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_OP_DISCARD? */ 433 __EE_TRIM, 434 /* explicit zero-out requested, or 435 * our lower level cannot handle trim, 436 * and we want to fall back to zeroout instead */ 437 __EE_ZEROOUT, 438 439 /* In case a barrier failed, 440 * we need to resubmit without the barrier flag. */ 441 __EE_RESUBMITTED, 442 443 /* we may have several bios per peer request. 444 * if any of those fail, we set this flag atomically 445 * from the endio callback */ 446 __EE_WAS_ERROR, 447 448 /* This ee has a pointer to a digest instead of a block id */ 449 __EE_HAS_DIGEST, 450 451 /* Conflicting local requests need to be restarted after this request */ 452 __EE_RESTART_REQUESTS, 453 454 /* The peer wants a write ACK for this (wire proto C) */ 455 __EE_SEND_WRITE_ACK, 456 457 /* Is set when net_conf had two_primaries set while creating this peer_req */ 458 __EE_IN_INTERVAL_TREE, 459 460 /* for debugfs: */ 461 /* has this been submitted, or does it still wait for something else? */ 462 __EE_SUBMITTED, 463 464 /* this is/was a write request */ 465 __EE_WRITE, 466 467 /* this is/was a write same request */ 468 __EE_WRITE_SAME, 469 470 /* this originates from application on peer 471 * (not some resync or verify or other DRBD internal request) */ 472 __EE_APPLICATION, 473 474 /* If it contains only 0 bytes, send back P_RS_DEALLOCATED */ 475 __EE_RS_THIN_REQ, 476 }; 477 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO) 478 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC) 479 #define EE_TRIM (1<<__EE_TRIM) 480 #define EE_ZEROOUT (1<<__EE_ZEROOUT) 481 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED) 482 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR) 483 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST) 484 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS) 485 #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK) 486 #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE) 487 #define EE_SUBMITTED (1<<__EE_SUBMITTED) 488 #define EE_WRITE (1<<__EE_WRITE) 489 #define EE_WRITE_SAME (1<<__EE_WRITE_SAME) 490 #define EE_APPLICATION (1<<__EE_APPLICATION) 491 #define EE_RS_THIN_REQ (1<<__EE_RS_THIN_REQ) 492 493 /* flag bits per device */ 494 enum { 495 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */ 496 MD_DIRTY, /* current uuids and flags not yet on disk */ 497 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */ 498 CL_ST_CHG_SUCCESS, 499 CL_ST_CHG_FAIL, 500 CRASHED_PRIMARY, /* This node was a crashed primary. 501 * Gets cleared when the state.conn 502 * goes into C_CONNECTED state. */ 503 CONSIDER_RESYNC, 504 505 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */ 506 507 BITMAP_IO, /* suspend application io; 508 once no more io in flight, start bitmap io */ 509 BITMAP_IO_QUEUED, /* Started bitmap IO */ 510 WAS_IO_ERROR, /* Local disk failed, returned IO error */ 511 WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */ 512 FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */ 513 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */ 514 RESIZE_PENDING, /* Size change detected locally, waiting for the response from 515 * the peer, if it changed there as well. */ 516 NEW_CUR_UUID, /* Create new current UUID when thawing IO */ 517 AL_SUSPENDED, /* Activity logging is currently suspended. */ 518 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */ 519 B_RS_H_DONE, /* Before resync handler done (already executed) */ 520 DISCARD_MY_DATA, /* discard_my_data flag per volume */ 521 READ_BALANCE_RR, 522 523 FLUSH_PENDING, /* if set, device->flush_jif is when we submitted that flush 524 * from drbd_flush_after_epoch() */ 525 526 /* cleared only after backing device related structures have been destroyed. */ 527 GOING_DISKLESS, /* Disk is being detached, because of io-error, or admin request. */ 528 529 /* to be used in drbd_device_post_work() */ 530 GO_DISKLESS, /* tell worker to schedule cleanup before detach */ 531 DESTROY_DISK, /* tell worker to close backing devices and destroy related structures. */ 532 MD_SYNC, /* tell worker to call drbd_md_sync() */ 533 RS_START, /* tell worker to start resync/OV */ 534 RS_PROGRESS, /* tell worker that resync made significant progress */ 535 RS_DONE, /* tell worker that resync is done */ 536 }; 537 538 struct drbd_bitmap; /* opaque for drbd_device */ 539 540 /* definition of bits in bm_flags to be used in drbd_bm_lock 541 * and drbd_bitmap_io and friends. */ 542 enum bm_flag { 543 /* currently locked for bulk operation */ 544 BM_LOCKED_MASK = 0xf, 545 546 /* in detail, that is: */ 547 BM_DONT_CLEAR = 0x1, 548 BM_DONT_SET = 0x2, 549 BM_DONT_TEST = 0x4, 550 551 /* so we can mark it locked for bulk operation, 552 * and still allow all non-bulk operations */ 553 BM_IS_LOCKED = 0x8, 554 555 /* (test bit, count bit) allowed (common case) */ 556 BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED, 557 558 /* testing bits, as well as setting new bits allowed, but clearing bits 559 * would be unexpected. Used during bitmap receive. Setting new bits 560 * requires sending of "out-of-sync" information, though. */ 561 BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED, 562 563 /* for drbd_bm_write_copy_pages, everything is allowed, 564 * only concurrent bulk operations are locked out. */ 565 BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED, 566 }; 567 568 struct drbd_work_queue { 569 struct list_head q; 570 spinlock_t q_lock; /* to protect the list. */ 571 wait_queue_head_t q_wait; 572 }; 573 574 struct drbd_socket { 575 struct mutex mutex; 576 struct socket *socket; 577 /* this way we get our 578 * send/receive buffers off the stack */ 579 void *sbuf; 580 void *rbuf; 581 }; 582 583 struct drbd_md { 584 u64 md_offset; /* sector offset to 'super' block */ 585 586 u64 la_size_sect; /* last agreed size, unit sectors */ 587 spinlock_t uuid_lock; 588 u64 uuid[UI_SIZE]; 589 u64 device_uuid; 590 u32 flags; 591 u32 md_size_sect; 592 593 s32 al_offset; /* signed relative sector offset to activity log */ 594 s32 bm_offset; /* signed relative sector offset to bitmap */ 595 596 /* cached value of bdev->disk_conf->meta_dev_idx (see below) */ 597 s32 meta_dev_idx; 598 599 /* see al_tr_number_to_on_disk_sector() */ 600 u32 al_stripes; 601 u32 al_stripe_size_4k; 602 u32 al_size_4k; /* cached product of the above */ 603 }; 604 605 struct drbd_backing_dev { 606 struct block_device *backing_bdev; 607 struct block_device *md_bdev; 608 struct drbd_md md; 609 struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */ 610 sector_t known_size; /* last known size of that backing device */ 611 }; 612 613 struct drbd_md_io { 614 struct page *page; 615 unsigned long start_jif; /* last call to drbd_md_get_buffer */ 616 unsigned long submit_jif; /* last _drbd_md_sync_page_io() submit */ 617 const char *current_use; 618 atomic_t in_use; 619 unsigned int done; 620 int error; 621 }; 622 623 struct bm_io_work { 624 struct drbd_work w; 625 char *why; 626 enum bm_flag flags; 627 int (*io_fn)(struct drbd_device *device); 628 void (*done)(struct drbd_device *device, int rv); 629 }; 630 631 struct fifo_buffer { 632 unsigned int head_index; 633 unsigned int size; 634 int total; /* sum of all values */ 635 int values[0]; 636 }; 637 extern struct fifo_buffer *fifo_alloc(int fifo_size); 638 639 /* flag bits per connection */ 640 enum { 641 NET_CONGESTED, /* The data socket is congested */ 642 RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */ 643 SEND_PING, 644 GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */ 645 CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */ 646 CONN_WD_ST_CHG_OKAY, 647 CONN_WD_ST_CHG_FAIL, 648 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */ 649 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */ 650 STATE_SENT, /* Do not change state/UUIDs while this is set */ 651 CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC) 652 * pending, from drbd worker context. 