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 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call"))) 59 #else 60 # define __protected_by(x) 61 # define __protected_read_by(x) 62 # define __protected_write_by(x) 63 # define __must_hold(x) 64 #endif 65 66 /* shared module parameters, defined in drbd_main.c */ 67 #ifdef CONFIG_DRBD_FAULT_INJECTION 68 extern int drbd_enable_faults; 69 extern int drbd_fault_rate; 70 #endif 71 72 extern unsigned int drbd_minor_count; 73 extern char drbd_usermode_helper[]; 74 extern int drbd_proc_details; 75 76 77 /* This is used to stop/restart our threads. 78 * Cannot use SIGTERM nor SIGKILL, since these 79 * are sent out by init on runlevel changes 80 * I choose SIGHUP for now. 81 */ 82 #define DRBD_SIGKILL SIGHUP 83 84 #define ID_IN_SYNC (4711ULL) 85 #define ID_OUT_OF_SYNC (4712ULL) 86 #define ID_SYNCER (-1ULL) 87 88 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL) 89 90 struct drbd_device; 91 struct drbd_connection; 92 93 #define __drbd_printk_device(level, device, fmt, args...) \ 94 dev_printk(level, disk_to_dev((device)->vdisk), fmt, ## args) 95 #define __drbd_printk_peer_device(level, peer_device, fmt, args...) \ 96 dev_printk(level, disk_to_dev((peer_device)->device->vdisk), fmt, ## args) 97 #define __drbd_printk_resource(level, resource, fmt, args...) \ 98 printk(level "drbd %s: " fmt, (resource)->name, ## args) 99 #define __drbd_printk_connection(level, connection, fmt, args...) \ 100 printk(level "drbd %s: " fmt, (connection)->resource->name, ## args) 101 102 void drbd_printk_with_wrong_object_type(void); 103 104 #define __drbd_printk_if_same_type(obj, type, func, level, fmt, args...) \ 105 (__builtin_types_compatible_p(typeof(obj), type) || \ 106 __builtin_types_compatible_p(typeof(obj), const type)), \ 107 func(level, (const type)(obj), fmt, ## args) 108 109 #define drbd_printk(level, obj, fmt, args...) \ 110 __builtin_choose_expr( \ 111 __drbd_printk_if_same_type(obj, struct drbd_device *, \ 112 __drbd_printk_device, level, fmt, ## args), \ 113 __builtin_choose_expr( \ 114 __drbd_printk_if_same_type(obj, struct drbd_resource *, \ 115 __drbd_printk_resource, level, fmt, ## args), \ 116 __builtin_choose_expr( \ 117 __drbd_printk_if_same_type(obj, struct drbd_connection *, \ 118 __drbd_printk_connection, level, fmt, ## args), \ 119 __builtin_choose_expr( \ 120 __drbd_printk_if_same_type(obj, struct drbd_peer_device *, \ 121 __drbd_printk_peer_device, level, fmt, ## args), \ 122 drbd_printk_with_wrong_object_type())))) 123 124 #define drbd_dbg(obj, fmt, args...) \ 125 drbd_printk(KERN_DEBUG, obj, fmt, ## args) 126 #define drbd_alert(obj, fmt, args...) \ 127 drbd_printk(KERN_ALERT, obj, fmt, ## args) 128 #define drbd_err(obj, fmt, args...) \ 129 drbd_printk(KERN_ERR, obj, fmt, ## args) 130 #define drbd_warn(obj, fmt, args...) \ 131 drbd_printk(KERN_WARNING, obj, fmt, ## args) 132 #define drbd_info(obj, fmt, args...) \ 133 drbd_printk(KERN_INFO, obj, fmt, ## args) 134 #define drbd_emerg(obj, fmt, args...) \ 135 drbd_printk(KERN_EMERG, obj, fmt, ## args) 136 137 #define dynamic_drbd_dbg(device, fmt, args...) \ 138 dynamic_dev_dbg(disk_to_dev(device->vdisk), fmt, ## args) 139 140 #define D_ASSERT(device, exp) do { \ 141 if (!(exp)) \ 142 drbd_err(device, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__); \ 143 } while (0) 144 145 /** 146 * expect - Make an assertion 147 * 148 * Unlike the assert macro, this macro returns a boolean result. 149 */ 150 #define expect(exp) ({ \ 151 bool _bool = (exp); \ 152 if (!_bool) \ 153 drbd_err(device, "ASSERTION %s FAILED in %s\n", \ 154 #exp, __func__); \ 155 _bool; \ 156 }) 157 158 /* Defines to control fault insertion */ 159 enum { 160 DRBD_FAULT_MD_WR = 0, /* meta data write */ 161 DRBD_FAULT_MD_RD = 1, /* read */ 162 DRBD_FAULT_RS_WR = 2, /* resync */ 163 DRBD_FAULT_RS_RD = 3, 164 DRBD_FAULT_DT_WR = 4, /* data */ 165 DRBD_FAULT_DT_RD = 5, 166 DRBD_FAULT_DT_RA = 6, /* data read ahead */ 167 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */ 168 DRBD_FAULT_AL_EE = 8, /* alloc ee */ 169 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */ 170 171 DRBD_FAULT_MAX, 172 }; 173 174 extern unsigned int 175 _drbd_insert_fault(struct drbd_device *device, unsigned int type); 176 177 static inline int 178 drbd_insert_fault(struct drbd_device *device, unsigned int type) { 179 #ifdef CONFIG_DRBD_FAULT_INJECTION 180 return drbd_fault_rate && 181 (drbd_enable_faults & (1<<type)) && 182 _drbd_insert_fault(device, type); 183 #else 184 return 0; 185 #endif 186 } 187 188 /* integer division, round _UP_ to the next integer */ 189 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0)) 190 /* usual integer division */ 191 #define div_floor(A, B) ((A)/(B)) 192 193 extern struct ratelimit_state drbd_ratelimit_state; 194 extern struct idr drbd_devices; /* RCU, updates: genl_lock() */ 195 extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */ 196 197 extern const char *cmdname(enum drbd_packet cmd); 198 199 /* for sending/receiving the bitmap, 200 * possibly in some encoding scheme */ 201 struct bm_xfer_ctx { 202 /* "const" 203 * stores total bits and long words 204 * of the bitmap, so we don't need to 205 * call the accessor functions over and again. */ 206 unsigned long bm_bits; 207 unsigned long bm_words; 208 /* during xfer, current position within the bitmap */ 209 unsigned long bit_offset; 210 unsigned long word_offset; 211 212 /* statistics; index: (h->command == P_BITMAP) */ 213 unsigned packets[2]; 214 unsigned bytes[2]; 215 }; 216 217 extern void INFO_bm_xfer_stats(struct drbd_device *device, 218 const char *direction, struct bm_xfer_ctx *c); 219 220 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c) 221 { 222 /* word_offset counts "native long words" (32 or 64 bit), 223 * aligned at 64 bit. 224 * Encoded packet may end at an unaligned bit offset. 225 * In case a fallback clear text packet is transmitted in 226 * between, we adjust this offset back to the last 64bit 227 * aligned "native long word", which makes coding and decoding 228 * the plain text bitmap much more convenient. */ 229 #if BITS_PER_LONG == 64 230 c->word_offset = c->bit_offset >> 6; 231 #elif BITS_PER_LONG == 32 232 c->word_offset = c->bit_offset >> 5; 233 c->word_offset &= ~(1UL); 234 #else 235 # error "unsupported BITS_PER_LONG" 236 #endif 237 } 238 239 extern unsigned int drbd_header_size(struct drbd_connection *connection); 240 241 /**********************************************************************/ 242 enum drbd_thread_state { 243 NONE, 244 RUNNING, 245 EXITING, 246 RESTARTING 247 }; 248 249 struct drbd_thread { 250 spinlock_t t_lock; 251 struct task_struct *task; 252 struct completion stop; 253 enum drbd_thread_state t_state; 254 int (*function) (struct drbd_thread *); 255 struct drbd_resource *resource; 256 struct drbd_connection *connection; 257 int reset_cpu_mask; 258 const char *name; 259 }; 260 261 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi) 262 { 263 /* THINK testing the t_state seems to be uncritical in all cases 264 * (but thread_{start,stop}), so we can read it *without* the lock. 265 * --lge */ 266 267 smp_rmb(); 268 return thi->t_state; 269 } 270 271 struct drbd_work { 272 struct list_head list; 273 int (*cb)(struct drbd_work *, int cancel); 274 }; 275 276 struct drbd_device_work { 277 struct drbd_work w; 278 struct drbd_device *device; 279 }; 280 281 #include "drbd_interval.h" 282 283 extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *); 284 285 extern void lock_all_resources(void); 286 extern void unlock_all_resources(void); 287 288 struct drbd_request { 289 struct drbd_work w; 290 struct drbd_device *device; 291 292 /* if local IO is not allowed, will be NULL. 293 * if local IO _is_ allowed, holds the locally submitted bio clone, 294 * or, after local IO completion, the ERR_PTR(error). 295 * see drbd_request_endio(). */ 296 struct bio *private_bio; 297 298 struct drbd_interval i; 299 300 /* epoch: used to check on "completion" whether this req was in 301 * the current epoch, and we therefore have to close it, 302 * causing a p_barrier packet to be send, starting a new epoch. 303 * 304 * This corresponds to "barrier" in struct p_barrier[_ack], 305 * and to "barrier_nr" in struct drbd_epoch (and various 306 * comments/function parameters/local variable names). 307 */ 308 unsigned int epoch; 309 310 struct list_head tl_requests; /* ring list in the transfer log */ 311 struct bio *master_bio; /* master bio pointer */ 312 313 /* see struct drbd_device */ 314 struct list_head req_pending_master_completion; 315 struct list_head req_pending_local; 316 317 /* for generic IO accounting */ 318 unsigned long start_jif; 319 320 /* for DRBD internal statistics */ 321 322 /* Minimal set of time stamps to determine if we wait for activity log 323 * transactions, local disk or peer. 32 bit "jiffies" are good enough, 324 * we don't expect a DRBD request to be stalled for several month. 325 */ 326 327 /* before actual request processing */ 328 unsigned long in_actlog_jif; 329 330 /* local disk */ 331 unsigned long pre_submit_jif; 332 333 /* per connection */ 334 unsigned long pre_send_jif; 335 unsigned long acked_jif; 336 unsigned long net_done_jif; 337 338 /* Possibly even more detail to track each phase: 339 * master_completion_jif 340 * how long did it take to complete the master bio 341 * (application visible latency) 342 * allocated_jif 343 * how long the master bio was blocked until we finally allocated 344 * a tracking struct 345 * in_actlog_jif 346 * how long did we wait for activity log transactions 347 * 348 * net_queued_jif 349 * when did we finally queue it for sending 350 * pre_send_jif 351 * when did we start sending it 352 * post_send_jif 353 * how long did we block in the network stack trying to send it 354 * acked_jif 355 * when did we receive (or fake, in protocol A) a remote ACK 356 * net_done_jif 357 * when did we receive final acknowledgement (P_BARRIER_ACK), 358 * or decide, e.g. on connection loss, that we do no longer expect 359 * anything from this peer for this request. 