| 1 | /* SPDX-License-Identifier: GPL-2.0-only */ |
| 2 | /* |
| 3 | drbd_int.h |
| 4 | |
| 5 | This file is part of DRBD by Philipp Reisner and Lars Ellenberg. |
| 6 | |
| 7 | Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. |
| 8 | Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. |
| 9 | Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. |
| 10 | |
| 11 | |
| 12 | */ |
| 13 | |
| 14 | #ifndef _DRBD_INT_H |
| 15 | #define _DRBD_INT_H |
| 16 | |
| 17 | #include <crypto/hash.h> |
| 18 | #include <linux/compiler.h> |
| 19 | #include <linux/types.h> |
| 20 | #include <linux/list.h> |
| 21 | #include <linux/sched/signal.h> |
| 22 | #include <linux/bitops.h> |
| 23 | #include <linux/slab.h> |
| 24 | #include <linux/ratelimit.h> |
| 25 | #include <linux/tcp.h> |
| 26 | #include <linux/mutex.h> |
| 27 | #include <linux/major.h> |
| 28 | #include <linux/blkdev.h> |
| 29 | #include <linux/backing-dev.h> |
| 30 | #include <linux/idr.h> |
| 31 | #include <linux/dynamic_debug.h> |
| 32 | #include <net/tcp.h> |
| 33 | #include <linux/lru_cache.h> |
| 34 | #include <linux/prefetch.h> |
| 35 | #include <linux/drbd_genl_api.h> |
| 36 | #include <linux/drbd.h> |
| 37 | #include <linux/drbd_config.h> |
| 38 | #include "drbd_strings.h" |
| 39 | #include "drbd_state.h" |
| 40 | #include "drbd_protocol.h" |
| 41 | #include "drbd_polymorph_printk.h" |
| 42 | |
| 43 | /* shared module parameters, defined in drbd_main.c */ |
| 44 | #ifdef CONFIG_DRBD_FAULT_INJECTION |
| 45 | extern int drbd_enable_faults; |
| 46 | extern int drbd_fault_rate; |
| 47 | #endif |
| 48 | |
| 49 | extern unsigned int drbd_minor_count; |
| 50 | extern char drbd_usermode_helper[]; |
| 51 | extern int drbd_proc_details; |
| 52 | |
| 53 | |
| 54 | /* This is used to stop/restart our threads. |
| 55 | * Cannot use SIGTERM nor SIGKILL, since these |
| 56 | * are sent out by init on runlevel changes |
| 57 | * I choose SIGHUP for now. |
| 58 | */ |
| 59 | #define DRBD_SIGKILL SIGHUP |
| 60 | |
| 61 | #define ID_IN_SYNC (4711ULL) |
| 62 | #define ID_OUT_OF_SYNC (4712ULL) |
| 63 | #define ID_SYNCER (-1ULL) |
| 64 | |
| 65 | #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL) |
| 66 | |
| 67 | struct drbd_device; |
| 68 | struct drbd_connection; |
| 69 | struct drbd_peer_device; |
| 70 | |
| 71 | /* Defines to control fault insertion */ |
| 72 | enum { |
| 73 | DRBD_FAULT_MD_WR = 0, /* meta data write */ |
| 74 | DRBD_FAULT_MD_RD = 1, /* read */ |
| 75 | DRBD_FAULT_RS_WR = 2, /* resync */ |
| 76 | DRBD_FAULT_RS_RD = 3, |
| 77 | DRBD_FAULT_DT_WR = 4, /* data */ |
| 78 | DRBD_FAULT_DT_RD = 5, |
| 79 | DRBD_FAULT_DT_RA = 6, /* data read ahead */ |
| 80 | DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */ |
| 81 | DRBD_FAULT_AL_EE = 8, /* alloc ee */ |
| 82 | DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */ |
| 83 | |
| 84 | DRBD_FAULT_MAX, |
| 85 | }; |
| 86 | |
| 87 | extern unsigned int |
| 88 | _drbd_insert_fault(struct drbd_device *device, unsigned int type); |
| 89 | |
| 90 | static inline int |
| 91 | drbd_insert_fault(struct drbd_device *device, unsigned int type) { |
| 92 | #ifdef CONFIG_DRBD_FAULT_INJECTION |
| 93 | return drbd_fault_rate && |
| 94 | (drbd_enable_faults & (1<<type)) && |
| 95 | _drbd_insert_fault(device, type); |
| 96 | #else |
| 97 | return 0; |
| 98 | #endif |
| 99 | } |
| 100 | |
| 101 | /* integer division, round _UP_ to the next integer */ |
| 102 | #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0)) |
| 103 | /* usual integer division */ |
| 104 | #define div_floor(A, B) ((A)/(B)) |
| 105 | |
| 106 | extern struct ratelimit_state drbd_ratelimit_state; |
| 107 | extern struct idr drbd_devices; /* RCU, updates: genl_lock() */ |
| 108 | extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */ |
| 109 | |
| 110 | extern const char *cmdname(enum drbd_packet cmd); |
| 111 | |
| 112 | /* for sending/receiving the bitmap, |
| 113 | * possibly in some encoding scheme */ |
| 114 | struct bm_xfer_ctx { |
| 115 | /* "const" |
| 116 | * stores total bits and long words |
| 117 | * of the bitmap, so we don't need to |
| 118 | * call the accessor functions over and again. */ |