653 * If set, bdi_write_congested() returns true, 654 * so shrink_page_list() would not recurse into, 655 * and potentially deadlock on, this drbd worker. 656 */ 657 DISCONNECT_SENT, 658 659 DEVICE_WORK_PENDING, /* tell worker that some device has pending work */ 660 }; 661 662 enum which_state { NOW, OLD = NOW, NEW }; 663 664 struct drbd_resource { 665 char *name; 666 #ifdef CONFIG_DEBUG_FS 667 struct dentry *debugfs_res; 668 struct dentry *debugfs_res_volumes; 669 struct dentry *debugfs_res_connections; 670 struct dentry *debugfs_res_in_flight_summary; 671 #endif 672 struct kref kref; 673 struct idr devices; /* volume number to device mapping */ 674 struct list_head connections; 675 struct list_head resources; 676 struct res_opts res_opts; 677 struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */ 678 struct mutex adm_mutex; /* mutex to serialize administrative requests */ 679 spinlock_t req_lock; 680 681 unsigned susp:1; /* IO suspended by user */ 682 unsigned susp_nod:1; /* IO suspended because no data */ 683 unsigned susp_fen:1; /* IO suspended because fence peer handler runs */ 684 685 enum write_ordering_e write_ordering; 686 687 cpumask_var_t cpu_mask; 688 }; 689 690 struct drbd_thread_timing_details 691 { 692 unsigned long start_jif; 693 void *cb_addr; 694 const char *caller_fn; 695 unsigned int line; 696 unsigned int cb_nr; 697 }; 698 699 struct drbd_connection { 700 struct list_head connections; 701 struct drbd_resource *resource; 702 #ifdef CONFIG_DEBUG_FS 703 struct dentry *debugfs_conn; 704 struct dentry *debugfs_conn_callback_history; 705 struct dentry *debugfs_conn_oldest_requests; 706 #endif 707 struct kref kref; 708 struct idr peer_devices; /* volume number to peer device mapping */ 709 enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */ 710 struct mutex cstate_mutex; /* Protects graceful disconnects */ 711 unsigned int connect_cnt; /* Inc each time a connection is established */ 712 713 unsigned long flags; 714 struct net_conf *net_conf; /* content protected by rcu */ 715 wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */ 716 717 struct sockaddr_storage my_addr; 718 int my_addr_len; 719 struct sockaddr_storage peer_addr; 720 int peer_addr_len; 721 722 struct drbd_socket data; /* data/barrier/cstate/parameter packets */ 723 struct drbd_socket meta; /* ping/ack (metadata) packets */ 724 int agreed_pro_version; /* actually used protocol version */ 725 u32 agreed_features; 726 unsigned long last_received; /* in jiffies, either socket */ 727 unsigned int ko_count; 728 729 struct list_head transfer_log; /* all requests not yet fully processed */ 730 731 struct crypto_shash *cram_hmac_tfm; 732 struct crypto_shash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */ 733 struct crypto_shash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */ 734 struct crypto_shash *csums_tfm; 735 struct crypto_shash *verify_tfm; 736 void *int_dig_in; 737 void *int_dig_vv; 738 739 /* receiver side */ 740 struct drbd_epoch *current_epoch; 741 spinlock_t epoch_lock; 742 unsigned int epochs; 743 atomic_t current_tle_nr; /* transfer log epoch number */ 744 unsigned current_tle_writes; /* writes seen within this tl epoch */ 745 746 unsigned long last_reconnect_jif; 747 /* empty member on older kernels without blk_start_plug() */ 748 struct blk_plug receiver_plug; 749 struct drbd_thread receiver; 750 struct drbd_thread worker; 751 struct drbd_thread ack_receiver; 752 struct workqueue_struct *ack_sender; 753 754 /* cached pointers, 755 * so we can look up the oldest pending requests more quickly. 756 * protected by resource->req_lock */ 757 struct drbd_request *req_next; /* DRBD 9: todo.req_next */ 758 struct drbd_request *req_ack_pending; 759 struct drbd_request *req_not_net_done; 760 761 /* sender side */ 762 struct drbd_work_queue sender_work; 763 764 #define DRBD_THREAD_DETAILS_HIST 16 765 unsigned int w_cb_nr; /* keeps counting up */ 766 unsigned int r_cb_nr; /* keeps counting up */ 767 struct drbd_thread_timing_details w_timing_details[DRBD_THREAD_DETAILS_HIST]; 768 struct drbd_thread_timing_details r_timing_details[DRBD_THREAD_DETAILS_HIST]; 769 770 struct { 771 unsigned long last_sent_barrier_jif; 772 773 /* whether this sender thread 774 * has processed a single write yet. */ 775 bool seen_any_write_yet; 776 777 /* Which barrier number to send with the next P_BARRIER */ 778 int current_epoch_nr; 779 780 /* how many write requests have been sent 781 * with req->epoch == current_epoch_nr. 782 * If none, no P_BARRIER will be sent. */ 783 unsigned current_epoch_writes; 784 } send; 785 }; 786 787 static inline bool has_net_conf(struct drbd_connection *connection) 788 { 789 bool has_net_conf; 790 791 rcu_read_lock(); 792 has_net_conf = rcu_dereference(connection->net_conf); 793 rcu_read_unlock(); 794 795 return has_net_conf; 796 } 797 798 void __update_timing_details( 799 struct drbd_thread_timing_details *tdp, 800 unsigned int *cb_nr, 801 void *cb, 802 const char *fn, const unsigned int line); 803 804 #define update_worker_timing_details(c, cb) \ 805 __update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ ) 806 #define update_receiver_timing_details(c, cb) \ 807 __update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ ) 808 809 struct submit_worker { 810 struct workqueue_struct *wq; 811 struct work_struct worker; 812 813 /* protected by ..->resource->req_lock */ 814 struct list_head writes; 815 }; 816 817 struct drbd_peer_device { 818 struct list_head peer_devices; 819 struct drbd_device *device; 820 struct drbd_connection *connection; 821 struct work_struct send_acks_work; 822 #ifdef CONFIG_DEBUG_FS 823 struct dentry *debugfs_peer_dev; 824 #endif 825 }; 826 827 struct drbd_device { 828 struct drbd_resource *resource; 829 struct list_head peer_devices; 830 struct list_head pending_bitmap_io; 831 832 unsigned long flush_jif; 833 #ifdef CONFIG_DEBUG_FS 834 struct dentry *debugfs_minor; 835 struct dentry *debugfs_vol; 836 struct dentry *debugfs_vol_oldest_requests; 837 struct dentry *debugfs_vol_act_log_extents; 838 struct dentry *debugfs_vol_resync_extents; 839 struct dentry *debugfs_vol_data_gen_id; 840 struct dentry *debugfs_vol_ed_gen_id; 841 #endif 842 843 unsigned int vnr; /* volume number within the connection */ 844 unsigned int minor; /* device minor number */ 845 846 struct kref kref; 847 848 /* things that are stored as / read from meta data on disk */ 849 unsigned long flags; 850 851 /* configured by drbdsetup */ 852 struct drbd_backing_dev *ldev __protected_by(local); 853 854 sector_t p_size; /* partner's disk size */ 855 struct request_queue *rq_queue; 856 struct block_device *this_bdev; 857 struct gendisk *vdisk; 858 859 unsigned long last_reattach_jif; 860 struct drbd_work resync_work; 861 struct drbd_work unplug_work; 862 struct timer_list resync_timer; 863 struct timer_list md_sync_timer; 864 struct timer_list start_resync_timer; 865 struct timer_list request_timer; 866 867 /* Used after attach while negotiating new disk state. */ 868 union drbd_state new_state_tmp; 869 870 union drbd_dev_state state; 871 wait_queue_head_t misc_wait; 872 wait_queue_head_t state_wait; /* upon each state change. */ 873 unsigned int send_cnt; 874 unsigned int recv_cnt; 875 unsigned int read_cnt; 876 unsigned int writ_cnt; 877 unsigned int al_writ_cnt; 878 unsigned int bm_writ_cnt; 879 atomic_t ap_bio_cnt; /* Requests we need to complete */ 880 atomic_t ap_actlog_cnt; /* Requests waiting for activity log */ 881 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */ 882 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */ 883 atomic_t unacked_cnt; /* Need to send replies for */ 884 atomic_t local_cnt; /* Waiting for local completion */ 885 atomic_t suspend_cnt; 886 887 /* Interval tree of pending local requests */ 888 struct rb_root read_requests; 889 struct rb_root write_requests; 890 891 /* for statistics and timeouts */ 892 /* [0] read, [1] write */ 893 struct list_head pending_master_completion[2]; 894 struct list_head