360 * 361 * pre_submit_jif 362 * post_sub_jif 363 * when did we start submiting to the lower level device, 364 * and how long did we block in that submit function 365 * local_completion_jif 366 * how long did it take the lower level device to complete this request 367 */ 368 369 370 /* once it hits 0, we may complete the master_bio */ 371 atomic_t completion_ref; 372 /* once it hits 0, we may destroy this drbd_request object */ 373 struct kref kref; 374 375 unsigned rq_state; /* see comments above _req_mod() */ 376 }; 377 378 struct drbd_epoch { 379 struct drbd_connection *connection; 380 struct list_head list; 381 unsigned int barrier_nr; 382 atomic_t epoch_size; /* increased on every request added. */ 383 atomic_t active; /* increased on every req. added, and dec on every finished. */ 384 unsigned long flags; 385 }; 386 387 /* Prototype declaration of function defined in drbd_receiver.c */ 388 int drbdd_init(struct drbd_thread *); 389 int drbd_asender(struct drbd_thread *); 390 391 /* drbd_epoch flag bits */ 392 enum { 393 DE_HAVE_BARRIER_NUMBER, 394 }; 395 396 enum epoch_event { 397 EV_PUT, 398 EV_GOT_BARRIER_NR, 399 EV_BECAME_LAST, 400 EV_CLEANUP = 32, /* used as flag */ 401 }; 402 403 struct digest_info { 404 int digest_size; 405 void *digest; 406 }; 407 408 struct drbd_peer_request { 409 struct drbd_work w; 410 struct drbd_peer_device *peer_device; 411 struct drbd_epoch *epoch; /* for writes */ 412 struct page *pages; 413 atomic_t pending_bios; 414 struct drbd_interval i; 415 /* see comments on ee flag bits below */ 416 unsigned long flags; 417 unsigned long submit_jif; 418 union { 419 u64 block_id; 420 struct digest_info *digest; 421 }; 422 }; 423 424 /* ee flag bits. 425 * While corresponding bios are in flight, the only modification will be 426 * set_bit WAS_ERROR, which has to be atomic. 427 * If no bios are in flight yet, or all have been completed, 428 * non-atomic modification to ee->flags is ok. 429 */ 430 enum { 431 __EE_CALL_AL_COMPLETE_IO, 432 __EE_MAY_SET_IN_SYNC, 433 434 /* is this a TRIM aka REQ_DISCARD? */ 435 __EE_IS_TRIM, 436 437 /* In case a barrier failed, 438 * we need to resubmit without the barrier flag. */ 439 __EE_RESUBMITTED, 440 441 /* we may have several bios per peer request. 442 * if any of those fail, we set this flag atomically 443 * from the endio callback */ 444 __EE_WAS_ERROR, 445 446 /* This ee has a pointer to a digest instead of a block id */ 447 __EE_HAS_DIGEST, 448 449 /* Conflicting local requests need to be restarted after this request */ 450 __EE_RESTART_REQUESTS, 451 452 /* The peer wants a write ACK for this (wire proto C) */ 453 __EE_SEND_WRITE_ACK, 454 455 /* Is set when net_conf had two_primaries set while creating this peer_req */ 456 __EE_IN_INTERVAL_TREE, 457 458 /* for debugfs: */ 459 /* has this been submitted, or does it still wait for something else? */ 460 __EE_SUBMITTED, 461 462 /* this is/was a write request */ 463 __EE_WRITE, 464 465 /* this is/was a write same request */ 466 __EE_WRITE_SAME, 467 468 /* this originates from application on peer 469 * (not some resync or verify or other DRBD internal request) */ 470 __EE_APPLICATION, 471 472 /* If it contains only 0 bytes, send back P_RS_DEALLOCATED */ 473 __EE_RS_THIN_REQ, 474 }; 475 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO) 476 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC) 477 #define EE_IS_TRIM (1<<__EE_IS_TRIM) 478 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED) 479 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR) 480 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST) 481 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS) 482 #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK) 483 #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE) 484 #define EE_SUBMITTED (1<<__EE_SUBMITTED) 485 #define EE_WRITE (1<<__EE_WRITE) 486 #define EE_WRITE_SAME (1<<__EE_WRITE_SAME) 487 #define EE_APPLICATION (1<<__EE_APPLICATION) 488 #define EE_RS_THIN_REQ (1<<__EE_RS_THIN_REQ) 489 490 /* flag bits per device */ 491 enum { 492 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */ 493 MD_DIRTY, /* current uuids and flags not yet on disk */ 494 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */ 495 CL_ST_CHG_SUCCESS, 496 CL_ST_CHG_FAIL, 497 CRASHED_PRIMARY, /* This node was a crashed primary. 498 * Gets cleared when the state.conn 499 * goes into C_CONNECTED state. */ 500 CONSIDER_RESYNC, 501 502 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */ 503 504 BITMAP_IO, /* suspend application io; 505 once no more io in flight, start bitmap io */ 506 BITMAP_IO_QUEUED, /* Started bitmap IO */ 507 WAS_IO_ERROR, /* Local disk failed, returned IO error */ 508 WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */ 509 FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */ 510 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */ 511 RESIZE_PENDING, /* Size change detected locally, waiting for the response from 512 * the peer, if it changed there as well. */ 513 NEW_CUR_UUID, /* Create new current UUID when thawing IO */ 514 AL_SUSPENDED, /* Activity logging is currently suspended. */ 515 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */ 516 B_RS_H_DONE, /* Before resync handler done (already executed) */ 517 DISCARD_MY_DATA, /* discard_my_data flag per volume */ 518 READ_BALANCE_RR, 519 520 FLUSH_PENDING, /* if set, device->flush_jif is when we submitted that flush 521 * from drbd_flush_after_epoch() */ 522 523 /* cleared only after backing device related structures have been destroyed. */ 524 GOING_DISKLESS, /* Disk is being detached, because of io-error, or admin request. */ 525 526 /* to be used in drbd_device_post_work() */ 527 GO_DISKLESS, /* tell worker to schedule cleanup before detach */ 528 DESTROY_DISK, /* tell worker to close backing devices and destroy related structures. */ 529 MD_SYNC, /* tell worker to call drbd_md_sync() */ 530 RS_START, /* tell worker to start resync/OV */ 531 RS_PROGRESS, /* tell worker that resync made significant progress */ 532 RS_DONE, /* tell worker that resync is done */ 533 }; 534 535 struct drbd_bitmap; /* opaque for drbd_device */ 536 537 /* definition of bits in bm_flags to be used in drbd_bm_lock 538 * and drbd_bitmap_io and friends. */ 539 enum bm_flag { 540 /* currently locked for bulk operation */ 541 BM_LOCKED_MASK = 0xf, 542 543 /* in detail, that is: */ 544 BM_DONT_CLEAR = 0x1, 545 BM_DONT_SET = 0x2, 546 BM_DONT_TEST = 0x4, 547 548 /* so we can mark it locked for bulk operation, 549 * and still allow all non-bulk operations */ 550 BM_IS_LOCKED = 0x8, 551 552 /* (test bit, count bit) allowed (common case) */ 553 BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED, 554 555 /* testing bits, as well as setting new bits allowed, but clearing bits 556 * would be unexpected. Used during bitmap receive. Setting new bits 557 * requires sending of "out-of-sync" information, though. */ 558 BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED, 559 560 /* for drbd_bm_write_copy_pages, everything is allowed, 561 * only concurrent bulk operations are locked out. */ 562 BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED, 563 }; 564 565 struct drbd_work_queue { 566 struct list_head q; 567 spinlock_t q_lock; /* to protect the list. */ 568 wait_queue_head_t q_wait; 569 }; 570 571 struct drbd_socket { 572 struct mutex mutex; 573 struct socket *socket; 574 /* this way we get our 575 * send/receive buffers off the stack */ 576 void *sbuf; 577 void *rbuf; 578 }; 579 580 struct drbd_md { 581 u64 md_offset; /* sector offset to 'super' block */ 582 583 u64 la_size_sect; /* last agreed size, unit sectors */ 584 spinlock_t uuid_lock; 585 u64 uuid[UI_SIZE]; 586 u64 device_uuid; 587 u32 flags; 588 u32 md_size_sect; 589 590 s32 al_offset; /* signed relative sector offset to activity log */ 591 s32 bm_offset; /* signed relative sector offset to bitmap */ 592 593 /* cached value of bdev->disk_conf->meta_dev_idx (see below) */ 594 s32 meta_dev_idx; 595 596 /* see al_tr_number_to_on_disk_sector() */ 597 u32 al_stripes; 598 u32 al_stripe_size_4k; 599 u32 al_size_4k; /* cached product of the above */ 600 }; 601 602 struct drbd_backing_dev { 603 struct block_device *backing_bdev; 604 struct block_device *md_bdev; 605 struct drbd_md md; 606 struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */ 607 sector_t known_size; /* last known size of that backing device */ 608 }; 609 610 struct drbd_md_io { 611 struct page *page; 612 unsigned long start_jif; /* last call to drbd_md_get_buffer */ 613 unsigned long submit_jif; /* last _drbd_md_sync_page_io() submit */ 614 const char *current_use; 615 atomic_t in_use; 616 unsigned int done; 617 int error; 618 }; 619 620 struct bm_io_work { 621 struct drbd_work w; 622 char *why; 623 enum bm_flag flags; 624 int (*io_fn)(struct drbd_device *device); 625 void (*done)(struct drbd_device *device, int rv); 626 }; 627 628 struct fifo_buffer { 629 unsigned int head_index; 630 unsigned int size; 631 int total; /* sum of all values */ 632 int values[0]; 633 }; 634 extern struct fifo_buffer *fifo_alloc(int fifo_size); 635 636 /* flag bits per connection */ 637 enum { 638 NET_CONGESTED, /* The data socket is congested */ 639 RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */ 640 SEND_PING, 641 GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */ 642 CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */ 643 CONN_WD_ST_CHG_OKAY, 644 CONN_WD_ST_CHG_FAIL, 645 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */ 646 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */ 647 STATE_SENT, /* Do not change state/UUIDs while this is set */ 648 CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC) 649 * pending, from drbd worker context. 650 * If set, bdi_write_congested() returns true, 651 * so shrink_page_list() would not recurse into, 652 * and potentially deadlock on, this drbd worker. 