| 119 | unsigned long bm_bits; |
| 120 | unsigned long bm_words; |
| 121 | /* during xfer, current position within the bitmap */ |
| 122 | unsigned long bit_offset; |
| 123 | unsigned long word_offset; |
| 124 | |
| 125 | /* statistics; index: (h->command == P_BITMAP) */ |
| 126 | unsigned packets[2]; |
| 127 | unsigned bytes[2]; |
| 128 | }; |
| 129 | |
| 130 | extern void INFO_bm_xfer_stats(struct drbd_peer_device *peer_device, |
| 131 | const char *direction, struct bm_xfer_ctx *c); |
| 132 | |
| 133 | static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c) |
| 134 | { |
| 135 | /* word_offset counts "native long words" (32 or 64 bit), |
| 136 | * aligned at 64 bit. |
| 137 | * Encoded packet may end at an unaligned bit offset. |
| 138 | * In case a fallback clear text packet is transmitted in |
| 139 | * between, we adjust this offset back to the last 64bit |
| 140 | * aligned "native long word", which makes coding and decoding |
| 141 | * the plain text bitmap much more convenient. */ |
| 142 | #if BITS_PER_LONG == 64 |
| 143 | c->word_offset = c->bit_offset >> 6; |
| 144 | #elif BITS_PER_LONG == 32 |
| 145 | c->word_offset = c->bit_offset >> 5; |
| 146 | c->word_offset &= ~(1UL); |
| 147 | #else |
| 148 | # error "unsupported BITS_PER_LONG" |
| 149 | #endif |
| 150 | } |
| 151 | |
| 152 | extern unsigned int drbd_header_size(struct drbd_connection *connection); |
| 153 | |
| 154 | /**********************************************************************/ |
| 155 | enum drbd_thread_state { |
| 156 | NONE, |
| 157 | RUNNING, |
| 158 | EXITING, |
| 159 | RESTARTING |
| 160 | }; |
| 161 | |
| 162 | struct drbd_thread { |
| 163 | spinlock_t t_lock; |
| 164 | struct task_struct *task; |
| 165 | struct completion stop; |
| 166 | enum drbd_thread_state t_state; |
| 167 | int (*function) (struct drbd_thread *); |
| 168 | struct drbd_resource *resource; |
| 169 | struct drbd_connection *connection; |
| 170 | int reset_cpu_mask; |
| 171 | const char *name; |
| 172 | }; |
| 173 | |
| 174 | static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi) |
| 175 | { |
| 176 | /* THINK testing the t_state seems to be uncritical in all cases |
| 177 | * (but thread_{start,stop}), so we can read it *without* the lock. |
| 178 | * --lge */ |
| 179 | |
| 180 | smp_rmb(); |
| 181 | return thi->t_state; |
| 182 | } |
| 183 | |
| 184 | struct drbd_work { |
| 185 | struct list_head list; |
| 186 | int (*cb)(struct drbd_work *, int cancel); |
| 187 | }; |
| 188 | |
| 189 | struct drbd_device_work { |
| 190 | struct drbd_work w; |
| 191 | struct drbd_device *device; |
| 192 | }; |
| 193 | |
| 194 | #include "drbd_interval.h" |
| 195 | |
| 196 | extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *); |
| 197 | |
| 198 | extern void lock_all_resources(void); |
| 199 | extern void unlock_all_resources(void); |
| 200 | |
| 201 | struct drbd_request { |
| 202 | struct drbd_work w; |
| 203 | struct drbd_device *device; |
| 204 | |
| 205 | /* if local IO is not allowed, will be NULL. |
| 206 | * if local IO _is_ allowed, holds the locally submitted bio clone, |
| 207 | * or, after local IO completion, the ERR_PTR(error). |
| 208 | * see drbd_request_endio(). */ |
| 209 | struct bio *private_bio; |
| 210 | |
| 211 | struct drbd_interval i; |
| 212 | |
| 213 | /* epoch: used to check on "completion" whether this req was in |
| 214 | * the current epoch, and we therefore have to close it, |
| 215 | * causing a p_barrier packet to be send, starting a new epoch. |
| 216 | * |
| 217 | * This corresponds to "barrier" in struct p_barrier[_ack], |
| 218 | * and to "barrier_nr" in struct drbd_epoch (and various |
| 219 | * comments/function parameters/local variable names). |
| 220 | */ |
| 221 | unsigned int epoch; |
| 222 | |
| 223 | struct list_head tl_requests; /* ring list in the transfer log */ |
| 224 | struct bio *master_bio; /* master bio pointer */ |
| 225 | |
| 226 | /* see struct drbd_device */ |
| 227 | struct list_head req_pending_master_completion; |
| 228 | struct list_head req_pending_local; |
| 229 | |
| 230 | /* for generic IO accounting */ |
| 231 | unsigned long start_jif; |
| 232 | |
| 233 | /* for DRBD internal statistics */ |