pending_completion[2]; 895 896 /* use checksums for *this* resync */ 897 bool use_csums; 898 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */ 899 unsigned long rs_total; 900 /* number of resync blocks that failed in this run */ 901 unsigned long rs_failed; 902 /* Syncer's start time [unit jiffies] */ 903 unsigned long rs_start; 904 /* cumulated time in PausedSyncX state [unit jiffies] */ 905 unsigned long rs_paused; 906 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */ 907 unsigned long rs_same_csum; 908 #define DRBD_SYNC_MARKS 8 909 #define DRBD_SYNC_MARK_STEP (3*HZ) 910 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */ 911 unsigned long rs_mark_left[DRBD_SYNC_MARKS]; 912 /* marks's time [unit jiffies] */ 913 unsigned long rs_mark_time[DRBD_SYNC_MARKS]; 914 /* current index into rs_mark_{left,time} */ 915 int rs_last_mark; 916 unsigned long rs_last_bcast; /* [unit jiffies] */ 917 918 /* where does the admin want us to start? (sector) */ 919 sector_t ov_start_sector; 920 sector_t ov_stop_sector; 921 /* where are we now? (sector) */ 922 sector_t ov_position; 923 /* Start sector of out of sync range (to merge printk reporting). */ 924 sector_t ov_last_oos_start; 925 /* size of out-of-sync range in sectors. */ 926 sector_t ov_last_oos_size; 927 unsigned long ov_left; /* in bits */ 928 929 struct drbd_bitmap *bitmap; 930 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */ 931 932 /* Used to track operations of resync... */ 933 struct lru_cache *resync; 934 /* Number of locked elements in resync LRU */ 935 unsigned int resync_locked; 936 /* resync extent number waiting for application requests */ 937 unsigned int resync_wenr; 938 939 int open_cnt; 940 u64 *p_uuid; 941 942 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */ 943 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */ 944 struct list_head done_ee; /* need to send P_WRITE_ACK */ 945 struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */ 946 struct list_head net_ee; /* zero-copy network send in progress */ 947 948 int next_barrier_nr; 949 struct list_head resync_reads; 950 atomic_t pp_in_use; /* allocated from page pool */ 951 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */ 952 wait_queue_head_t ee_wait; 953 struct drbd_md_io md_io; 954 spinlock_t al_lock; 955 wait_queue_head_t al_wait; 956 struct lru_cache *act_log; /* activity log */ 957 unsigned int al_tr_number; 958 int al_tr_cycle; 959 wait_queue_head_t seq_wait; 960 atomic_t packet_seq; 961 unsigned int peer_seq; 962 spinlock_t peer_seq_lock; 963 unsigned long comm_bm_set; /* communicated number of set bits. */ 964 struct bm_io_work bm_io_work; 965 u64 ed_uuid; /* UUID of the exposed data */ 966 struct mutex own_state_mutex; 967 struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */ 968 char congestion_reason; /* Why we where congested... */ 969 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */ 970 atomic_t rs_sect_ev; /* for submitted resync data rate, both */ 971 int rs_last_sect_ev; /* counter to compare with */ 972 int rs_last_events; /* counter of read or write "events" (unit sectors) 973 * on the lower level device when we last looked. */ 974 int c_sync_rate; /* current resync rate after syncer throttle magic */ 975 struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */ 976 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */ 977 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */ 978 unsigned int peer_max_bio_size; 979 unsigned int local_max_bio_size; 980 981 /* any requests that would block in drbd_make_request() 982 * are deferred to this single-threaded work queue */ 983 struct submit_worker submit; 984 }; 985 986 struct drbd_bm_aio_ctx { 987 struct drbd_device *device; 988 struct list_head list; /* on device->pending_bitmap_io */; 989 unsigned long start_jif; 990 atomic_t in_flight; 991 unsigned int done; 992 unsigned flags; 993 #define BM_AIO_COPY_PAGES 1 994 #define BM_AIO_WRITE_HINTED 2 995 #define BM_AIO_WRITE_ALL_PAGES 4 996 #define BM_AIO_READ 8 997 int error; 998 struct kref kref; 999 }; 1000 1001 struct drbd_config_context { 1002 /* assigned from drbd_genlmsghdr */ 1003 unsigned int minor; 1004 /* assigned from request attributes, if present */ 1005 unsigned int volume; 1006 #define VOLUME_UNSPECIFIED (-1U) 1007 /* pointer into the request skb, 1008 * limited lifetime! */ 1009 char *resource_name; 1010 struct nlattr *my_addr; 1011 struct nlattr *peer_addr; 1012 1013 /* reply buffer */ 1014 struct sk_buff *reply_skb; 1015 /* pointer into reply buffer */ 1016 struct drbd_genlmsghdr *reply_dh; 1017 /* resolved from attributes, if possible */ 1018 struct drbd_device *device; 1019 struct drbd_resource *resource; 1020 struct drbd_connection *connection; 1021 }; 1022 1023 static inline struct drbd_device *minor_to_device(unsigned int minor) 1024 { 1025 return (struct drbd_device *)idr_find(&drbd_devices, minor); 1026 } 1027 1028 static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device) 1029 { 1030 return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices); 1031 } 1032 1033 static inline struct drbd_peer_device * 1034 conn_peer_device(struct drbd_connection *connection, int volume_number) 1035 { 1036 return idr_find(&connection->peer_devices, volume_number); 1037 } 1038 1039 #define for_each_resource(resource, _resources) \ 1040 list_for_each_entry(resource, _resources, resources) 1041 1042 #define for_each_resource_rcu(resource, _resources) \ 1043 list_for_each_entry_rcu(resource, _resources, resources) 1044 1045 #define for_each_resource_safe(resource, tmp, _resources) \ 1046 list_for_each_entry_safe(resource, tmp, _resources, resources) 1047 1048 #define for_each_connection(connection, resource) \ 1049 list_for_each_entry(connection, &resource->connections, connections) 1050 1051 #define for_each_connection_rcu(connection, resource) \ 1052 list_for_each_entry_rcu(connection, &resource->connections, connections) 1053 1054 #define for_each_connection_safe(connection, tmp, resource) \ 1055 list_for_each_entry_safe(connection, tmp, &resource->connections, connections) 1056 1057 #define for_each_peer_device(peer_device, device) \ 1058 list_for_each_entry(peer_device, &device->peer_devices, peer_devices) 1059 1060 #define for_each_peer_device_rcu(peer_device, device) \ 1061 list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices) 1062 1063 #define for_each_peer_device_safe(peer_device, tmp, device) \ 1064 list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices) 1065 1066 static inline unsigned int device_to_minor(struct drbd_device *device) 1067 { 1068 return device->minor; 1069 } 1070 1071 /* 1072 * function declarations 1073 *************************/ 1074 1075 /* drbd_main.c */ 1076 1077 enum dds_flags { 1078 DDSF_FORCED = 1, 1079 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */ 1080 }; 1081 1082 extern void drbd_init_set_defaults(struct drbd_device *device); 1083 extern int drbd_thread_start(struct drbd_thread *thi); 1084 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait); 1085 #ifdef CONFIG_SMP 1086 extern void drbd_thread_current_set_cpu(struct drbd_thread *thi); 1087 #else 1088 #define drbd_thread_current_set_cpu(A) ({}) 1089 #endif 1090 extern void tl_release(struct drbd_connection *, unsigned int barrier_nr, 1091 unsigned int set_size); 1092 extern void tl_clear(struct drbd_connection *); 1093 extern void drbd_free_sock(struct drbd_connection *connection); 1094 extern int drbd_send(struct drbd_connection *connection, struct socket *sock, 1095 void *buf, size_t size, unsigned msg_flags); 1096 extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t, 1097 unsigned); 1098 1099 extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd); 1100 extern int drbd_send_protocol(struct drbd_connection *connection); 1101 extern int drbd_send_uuids(struct drbd_peer_device *); 1102 extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *); 1103 extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *); 1104 extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags); 1105 extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s); 1106 extern int drbd_send_current_state(struct drbd_peer_device *); 1107 extern int drbd_send_sync_param(struct drbd_peer_device *); 1108 extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr, 1109 u32 set_size); 1110 extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet, 1111 struct drbd_peer_request *); 1112 extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet, 1113 struct p_block_req *rp); 1114 extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet, 1115 struct p_data *dp, int data_size); 1116 extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet, 1117 sector_t sector, int blksize, u64 block_id); 1118 extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *); 1119 extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet, 1120 struct drbd_peer_request *); 1121 extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req); 1122 extern int drbd_send_drequest(struct drbd_peer_device *, int cmd, 1123 sector_t sector, int size, u64 block_id); 1124 extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector, 1125 int size, void *digest, int digest_size, 1126 enum drbd_packet cmd); 1127 extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size); 1128 1129 extern int drbd_send_bitmap(struct drbd_device *device); 1130 extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode); 1131 extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode); 1132 extern int drbd_send_rs_deallocated(struct drbd_peer_device *, struct drbd_peer_request *); 1133 extern void drbd_backing_dev_free(struct drbd_device *device, struct drbd_backing_dev *ldev); 1134 extern void drbd_device_cleanup(struct drbd_device *device); 1135 extern void drbd_print_uuids(struct drbd_device *device, const char *text); 1136 extern void drbd_queue_unplug(struct drbd_device *device); 1137 1138 extern void conn_md_sync(struct drbd_connection *connection); 1139 extern void drbd_md_write(struct drbd_device *device, void *buffer); 1140 extern void drbd_md_sync(struct drbd_device *device); 1141 extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev); 1142 extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); 1143 extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); 1144 extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local); 1145 extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local); 1146 extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local); 1147 extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); 1148 extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local); 1149 extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local); 1150 extern int drbd_md_test_flag(struct drbd_backing_dev *, int); 1151 extern void drbd_md_mark_dirty(struct drbd_device *device); 1152 extern void drbd_queue_bitmap_io(struct drbd_device *device, 1153 int (*io_fn)(struct drbd_device *), 1154 void (*done)(struct drbd_device *, int), 1155 char *why, enum bm_flag flags); 1156 extern int drbd_bitmap_io(struct drbd_device *device, 1157 int (*io_fn)(struct drbd_device *), 1158 char *why, enum bm_flag flags); 1159 extern int drbd_bitmap_io_from_worker(struct drbd_device *device, 1160 int (*io_fn)(struct drbd_device *), 1161 char *why, enum bm_flag flags); 1162 extern int drbd_bmio_set_n_write(struct drbd_device *device) __must_hold(local); 1163 extern int drbd_bmio_clear_n_write(struct drbd_device *device) __must_hold(local); 1164 1165 /* Meta data layout 1166 * 1167 * We currently have two possible layouts. 1168 * Offsets in (512 byte) sectors. 1169 * external: 1170 * |----------- md_size_sect ------------------| 1171 * [ 4k superblock ][ activity log ][ Bitmap ] 1172 * | al_offset == 8 | 1173 * | bm_offset = al_offset + X | 1174 * ==> bitmap sectors = md_size_sect - bm_offset 1175 * 1176 * Variants: 1177 * old, indexed fixed size meta data: 1178 * 1179 * internal: 1180 * |----------- md_size_sect ------------------| 1181 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*] 1182 * | al_offset < 0 | 1183 * | bm_offset = al_offset - Y | 1184 * ==> bitmap sectors = Y = al_offset - bm_offset 1185 * 1186 * [padding*] are zero or up to 7 unused 512 Byte sectors to the 1187 * end of the device, so that the [4k superblock] will be 4k aligned. 1188 * 1189 * The activity log consists of 4k transaction blocks, 1190 * which are written in a ring-buffer, or striped ring-buffer like fashion, 1191 * which are writtensize used to be fixed 32kB, 1192 * but is about to become configurable. 1193 */ 1194 1195 /* Our old fixed size meta data layout 1196 * allows up to about 3.8TB, so if you want more, 1197 * you need to use the "flexible" meta data format. */ 1198 #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */ 1199 #define MD_4kB_SECT 8 1200 #define MD_32kB_SECT 64 1201 1202 /* One activity log extent represents 4M of storage */ 1203 #define AL_EXTENT_SHIFT 22 1204 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT) 1205 1206 /* We could make these currently hardcoded constants configurable 1207 * variables at create-md time (or even re-configurable at runtime?). 1208 * Which will require some more changes to the DRBD "super block" 1209 * and attach code. 1210 * 1211 * updates per transaction: 1212 * This many changes to the active set can be logged with one transaction. 1213 * This number is arbitrary. 1214 * context per transaction: 1215 * This many context extent numbers are logged with each transaction. 1216 * This number is resulting from the transaction block size (4k), the layout 1217 * of the transaction header, and the number of updates per transaction. 1218 * See drbd_actlog.c:struct al_transaction_on_disk 1219 * */ 1220 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary 1221 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4 1222 1223 #if BITS_PER_LONG == 32 1224 #define LN2_BPL 5 1225 #define cpu_to_lel(A) cpu_to_le32(A) 1226 #define lel_to_cpu(A) le32_to_cpu(A) 1227 #elif BITS_PER_LONG == 64 1228 #define LN2_BPL 6 1229 #define cpu_to_lel(A) cpu_to_le64(A) 1230 #define lel_to_cpu(A) le64_to_cpu(A) 1231 #else 1232 #error "LN2 of BITS_PER_LONG unknown!" 1233 #endif 1234 1235 /* resync bitmap */ 1236 /* 16MB sized 'bitmap extent' to track syncer usage */ 1237 struct bm_extent { 1238 int rs_left; /* number of bits set (out of sync) in this extent. */ 1239 int rs_failed; /* number of failed resync requests in this extent. */ 1240 unsigned long flags; 1241 struct lc_element lce; 1242 }; 1243 1244 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */ 1245 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */ 1246 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */ 1247 1248 /* drbd_bitmap.c */ 1249 /* 1250 * We need to store one bit for a block. 1251 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap. 1252 * Bit 0 ==> local node thinks this block is binary identical on both nodes 1253 * Bit 1 ==> local node thinks this block needs to be synced. 1254 */ 1255 1256 #define SLEEP_TIME (HZ/10) 1257 1258 /* We do bitmap IO in units of 4k blocks. 