653 */ 654 DISCONNECT_SENT, 655 656 DEVICE_WORK_PENDING, /* tell worker that some device has pending work */ 657 }; 658 659 enum which_state { NOW, OLD = NOW, NEW }; 660 661 struct drbd_resource { 662 char *name; 663 #ifdef CONFIG_DEBUG_FS 664 struct dentry *debugfs_res; 665 struct dentry *debugfs_res_volumes; 666 struct dentry *debugfs_res_connections; 667 struct dentry *debugfs_res_in_flight_summary; 668 #endif 669 struct kref kref; 670 struct idr devices; /* volume number to device mapping */ 671 struct list_head connections; 672 struct list_head resources; 673 struct res_opts res_opts; 674 struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */ 675 struct mutex adm_mutex; /* mutex to serialize administrative requests */ 676 spinlock_t req_lock; 677 678 unsigned susp:1; /* IO suspended by user */ 679 unsigned susp_nod:1; /* IO suspended because no data */ 680 unsigned susp_fen:1; /* IO suspended because fence peer handler runs */ 681 682 enum write_ordering_e write_ordering; 683 684 cpumask_var_t cpu_mask; 685 }; 686 687 struct drbd_thread_timing_details 688 { 689 unsigned long start_jif; 690 void *cb_addr; 691 const char *caller_fn; 692 unsigned int line; 693 unsigned int cb_nr; 694 }; 695 696 struct drbd_connection { 697 struct list_head connections; 698 struct drbd_resource *resource; 699 #ifdef CONFIG_DEBUG_FS 700 struct dentry *debugfs_conn; 701 struct dentry *debugfs_conn_callback_history; 702 struct dentry *debugfs_conn_oldest_requests; 703 #endif 704 struct kref kref; 705 struct idr peer_devices; /* volume number to peer device mapping */ 706 enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */ 707 struct mutex cstate_mutex; /* Protects graceful disconnects */ 708 unsigned int connect_cnt; /* Inc each time a connection is established */ 709 710 unsigned long flags; 711 struct net_conf *net_conf; /* content protected by rcu */ 712 wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */ 713 714 struct sockaddr_storage my_addr; 715 int my_addr_len; 716 struct sockaddr_storage peer_addr; 717 int peer_addr_len; 718 719 struct drbd_socket data; /* data/barrier/cstate/parameter packets */ 720 struct drbd_socket meta; /* ping/ack (metadata) packets */ 721 int agreed_pro_version; /* actually used protocol version */ 722 u32 agreed_features; 723 unsigned long last_received; /* in jiffies, either socket */ 724 unsigned int ko_count; 725 726 struct list_head transfer_log; /* all requests not yet fully processed */ 727 728 struct crypto_shash *cram_hmac_tfm; 729 struct crypto_ahash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */ 730 struct crypto_ahash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */ 731 struct crypto_ahash *csums_tfm; 732 struct crypto_ahash *verify_tfm; 733 void *int_dig_in; 734 void *int_dig_vv; 735 736 /* receiver side */ 737 struct drbd_epoch *current_epoch; 738 spinlock_t epoch_lock; 739 unsigned int epochs; 740 atomic_t current_tle_nr; /* transfer log epoch number */ 741 unsigned current_tle_writes; /* writes seen within this tl epoch */ 742 743 unsigned long last_reconnect_jif; 744 /* empty member on older kernels without blk_start_plug() */ 745 struct blk_plug receiver_plug; 746 struct drbd_thread receiver; 747 struct drbd_thread worker; 748 struct drbd_thread ack_receiver; 749 struct workqueue_struct *ack_sender; 750 751 /* cached pointers, 752 * so we can look up the oldest pending requests more quickly. 753 * protected by resource->req_lock */ 754 struct drbd_request *req_next; /* DRBD 9: todo.req_next */ 755 struct drbd_request *req_ack_pending; 756 struct drbd_request *req_not_net_done; 757 758 /* sender side */ 759 struct drbd_work_queue sender_work; 760 761 #define DRBD_THREAD_DETAILS_HIST 16 762 unsigned int w_cb_nr; /* keeps counting up */ 763 unsigned int r_cb_nr; /* keeps counting up */ 764 struct drbd_thread_timing_details w_timing_details[DRBD_THREAD_DETAILS_HIST]; 765 struct drbd_thread_timing_details r_timing_details[DRBD_THREAD_DETAILS_HIST]; 766 767 struct { 768 unsigned long last_sent_barrier_jif; 769 770 /* whether this sender thread 771 * has processed a single write yet. */ 772 bool seen_any_write_yet; 773 774 /* Which barrier number to send with the next P_BARRIER */ 775 int current_epoch_nr; 776 777 /* how many write requests have been sent 778 * with req->epoch == current_epoch_nr. 779 * If none, no P_BARRIER will be sent. */ 780 unsigned current_epoch_writes; 781 } send; 782 }; 783 784 static inline bool has_net_conf(struct drbd_connection *connection) 785 { 786 bool has_net_conf; 787 788 rcu_read_lock(); 789 has_net_conf = rcu_dereference(connection->net_conf); 790 rcu_read_unlock(); 791 792 return has_net_conf; 793 } 794 795 void __update_timing_details( 796 struct drbd_thread_timing_details *tdp, 797 unsigned int *cb_nr, 798 void *cb, 799 const char *fn, const unsigned int line); 800 801 #define update_worker_timing_details(c, cb) \ 802 __update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ ) 803 #define update_receiver_timing_details(c, cb) \ 804 __update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ ) 805 806 struct submit_worker { 807 struct workqueue_struct *wq; 808 struct work_struct worker; 809 810 /* protected by ..->resource->req_lock */ 811 struct list_head writes; 812 }; 813 814 struct drbd_peer_device { 815 struct list_head peer_devices; 816 struct drbd_device *device; 817 struct drbd_connection *connection; 818 struct work_struct send_acks_work; 819 #ifdef CONFIG_DEBUG_FS 820 struct dentry *debugfs_peer_dev; 821 #endif 822 }; 823 824 struct drbd_device { 825 struct drbd_resource *resource; 826 struct list_head peer_devices; 827 struct list_head pending_bitmap_io; 828 829 unsigned long flush_jif; 830 #ifdef CONFIG_DEBUG_FS 831 struct dentry *debugfs_minor; 832 struct dentry *debugfs_vol; 833 struct dentry *debugfs_vol_oldest_requests; 834 struct dentry *debugfs_vol_act_log_extents; 835 struct dentry *debugfs_vol_resync_extents; 836 struct dentry *debugfs_vol_data_gen_id; 837 struct dentry *debugfs_vol_ed_gen_id; 838 #endif 839 840 unsigned int vnr; /* volume number within the connection */ 841 unsigned int minor; /* device minor number */ 842 843 struct kref kref; 844 845 /* things that are stored as / read from meta data on disk */ 846 unsigned long flags; 847 848 /* configured by drbdsetup */ 849 struct drbd_backing_dev *ldev __protected_by(local); 850 851 sector_t p_size; /* partner's disk size */ 852 struct request_queue *rq_queue; 853 struct block_device *this_bdev; 854 struct gendisk *vdisk; 855 856 unsigned long last_reattach_jif; 857 struct drbd_work resync_work; 858 struct drbd_work unplug_work; 859 struct timer_list resync_timer; 860 struct timer_list md_sync_timer; 861 struct timer_list start_resync_timer; 862 struct timer_list request_timer; 863 864 /* Used after attach while negotiating new disk state. */ 865 union drbd_state new_state_tmp; 866 867 union drbd_dev_state state; 868 wait_queue_head_t misc_wait; 869 wait_queue_head_t state_wait; /* upon each state change. */ 870 unsigned int send_cnt; 871 unsigned int recv_cnt; 872 unsigned int read_cnt; 873 unsigned int writ_cnt; 874 unsigned int al_writ_cnt; 875 unsigned int bm_writ_cnt; 876 atomic_t ap_bio_cnt; /* Requests we need to complete */ 877 atomic_t ap_actlog_cnt; /* Requests waiting for activity log */ 878 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */ 879 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */ 880 atomic_t unacked_cnt; /* Need to send replies for */ 881 atomic_t local_cnt; /* Waiting for local completion */ 882 atomic_t suspend_cnt; 883 884 /* Interval tree of pending local requests */ 885 struct rb_root read_requests; 886 struct rb_root write_requests; 887 888 /* for statistics and timeouts */ 889 /* [0] read, [1] write */ 890 struct list_head pending_master_completion[2]; 891 struct list_head pending_completion[2]; 892 893 /* use checksums for *this* resync */ 894 bool use_csums; 895 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */ 896 unsigned long rs_total; 897 /* number of resync blocks that failed in this run */ 898 unsigned long rs_failed; 899 /* Syncer's start time [unit jiffies] */ 900 unsigned long rs_start; 901 /* cumulated time in PausedSyncX state [unit jiffies] */ 902 unsigned long rs_paused; 903 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */ 904 unsigned long rs_same_csum; 905 #define DRBD_SYNC_MARKS 8 906 #define DRBD_SYNC_MARK_STEP (3*HZ) 907 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */ 908 unsigned long rs_mark_left[DRBD_SYNC_MARKS]; 909 /* marks's time [unit jiffies] */ 910 unsigned long rs_mark_time[DRBD_SYNC_MARKS]; 911 /* current index into rs_mark_{left,time} */ 912 int rs_last_mark; 913 unsigned long rs_last_bcast; /* [unit jiffies] */ 914 915 /* where does the admin want us to start? (sector) */ 916 sector_t ov_start_sector; 917 sector_t ov_stop_sector; 918 /* where are we now? (sector) */ 919 sector_t ov_position; 920 /* Start sector of out of sync range (to merge printk reporting). */ 921 sector_t ov_last_oos_start; 922 /* size of out-of-sync range in sectors. */ 923 sector_t ov_last_oos_size; 924 unsigned long ov_left; /* in bits */ 925 926 struct drbd_bitmap *bitmap; 927 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */ 928 929 /* Used to track operations of resync... */ 930 struct lru_cache *resync; 931 /* Number of locked elements in resync LRU */ 932 unsigned int resync_locked; 933 /* resync extent number waiting for application requests */ 934 unsigned int resync_wenr; 935 936 int open_cnt; 937 u64 *p_uuid; 938 939 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */ 940 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */ 941 struct list_head done_ee; /* need to send P_WRITE_ACK */ 942 struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */ 943 struct list_head net_ee; /* zero-copy network send in progress */ 944 945 int next_barrier_nr; 946 struct list_head resync_reads; 947 atomic_t pp_in_use; /* allocated from page pool */ 948 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */ 949 wait_queue_head_t ee_wait; 950 struct drbd_md_io md_io; 951 spinlock_t al_lock; 952 wait_queue_head_t al_wait; 953 struct lru_cache *act_log; /* activity log */ 954 unsigned int al_tr_number; 955 int al_tr_cycle; 956 wait_queue_head_t seq_wait; 957 atomic_t packet_seq; 958 unsigned int peer_seq; 959 spinlock_t peer_seq_lock; 960 unsigned long comm_bm_set; /* communicated number of set bits. */ 961 struct bm_io_work bm_io_work; 962 u64 ed_uuid; /* UUID of the exposed data */ 963 struct mutex own_state_mutex; 964 struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */ 965 char congestion_reason; /* Why we where congested... */ 966 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */ 967 atomic_t rs_sect_ev; /* for submitted resync data rate, both */ 968 int rs_last_sect_ev; /* counter to compare with */ 969 int rs_last_events; /* counter of read or write "events" (unit sectors) 970 * on the lower level device when we last looked. */ 971 int c_sync_rate; /* current resync rate after syncer throttle magic */ 972 struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */ 973 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */ 974 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */ 975 unsigned int peer_max_bio_size; 976 unsigned int local_max_bio_size; 977 978 /* any requests that would block in drbd_make_request() 979 * are deferred to this single-threaded work queue */ 980 struct submit_worker submit; 981 }; 982 983 struct drbd_bm_aio_ctx { 984 struct drbd_device *device; 985 struct list_head list; /* on device->pending_bitmap_io */; 986 unsigned long start_jif; 987 atomic_t in_flight; 988 unsigned int done; 989 unsigned flags; 990 #define BM_AIO_COPY_PAGES 1 991 #define BM_AIO_WRITE_HINTED 2 992 #define BM_AIO_WRITE_ALL_PAGES 4 993 #define BM_AIO_READ 8 994 int error; 995 struct kref kref; 996 }; 997 998 struct drbd_config_context { 999 /* assigned from drbd_genlmsghdr */ 1000 unsigned int minor; 1001 /* assigned from request attributes, if present */ 1002 unsigned int volume; 1003 #define VOLUME_UNSPECIFIED (-1U) 1004 /* pointer into the request skb, 1005 * limited lifetime! */ 1006 char *resource_name; 1007 struct nlattr *my_addr; 