int drbd_resync_finished(struct drbd_device *device); 1513 /* maybe rather drbd_main.c ? */ 1514 extern void *drbd_md_get_buffer(struct drbd_device *device, const char *intent); 1515 extern void drbd_md_put_buffer(struct drbd_device *device); 1516 extern int drbd_md_sync_page_io(struct drbd_device *device, 1517 struct drbd_backing_dev *bdev, sector_t sector, int op); 1518 extern void drbd_ov_out_of_sync_found(struct drbd_device *, sector_t, int); 1519 extern void wait_until_done_or_force_detached(struct drbd_device *device, 1520 struct drbd_backing_dev *bdev, unsigned int *done); 1521 extern void drbd_rs_controller_reset(struct drbd_device *device); 1522 1523 static inline void ov_out_of_sync_print(struct drbd_device *device) 1524 { 1525 if (device->ov_last_oos_size) { 1526 drbd_err(device, "Out of sync: start=%llu, size=%lu (sectors)\n", 1527 (unsigned long long)device->ov_last_oos_start, 1528 (unsigned long)device->ov_last_oos_size); 1529 } 1530 device->ov_last_oos_size = 0; 1531 } 1532 1533 1534 extern void drbd_csum_bio(struct crypto_ahash *, struct bio *, void *); 1535 extern void drbd_csum_ee(struct crypto_ahash *, struct drbd_peer_request *, void *); 1536 /* worker callbacks */ 1537 extern int w_e_end_data_req(struct drbd_work *, int); 1538 extern int w_e_end_rsdata_req(struct drbd_work *, int); 1539 extern int w_e_end_csum_rs_req(struct drbd_work *, int); 1540 extern int w_e_end_ov_reply(struct drbd_work *, int); 1541 extern int w_e_end_ov_req(struct drbd_work *, int); 1542 extern int w_ov_finished(struct drbd_work *, int); 1543 extern int w_resync_timer(struct drbd_work *, int); 1544 extern int w_send_write_hint(struct drbd_work *, int); 1545 extern int w_send_dblock(struct drbd_work *, int); 1546 extern int w_send_read_req(struct drbd_work *, int); 1547 extern int w_e_reissue(struct drbd_work *, int); 1548 extern int w_restart_disk_io(struct drbd_work *, int); 1549 extern int w_send_out_of_sync(struct drbd_work *, int); 1550 extern int w_start_resync(struct drbd_work *, int); 1551 1552 extern void resync_timer_fn(struct timer_list *t); 1553 extern void start_resync_timer_fn(struct timer_list *t); 1554 1555 extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req); 1556 1557 /* drbd_receiver.c */ 1558 extern int drbd_receiver(struct drbd_thread *thi); 1559 extern int drbd_ack_receiver(struct drbd_thread *thi); 1560 extern void drbd_send_ping_wf(struct work_struct *ws); 1561 extern void drbd_send_acks_wf(struct work_struct *ws); 1562 extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device); 1563 extern bool drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector, 1564 bool throttle_if_app_is_waiting); 1565 extern int drbd_submit_peer_request(struct drbd_device *, 1566 struct drbd_peer_request *, const unsigned, 1567 const unsigned, const int); 1568 extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *); 1569 extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64, 1570 sector_t, unsigned int, 1571 unsigned int, 1572 gfp_t) __must_hold(local); 1573 extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *, 1574 int); 1575 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0) 1576 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1) 1577 extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool); 1578 extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled); 1579 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed); 1580 extern int drbd_connected(struct drbd_peer_device *); 1581 1582 static inline void drbd_tcp_cork(struct socket *sock) 1583 { 1584 int val = 1; 1585 (void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK, 1586 (char*)&val, sizeof(val)); 1587 } 1588 1589 static inline void drbd_tcp_uncork(struct socket *sock) 1590 { 1591 int val = 0; 1592 (void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK, 