1259 * We also still have a hardcoded 4k per bit relation. */ 1260 #define BM_BLOCK_SHIFT 12 /* 4k per bit */ 1261 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT) 1262 /* mostly arbitrarily set the represented size of one bitmap extent, 1263 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap 1264 * at 4k per bit resolution) */ 1265 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */ 1266 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT) 1267 1268 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12) 1269 #error "HAVE YOU FIXED drbdmeta AS WELL??" 1270 #endif 1271 1272 /* thus many _storage_ sectors are described by one bit */ 1273 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9)) 1274 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9)) 1275 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1) 1276 1277 /* bit to represented kilo byte conversion */ 1278 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10)) 1279 1280 /* in which _bitmap_ extent (resp. sector) the bit for a certain 1281 * _storage_ sector is located in */ 1282 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9)) 1283 #define BM_BIT_TO_EXT(x) ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT)) 1284 1285 /* first storage sector a bitmap extent corresponds to */ 1286 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9)) 1287 /* how much _storage_ sectors we have per bitmap extent */ 1288 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1) 1289 /* how many bits are covered by one bitmap extent (resync extent) */ 1290 #define BM_BITS_PER_EXT (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT)) 1291 1292 #define BM_BLOCKS_PER_BM_EXT_MASK (BM_BITS_PER_EXT - 1) 1293 1294 1295 /* in one sector of the bitmap, we have this many activity_log extents. */ 1296 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT)) 1297 1298 /* the extent in "PER_EXTENT" below is an activity log extent 1299 * we need that many (long words/bytes) to store the bitmap 1300 * of one AL_EXTENT_SIZE chunk of storage. 1301 * we can store the bitmap for that many AL_EXTENTS within 1302 * one sector of the _on_disk_ bitmap: 1303 * bit 0 bit 37 bit 38 bit (512*8)-1 1304 * ...|........|........|.. // ..|........| 1305 * sect. 0 `296 `304 ^(512*8*8)-1 1306 * 1307 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG ) 1308 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128 1309 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4 1310 */ 1311 1312 #define DRBD_MAX_SECTORS_32 (0xffffffffLU) 1313 /* we have a certain meta data variant that has a fixed on-disk size of 128 1314 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity 1315 * log, leaving this many sectors for the bitmap. 1316 */ 1317 1318 #define DRBD_MAX_SECTORS_FIXED_BM \ 1319 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9))) 1320 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM 1321 /* 16 TB in units of sectors */ 1322 #if BITS_PER_LONG == 32 1323 /* adjust by one page worth of bitmap, 1324 * so we won't wrap around in drbd_bm_find_next_bit. 1325 * you should use 64bit OS for that much storage, anyways. */ 1326 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff) 1327 #else 1328 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */ 1329 #define DRBD_MAX_SECTORS_FLEX (1UL << 51) 1330 /* corresponds to (1UL << 38) bits right now. */ 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_shash *, struct bio *, void *); 1535 extern void drbd_csum_ee(struct crypto_shash *, struct drbd_peer_request *, 1536 void *); 1537 /* worker callbacks */ 1538 extern int w_e_end_data_req(struct drbd_work *, int); 1539 extern int w_e_end_rsdata_req(struct drbd_work *, int); 1540 extern int w_e_end_csum_rs_req(struct drbd_work *, int); 1541 extern int w_e_end_ov_reply(struct drbd_work *, int); 1542 extern int w_e_end_ov_req(struct drbd_work *, int); 1543 extern int w_ov_finished(struct drbd_work *, int); 1544 extern int w_resync_timer(struct drbd_work *, int); 1545 extern int w_send_write_hint(struct drbd_work *, int); 1546 extern int w_send_dblock(struct drbd_work *, int); 1547 extern int w_send_read_req(struct drbd_work *, int); 1548 extern int w_e_reissue(struct drbd_work *, int); 1549 extern int w_restart_disk_io(struct drbd_work *, int); 1550 extern int w_send_out_of_sync(struct drbd_work *, int); 1551 extern int w_start_resync(struct drbd_work *, int); 1552 1553 extern void resync_timer_fn(struct timer_list *t); 1554 extern void start_resync_timer_fn(struct timer_list *t); 1555 1556 extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req); 1557 1558 /* drbd_receiver.c */ 1559 extern int drbd_issue_discard_or_zero_out(struct drbd_device *device, 1560 sector_t start, unsigned int nr_sectors, int flags); 1561 extern int drbd_receiver(struct drbd_thread *thi); 1562 extern int drbd_ack_receiver(struct drbd_thread *thi); 1563 extern void drbd_send_ping_wf(struct work_struct *ws); 1564 extern void drbd_send_acks_wf(struct work_struct *ws); 1565 extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device); 1566 extern bool drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector, 1567 bool throttle_if_app_is_waiting); 1568 extern int drbd_submit_peer_request(struct drbd_device *, 1569 struct drbd_peer_request *, const unsigned, 1570 const unsigned, const int); 1571 extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *); 1572 extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64, 1573 sector_t, unsigned int, 1574 unsigned int, 1575 gfp_t) __must_hold(local); 1576 extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *, 1577 int); 1578 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0) 1579 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1) 1580 extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool); 1581 extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled); 1582 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed); 1583 extern int drbd_connected(struct drbd_peer_device *); 1584 1585 static inline void drbd_tcp_cork(struct socket *sock) 1586 { 1587 int val = 1; 1588 (void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK, 1589 (char*)&val, sizeof(val)); 1590 } 1591 1592 static inline void drbd_tcp_uncork(struct socket *sock) 1593 { 1594 int val = 0; 1595 (void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK, 1596 (char*)&val, sizeof(val)); 1597 } 1598 1599 static inline void drbd_tcp_nodelay(struct socket *sock) 1600 { 1601 int val = 1; 1602 (void) kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, 1603 (char*)&val, sizeof(val)); 1604 } 1605 1606 static inline void drbd_tcp_quickack(struct socket *sock) 1607 { 1608 int val = 2; 1609 (void) kernel_setsockopt(sock, SOL_TCP, TCP_QUICKACK, 1610 (char*)&val, sizeof(val)); 1611 } 1612 1613 /* sets the number of 512 byte sectors of our virtual device */ 1614 void drbd_set_my_capacity(struct drbd_device *device, sector_t size); 1615 1616 /* 1617 * used to submit our private bio 1618 */ 1619 static inline void drbd_generic_make_request(struct drbd_device *device, 1620 int fault_type, struct bio *bio) 1621 { 1622 __release(local); 1623 if (!bio->bi_disk) { 1624 drbd_err(device, "drbd_generic_make_request: bio->bi_disk == NULL\n"); 1625 bio->bi_status = BLK_STS_IOERR; 1626 bio_endio(bio); 1627 return; 1628 } 1629 1630 if (drbd_insert_fault(device, fault_type)) 1631 bio_io_error(bio); 1632 else 1633 generic_make_request(bio); 1634 } 1635 1636 void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev, 1637 enum write_ordering_e wo); 1638 1639 /* drbd_proc.c */ 1640 extern struct proc_dir_entry *drbd_proc; 1641 int drbd_seq_show(struct seq_file *seq, void *v); 1642 1643 /* drbd_actlog.c */ 1644 extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i); 1645 extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i); 1646 extern void drbd_al_begin_io_commit(struct drbd_device *device); 1647 extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i); 1648 extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i); 1649 extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i); 1650 extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector); 1651 extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector); 1652 extern int drbd_try_rs_begin_io(struct drbd_device *device, sector_t sector); 1653 extern void drbd_rs_cancel_all(struct drbd_device *device); 1654 extern int drbd_rs_del_all(struct drbd_device *device); 1655 extern void drbd_rs_failed_io(struct drbd_device *device, 1656 sector_t sector, int size); 1657 extern void drbd_advance_rs_marks(struct drbd_device *device, unsigned long still_to_go); 1658 1659 enum update_sync_bits_mode { RECORD_RS_FAILED, SET_OUT_OF_SYNC, SET_IN_SYNC }; 1660 extern int __drbd_change_sync(struct drbd_device *device, sector_t sector, int size, 1661 enum update_sync_bits_mode