1008 struct nlattr *peer_addr; 1009 1010 /* reply buffer */ 1011 struct sk_buff *reply_skb; 1012 /* pointer into reply buffer */ 1013 struct drbd_genlmsghdr *reply_dh; 1014 /* resolved from attributes, if possible */ 1015 struct drbd_device *device; 1016 struct drbd_resource *resource; 1017 struct drbd_connection *connection; 1018 }; 1019 1020 static inline struct drbd_device *minor_to_device(unsigned int minor) 1021 { 1022 return (struct drbd_device *)idr_find(&drbd_devices, minor); 1023 } 1024 1025 static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device) 1026 { 1027 return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices); 1028 } 1029 1030 static inline struct drbd_peer_device * 1031 conn_peer_device(struct drbd_connection *connection, int volume_number) 1032 { 1033 return idr_find(&connection->peer_devices, volume_number); 1034 } 1035 1036 #define for_each_resource(resource, _resources) \ 1037 list_for_each_entry(resource, _resources, resources) 1038 1039 #define for_each_resource_rcu(resource, _resources) \ 1040 list_for_each_entry_rcu(resource, _resources, resources) 1041 1042 #define for_each_resource_safe(resource, tmp, _resources) \ 1043 list_for_each_entry_safe(resource, tmp, _resources, resources) 1044 1045 #define for_each_connection(connection, resource) \ 1046 list_for_each_entry(connection, &resource->connections, connections) 1047 1048 #define for_each_connection_rcu(connection, resource) \ 1049 list_for_each_entry_rcu(connection, &resource->connections, connections) 1050 1051 #define for_each_connection_safe(connection, tmp, resource) \ 1052 list_for_each_entry_safe(connection, tmp, &resource->connections, connections) 1053 1054 #define for_each_peer_device(peer_device, device) \ 1055 list_for_each_entry(peer_device, &device->peer_devices, peer_devices) 1056 1057 #define for_each_peer_device_rcu(peer_device, device) \ 1058 list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices) 1059 1060 #define for_each_peer_device_safe(peer_device, tmp, device) \ 1061 list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices) 1062 1063 static inline unsigned int device_to_minor(struct drbd_device *device) 1064 { 1065 return device->minor; 1066 } 1067 1068 /* 1069 * function declarations 1070 *************************/ 1071 1072 /* drbd_main.c */ 1073 1074 enum dds_flags { 1075 DDSF_FORCED = 1, 1076 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */ 1077 }; 1078 1079 extern void drbd_init_set_defaults(struct drbd_device *device); 1080 extern int drbd_thread_start(struct drbd_thread *thi); 1081 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait); 1082 #ifdef CONFIG_SMP 1083 extern void drbd_thread_current_set_cpu(struct drbd_thread *thi); 1084 #else 1085 #define drbd_thread_current_set_cpu(A) ({}) 1086 #endif 1087 extern void tl_release(struct drbd_connection *, unsigned int barrier_nr, 1088 unsigned int set_size); 1089 extern void tl_clear(struct drbd_connection *); 1090 extern void drbd_free_sock(struct drbd_connection *connection); 1091 extern int drbd_send(struct drbd_connection *connection, struct socket *sock, 1092 void *buf, size_t size, unsigned msg_flags); 1093 extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t, 1094 unsigned); 1095 1096 extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd); 1097 extern int drbd_send_protocol(struct drbd_connection *connection); 1098 extern int drbd_send_uuids(struct drbd_peer_device *); 1099 extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *); 1100 extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *); 1101 extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags); 1102 extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s); 1103 extern int drbd_send_current_state(struct drbd_peer_device *); 1104 extern int drbd_send_sync_param(struct drbd_peer_device *); 1105 extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr, 1106 u32 set_size); 1107 extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet, 1108 struct drbd_peer_request *); 1109 extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet, 1110 struct p_block_req *rp); 1111 extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet, 1112 struct p_data *dp, int data_size); 1113 extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet, 1114 sector_t sector, int blksize, u64 block_id); 1115 extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *); 1116 extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet, 1117 struct drbd_peer_request *); 1118 extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req); 1119 extern int drbd_send_drequest(struct drbd_peer_device *, int cmd, 1120 sector_t sector, int size, u64 block_id); 1121 extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector, 1122 int size, void *digest, int digest_size, 1123 enum drbd_packet cmd); 1124 extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size); 1125 1126 extern int drbd_send_bitmap(struct drbd_device *device); 1127 extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode); 1128 extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode); 1129 extern int drbd_send_rs_deallocated(struct drbd_peer_device *, struct drbd_peer_request *); 1130 extern void drbd_backing_dev_free(struct drbd_device *device, struct drbd_backing_dev *ldev); 1131 extern void drbd_device_cleanup(struct drbd_device *device); 1132 extern void drbd_print_uuids(struct drbd_device *device, const char *text); 1133 extern void drbd_queue_unplug(struct drbd_device *device); 1134 1135 extern void conn_md_sync(struct drbd_connection *connection); 1136 extern void drbd_md_write(struct drbd_device *device, void *buffer); 1137 extern void drbd_md_sync(struct drbd_device *device); 1138 extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev); 1139 extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); 1140 extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); 1141 extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local); 1142 extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local); 1143 extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local); 1144 extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); 1145 extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local); 1146 extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local); 1147 extern int drbd_md_test_flag(struct drbd_backing_dev *, int); 1148 extern void drbd_md_mark_dirty(struct drbd_device *device); 1149 extern void drbd_queue_bitmap_io(struct drbd_device *device, 1150 int (*io_fn)(struct drbd_device *), 1151 void (*done)(struct drbd_device *, int), 1152 char *why, enum bm_flag flags); 1153 extern int drbd_bitmap_io(struct drbd_device *device, 1154 int (*io_fn)(struct drbd_device *), 1155 char *why, enum bm_flag flags); 1156 extern int drbd_bitmap_io_from_worker(struct drbd_device *device, 1157 int (*io_fn)(struct drbd_device *), 1158 char *why, enum bm_flag flags); 1159 extern int drbd_bmio_set_n_write(struct drbd_device *device) __must_hold(local); 1160 extern int drbd_bmio_clear_n_write(struct drbd_device *device) __must_hold(local); 1161 1162 /* Meta data layout 1163 * 1164 * We currently have two possible layouts. 1165 * Offsets in (512 byte) sectors. 1166 * external: 1167 * |----------- md_size_sect ------------------| 1168 * [ 4k superblock ][ activity log ][ Bitmap ] 1169 * | al_offset == 8 | 1170 * | bm_offset = al_offset + X | 1171 * ==> bitmap sectors = md_size_sect - bm_offset 1172 * 1173 * Variants: 1174 * old, indexed fixed size meta data: 1175 * 1176 * internal: 1177 * |----------- md_size_sect ------------------| 1178 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*] 1179 * | al_offset < 0 | 1180 * | bm_offset = al_offset - Y | 1181 * ==> bitmap sectors = Y = al_offset - bm_offset 1182 * 1183 * [padding*] are zero or up to 7 unused 512 Byte sectors to the 1184 * end of the device, so that the [4k superblock] will be 4k aligned. 1185 * 1186 * The activity log consists of 4k transaction blocks, 1187 * which are written in a ring-buffer, or striped ring-buffer like fashion, 1188 * which are writtensize used to be fixed 32kB, 1189 * but is about to become configurable. 1190 */ 1191 1192 /* Our old fixed size meta data layout 1193 * allows up to about 3.8TB, so if you want more, 1194 * you need to use the "flexible" meta data format. */ 1195 #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */ 1196 #define MD_4kB_SECT 8 1197 #define MD_32kB_SECT 64 1198 1199 /* One activity log extent represents 4M of storage */ 1200 #define AL_EXTENT_SHIFT 22 1201 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT) 1202 1203 /* We could make these currently hardcoded constants configurable 1204 * variables at create-md time (or even re-configurable at runtime?). 1205 * Which will require some more changes to the DRBD "super block" 1206 * and attach code. 1207 * 1208 * updates per transaction: 1209 * This many changes to the active set can be logged with one transaction. 1210 * This number is arbitrary. 1211 * context per transaction: 1212 * This many context extent numbers are logged with each transaction. 1213 * This number is resulting from the transaction block size (4k), the layout 1214 * of the transaction header, and the number of updates per transaction. 1215 * See drbd_actlog.c:struct al_transaction_on_disk 1216 * */ 1217 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary 1218 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4 1219 1220 #if BITS_PER_LONG == 32 1221 #define LN2_BPL 5 1222 #define cpu_to_lel(A) cpu_to_le32(A) 1223 #define lel_to_cpu(A) le32_to_cpu(A) 1224 #elif BITS_PER_LONG == 64 1225 #define LN2_BPL 6 1226 #define cpu_to_lel(A) cpu_to_le64(A) 1227 #define lel_to_cpu(A) le64_to_cpu(A) 1228 #else 1229 #error "LN2 of BITS_PER_LONG unknown!" 1230 #endif 1231 1232 /* resync bitmap */ 1233 /* 16MB sized 'bitmap extent' to track syncer usage */ 1234 struct bm_extent { 1235 int rs_left; /* number of bits set (out of sync) in this extent. */ 1236 int rs_failed; /* number of failed resync requests in this extent. */ 1237 unsigned long flags; 1238 struct lc_element lce; 1239 }; 1240 1241 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */ 1242 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */ 1243 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */ 1244 1245 /* drbd_bitmap.c */ 1246 /* 1247 * We need to store one bit for a block. 1248 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap. 1249 * Bit 0 ==> local node thinks this block is binary identical on both nodes 1250 * Bit 1 ==> local node thinks this block needs to be synced. 1251 */ 1252 1253 #define SLEEP_TIME (HZ/10) 1254 1255 /* We do bitmap IO in units of 4k blocks. 