1593 (char*)&val, sizeof(val)); 1594 } 1595 1596 static inline void drbd_tcp_nodelay(struct socket *sock) 1597 { 1598 int val = 1; 1599 (void) kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, 1600 (char*)&val, sizeof(val)); 1601 } 1602 1603 static inline void drbd_tcp_quickack(struct socket *sock) 1604 { 1605 int val = 2; 1606 (void) kernel_setsockopt(sock, SOL_TCP, TCP_QUICKACK, 1607 (char*)&val, sizeof(val)); 1608 } 1609 1610 /* sets the number of 512 byte sectors of our virtual device */ 1611 static inline void drbd_set_my_capacity(struct drbd_device *device, 1612 sector_t size) 1613 { 1614 /* set_capacity(device->this_bdev->bd_disk, size); */ 1615 set_capacity(device->vdisk, size); 1616 device->this_bdev->bd_inode->i_size = (loff_t)size << 9; 1617 } 1618 1619 /* 1620 * used to submit our private bio 1621 */ 1622 static inline void drbd_generic_make_request(struct drbd_device *device, 1623 int fault_type, struct bio *bio) 1624 { 1625 __release(local); 1626 if (!bio->bi_disk) { 1627 drbd_err(device, "drbd_generic_make_request: bio->bi_disk == NULL\n"); 1628 bio->bi_status = BLK_STS_IOERR; 1629 bio_endio(bio); 1630 return; 1631 } 1632 1633 if (drbd_insert_fault(device, fault_type)) 1634 bio_io_error(bio); 1635 else 1636 generic_make_request(bio); 1637 } 1638 1639 void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev, 1640 enum write_ordering_e wo); 1641 1642 /* drbd_proc.c */ 1643 extern struct proc_dir_entry *drbd_proc; 1644 int drbd_seq_show(struct seq_file *seq, void *v); 1645 1646 /* drbd_actlog.c */ 1647 extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i); 1648 extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i); 1649 extern void drbd_al_begin_io_commit(struct drbd_device *device); 1650 extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i); 1651 extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i); 1652 extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i); 1653 extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector); 1654 extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector); 1655 extern int drbd_try_rs_begin_io(struct drbd_device *device, sector_t sector); 1656 extern void drbd_rs_cancel_all(struct drbd_device *device); 1657 extern int drbd_rs_del_all(struct drbd_device *device); 1658 extern void drbd_rs_failed_io(struct drbd_device *device, 1659 sector_t sector, int size); 1660 extern void drbd_advance_rs_marks(struct drbd_device *device, unsigned long still_to_go); 1661 1662 enum update_sync_bits_mode { RECORD_RS_FAILED, SET_OUT_OF_SYNC, SET_IN_SYNC }; 1663 extern int __drbd_change_sync(struct drbd_device *device, sector_t sector, int size, 1664 enum update_sync_bits_mode mode); 1665 #define drbd_set_in_sync(device, sector, size) \ 1666 __drbd_change_sync(device, sector, size, SET_IN_SYNC) 1667 #define drbd_set_out_of_sync(device, sector, size) \ 1668 __drbd_change_sync(device, sector, size, SET_OUT_OF_SYNC) 1669 #define drbd_rs_failed_io(device, sector, size) \ 1670 __drbd_change_sync(device, sector, size, RECORD_RS_FAILED) 1671 extern void drbd_al_shrink(struct drbd_device *device); 1672 extern int drbd_al_initialize(struct drbd_device *, void *); 1673 1674 /* drbd_nl.c */ 1675 /* state info broadcast */ 1676 struct sib_info { 1677 enum drbd_state_info_bcast_reason sib_reason; 1678 union { 1679 struct { 1680 char *helper_name; 1681 unsigned helper_exit_code; 1682 }; 1683 struct { 1684 union drbd_state os; 1685 union drbd_state ns; 1686 }; 1687 }; 1688 }; 1689 void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib); 1690 1691 extern void notify_resource_state(struct sk_buff *, 1692 unsigned int, 1693 struct drbd_resource *, 1694 struct resource_info *, 1695 enum drbd_notification_type); 1696 extern void notify_device_state(struct