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