mode); 1662 #define drbd_set_in_sync(device, sector, size) \ 1663 __drbd_change_sync(device, sector, size, SET_IN_SYNC) 1664 #define drbd_set_out_of_sync(device, sector, size) \ 1665 __drbd_change_sync(device, sector, size, SET_OUT_OF_SYNC) 1666 #define drbd_rs_failed_io(device, sector, size) \ 1667 __drbd_change_sync(device, sector, size, RECORD_RS_FAILED) 1668 extern void drbd_al_shrink(struct drbd_device *device); 1669 extern int drbd_al_initialize(struct drbd_device *, void *); 1670 1671 /* drbd_nl.c */ 1672 /* state info broadcast */ 1673 struct sib_info { 1674 enum drbd_state_info_bcast_reason sib_reason; 1675 union { 1676 struct { 1677 char *helper_name; 1678 unsigned helper_exit_code; 1679 }; 1680 struct { 1681 union drbd_state os; 1682 union drbd_state ns; 1683 }; 1684 }; 1685 }; 1686 void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib); 1687 1688 extern void notify_resource_state(struct sk_buff *, 1689 unsigned int, 1690 struct drbd_resource *, 1691 struct resource_info *, 1692 enum drbd_notification_type); 1693 extern void notify_device_state(struct sk_buff *, 1694 unsigned int, 1695 struct drbd_device *, 1696 struct device_info *, 1697 enum drbd_notification_type); 1698 extern void notify_connection_state(struct sk_buff *, 1699 unsigned int, 1700 struct drbd_connection *, 1701 struct connection_info *, 1702 enum drbd_notification_type); 1703 extern void notify_peer_device_state(struct sk_buff *, 1704 unsigned int, 1705 struct drbd_peer_device *, 1706 struct peer_device_info *, 1707 enum drbd_notification_type); 1708 extern void notify_helper(enum drbd_notification_type, struct drbd_device *, 1709 struct drbd_connection *, const char *, int); 1710 1711 /* 1712 * inline helper functions 1713 *************************/ 1714 1715 /* see also page_chain_add and friends in drbd_receiver.c */ 1716 static inline struct page *page_chain_next(struct page *page) 1717 { 1718 return (struct page *)page_private(page); 1719 } 1720 #define page_chain_for_each(page) \ 1721 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \ 1722 page = page_chain_next(page)) 1723 #define page_chain_for_each_safe(page, n) \ 1724 for (; page && ({ n = page_chain_next(page); 1; }); page = n) 1725 1726 1727 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req) 1728 { 1729 struct page *page = peer_req->pages; 1730 page_chain_for_each(page) { 1731 if (page_count(page) > 1) 1732 return 1; 1733 } 1734 return 0; 1735 } 1736 1737 static inline union drbd_state drbd_read_state(struct drbd_device *device) 1738 { 1739 struct drbd_resource *resource = device->resource; 1740 union drbd_state rv; 1741 1742 rv.i = device->state.i; 1743 rv.susp = resource->susp; 1744 rv.susp_nod = resource->susp_nod; 1745 rv.susp_fen = resource->susp_fen; 1746 1747 return rv; 1748 } 1749 1750 enum drbd_force_detach_flags { 1751 DRBD_READ_ERROR, 1752 DRBD_WRITE_ERROR, 1753 DRBD_META_IO_ERROR, 1754 DRBD_FORCE_DETACH, 1755 }; 1756 1757 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__) 1758 static inline void __drbd_chk_io_error_(struct drbd_device *device, 1759 enum drbd_force_detach_flags df, 1760 const char *where) 1761 { 1762 enum drbd_io_error_p ep; 1763 1764 rcu_read_lock(); 1765 ep = rcu_dereference(device->ldev->disk_conf)->on_io_error; 1766 rcu_read_unlock(); 1767 switch (ep) { 1768 case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */ 1769 if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) { 1770 if (__ratelimit(&drbd_ratelimit_state)) 1771 drbd_err(device, "Local IO failed in %s.\n", where); 1772 if (device->state.disk > D_INCONSISTENT) 1773 _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL); 1774 break; 1775 } 1776 /* fall through - for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */ 1777 case EP_DETACH: 1778 case EP_CALL_HELPER: 1779 /* Remember whether we saw a READ or WRITE error. 1780 * 1781 * Recovery of the affected area for WRITE failure is covered 1782 * by the activity log. 1783 * READ errors may fall outside that area though. Certain READ 1784 * errors can be "healed" by writing good data to the affected 1785 * blocks, which triggers block re-allocation in lower layers. 1786 * 1787 * If we can not write the bitmap after a READ error, 1788 * we may need to trigger a full sync (see w_go_diskless()). 1789 * 1790 * Force-detach is not really an IO error, but rather a 1791 * desperate measure to try to deal with a completely 1792 * unresponsive lower level IO stack. 1793 * Still it should be treated as a WRITE error. 1794 * 1795 * Meta IO error is always WRITE error: 1796 * we read meta data only once during attach, 1797 * which will fail in case of errors. 1798 */ 1799 set_bit(WAS_IO_ERROR, &device->flags); 1800 if (df == DRBD_READ_ERROR) 1801 set_bit(WAS_READ_ERROR, &device->flags); 1802 if (df == DRBD_FORCE_DETACH) 1803 set_bit(FORCE_DETACH, &device->flags); 1804 if (device->state.disk > D_FAILED) { 1805 _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL); 1806 drbd_err(device, 1807 "Local IO failed in %s. Detaching...\n", where); 1808 } 1809 break; 1810 } 1811 } 1812 1813 /** 1814 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers 1815 * @device: DRBD device. 1816 * @error: Error code passed to the IO completion callback 1817 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data 1818 * 1819 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED) 1820 */ 1821 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__) 1822 static inline void drbd_chk_io_error_(struct drbd_device *device, 1823 int error, enum drbd_force_detach_flags forcedetach, const char *where) 1824 { 1825 if (error) { 1826 unsigned long flags; 1827 spin_lock_irqsave(&device->resource->req_lock, flags); 1828 __drbd_chk_io_error_(device, forcedetach, where); 1829 spin_unlock_irqrestore(&device->resource->req_lock, flags); 1830 } 1831 } 1832 1833 1834 /** 1835 * drbd_md_first_sector() - Returns the first sector number of the meta data area 1836 * @bdev: Meta data block device. 1837 * 1838 * BTW, for internal meta data, this happens to be the maximum capacity 1839 * we could agree upon with our peer node. 1840 */ 1841 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev) 1842 { 1843 switch (bdev->md.meta_dev_idx) { 1844 case DRBD_MD_INDEX_INTERNAL: 1845 case DRBD_MD_INDEX_FLEX_INT: 1846 return bdev->md.md_offset + bdev->md.bm_offset; 1847 case DRBD_MD_INDEX_FLEX_EXT: 1848 default: 1849 return bdev->md.md_offset; 1850 } 1851 } 1852 1853 /** 1854 * drbd_md_last_sector() - Return the last sector number of the meta data area 1855 * @bdev: Meta data block device. 1856 */ 1857 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev) 1858 { 1859 switch (bdev->md.meta_dev_idx) { 1860 case DRBD_MD_INDEX_INTERNAL: 1861 case DRBD_MD_INDEX_FLEX_INT: 1862 return bdev->md.md_offset + MD_4kB_SECT -1; 1863 case DRBD_MD_INDEX_FLEX_EXT: 1864 default: 1865 return bdev->md.md_offset + bdev->md.md_size_sect -1; 1866 } 1867 } 1868 1869 /* Returns the number of 512 byte sectors of the device */ 1870 static inline sector_t drbd_get_capacity(struct block_device *bdev) 1871 { 1872 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */ 1873 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0; 1874 } 1875 1876 /** 1877 * drbd_get_max_capacity() - Returns the capacity we announce to out peer 1878 * @bdev: Meta data block device. 1879 * 1880 * returns the capacity we announce to out peer. we clip ourselves at the 1881 * various MAX_SECTORS, because if we don't, current implementation will 1882 * oops sooner or later 1883 */ 1884 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev) 1885 { 1886 sector_t s; 1887 1888 switch (bdev->md.meta_dev_idx) { 1889 case DRBD_MD_INDEX_INTERNAL: 1890 case DRBD_MD_INDEX_FLEX_INT: 1891 s = drbd_get_capacity(bdev->backing_bdev) 1892 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX, 1893 drbd_md_first_sector(bdev)) 1894 : 0; 1895 break; 1896 case DRBD_MD_INDEX_FLEX_EXT: 1897 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX, 1898 drbd_get_capacity(bdev->backing_bdev)); 1899 /* clip at maximum size the meta device can support */ 1900 s = min_t(sector_t, s, 1901 BM_EXT_TO_SECT(bdev->md.md_size_sect 1902 - bdev->md.bm_offset)); 1903 break; 1904 default: 1905 s = min_t(sector_t, DRBD_MAX_SECTORS, 1906 drbd_get_capacity(bdev->backing_bdev)); 1907 } 1908 return s; 1909 } 1910 1911 /** 1912 * drbd_md_ss() - Return the sector number of our meta data super block 1913 * @bdev: Meta data block device. 