sk_buff *, 1697 unsigned int, 1698 struct drbd_device *, 1699 struct device_info *, 1700 enum drbd_notification_type); 1701 extern void notify_connection_state(struct sk_buff *, 1702 unsigned int, 1703 struct drbd_connection *, 1704 struct connection_info *, 1705 enum drbd_notification_type); 1706 extern void notify_peer_device_state(struct sk_buff *, 1707 unsigned int, 1708 struct drbd_peer_device *, 1709 struct peer_device_info *, 1710 enum drbd_notification_type); 1711 extern void notify_helper(enum drbd_notification_type, struct drbd_device *, 1712 struct drbd_connection *, const char *, int); 1713 1714 /* 1715 * inline helper functions 1716 *************************/ 1717 1718 /* see also page_chain_add and friends in drbd_receiver.c */ 1719 static inline struct page *page_chain_next(struct page *page) 1720 { 1721 return (struct page *)page_private(page); 1722 } 1723 #define page_chain_for_each(page) \ 1724 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \ 1725 page = page_chain_next(page)) 1726 #define page_chain_for_each_safe(page, n) \ 1727 for (; page && ({ n = page_chain_next(page); 1; }); page = n) 1728 1729 1730 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req) 1731 { 1732 struct page *page = peer_req->pages; 1733 page_chain_for_each(page) { 1734 if (page_count(page) > 1) 1735 return 1; 1736 } 1737 return 0; 1738 } 1739 1740 static inline union drbd_state drbd_read_state(struct drbd_device *device) 1741 { 1742 struct drbd_resource *resource = device->resource; 1743 union drbd_state rv; 1744 1745 rv.i = device->state.i; 1746 rv.susp = resource->susp; 1747 rv.susp_nod = resource->susp_nod; 1748 rv.susp_fen = resource->susp_fen; 1749 1750 return rv; 1751 } 1752 1753 enum drbd_force_detach_flags { 1754 DRBD_READ_ERROR, 1755 DRBD_WRITE_ERROR, 1756 DRBD_META_IO_ERROR, 1757 DRBD_FORCE_DETACH, 1758 }; 1759 1760 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__) 1761 static inline void __drbd_chk_io_error_(struct drbd_device *device, 1762 enum drbd_force_detach_flags df, 1763 const char *where) 1764 { 1765 enum drbd_io_error_p ep; 1766 1767 rcu_read_lock(); 1768 ep = rcu_dereference(device->ldev->disk_conf)->on_io_error; 1769 rcu_read_unlock(); 1770 switch (ep) { 1771 case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */ 1772 if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) { 1773 if (__ratelimit(&drbd_ratelimit_state)) 1774 drbd_err(device, "Local IO failed in %s.\n", where); 1775 if (device->state.disk > D_INCONSISTENT) 1776 _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL); 1777 break; 1778 } 1779 /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */ 1780 case EP_DETACH: 1781 case EP_CALL_HELPER: 1782 /* Remember whether we saw a READ or WRITE error. 1783 * 1784 * Recovery of the affected area for WRITE failure is covered 1785 * by the activity log. 1786 * READ errors may fall outside that area though. Certain READ 1787 * errors can be "healed" by writing good data to the affected 1788 * blocks, which triggers block re-allocation in lower layers. 1789 * 1790 * If we can not write the bitmap after a READ error, 1791 * we may need to trigger a full sync (see w_go_diskless()). 1792 * 1793 * Force-detach is not really an IO error, but rather a 1794 * desperate measure to try to deal with a completely 1795 * unresponsive lower level IO stack. 1796 * Still it should be treated as a WRITE error. 1797 * 1798 * Meta IO error is always WRITE error: 1799 * we read meta data only once during attach, 1800 * which will fail in case of errors. 1801 */ 1802 set_bit(WAS_IO_ERROR, &device->flags); 1803 if (df == DRBD_READ_ERROR) 1804 set_bit(WAS_READ_ERROR, &device->flags); 1805 if (df == DRBD_FORCE_DETACH) 1806 set_bit(FORCE_DETACH, &device->flags); 1807 if (device->state.disk > D_FAILED) { 1808 _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL); 1809 drbd_err(device, 1810 "Local IO failed in %s. Detaching...