1256 * We also still have a hardcoded 4k per bit relation. */ 1257 #define BM_BLOCK_SHIFT 12 /* 4k per bit */ 1258 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT) 1259 /* mostly arbitrarily set the represented size of one bitmap extent, 1260 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap 1261 * at 4k per bit resolution) */ 1262 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */ 1263 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT) 1264 1265 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12) 1266 #error "HAVE YOU FIXED drbdmeta AS WELL??" 1267 #endif 1268 1269 /* thus many _storage_ sectors are described by one bit */ 1270 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9)) 1271 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9)) 1272 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1) 1273 1274 /* bit to represented kilo byte conversion */ 1275 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10)) 1276 1277 /* in which _bitmap_ extent (resp. sector) the bit for a certain 1278 * _storage_ sector is located in */ 1279 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9)) 1280 #define BM_BIT_TO_EXT(x) ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT)) 1281 1282 /* first storage sector a bitmap extent corresponds to */ 1283 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9)) 1284 /* how much _storage_ sectors we have per bitmap extent */ 1285 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1) 1286 /* how many bits are covered by one bitmap extent (resync extent) */ 1287 #define BM_BITS_PER_EXT (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT)) 1288 1289 #define BM_BLOCKS_PER_BM_EXT_MASK (BM_BITS_PER_EXT - 1) 1290 1291 1292 /* in one sector of the bitmap, we have this many activity_log extents. */ 1293 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT)) 1294 1295 /* the extent in "PER_EXTENT" below is an activity log extent 1296 * we need that many (long words/bytes) to store the bitmap 1297 * of one AL_EXTENT_SIZE chunk of storage. 1298 * we can store the bitmap for that many AL_EXTENTS within 1299 * one sector of the _on_disk_ bitmap: 1300 * bit 0 bit 37 bit 38 bit (512*8)-1 1301 * ...|........|........|.. // ..|........| 1302 * sect. 0 `296 `304 ^(512*8*8)-1 1303 * 1304 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG ) 1305 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128 1306 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4 1307 */ 1308 1309 #define DRBD_MAX_SECTORS_32 (0xffffffffLU) 1310 /* we have a certain meta data variant that has a fixed on-disk size of 128 1311 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity 1312 * log, leaving this many sectors for the bitmap. 1313 */ 1314 1315 #define DRBD_MAX_SECTORS_FIXED_BM \ 1316 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9))) 1317 #if !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32 1318 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32 1319 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32 1320 #else 1321 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM 1322 /* 16 TB in units of sectors */ 1323 #if BITS_PER_LONG == 32 1324 /* adjust by one page worth of bitmap, 1325 * so we won't wrap around in drbd_bm_find_next_bit. 1326 * you should use 64bit OS for that much storage, anyways. */ 1327 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff) 1328 #else 1329 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */ 1330 #define DRBD_MAX_SECTORS_FLEX (1UL << 51) 1331 /* corresponds to (1UL << 38) bits right now. */ 1332 #endif 1333 #endif 1334 1335 /* Estimate max bio size as 256 * PAGE_SIZE, 1336 * so for typical PAGE_SIZE of 4k, that is (1<<20) Byte. 1337 * Since we may live in a mixed-platform cluster, 1338 * we limit us to a platform agnostic constant here for now. 1339 * A followup commit may allow even bigger BIO sizes, 1340 * once we thought that through. */ 1341 #define DRBD_MAX_BIO_SIZE (1U << 20) 1342 #if DRBD_MAX_BIO_SIZE > (BIO_MAX_PAGES << PAGE_SHIFT) 1343 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE 1344 #endif 1345 #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */ 1346 1347 #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */ 1348 #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */ 1349 1350 /* For now, don't allow more than half of what we can "activate" in one 1351 * activity log transaction to be discarded in one go. We may need to rework 1352 * drbd_al_begin_io() to allow for even larger discard ranges */ 1353 #define DRBD_MAX_BATCH_BIO_SIZE (AL_UPDATES_PER_TRANSACTION/2*AL_EXTENT_SIZE) 1354 #define DRBD_MAX_BBIO_SECTORS (DRBD_MAX_BATCH_BIO_SIZE >> 9) 1355 1356 extern int drbd_bm_init(struct drbd_device *device); 1357 extern int drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits); 1358 extern void drbd_bm_cleanup(struct drbd_device *device); 1359 extern void drbd_bm_set_all(struct drbd_device *device); 1360 extern void drbd_bm_clear_all(struct drbd_device *device); 1361 /* set/clear/test only a few bits at a time */ 1362 extern int drbd_bm_set_bits( 1363 struct drbd_device *device, unsigned long s, unsigned long e); 1364 extern int drbd_bm_clear_bits( 1365 struct drbd_device *device, unsigned long s, unsigned long e); 1366 extern int drbd_bm_count_bits( 1367 struct drbd_device *device, const unsigned long s, const unsigned long e); 1368 /* bm_set_bits variant for use while holding drbd_bm_lock, 1369 * may process the whole bitmap in one go */ 1370 extern void _drbd_bm_set_bits(struct drbd_device *device, 1371 const unsigned long s, const unsigned long e); 1372 extern int drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr); 1373 extern int drbd_bm_e_weight(struct drbd_device *device, unsigned long enr); 1374 extern int drbd_bm_read(struct drbd_device *device) __must_hold(local); 1375 extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr); 1376 extern int drbd_bm_write(struct drbd_device *device) __must_hold(local); 1377 extern void drbd_bm_reset_al_hints(struct drbd_device *device) __must_hold(local); 1378 extern int drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local); 1379 extern int drbd_bm_write_lazy(struct drbd_device *device, unsigned upper_idx) __must_hold(local); 1380 extern int drbd_bm_write_all(struct drbd_device *device) __must_hold(local); 1381 extern int drbd_bm_write_copy_pages(struct drbd_device *device) __must_hold(local); 1382 extern size_t drbd_bm_words(struct drbd_device *device); 1383 extern unsigned long drbd_bm_bits(struct drbd_device *device); 1384 extern sector_t drbd_bm_capacity(struct drbd_device *device); 1385 1386 #define DRBD_END_OF_BITMAP (~(unsigned long)0) 1387 extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); 1388 /* bm_find_next variants for use while you hold drbd_bm_lock() */ 1389 extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); 1390 extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo); 1391 extern unsigned long _drbd_bm_total_weight(struct drbd_device *device); 1392 extern unsigned long drbd_bm_total_weight(struct drbd_device *device); 1393 /* for receive_bitmap */ 1394 extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset, 1395 size_t number, unsigned long *buffer); 1396 /* for _drbd_send_bitmap */ 1397 extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset, 1398 size_t number, unsigned long *buffer); 1399 1400 extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags); 1401 extern void drbd_bm_unlock(struct drbd_device *device); 1402 /* drbd_main.c */ 1403 1404 extern struct kmem_cache *drbd_request_cache; 1405 extern struct kmem_cache *drbd_ee_cache; /* peer requests */ 1406 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */ 1407 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */ 1408 extern mempool_t *drbd_request_mempool; 1409 extern mempool_t *drbd_ee_mempool; 1410 1411 /* drbd's page pool, used to buffer data received from the peer, 1412 * or data requested by the peer. 1413 * 1414 * This does not have an emergency reserve. 1415 * 1416 * When allocating from this pool, it first takes pages from the pool. 1417 * Only if the pool is depleted will try to allocate from the system. 1418 * 1419 * The assumption is that pages taken from this pool will be processed, 1420 * and given back, "quickly", and then can be recycled, so we can avoid 1421 * frequent calls to alloc_page(), and still will be able to make progress even 1422 * under memory pressure. 1423 */ 1424 extern struct page *drbd_pp_pool; 1425 extern spinlock_t drbd_pp_lock; 1426 extern int drbd_pp_vacant; 1427 extern wait_queue_head_t drbd_pp_wait; 1428 1429 /* We also need a standard (emergency-reserve backed) page pool 1430 * for meta data IO (activity log, bitmap). 1431 * We can keep it global, as long as it is used as "N pages at a time". 1432 * 128 should be plenty, currently we probably can get away with as few as 1. 1433 */ 1434 #define DRBD_MIN_POOL_PAGES 128 1435 extern mempool_t *drbd_md_io_page_pool; 1436 1437 /* We also need to make sure we get a bio 1438 * when we need it for housekeeping purposes */ 1439 extern struct bio_set *drbd_md_io_bio_set; 1440 /* to allocate from that set */ 1441 extern struct bio *bio_alloc_drbd(gfp_t gfp_mask); 1442 1443 /* And a bio_set for cloning */ 1444 extern struct bio_set *drbd_io_bio_set; 1445 1446 extern struct mutex resources_mutex; 1447 1448 extern int conn_lowest_minor(struct drbd_connection *connection); 1449 extern enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor); 1450 extern void drbd_destroy_device(struct kref *kref); 1451 extern void drbd_delete_device(struct drbd_device *device); 1452 1453 extern struct drbd_resource *drbd_create_resource(const char *name); 1454 extern void drbd_free_resource(struct drbd_resource *resource); 1455 1456 extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts); 1457 extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts); 1458 extern void drbd_destroy_connection(struct kref *kref); 1459 extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len, 1460 void *peer_addr, int peer_addr_len); 1461 extern struct drbd_resource *drbd_find_resource(const char *name); 1462 extern void drbd_destroy_resource(struct kref *kref); 1463 extern void conn_free_crypto(struct drbd_connection *connection); 1464 1465 /* drbd_req */ 1466 extern void do_submit(struct work_struct *ws); 1467 extern void __drbd_make_request(struct drbd_device *, struct bio *, unsigned long); 1468 extern blk_qc_t drbd_make_request(struct request_queue *q, struct bio *bio); 1469 extern int drbd_read_remote(struct drbd_device *device, struct drbd_request *req); 1470 extern int is_valid_ar_handle(struct drbd_request *, sector_t); 1471 1472 1473 /* drbd_nl.c */ 1474 1475 extern struct mutex notification_mutex; 1476 