1914 */ 1915 static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev) 1916 { 1917 const int meta_dev_idx = bdev->md.meta_dev_idx; 1918 1919 if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT) 1920 return 0; 1921 1922 /* Since drbd08, internal meta data is always "flexible". 1923 * position: last 4k aligned block of 4k size */ 1924 if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL || 1925 meta_dev_idx == DRBD_MD_INDEX_FLEX_INT) 1926 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8; 1927 1928 /* external, some index; this is the old fixed size layout */ 1929 return MD_128MB_SECT * bdev->md.meta_dev_idx; 1930 } 1931 1932 static inline void 1933 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w) 1934 { 1935 unsigned long flags; 1936 spin_lock_irqsave(&q->q_lock, flags); 1937 list_add_tail(&w->list, &q->q); 1938 spin_unlock_irqrestore(&q->q_lock, flags); 1939 wake_up(&q->q_wait); 1940 } 1941 1942 static inline void 1943 drbd_queue_work_if_unqueued(struct drbd_work_queue *q, struct drbd_work *w) 1944 { 1945 unsigned long flags; 1946 spin_lock_irqsave(&q->q_lock, flags); 1947 if (list_empty_careful(&w->list)) 1948 list_add_tail(&w->list, &q->q); 1949 spin_unlock_irqrestore(&q->q_lock, flags); 1950 wake_up(&q->q_wait); 1951 } 1952 1953 static inline void 1954 drbd_device_post_work(struct drbd_device *device, int work_bit) 1955 { 1956 if (!test_and_set_bit(work_bit, &device->flags)) { 1957 struct drbd_connection *connection = 1958 first_peer_device(device)->connection; 1959 struct drbd_work_queue *q = &connection->sender_work; 1960 if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags)) 1961 wake_up(&q->q_wait); 1962 } 1963 } 1964 1965 extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue); 1966 1967 /* To get the ack_receiver out of the blocking network stack, 1968 * so it can change its sk_rcvtimeo from idle- to ping-timeout, 1969 * and send a ping, we need to send a signal. 1970 * Which signal we send is irrelevant. */ 1971 static inline void wake_ack_receiver(struct drbd_connection *connection) 1972 { 1973 struct task_struct *task = connection->ack_receiver.task; 1974 if (task && get_t_state(&connection->ack_receiver) == RUNNING) 1975 force_sig(SIGXCPU, task); 1976 } 1977 1978 static inline void request_ping(struct drbd_connection *connection) 1979 { 1980 set_bit(SEND_PING, &connection->flags); 1981 wake_ack_receiver(connection); 1982 } 1983 1984 extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *); 1985 extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *); 1986 extern int conn_send_command(struct drbd_connection *, struct drbd_socket *, 1987 enum drbd_packet, unsigned int, void *, 1988 unsigned int); 1989 extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *, 1990 enum drbd_packet, unsigned int, void *, 1991 unsigned int); 1992 1993 extern int drbd_send_ping(struct drbd_connection *connection); 1994 extern int drbd_send_ping_ack(struct drbd_connection *connection); 1995 extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state); 1996 extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state); 1997 1998 static inline void drbd_thread_stop(struct drbd_thread *thi) 1999 { 2000 _drbd_thread_stop(thi, false, true); 2001 } 2002 2003 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi) 2004 { 2005 _drbd_thread_stop(thi, false, false); 2006 } 2007 2008 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi) 2009 { 2010 _drbd_thread_stop(thi, true, false); 2011 } 2012 2013 /* counts how many answer packets packets we expect from our peer, 2014 * for either explicit application requests, 2015 * or implicit barrier packets as necessary. 2016 * increased: 2017 * w_send_barrier 2018 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ); 2019 * it is much easier and equally valid to count what we queue for the 2020 * worker, even before it actually was queued or send. 2021 * (drbd_make_request_common; recovery path on read io-error) 2022 * decreased: 2023 * got_BarrierAck (respective tl_clear, tl_clear_barrier) 2024 * _req_mod(req, DATA_RECEIVED) 2025 * [from receive_DataReply] 2026 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED) 2027 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)] 2028 * for some reason it is NOT decreased in got_NegAck, 2029 * but in the resulting cleanup code from report_params. 2030 * we should try to remember the reason for that... 2031 * _req_mod(req, SEND_FAILED or SEND_CANCELED) 2032 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING) 2033 * [from tl_clear_barrier] 2034 */ 2035 static inline void inc_ap_pending(struct drbd_device *device) 2036 { 2037 atomic_inc(&device->ap_pending_cnt); 2038 } 2039 2040 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \ 2041 if (atomic_read(&device->which) < 0) \ 2042 drbd_err(device, "in %s:%d: " #which " = %d < 0 !\n", \ 2043 func, line, \ 2044 atomic_read(&device->which)) 2045 2046 #define dec_ap_pending(device) _dec_ap_pending(device, __func__, __LINE__) 2047 static inline void _dec_ap_pending(struct drbd_device *device, const char *func, int line) 2048 { 2049 if (atomic_dec_and_test(&device->ap_pending_cnt)) 2050 wake_up(&device->misc_wait); 2051 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line); 2052 } 2053 2054 /* counts how many resync-related answers we still expect from the peer 2055 * increase decrease 2056 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY) 2057 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER) 2058 * (or P_NEG_ACK with ID_SYNCER) 2059 */ 2060 static inline void inc_rs_pending(struct drbd_device *device) 2061 { 2062 atomic_inc(&device->rs_pending_cnt); 2063 } 2064 2065 #define dec_rs_pending(device) _dec_rs_pending(device, __func__, __LINE__) 2066 static inline void _dec_rs_pending(struct drbd_device *device, const char *func, int line) 2067 { 2068 atomic_dec(&device->rs_pending_cnt); 2069 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line); 2070 } 2071 2072 /* counts how many answers we still need to send to the peer. 2073 * increased on 2074 * receive_Data unless protocol A; 2075 * we need to send a P_RECV_ACK (proto B) 2076 * or P_WRITE_ACK (proto C) 2077 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK 2078 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA 2079 * receive_Barrier_* we need to send a P_BARRIER_ACK 2080 */ 2081 static inline void inc_unacked(struct drbd_device *device) 2082 { 2083 atomic_inc(&device->unacked_cnt); 2084 } 2085 2086 #define dec_unacked(device) _dec_unacked(device, __func__, __LINE__) 2087 static inline void _dec_unacked(struct drbd_device *device, const char *func, int line) 2088 { 2089 atomic_dec(&device->unacked_cnt); 2090 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line); 2091 } 2092 2093 #define sub_unacked(device, n) _sub_unacked(device, n, __func__, __LINE__) 2094 static inline void _sub_unacked(struct drbd_device *device, int n, const char *func, int line) 2095 { 2096 atomic_sub(n, &device->unacked_cnt); 2097 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line); 2098 } 2099 2100 static inline bool is_sync_target_state(enum drbd_conns connection_state) 2101 { 2102 return connection_state == C_SYNC_TARGET || 2103 connection_state == C_PAUSED_SYNC_T; 2104 } 2105 2106 static inline bool is_sync_source_state(enum drbd_conns connection_state) 2107 { 2108 return connection_state == C_SYNC_SOURCE || 2109 connection_state == C_PAUSED_SYNC_S; 2110 } 2111 2112 static inline bool is_sync_state(enum drbd_conns connection_state) 2113 { 2114 return is_sync_source_state(connection_state) || 2115 is_sync_target_state(connection_state); 2116 } 2117 2118 /** 2119 * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev 2120 * @_device: DRBD device. 