\n", where); 1811 } 1812 break; 1813 } 1814 } 1815 1816 /** 1817 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers 1818 * @device: DRBD device. 1819 * @error: Error code passed to the IO completion callback 1820 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data 1821 * 1822 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED) 1823 */ 1824 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__) 1825 static inline void drbd_chk_io_error_(struct drbd_device *device, 1826 int error, enum drbd_force_detach_flags forcedetach, const char *where) 1827 { 1828 if (error) { 1829 unsigned long flags; 1830 spin_lock_irqsave(&device->resource->req_lock, flags); 1831 __drbd_chk_io_error_(device, forcedetach, where); 1832 spin_unlock_irqrestore(&device->resource->req_lock, flags); 1833 } 1834 } 1835 1836 1837 /** 1838 * drbd_md_first_sector() - Returns the first sector number of the meta data area 1839 * @bdev: Meta data block device. 1840 * 1841 * BTW, for internal meta data, this happens to be the maximum capacity 1842 * we could agree upon with our peer node. 1843 */ 1844 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev) 1845 { 1846 switch (bdev->md.meta_dev_idx) { 1847 case DRBD_MD_INDEX_INTERNAL: 1848 case DRBD_MD_INDEX_FLEX_INT: 1849 return bdev->md.md_offset + bdev->md.bm_offset; 1850 case DRBD_MD_INDEX_FLEX_EXT: 1851 default: 1852 return bdev->md.md_offset; 1853 } 1854 } 1855 1856 /** 1857 * drbd_md_last_sector() - Return the last sector number of the meta data area 1858 * @bdev: Meta data block device. 1859 */ 1860 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev) 1861 { 1862 switch (bdev->md.meta_dev_idx) { 1863 case DRBD_MD_INDEX_INTERNAL: 1864 case DRBD_MD_INDEX_FLEX_INT: 1865 return bdev->md.md_offset + MD_4kB_SECT -1; 1866 case DRBD_MD_INDEX_FLEX_EXT: 1867 default: 1868 return bdev->md.md_offset + bdev->md.md_size_sect -1; 1869 } 1870 } 1871 1872 /* Returns the number of 512 byte sectors of the device */ 1873 static inline sector_t drbd_get_capacity(struct block_device *bdev) 1874 { 1875 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */ 1876 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0; 1877 } 1878 1879 /** 1880 * drbd_get_max_capacity() - Returns the capacity we announce to out peer 1881 * @bdev: Meta data block device. 1882 * 1883 * returns the capacity we announce to out peer. we clip ourselves at the 1884 * various MAX_SECTORS, because if we don't, current implementation will 1885 * oops sooner or later 1886 */ 1887 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev) 1888 { 1889 sector_t s; 1890 1891 switch (bdev->md.meta_dev_idx) { 1892 case DRBD_MD_INDEX_INTERNAL: 1893 case DRBD_MD_INDEX_FLEX_INT: 1894 s = drbd_get_capacity(bdev->backing_bdev) 1895 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX, 1896 drbd_md_first_sector(bdev)) 1897 : 0; 1898 break; 1899 case DRBD_MD_INDEX_FLEX_EXT: 1900 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX, 1901 drbd_get_capacity(bdev->backing_bdev)); 1902 /* clip at maximum size the meta device can support */ 1903 s = min_t(sector_t, s, 1904 BM_EXT_TO_SECT(bdev->md.md_size_sect 1905 - bdev->md.bm_offset)); 1906 break; 1907 default: 1908 s = min_t(sector_t, DRBD_MAX_SECTORS, 1909 drbd_get_capacity(bdev->backing_bdev)); 1910 } 1911 return s; 1912 } 1913 1914 /** 1915 * drbd_md_ss() - Return the sector number of our meta data super block 1916 * @bdev: Meta data block device. 1917 */ 1918 static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev) 1919 { 1920 const int meta_dev_idx = bdev->md.meta_dev_idx; 1921 1922 if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT) 1923 return 0; 1924 1925 /* Since drbd08, internal meta data is always "flexible". 