1477 extern void drbd_suspend_io(struct drbd_device *device); 1478 extern void drbd_resume_io(struct drbd_device *device); 1479 extern char *ppsize(char *buf, unsigned long long size); 1480 extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int); 1481 enum determine_dev_size { 1482 DS_ERROR_SHRINK = -3, 1483 DS_ERROR_SPACE_MD = -2, 1484 DS_ERROR = -1, 1485 DS_UNCHANGED = 0, 1486 DS_SHRUNK = 1, 1487 DS_GREW = 2, 1488 DS_GREW_FROM_ZERO = 3, 1489 }; 1490 extern enum determine_dev_size 1491 drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local); 1492 extern void resync_after_online_grow(struct drbd_device *); 1493 extern void drbd_reconsider_queue_parameters(struct drbd_device *device, 1494 struct drbd_backing_dev *bdev, struct o_qlim *o); 1495 extern enum drbd_state_rv drbd_set_role(struct drbd_device *device, 1496 enum drbd_role new_role, 1497 int force); 1498 extern bool conn_try_outdate_peer(struct drbd_connection *connection); 1499 extern void conn_try_outdate_peer_async(struct drbd_connection *connection); 1500 extern enum drbd_peer_state conn_khelper(struct drbd_connection *connection, char *cmd); 1501 extern int drbd_khelper(struct drbd_device *device, char *cmd); 1502 1503 /* drbd_worker.c */ 1504 /* bi_end_io handlers */ 1505 extern void drbd_md_endio(struct bio *bio); 1506 extern void drbd_peer_request_endio(struct bio *bio); 1507 extern void drbd_request_endio(struct bio *bio); 1508 extern int drbd_worker(struct drbd_thread *thi); 1509 enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor); 1510 void drbd_resync_after_changed(struct drbd_device *device); 1511 extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side); 1512 extern void resume_next_sg(struct drbd_device *device); 1513 extern void suspend_other_sg(struct drbd_device *device); 1514 extern int drbd_resync_finished(struct drbd_device *device); 1515 /* maybe rather drbd_main.c ? */ 1516 extern void *drbd_md_get_buffer(struct drbd_device *device, const char *intent); 1517 extern void drbd_md_put_buffer(struct drbd_device *device); 1518 extern int drbd_md_sync_page_io(struct drbd_device *device, 1519 struct drbd_backing_dev *bdev, sector_t sector, int op); 1520 extern void drbd_ov_out_of_sync_found(struct drbd_device *, sector_t, int); 1521 extern void wait_until_done_or_force_detached(struct drbd_device *device, 1522 struct drbd_backing_dev *bdev, unsigned int *done); 1523 extern void drbd_rs_controller_reset(struct drbd_device *device); 1524 1525 static inline void ov_out_of_sync_print(struct drbd_device *device) 1526 { 1527 if (device->ov_last_oos_size) { 1528 drbd_err(device, "Out of sync: start=%llu, size=%lu (sectors)\n", 1529 (unsigned long long)device->ov_last_oos_start, 1530 (unsigned long)device->ov_last_oos_size); 1531 } 1532 device->ov_last_oos_size = 0; 1533 } 1534 1535 1536 extern void drbd_csum_bio(struct crypto_ahash *, struct bio *, void *); 1537 extern void drbd_csum_ee(struct crypto_ahash *, struct drbd_peer_request *, void *); 1538 /* worker callbacks */ 1539 extern int w_e_end_data_req(struct drbd_work *, int); 1540 extern int w_e_end_rsdata_req(struct drbd_work *, int); 1541 extern int w_e_end_csum_rs_req(struct drbd_work *, int); 1542 extern int w_e_end_ov_reply(struct drbd_work *, int); 1543 extern int w_e_end_ov_req(struct drbd_work *, int); 1544 extern int w_ov_finished(struct drbd_work *, int); 1545 extern int w_resync_timer(struct drbd_work *, int); 1546 extern int w_send_write_hint(struct drbd_work *, int); 1547 extern int w_send_dblock(struct drbd_work *, int); 1548 extern int w_send_read_req(struct drbd_work *, int); 1549 extern int w_e_reissue(struct drbd_work *, int); 1550 extern int w_restart_disk_io(struct drbd_work *, int); 1551 extern int w_send_out_of_sync(struct drbd_work *, int); 1552 extern int w_start_resync(struct drbd_work *, int); 1553 1554 extern void resync_timer_fn(struct timer_list *t); 1555 extern void start_resync_timer_fn(struct timer_list *t); 1556 1557 extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req); 1558 1559 /* drbd_receiver.c */ 1560 extern int drbd_receiver(struct drbd_thread *thi); 1561 extern int drbd_ack_receiver(struct drbd_thread *thi); 1562 extern void drbd_send_ping_wf(struct work_struct *ws); 1563 extern void drbd_send_acks_wf(struct work_struct *ws); 1564 extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device); 1565 extern bool drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector, 1566 bool throttle_if_app_is_waiting); 1567 extern int drbd_submit_peer_request(struct drbd_device *, 1568 struct drbd_peer_request *, const unsigned, 1569 const unsigned, const int); 1570 extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *); 1571 extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64, 1572 sector_t, unsigned int, 1573 unsigned int, 1574 gfp_t) __must_hold(local); 1575 extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *, 1576 int); 1577 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0) 1578 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1) 1579 extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool); 1580 extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled); 1581 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed); 1582 extern int drbd_connected(struct drbd_peer_device *); 1583 1584 static inline void drbd_tcp_cork(struct socket *sock) 1585 { 1586 int val = 1; 1587 (void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK, 1588 (char*)&val, sizeof(val)); 1589 } 1590 1591 static inline void drbd_tcp_uncork(struct socket *sock) 1592 { 1593 int val = 0; 1594 (void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK, 1595 (char*)&val, sizeof(val)); 1596 } 1597 1598 static inline void drbd_tcp_nodelay(struct socket *sock) 1599 { 1600 int val = 1; 1601 (void) kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, 1602 (char*)&val, sizeof(val)); 1603 } 1604 1605 static inline void drbd_tcp_quickack(struct socket *sock) 1606 { 1607 int val = 2; 1608 (void) kernel_setsockopt(sock, SOL_TCP, TCP_QUICKACK, 1609 (char*)&val, sizeof(val)); 1610 } 1611 1612 /* sets the number of 512 byte sectors of our virtual device */ 1613 static inline void drbd_set_my_capacity(struct drbd_device *device, 1614 sector_t size) 1615 { 1616 /* set_capacity(device->this_bdev->bd_disk, size); */ 1617 set_capacity(device->vdisk, size); 1618 device->this_bdev->bd_inode->i_size = (loff_t)size << 9; 1619 } 1620 1621 /* 1622 * used to submit our private bio 1623 */ 1624 static inline void drbd_generic_make_request(struct drbd_device *device, 1625 int fault_type, struct bio *bio) 1626 { 1627 __release(local); 1628 if (!bio->bi_disk) { 1629 drbd_err(device, "drbd_generic_make_request: bio->bi_disk == NULL\n"); 1630 bio->bi_status = BLK_STS_IOERR; 1631 bio_endio(bio); 1632 return; 1633 } 1634 1635 if (drbd_insert_fault(device, fault_type)) 1636 bio_io_error(bio); 1637 else 1638 generic_make_request(bio); 1639 } 1640 1641 void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev, 1642 enum write_ordering_e wo); 1643 1644 /* drbd_proc.c */ 1645 extern struct proc_dir_entry *drbd_proc; 1646 extern const struct file_operations drbd_proc_fops; 1647 1648 /* drbd_actlog.c */ 1649 extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i); 1650 extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i); 1651 extern void drbd_al_begin_io_commit(struct drbd_device *device); 1652 extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i); 1653 extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i); 1654 extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i); 1655 extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector); 1656 extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector); 1657 extern int drbd_try_rs_begin_io(struct drbd_device *device, sector_t sector); 1658 extern void drbd_rs_cancel_all(struct drbd_device *device); 1659 extern int drbd_rs_del_all(struct drbd_device *device); 1660 extern void drbd_rs_failed_io(struct drbd_device *device, 1661 sector_t sector, int size); 1662 extern void drbd_advance_rs_marks(struct drbd_device *device, unsigned long still_to_go); 1663 1664 enum update_sync_bits_mode { RECORD_RS_FAILED, SET_OUT_OF_SYNC, SET_IN_SYNC }; 1665 extern int __drbd_change_sync(struct drbd_device *device, sector_t sector, int size, 1666 enum update_sync_bits_mode mode); 1667 #define drbd_set_in_sync(device, sector, size) \ 1668 __drbd_change_sync(device, sector, size, SET_IN_SYNC) 1669 #define drbd_set_out_of_sync(device, sector, size) \ 1670 __drbd_change_sync(device, sector, size, SET_OUT_OF_SYNC) 1671 #define drbd_rs_failed_io(device, sector, size) \ 1672 __drbd_change_sync(device, sector, size, RECORD_RS_FAILED) 1673 extern void drbd_al_shrink(struct drbd_device *device); 1674 extern int drbd_al_initialize(struct drbd_device *, void *); 1675 1676 /* drbd_nl.c */ 1677 /* state info broadcast */ 1678 struct sib_info { 1679 enum drbd_state_info_bcast_reason sib_reason; 1680 union { 1681 struct { 1682 char *helper_name; 1683 unsigned helper_exit_code; 1684 }; 1685 struct { 1686 union drbd_state os; 1687 union drbd_state ns; 1688 }; 1689 }; 1690 }; 1691 void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib); 1692 1693 extern void notify_resource_state(struct sk_buff *, 1694 unsigned int, 1695 struct drbd_resource *, 1696 struct resource_info *, 1697 enum drbd_notification_type); 1698 extern void notify_device_state(struct sk_buff *, 1699 unsigned int, 1700 struct drbd_device *, 1701 struct device_info *, 1702 enum drbd_notification_type); 1703 extern void notify_connection_state(struct sk_buff *, 1704 unsigned int, 1705 struct drbd_connection *, 1706 struct connection_info *, 1707 enum drbd_notification_type); 1708 extern void notify_peer_device_state(struct sk_buff *, 1709 unsigned int, 1710 struct drbd_peer_device *, 1711 struct peer_device_info *, 1712 enum drbd_notification_type); 1713 extern void notify_helper(enum drbd_notification_type, struct drbd_device *, 1714 struct drbd_connection *, const char *, int); 1715 1716 /* 1717 * inline helper functions 1718 *************************/ 1719 1720 /* see also page_chain_add and friends in drbd_receiver.c */ 1721 static inline struct page *page_chain_next(struct page *page) 1722 { 1723 return (struct page *)page_private(page); 1724 } 1725 #define page_chain_for_each(page) \ 1726 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \ 1727 page = page_chain_next(page)) 1728 #define page_chain_for_each_safe(page, n) \ 1729 for (; page && ({ n = page_chain_next(page); 1; }); page = n) 1730 1731 1732 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req) 1733 { 1734 struct page *page = peer_req->pages; 1735 page_chain_for_each(page) { 1736 if (page_count(page) > 1) 1737 return 1; 1738 } 1739 return 0; 1740 } 1741 1742 static inline union drbd_state drbd_read_state(struct drbd_device *device) 1743 { 1744 struct drbd_resource *resource = device->resource; 1745 union drbd_state rv; 1746 1747 rv.i = device->state.i; 1748 rv.susp = resource->susp; 1749 rv.susp_nod = resource->susp_nod; 1750 rv.susp_fen = resource->susp_fen; 1751 1752 return rv; 1753 } 1754 1755 enum drbd_force_detach_flags { 1756 DRBD_READ_ERROR, 1757 DRBD_WRITE_ERROR, 1758 DRBD_META_IO_ERROR, 1759 DRBD_FORCE_DETACH, 1760 }; 1761 1762 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__) 1763 static inline void __drbd_chk_io_error_(struct drbd_device *device, 1764 enum drbd_force_detach_flags df, 1765 const char *where) 1766 { 1767 enum drbd_io_error_p ep; 1768 1769 rcu_read_lock(); 1770 ep = rcu_dereference(device->ldev->disk_conf)->on_io_error; 1771 rcu_read_unlock(); 1772 switch (ep) { 1773 case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */ 1774 if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) { 1775 if (__ratelimit(&drbd_ratelimit_state)) 1776 drbd_err(device, "Local IO failed in %s.