2121 * @_min_state: Minimum device state required for success. 2122 * 2123 * You have to call put_ldev() when finished working with device->ldev. 2124 */ 2125 #define get_ldev_if_state(_device, _min_state) \ 2126 (_get_ldev_if_state((_device), (_min_state)) ? \ 2127 ({ __acquire(x); true; }) : false) 2128 #define get_ldev(_device) get_ldev_if_state(_device, D_INCONSISTENT) 2129 2130 static inline void put_ldev(struct drbd_device *device) 2131 { 2132 enum drbd_disk_state disk_state = device->state.disk; 2133 /* We must check the state *before* the atomic_dec becomes visible, 2134 * or we have a theoretical race where someone hitting zero, 2135 * while state still D_FAILED, will then see D_DISKLESS in the 2136 * condition below and calling into destroy, where he must not, yet. */ 2137 int i = atomic_dec_return(&device->local_cnt); 2138 2139 /* This may be called from some endio handler, 2140 * so we must not sleep here. */ 2141 2142 __release(local); 2143 D_ASSERT(device, i >= 0); 2144 if (i == 0) { 2145 if (disk_state == D_DISKLESS) 2146 /* even internal references gone, safe to destroy */ 2147 drbd_device_post_work(device, DESTROY_DISK); 2148 if (disk_state == D_FAILED) 2149 /* all application IO references gone. */ 2150 if (!test_and_set_bit(GOING_DISKLESS, &device->flags)) 2151 drbd_device_post_work(device, GO_DISKLESS); 2152 wake_up(&device->misc_wait); 2153 } 2154 } 2155 2156 #ifndef __CHECKER__ 2157 static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins) 2158 { 2159 int io_allowed; 2160 2161 /* never get a reference while D_DISKLESS */ 2162 if (device->state.disk == D_DISKLESS) 2163 return 0; 2164 2165 atomic_inc(&device->local_cnt); 2166 io_allowed = (device->state.disk >= mins); 2167 if (!io_allowed) 2168 put_ldev(device); 2169 return io_allowed; 2170 } 2171 #else 2172 extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins); 2173 #endif 2174 2175 /* this throttles on-the-fly application requests 2176 * according to max_buffers settings; 2177 * maybe re-implement using semaphores? */ 2178 static inline int drbd_get_max_buffers(struct drbd_device *device) 2179 { 2180 struct net_conf *nc; 2181 int mxb; 2182 2183 rcu_read_lock(); 2184 nc = rcu_dereference(first_peer_device(device)->connection->net_conf); 2185 mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */ 2186 rcu_read_unlock(); 2187 2188 return mxb; 2189 } 2190 2191 static inline int drbd_state_is_stable(struct drbd_device *device) 2192 { 2193 union drbd_dev_state s = device->state; 2194 2195 /* DO NOT add a default clause, we want the compiler to warn us 2196 * for any newly introduced state we may have forgotten to add here */ 2197 2198 switch ((enum drbd_conns)s.conn) { 2199 /* new io only accepted when there is no connection, ... */ 2200 case C_STANDALONE: 2201 case C_WF_CONNECTION: 2202 /* ... or there is a well established connection. */ 2203 case C_CONNECTED: 2204 case C_SYNC_SOURCE: 2205 case C_SYNC_TARGET: 2206 case C_VERIFY_S: 2207 case C_VERIFY_T: 2208 case C_PAUSED_SYNC_S: 2209 case C_PAUSED_SYNC_T: 2210 case C_AHEAD: 2211 case C_BEHIND: 2212 /* transitional states, IO allowed */ 2213 case C_DISCONNECTING: 2214 case C_UNCONNECTED: 2215 case C_TIMEOUT: 2216 case C_BROKEN_PIPE: 2217 case C_NETWORK_FAILURE: 2218 case C_PROTOCOL_ERROR: 2219 case C_TEAR_DOWN: 2220 case C_WF_REPORT_PARAMS: 2221 case C_STARTING_SYNC_S: 2222 case C_STARTING_SYNC_T: 2223 break; 2224 2225 /* Allow IO in BM exchange states with new protocols */ 2226 case C_WF_BITMAP_S: 2227 if (first_peer_device(device)->connection->agreed_pro_version < 96) 2228 return 0; 2229 break; 2230 2231 /* no new io accepted in these states */ 2232 case C_WF_BITMAP_T: 2233 case C_WF_SYNC_UUID: 2234 case C_MASK: 2235 /* not "stable" */ 2236 return 0; 2237 } 2238 2239 switch ((enum drbd_disk_state)s.disk) { 2240 case D_DISKLESS: 2241 case D_INCONSISTENT: 2242 case D_OUTDATED: 2243 case D_CONSISTENT: 2244 case D_UP_TO_DATE: 2245 case D_FAILED: 2246 /* disk state is stable as well. */ 2247 break; 2248 2249 /* no new io accepted during transitional states */ 2250 case D_ATTACHING: 2251 case D_NEGOTIATING: 2252 case D_UNKNOWN: 2253 case D_MASK: 2254 /* not "stable" */ 2255 return 0; 2256 } 2257 2258 return 1; 2259 } 2260 2261 static inline int drbd_suspended(struct drbd_device *device) 2262 { 2263 struct drbd_resource *resource = device->resource; 2264 2265 return resource->susp || resource->susp_fen || resource->susp_nod; 2266 } 2267 2268 static inline bool may_inc_ap_bio(struct drbd_device *device) 2269 { 2270 int mxb = drbd_get_max_buffers(device); 2271 2272 if (drbd_suspended(device)) 2273 return false; 2274 if (atomic_read(&device->suspend_cnt)) 2275 return false; 2276 2277 /* to avoid potential deadlock or bitmap corruption, 2278 * in various places, we only allow new application io 2279 * to start during "stable" states. */ 2280 2281 /* no new io accepted when attaching or detaching the disk */ 2282 if (!drbd_state_is_stable(device)) 2283 return false; 2284 2285 /* since some older kernels don't have atomic_add_unless, 2286 * and we are within the spinlock anyways, we have this workaround. */ 2287 if (atomic_read(&device->ap_bio_cnt) > mxb) 2288 return false; 2289 if (test_bit(BITMAP_IO, &device->flags)) 2290 return false; 2291 return true; 2292 } 2293 2294 static inline bool inc_ap_bio_cond(struct drbd_device *device) 2295 { 2296 bool rv = false; 2297 2298 spin_lock_irq(&device->resource->req_lock); 2299 rv = may_inc_ap_bio(device); 2300 if (rv) 2301 atomic_inc(&device->ap_bio_cnt); 2302 spin_unlock_irq(&device->resource->req_lock); 2303 2304 return rv; 2305 } 2306 2307 static inline void inc_ap_bio(struct drbd_device *device) 2308 { 2309 /* we wait here 2310 * as long as the device is suspended 2311 * until the bitmap is no longer on the fly during connection 2312 * handshake as long as we would exceed the max_buffer limit. 2313 * 2314 * to avoid races with the reconnect code, 2315 * we need to atomic_inc within the spinlock. */ 2316 2317 wait_event(device->misc_wait, inc_ap_bio_cond(device)); 2318 } 2319 2320 static inline void dec_ap_bio(struct drbd_device *device) 2321 { 2322 int mxb = drbd_get_max_buffers(device); 2323 int ap_bio = atomic_dec_return(&device->ap_bio_cnt); 2324 2325 D_ASSERT(device, ap_bio >= 0); 2326 2327 if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) { 2328 if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags)) 2329 drbd_queue_work(&first_peer_device(device)-> 2330 connection->sender_work, 2331 &device->bm_io_work.w); 2332 } 2333 2334 /* this currently does wake_up for every dec_ap_bio! 2335 * maybe rather introduce some type of hysteresis? 2336 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */ 2337 if (ap_bio < mxb) 2338 wake_up(&device->misc_wait); 2339 } 2340 2341 static inline bool verify_can_do_stop_sector(struct drbd_device *device) 2342 { 2343 return first_peer_device(device)->connection->agreed_pro_version >= 97 && 2344 first_peer_device(device)->connection->agreed_pro_version != 100; 2345 } 2346 2347 static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val) 2348 { 2349 int changed = device->ed_uuid != val; 2350 device->ed_uuid = val; 2351 return changed; 2352 } 2353 2354 static inline int drbd_queue_order_type(struct drbd_device *device) 2355 { 2356 /* sorry, we currently have no working implementation 2357 * of distributed TCQ stuff */ 2358 #ifndef QUEUE_ORDERED_NONE 2359 #define QUEUE_ORDERED_NONE 0 2360 #endif 2361 return QUEUE_ORDERED_NONE; 2362 } 2363 2364 static inline struct drbd_connection *first_connection(struct drbd_resource *resource) 2365 { 2366 return list_first_entry_or_null(&resource->connections, 2367 struct drbd_connection, connections); 2368 } 2369 2370 #endif 2371