1926 * position: last 4k aligned block of 4k size */ 1927 if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL || 1928 meta_dev_idx == DRBD_MD_INDEX_FLEX_INT) 1929 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8; 1930 1931 /* external, some index; this is the old fixed size layout */ 1932 return MD_128MB_SECT * bdev->md.meta_dev_idx; 1933 } 1934 1935 static inline void 1936 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w) 1937 { 1938 unsigned long flags; 1939 spin_lock_irqsave(&q->q_lock, flags); 1940 list_add_tail(&w->list, &q->q); 1941 spin_unlock_irqrestore(&q->q_lock, flags); 1942 wake_up(&q->q_wait); 1943 } 1944 1945 static inline void 1946 drbd_queue_work_if_unqueued(struct drbd_work_queue *q, struct drbd_work *w) 1947 { 1948 unsigned long flags; 1949 spin_lock_irqsave(&q->q_lock, flags); 1950 if (list_empty_careful(&w->list)) 1951 list_add_tail(&w->list, &q->q); 1952 spin_unlock_irqrestore(&q->q_lock, flags); 1953 wake_up(&q->q_wait); 1954 } 1955 1956 static inline void 1957 drbd_device_post_work(struct drbd_device *device, int work_bit) 1958 { 1959 if (!test_and_set_bit(work_bit, &device->flags)) { 1960 struct drbd_connection *connection = 1961 first_peer_device(device)->connection; 1962 struct drbd_work_queue *q = &connection->sender_work; 1963 if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags)) 1964 wake_up(&q->q_wait); 1965 } 1966 } 1967 1968 extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue); 1969 1970 /* To get the ack_receiver out of the blocking network stack, 1971 * so it can change its sk_rcvtimeo from idle- to ping-timeout, 1972 * and send a ping, we need to send a signal. 1973 * Which signal we send is irrelevant. */ 1974 static inline void wake_ack_receiver(struct drbd_connection *connection) 1975 { 1976 struct task_struct *task = connection->ack_receiver.task; 1977 if (task && get_t_state(&connection->ack_receiver) == RUNNING) 1978 force_sig(SIGXCPU, task); 1979 } 1980 1981 static inline void request_ping(struct drbd_connection *connection) 1982 { 1983 set_bit(SEND_PING, &connection->flags); 1984 wake_ack_receiver(connection); 1985 } 1986 1987 extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *); 1988 extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *); 1989 extern int conn_send_command(struct drbd_connection *, struct drbd_socket *, 1990 enum drbd_packet, unsigned int, void *, 1991 unsigned int); 1992 extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *, 1993 enum drbd_packet, unsigned int, void *, 1994 unsigned int); 1995 1996 extern int drbd_send_ping(struct drbd_connection *connection); 1997 extern int drbd_send_ping_ack(struct drbd_connection *connection); 1998 extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state); 1999 extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state); 2000 2001 static inline void drbd_thread_stop(struct drbd_thread *thi) 2002 { 2003 _drbd_thread_stop(thi, false, true); 2004 } 2005 2006 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi) 2007 { 2008 _drbd_thread_stop(thi, false, false); 2009 } 2010 2011 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi) 2012 { 2013 _drbd_thread_stop(thi, true, false); 2014 } 2015 2016 /* counts how many answer packets packets we expect from our peer, 2017 * for either explicit application requests, 2018 * or implicit barrier packets as necessary. 2019 * increased: 2020 * w_send_barrier 2021 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ); 2022 * it is much easier and equally valid to count what we queue for the 2023 * worker, even before it actually was queued or send. 2024 * (drbd_make_request_common; recovery path on read io-error) 2025 * decreased: 2026 * got_BarrierAck (respective tl_clear, tl_clear_barrier) 2027 * _req_mod(req, DATA_RECEIVED) 2028 * [from receive_DataReply] 2029 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED) 2030 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)] 2031 * for some reason it is NOT decreased in got_NegAck, 2032 * but in the resulting cleanup code from report_params. 