\n", where); 1777 if (device->state.disk > D_INCONSISTENT) 1778 _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL); 1779 break; 1780 } 1781 /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */ 1782 case EP_DETACH: 1783 case EP_CALL_HELPER: 1784 /* Remember whether we saw a READ or WRITE error. 1785 * 1786 * Recovery of the affected area for WRITE failure is covered 1787 * by the activity log. 1788 * READ errors may fall outside that area though. Certain READ 1789 * errors can be "healed" by writing good data to the affected 1790 * blocks, which triggers block re-allocation in lower layers. 1791 * 1792 * If we can not write the bitmap after a READ error, 1793 * we may need to trigger a full sync (see w_go_diskless()). 1794 * 1795 * Force-detach is not really an IO error, but rather a 1796 * desperate measure to try to deal with a completely 1797 * unresponsive lower level IO stack. 1798 * Still it should be treated as a WRITE error. 1799 * 1800 * Meta IO error is always WRITE error: 1801 * we read meta data only once during attach, 1802 * which will fail in case of errors. 1803 */ 1804 set_bit(WAS_IO_ERROR, &device->flags); 1805 if (df == DRBD_READ_ERROR) 1806 set_bit(WAS_READ_ERROR, &device->flags); 1807 if (df == DRBD_FORCE_DETACH) 1808 set_bit(FORCE_DETACH, &device->flags); 1809 if (device->state.disk > D_FAILED) { 1810 _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL); 1811 drbd_err(device, 1812 "Local IO failed in %s. Detaching...\n", where); 1813 } 1814 break; 1815 } 1816 } 1817 1818 /** 1819 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers 1820 * @device: DRBD device. 1821 * @error: Error code passed to the IO completion callback 1822 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data 1823 * 1824 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED) 1825 */ 1826 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__) 1827 static inline void drbd_chk_io_error_(struct drbd_device *device, 1828 int error, enum drbd_force_detach_flags forcedetach, const char *where) 1829 { 1830 if (error) { 1831 unsigned long flags; 1832 spin_lock_irqsave(&device->resource->req_lock, flags); 1833 __drbd_chk_io_error_(device, forcedetach, where); 1834 spin_unlock_irqrestore(&device->resource->req_lock, flags); 1835 } 1836 } 1837 1838 1839 /** 1840 * drbd_md_first_sector() - Returns the first sector number of the meta data area 1841 * @bdev: Meta data block device. 1842 * 1843 * BTW, for internal meta data, this happens to be the maximum capacity 1844 * we could agree upon with our peer node. 1845 */ 1846 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev) 1847 { 1848 switch (bdev->md.meta_dev_idx) { 1849 case DRBD_MD_INDEX_INTERNAL: 1850 case DRBD_MD_INDEX_FLEX_INT: 1851 return bdev->md.md_offset + bdev->md.bm_offset; 1852 case DRBD_MD_INDEX_FLEX_EXT: 1853 default: 1854 return bdev->md.md_offset; 1855 } 1856 } 1857 1858 /** 1859 * drbd_md_last_sector() - Return the last sector number of the meta data area 1860 * @bdev: Meta data block device. 1861 */ 1862 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev) 1863 { 1864 switch (bdev->md.meta_dev_idx) { 1865 case DRBD_MD_INDEX_INTERNAL: 1866 case DRBD_MD_INDEX_FLEX_INT: 1867 return bdev->md.md_offset + MD_4kB_SECT -1; 1868 case DRBD_MD_INDEX_FLEX_EXT: 1869 default: 1870 return bdev->md.md_offset + bdev->md.md_size_sect -1; 1871 } 1872 } 1873 1874 /* Returns the number of 512 byte sectors of the device */ 1875 static inline sector_t drbd_get_capacity(struct block_device *bdev) 1876 { 1877 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */ 1878 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0; 1879 } 1880 1881 /** 1882 * drbd_get_max_capacity() - Returns the capacity we announce to out peer 1883 * @bdev: Meta data block device. 1884 * 1885 * returns the capacity we announce to out peer. we clip ourselves at the 1886 * various MAX_SECTORS, because if we don't, current implementation will 1887 * oops sooner or later 1888 */ 1889 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev) 1890 { 1891 sector_t s; 1892 1893 switch (bdev->md.meta_dev_idx) { 1894 case DRBD_MD_INDEX_INTERNAL: 1895 case DRBD_MD_INDEX_FLEX_INT: 1896 s = drbd_get_capacity(bdev->backing_bdev) 1897 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX, 1898 drbd_md_first_sector(bdev)) 1899 : 0; 1900 break; 1901 case DRBD_MD_INDEX_FLEX_EXT: 1902 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX, 1903 drbd_get_capacity(bdev->backing_bdev)); 1904 /* clip at maximum size the meta device can support */ 1905 s = min_t(sector_t, s, 1906 BM_EXT_TO_SECT(bdev->md.md_size_sect 1907 - bdev->md.bm_offset)); 1908 break; 1909 default: 1910 s = min_t(sector_t, DRBD_MAX_SECTORS, 1911 drbd_get_capacity(bdev->backing_bdev)); 1912 } 1913 return s; 1914 } 1915 1916 /** 1917 * drbd_md_ss() - Return the sector number of our meta data super block 1918 * @bdev: Meta data block device. 1919 */ 1920 static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev) 1921 { 1922 const int meta_dev_idx = bdev->md.meta_dev_idx; 1923 1924 if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT) 1925 return 0; 1926 1927 /* Since drbd08, internal meta data is always "flexible". 1928 * position: last 4k aligned block of 4k size */ 1929 if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL || 1930 meta_dev_idx == DRBD_MD_INDEX_FLEX_INT) 1931 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8; 1932 1933 /* external, some index; this is the old fixed size layout */ 1934 return MD_128MB_SECT * bdev->md.meta_dev_idx; 1935 } 1936 1937 static inline void 1938 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w) 1939 { 1940 unsigned long flags; 1941 spin_lock_irqsave(&q->q_lock, flags); 1942 list_add_tail(&w->list, &q->q); 1943 spin_unlock_irqrestore(&q->q_lock, flags); 1944 wake_up(&q->q_wait); 1945 } 1946 1947 static inline void 1948 drbd_queue_work_if_unqueued(struct drbd_work_queue *q, struct drbd_work *w) 1949 { 1950 unsigned long flags; 1951 spin_lock_irqsave(&q->q_lock, flags); 1952 if (list_empty_careful(&w->list)) 1953 list_add_tail(&w->list, &q->q); 1954 spin_unlock_irqrestore(&q->q_lock, flags); 1955 wake_up(&q->q_wait); 1956 } 1957 1958 static inline void 1959 drbd_device_post_work(struct drbd_device *device, int work_bit) 1960 { 1961 if (!test_and_set_bit(work_bit, &device->flags)) { 1962 struct drbd_connection *connection = 1963 first_peer_device(device)->connection; 1964 struct drbd_work_queue *q = &connection->sender_work; 1965 if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags)) 1966 wake_up(&q->q_wait); 1967 } 1968 } 1969 1970 extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue); 1971 1972 /* To get the ack_receiver out of the blocking network stack, 1973 * so it can change its sk_rcvtimeo from idle- to ping-timeout, 1974 * and send a ping, we need to send a signal. 1975 * Which signal we send is irrelevant. */ 1976 static inline void wake_ack_receiver(struct drbd_connection *connection) 1977 { 1978 struct task_struct *task = connection->ack_receiver.task; 1979 if (task && get_t_state(&connection->ack_receiver) == RUNNING) 1980 force_sig(SIGXCPU, task); 1981 } 1982 1983 static inline void request_ping(struct drbd_connection *connection) 1984 { 1985 set_bit(SEND_PING, &connection->flags); 1986 wake_ack_receiver(connection); 1987 } 1988 1989 extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *); 1990 extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *); 1991 extern int conn_send_command(struct drbd_connection *, struct drbd_socket *, 1992 enum drbd_packet, unsigned int, void *, 1993 unsigned int); 1994 extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *, 1995 enum drbd_packet, unsigned int, void *, 1996 unsigned int); 1997 1998 extern int drbd_send_ping(struct drbd_connection *connection); 1999 extern int drbd_send_ping_ack(struct drbd_connection *connection); 2000 extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state); 2001 extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state); 2002 2003 static inline void drbd_thread_stop(struct drbd_thread *thi) 2004 { 2005 _drbd_thread_stop(thi, false, true); 2006 } 2007 2008 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi) 2009 { 2010 _drbd_thread_stop(thi, false, false); 2011 } 2012 2013 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi) 2014 { 2015 _drbd_thread_stop(thi, true, false); 2016 } 2017 2018 /* counts how many answer packets packets we expect from our peer, 2019 * for either explicit application requests, 2020 * or implicit barrier packets as necessary. 2021 * increased: 2022 * w_send_barrier 2023 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ); 2024 * it is much easier and equally valid to count what we queue for the 2025 * worker, even before it actually was queued or send. 2026 * (drbd_make_request_common; recovery path on read io-error) 2027 * decreased: 2028 * got_BarrierAck (respective tl_clear, tl_clear_barrier) 2029 * _req_mod(req, DATA_RECEIVED) 2030 * [from receive_DataReply] 2031 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED) 2032 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)] 2033 * for some reason it is NOT decreased in got_NegAck, 2034 * but in the resulting cleanup code from report_params. 2035 * we should try to remember the reason for that... 2036 * _req_mod(req, SEND_FAILED or SEND_CANCELED) 2037 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING) 2038 * [from tl_clear_barrier] 2039 */ 2040 static inline void inc_ap_pending(struct drbd_device *device) 2041 { 2042 atomic_inc(&device->ap_pending_cnt); 2043 } 2044 2045 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \ 2046 if (atomic_read(&device->which) < 0) \ 2047 drbd_err(device, "in %s:%d: " #which " = %d < 0 !