2033 * we should try to remember the reason for that... 2034 * _req_mod(req, SEND_FAILED or SEND_CANCELED) 2035 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING) 2036 * [from tl_clear_barrier] 2037 */ 2038 static inline void inc_ap_pending(struct drbd_device *device) 2039 { 2040 atomic_inc(&device->ap_pending_cnt); 2041 } 2042 2043 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \ 2044 if (atomic_read(&device->which) < 0) \ 2045 drbd_err(device, "in %s:%d: " #which " = %d < 0 !\n", \ 2046 func, line, \ 2047 atomic_read(&device->which)) 2048 2049 #define dec_ap_pending(device) _dec_ap_pending(device, __func__, __LINE__) 2050 static inline void _dec_ap_pending(struct drbd_device *device, const char *func, int line) 2051 { 2052 if (atomic_dec_and_test(&device->ap_pending_cnt)) 2053 wake_up(&device->misc_wait); 2054 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line); 2055 } 2056 2057 /* counts how many resync-related answers we still expect from the peer 2058 * increase decrease 2059 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY) 2060 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER) 2061 * (or P_NEG_ACK with ID_SYNCER) 2062 */ 2063 static inline void inc_rs_pending(struct drbd_device *device) 2064 { 2065 atomic_inc(&device->rs_pending_cnt); 2066 } 2067 2068 #define dec_rs_pending(device) _dec_rs_pending(device, __func__, __LINE__) 2069 static inline void _dec_rs_pending(struct drbd_device *device, const char *func, int line) 2070 { 2071 atomic_dec(&device->rs_pending_cnt); 2072 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line); 2073 } 2074 2075 /* counts how many answers we still need to send to the peer. 2076 * increased on 2077 * receive_Data unless protocol A; 2078 * we need to send a P_RECV_ACK (proto B) 2079 * or P_WRITE_ACK (proto C) 2080 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK 2081 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA 2082 * receive_Barrier_* we need to send a P_BARRIER_ACK 2083 */ 2084 static inline void inc_unacked(struct drbd_device *device) 2085 { 2086 atomic_inc(&device->unacked_cnt); 2087 } 2088 2089 #define dec_unacked(device) _dec_unacked(device, __func__, __LINE__) 2090 static inline void _dec_unacked(struct drbd_device *device, const char *func, int line) 2091 { 2092 atomic_dec(&device->unacked_cnt); 2093 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line); 2094 } 2095 2096 #define sub_unacked(device, n) _sub_unacked(device, n, __func__, __LINE__) 2097 static inline void _sub_unacked(struct drbd_device *device, int n, const char *func, int line) 2098 { 2099 atomic_sub(n, &device->unacked_cnt); 2100 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line); 2101 } 2102 2103 static inline bool is_sync_target_state(enum drbd_conns connection_state) 2104 { 2105 return connection_state == C_SYNC_TARGET || 2106 connection_state == C_PAUSED_SYNC_T; 2107 } 2108 2109 static inline bool is_sync_source_state(enum drbd_conns connection_state) 2110 { 2111 return connection_state == C_SYNC_SOURCE || 2112 connection_state == C_PAUSED_SYNC_S; 2113 } 2114 2115 static inline bool is_sync_state(enum drbd_conns connection_state) 2116 { 2117 return is_sync_source_state(connection_state) || 2118 is_sync_target_state(connection_state); 2119 } 2120 2121 /** 2122 * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev 2123 * @_device: DRBD device. 2124 * @_min_state: Minimum device state required for success. 2125 * 2126 * You have to call put_ldev() when finished working with device->ldev. 2127 */ 2128 #define get_ldev_if_state(_device, _min_state) \ 2129 (_get_ldev_if_state((_device), (_min_state)) ? \ 2130 ({ __acquire(x); true; }) : false) 2131 #define get_ldev(_device) get_ldev_if_state(_device, D_INCONSISTENT) 2132 2133 static inline void put_ldev(struct drbd_device *device) 2134 { 2135 enum drbd_disk_state disk_state = device->state.disk; 2136 /* We must check the state *before* the atomic_dec becomes visible, 2137 * or we have a theoretical race where someone hitting zero, 2138 * while state still D_FAILED, will then see D_DISKLESS in the 2139 * condition below and calling into destroy, where he must not, yet. */ 2140 int i = atomic_dec_return(&device->local_cnt); 2141 2142 /* This may be called from some endio handler, 2143 * so we must not sleep here. */ 2144 2145 __release(local); 2146 D_ASSERT(device, i >= 0); 2147 if (i == 0) { 2148 if (disk_state == D_DISKLESS) 2149 /* even internal references gone, safe to destroy */ 2150 drbd_device_post_work(device, DESTROY_DISK); 2151 if (disk_state == D_FAILED) 2152 /* all application IO references gone. */ 2153 if (!test_and_set_bit(GOING_DISKLESS, &device->flags)) 2154 drbd_device_post_work(device, GO_DISKLESS); 2155 wake_up(&device->misc_wait); 2156 } 2157 } 2158 2159 #ifndef __CHECKER__ 2160 static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins) 2161 { 2162 int io_allowed; 2163 2164 /* never get a reference while D_DISKLESS */ 2165 if (device->state.disk == D_DISKLESS) 2166 return 0; 2167 2168 atomic_inc(&device->local_cnt); 2169 io_allowed = (device->state.disk >= mins); 2170 if (!io_allowed) 2171 put_ldev(device); 2172 return io_allowed; 2173 } 2174 #else 2175 extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins); 2176 #endif 2177 2178 /* this throttles on-the-fly application requests 2179 * according to max_buffers settings; 2180 * maybe re-implement using semaphores? */ 2181 static inline int drbd_get_max_buffers(struct drbd_device *device) 2182 { 2183 struct net_conf *nc; 2184 int mxb; 2185 2186 rcu_read_lock(); 2187 nc = rcu_dereference(first_peer_device(device)->connection->net_conf); 2188 mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */ 2189 rcu_read_unlock(); 2190 2191 return mxb; 2192 } 2193 2194 static inline int drbd_state_is_stable(struct drbd_device *device) 2195 { 2196 union drbd_dev_state s = device->state; 2197 2198 /* DO NOT add a default clause, we want the compiler to warn us 2199 * for any newly introduced state we may have forgotten to add here */ 2200 2201 switch ((enum drbd_conns)s.conn) { 2202 /* new io only accepted when there is no connection, ... */ 2203 case C_STANDALONE: 2204 case C_WF_CONNECTION: 2205 /* ... or there is a well established connection. */ 2206 case C_CONNECTED: 2207 case C_SYNC_SOURCE: 2208 case C_SYNC_TARGET: 2209 case C_VERIFY_S: 2210 case C_VERIFY_T: 2211 case C_PAUSED_SYNC_S: 2212 case C_PAUSED_SYNC_T: 2213 case C_AHEAD: 2214 case C_BEHIND: 2215 /* transitional states, IO allowed */ 2216 case C_DISCONNECTING: 2217 case C_UNCONNECTED: 2218 case C_TIMEOUT: 2219 case C_BROKEN_PIPE: 2220 case C_NETWORK_FAILURE: 2221 case C_PROTOCOL_ERROR: 2222 case C_TEAR_DOWN: 2223 case C_WF_REPORT_PARAMS: 2224 case C_STARTING_SYNC_S: 2225 case C_STARTING_SYNC_T: 2226 break; 2227 2228 /* Allow IO in BM exchange states with new protocols */ 2229 case C_WF_BITMAP_S: 2230 if (first_peer_device(device)->connection->agreed_pro_version < 96) 2231 return 0; 2232 break; 2233 2234 /* no new io accepted in these states */ 2235 case C_WF_BITMAP_T: 2236 case C_WF_SYNC_UUID: 2237 case C_MASK: 2238 /* not "stable" */ 2239 return 0; 2240 } 2241 2242 switch ((enum drbd_disk_state)s.disk) { 2243 case D_DISKLESS: 2244 case D_INCONSISTENT: 2245 case D_OUTDATED: 2246 case D_CONSISTENT: 2247 case D_UP_TO_DATE: 