\n", \ 2048 func, line, \ 2049 atomic_read(&device->which)) 2050 2051 #define dec_ap_pending(device) _dec_ap_pending(device, __func__, __LINE__) 2052 static inline void _dec_ap_pending(struct drbd_device *device, const char *func, int line) 2053 { 2054 if (atomic_dec_and_test(&device->ap_pending_cnt)) 2055 wake_up(&device->misc_wait); 2056 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line); 2057 } 2058 2059 /* counts how many resync-related answers we still expect from the peer 2060 * increase decrease 2061 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY) 2062 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER) 2063 * (or P_NEG_ACK with ID_SYNCER) 2064 */ 2065 static inline void inc_rs_pending(struct drbd_device *device) 2066 { 2067 atomic_inc(&device->rs_pending_cnt); 2068 } 2069 2070 #define dec_rs_pending(device) _dec_rs_pending(device, __func__, __LINE__) 2071 static inline void _dec_rs_pending(struct drbd_device *device, const char *func, int line) 2072 { 2073 atomic_dec(&device->rs_pending_cnt); 2074 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line); 2075 } 2076 2077 /* counts how many answers we still need to send to the peer. 2078 * increased on 2079 * receive_Data unless protocol A; 2080 * we need to send a P_RECV_ACK (proto B) 2081 * or P_WRITE_ACK (proto C) 2082 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK 2083 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA 2084 * receive_Barrier_* we need to send a P_BARRIER_ACK 2085 */ 2086 static inline void inc_unacked(struct drbd_device *device) 2087 { 2088 atomic_inc(&device->unacked_cnt); 2089 } 2090 2091 #define dec_unacked(device) _dec_unacked(device, __func__, __LINE__) 2092 static inline void _dec_unacked(struct drbd_device *device, const char *func, int line) 2093 { 2094 atomic_dec(&device->unacked_cnt); 2095 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line); 2096 } 2097 2098 #define sub_unacked(device, n) _sub_unacked(device, n, __func__, __LINE__) 2099 static inline void _sub_unacked(struct drbd_device *device, int n, const char *func, int line) 2100 { 2101 atomic_sub(n, &device->unacked_cnt); 2102 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line); 2103 } 2104 2105 static inline bool is_sync_target_state(enum drbd_conns connection_state) 2106 { 2107 return connection_state == C_SYNC_TARGET || 2108 connection_state == C_PAUSED_SYNC_T; 2109 } 2110 2111 static inline bool is_sync_source_state(enum drbd_conns connection_state) 2112 { 2113 return connection_state == C_SYNC_SOURCE || 2114 connection_state == C_PAUSED_SYNC_S; 2115 } 2116 2117 static inline bool is_sync_state(enum drbd_conns connection_state) 2118 { 2119 return is_sync_source_state(connection_state) || 2120 is_sync_target_state(connection_state); 2121 } 2122 2123 /** 2124 * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev 2125 * @_device: DRBD device. 2126 * @_min_state: Minimum device state required for success. 2127 * 2128 * You have to call put_ldev() when finished working with device->ldev. 2129 */ 2130 #define get_ldev_if_state(_device, _min_state) \ 2131 (_get_ldev_if_state((_device), (_min_state)) ? \ 2132 ({ __acquire(x); true; }) : false) 2133 #define get_ldev(_device) get_ldev_if_state(_device, D_INCONSISTENT) 2134 2135 static inline void put_ldev(struct drbd_device *device) 2136 { 2137 enum drbd_disk_state disk_state = device->state.disk; 2138 /* We must check the state *before* the atomic_dec becomes visible, 2139 * or we have a theoretical race where someone hitting zero, 2140 * while state still D_FAILED, will then see D_DISKLESS in the 2141 * condition below and calling into destroy, where he must not, yet. */ 2142 int i = atomic_dec_return(&device->local_cnt); 2143 2144 /* This may be called from some endio handler, 2145 * so we must not sleep here. */ 2146 2147 __release(local); 2148 D_ASSERT(device, i >= 0); 2149 if (i == 0) { 2150 if (disk_state == D_DISKLESS) 2151 /* even internal references gone, safe to destroy */ 2152 drbd_device_post_work(device, DESTROY_DISK); 2153 if (disk_state == D_FAILED) 2154 /* all application IO references gone. */ 2155 if (!test_and_set_bit(GOING_DISKLESS, &device->flags)) 2156 drbd_device_post_work(device, GO_DISKLESS); 2157 wake_up(&device->misc_wait); 2158 } 2159 } 2160 2161 #ifndef __CHECKER__ 2162 static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins) 2163 { 2164 int io_allowed; 2165 2166 /* never get a reference while D_DISKLESS */ 2167 if (device->state.disk == D_DISKLESS) 2168 return 0; 2169 2170 atomic_inc(&device->local_cnt); 2171 io_allowed = (device->state.disk >= mins); 2172 if (!io_allowed) 2173 put_ldev(device); 2174 return io_allowed; 2175 } 2176 #else 2177 extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins); 2178 #endif 2179 2180 /* this throttles on-the-fly application requests 2181 * according to max_buffers settings; 2182 * maybe re-implement using semaphores? */ 2183 static inline int drbd_get_max_buffers(struct drbd_device *device) 2184 { 2185 struct net_conf *nc; 2186 int mxb; 2187 2188 rcu_read_lock(); 2189 nc = rcu_dereference(first_peer_device(device)->connection->net_conf); 2190 mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */ 2191 rcu_read_unlock(); 2192 2193 return mxb; 2194 } 2195 2196 static inline int drbd_state_is_stable(struct drbd_device *device) 2197 { 2198 union drbd_dev_state s = device->state; 2199 2200 /* DO NOT add a default clause, we want the compiler to warn us 2201 * for any newly introduced state we may have forgotten to add here */ 2202 2203 switch ((enum drbd_conns)s.conn) { 2204 /* new io only accepted when there is no connection, ... */ 2205 case C_STANDALONE: 2206 case C_WF_CONNECTION: 2207 /* ... or there is a well established connection. */ 2208 case C_CONNECTED: 2209 case C_SYNC_SOURCE: 2210 case C_SYNC_TARGET: 2211 case C_VERIFY_S: 2212 case C_VERIFY_T: 2213 case C_PAUSED_SYNC_S: 2214 case C_PAUSED_SYNC_T: 2215 case C_AHEAD: 2216 case C_BEHIND: 2217 /* transitional states, IO allowed */ 2218 case C_DISCONNECTING: 2219 case C_UNCONNECTED: 2220 case C_TIMEOUT: 2221 case C_BROKEN_PIPE: 2222 case C_NETWORK_FAILURE: 2223 case C_PROTOCOL_ERROR: 2224 case C_TEAR_DOWN: 2225 case C_WF_REPORT_PARAMS: 2226 case C_STARTING_SYNC_S: 2227 case C_STARTING_SYNC_T: 2228 break; 2229 2230 /* Allow IO in BM exchange states with new protocols */ 2231 case C_WF_BITMAP_S: 2232 if (first_peer_device(device)->connection->agreed_pro_version < 96) 2233 return 0; 2234 break; 2235 2236 /* no new io accepted in these states */ 2237 case C_WF_BITMAP_T: 2238 case C_WF_SYNC_UUID: 2239 case C_MASK: 2240 /* not "stable" */ 2241 return 0; 2242 } 2243 2244 switch ((enum drbd_disk_state)s.disk) { 2245 case D_DISKLESS: 2246 case D_INCONSISTENT: 2247 case D_OUTDATED: 2248 case D_CONSISTENT: 2249 case D_UP_TO_DATE: 2250 case D_FAILED: 2251 /* disk state is stable as well. */ 2252 break; 2253 2254 /* no new io accepted during transitional states */ 2255 case D_ATTACHING: 2256 case D_NEGOTIATING: 2257 case D_UNKNOWN: 2258 case D_MASK: 2259 /* not "stable" */ 2260 return 0; 2261 } 2262 2263 return 1; 2264 } 2265 2266 static inline int drbd_suspended(struct drbd_device *device) 2267 { 2268 struct drbd_resource *resource = device->resource; 2269 2270 return resource->susp || resource->susp_fen || resource->susp_nod; 2271 } 2272 2273 static inline bool may_inc_ap_bio(struct drbd_device *device) 2274 { 2275 int mxb = drbd_get_max_buffers(device); 2276 2277 if (drbd_suspended(device)) 2278 return false; 2279 if (atomic_read(&device->suspend_cnt)) 2280 return false; 2281 2282 /* to avoid potential deadlock or bitmap corruption, 2283 * in various places, we only allow new application io 2284 * to start during "stable" states. */ 2285 2286 /* no new io accepted when attaching or detaching the disk */ 2287 if (!drbd_state_is_stable(device)) 2288 return false; 2289 2290 /* since some older kernels don't have atomic_add_unless, 2291 * and we are within the spinlock anyways, we have this workaround. */ 2292 if (atomic_read(&device->ap_bio_cnt) > mxb) 2293 return false; 2294 if (test_bit(BITMAP_IO, &device->flags)) 2295 return false; 2296 return true; 2297 } 2298 2299 static inline bool inc_ap_bio_cond(struct drbd_device *device) 2300 { 2301 bool rv = false; 2302 2303 spin_lock_irq(&device->resource->req_lock); 2304 rv = may_inc_ap_bio(device); 2305 if (rv) 2306 atomic_inc(&device->ap_bio_cnt); 2307 spin_unlock_irq(&device->resource->req_lock); 2308 2309 return rv; 2310 } 2311 2312 static inline void inc_ap_bio(struct drbd_device *device) 2313 { 2314 /* we wait here 2315 * as long as the device is suspended 2316 * until the bitmap is no longer on the fly during connection 2317 * handshake as long as we would exceed the max_buffer limit. 2318 * 2319 * to avoid races with the reconnect code, 2320 * we need to atomic_inc within the spinlock. */ 2321 2322 wait_event(device->misc_wait, inc_ap_bio_cond(device)); 2323 } 2324 2325 static inline void dec_ap_bio(struct drbd_device *device) 2326 { 2327 int mxb = drbd_get_max_buffers(device); 2328 int ap_bio = atomic_dec_return(&device->ap_bio_cnt); 2329 2330 D_ASSERT(device, ap_bio >= 0); 2331 2332 if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) { 2333 if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags)) 2334 drbd_queue_work(&first_peer_device(device)-> 2335 connection->sender_work, 2336 &device->bm_io_work.w); 2337 } 2338 2339 /* this currently does wake_up for every dec_ap_bio! 2340 * maybe rather introduce some type of hysteresis? 2341 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */ 2342 if (ap_bio < mxb) 2343 wake_up(&device->misc_wait); 2344 } 2345 2346 static inline bool verify_can_do_stop_sector(struct drbd_device *device) 2347 { 2348 return first_peer_device(device)->connection->agreed_pro_version >= 97 && 2349 first_peer_device(device)->connection->agreed_pro_version != 100; 2350 } 2351 2352 static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val) 2353 { 2354 int changed = device->ed_uuid != val; 2355 device->ed_uuid = val; 2356 return changed; 2357 } 2358 2359 static inline int drbd_queue_order_type(struct drbd_device *device) 2360 { 2361 /* sorry, we currently have no working implementation 2362 * of distributed TCQ stuff */ 2363 #ifndef QUEUE_ORDERED_NONE 2364 #define QUEUE_ORDERED_NONE 0 2365 #endif 2366 return QUEUE_ORDERED_NONE; 2367 } 2368 2369 static inline struct drbd_connection *first_connection(struct drbd_resource *resource) 2370 { 2371 return list_first_entry_or_null(&resource->connections, 2372 struct drbd_connection, connections); 2373 } 2374 2375 #endif 2376