2248 case D_FAILED: 2249 /* disk state is stable as well. */ 2250 break; 2251 2252 /* no new io accepted during transitional states */ 2253 case D_ATTACHING: 2254 case D_NEGOTIATING: 2255 case D_UNKNOWN: 2256 case D_MASK: 2257 /* not "stable" */ 2258 return 0; 2259 } 2260 2261 return 1; 2262 } 2263 2264 static inline int drbd_suspended(struct drbd_device *device) 2265 { 2266 struct drbd_resource *resource = device->resource; 2267 2268 return resource->susp || resource->susp_fen || resource->susp_nod; 2269 } 2270 2271 static inline bool may_inc_ap_bio(struct drbd_device *device) 2272 { 2273 int mxb = drbd_get_max_buffers(device); 2274 2275 if (drbd_suspended(device)) 2276 return false; 2277 if (atomic_read(&device->suspend_cnt)) 2278 return false; 2279 2280 /* to avoid potential deadlock or bitmap corruption, 2281 * in various places, we only allow new application io 2282 * to start during "stable" states. */ 2283 2284 /* no new io accepted when attaching or detaching the disk */ 2285 if (!drbd_state_is_stable(device)) 2286 return false; 2287 2288 /* since some older kernels don't have atomic_add_unless, 2289 * and we are within the spinlock anyways, we have this workaround. */ 2290 if (atomic_read(&device->ap_bio_cnt) > mxb) 2291 return false; 2292 if (test_bit(BITMAP_IO, &device->flags)) 2293 return false; 2294 return true; 2295 } 2296 2297 static inline bool inc_ap_bio_cond(struct drbd_device *device) 2298 { 2299 bool rv = false; 2300 2301 spin_lock_irq(&device->resource->req_lock); 2302 rv = may_inc_ap_bio(device); 2303 if (rv) 2304 atomic_inc(&device->ap_bio_cnt); 2305 spin_unlock_irq(&device->resource->req_lock); 2306 2307 return rv; 2308 } 2309 2310 static inline void inc_ap_bio(struct drbd_device *device) 2311 { 2312 /* we wait here 2313 * as long as the device is suspended 2314 * until the bitmap is no longer on the fly during connection 2315 * handshake as long as we would exceed the max_buffer limit. 2316 * 2317 * to avoid races with the reconnect code, 2318 * we need to atomic_inc within the spinlock. */ 2319 2320 wait_event(device->misc_wait, inc_ap_bio_cond(device)); 2321 } 2322 2323 static inline void dec_ap_bio(struct drbd_device *device) 2324 { 2325 int mxb = drbd_get_max_buffers(device); 2326 int ap_bio = atomic_dec_return(&device->ap_bio_cnt); 2327 2328 D_ASSERT(device, ap_bio >= 0); 2329 2330 if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) { 2331 if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags)) 2332 drbd_queue_work(&first_peer_device(device)-> 2333 connection->sender_work, 2334 &device->bm_io_work.w); 2335 } 2336 2337 /* this currently does wake_up for every dec_ap_bio! 2338 * maybe rather introduce some type of hysteresis? 2339 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */ 2340 if (ap_bio < mxb) 2341 wake_up(&device->misc_wait); 2342 } 2343 2344 static inline bool verify_can_do_stop_sector(struct drbd_device *device) 2345 { 2346 return first_peer_device(device)->connection->agreed_pro_version >= 97 && 2347 first_peer_device(device)->connection->agreed_pro_version != 100; 2348 } 2349 2350 static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val) 2351 { 2352 int changed = device->ed_uuid != val; 2353 device->ed_uuid = val; 2354 return changed; 2355 } 2356 2357 static inline int drbd_queue_order_type(struct drbd_device *device) 2358 { 2359 /* sorry, we currently have no working implementation 2360 * of distributed TCQ stuff */ 2361 #ifndef QUEUE_ORDERED_NONE 2362 #define QUEUE_ORDERED_NONE 0 2363 #endif 2364 return QUEUE_ORDERED_NONE; 2365 } 2366 2367 static inline struct drbd_connection *first_connection(struct drbd_resource *resource) 2368 { 2369 return list_first_entry_or_null(&resource->connections, 2370 struct drbd_connection, connections); 2371 } 2372 2373 #endif 2374