| 234 | |
| 235 | /* Minimal set of time stamps to determine if we wait for activity log |
| 236 | * transactions, local disk or peer. 32 bit "jiffies" are good enough, |
| 237 | * we don't expect a DRBD request to be stalled for several month. |
| 238 | */ |
| 239 | |
| 240 | /* before actual request processing */ |
| 241 | unsigned long in_actlog_jif; |
| 242 | |
| 243 | /* local disk */ |
| 244 | unsigned long pre_submit_jif; |
| 245 | |
| 246 | /* per connection */ |
| 247 | unsigned long pre_send_jif; |
| 248 | unsigned long acked_jif; |
| 249 | unsigned long net_done_jif; |
| 250 | |
| 251 | /* Possibly even more detail to track each phase: |
| 252 | * master_completion_jif |
| 253 | * how long did it take to complete the master bio |
| 254 | * (application visible latency) |
| 255 | * allocated_jif |
| 256 | * how long the master bio was blocked until we finally allocated |
| 257 | * a tracking struct |
| 258 | * in_actlog_jif |
| 259 | * how long did we wait for activity log transactions |
| 260 | * |
| 261 | * net_queued_jif |
| 262 | * when did we finally queue it for sending |
| 263 | * pre_send_jif |
| 264 | * when did we start sending it |
| 265 | * post_send_jif |
| 266 | * how long did we block in the network stack trying to send it |
| 267 | * acked_jif |
| 268 | * when did we receive (or fake, in protocol A) a remote ACK |
| 269 | * net_done_jif |
| 270 | * when did we receive final acknowledgement (P_BARRIER_ACK), |
| 271 | * or decide, e.g. on connection loss, that we do no longer expect |
| 272 | * anything from this peer for this request. |
| 273 | * |
| 274 | * pre_submit_jif |
| 275 | * post_sub_jif |
| 276 | * when did we start submiting to the lower level device, |
| 277 | * and how long did we block in that submit function |
| 278 | * local_completion_jif |
| 279 | * how long did it take the lower level device to complete this request |
| 280 | */ |
| 281 | |
| 282 | |
| 283 | /* once it hits 0, we may complete the master_bio */ |
| 284 | atomic_t completion_ref; |
| 285 | /* once it hits 0, we may destroy this drbd_request object */ |
| 286 | struct kref kref; |
| 287 | |
| 288 | unsigned rq_state; /* see comments above _req_mod() */ |
| 289 | }; |
| 290 | |
| 291 | struct drbd_epoch { |
| 292 | struct drbd_connection *connection; |
| 293 | struct list_head list; |
| 294 | unsigned int barrier_nr; |
| 295 | atomic_t epoch_size; /* increased on every request added. */ |
| 296 | atomic_t active; /* increased on every req. added, and dec on every finished. */ |
| 297 | unsigned long flags; |
| 298 | }; |
| 299 | |
| 300 | /* drbd_epoch flag bits */ |
| 301 | enum { |
| 302 | DE_HAVE_BARRIER_NUMBER, |
| 303 | }; |
| 304 | |
| 305 | enum epoch_event { |
| 306 | EV_PUT, |
| 307 | EV_GOT_BARRIER_NR, |
| 308 | EV_BECAME_LAST, |
| 309 | EV_CLEANUP = 32, /* used as flag */ |
| 310 | }; |
| 311 | |
| 312 | struct digest_info { |
| 313 | int digest_size; |
| 314 | void *digest; |
| 315 | }; |
| 316 | |
| 317 | struct drbd_peer_request { |
| 318 | struct drbd_work w; |
| 319 | struct drbd_peer_device *peer_device; |
| 320 | struct drbd_epoch *epoch; /* for writes */ |
| 321 | struct page *pages; |
| 322 | blk_opf_t opf; |
| 323 | atomic_t pending_bios; |
| 324 | struct drbd_interval i; |
| 325 | /* see comments on ee flag bits below */ |
| 326 | unsigned long flags; |
| 327 | unsigned long submit_jif; |
| 328 | union { |
| 329 | u64 block_id; |
| 330 | struct digest_info *digest; |
| 331 | }; |
| 332 | }; |
| 333 | |
| 334 | /* Equivalent to bio_op and req_op. */ |
| 335 | #define peer_req_op(peer_req) \ |
| 336 | ((peer_req)->opf & REQ_OP_MASK) |
| 337 | |
| 338 | /* ee flag bits. |
| 339 | * While corresponding bios are in flight, the only modification will be |
| 340 | * set_bit WAS_ERROR, which has to be atomic. |
| 341 | * If no bios are in flight yet, or all have been completed, |
| 342 | * non-atomic modification to ee->flags is ok. |
| 343 | */ |
| 344 | enum { |
| 345 | __EE_CALL_AL_COMPLETE_IO, |
| 346 | __EE_MAY_SET_IN_SYNC, |
| 347 | |
| 348 | /* is this a TRIM aka REQ_OP_DISCARD? */ |
| 349 | __EE_TRIM, |
| 350 | /* explicit zero-out requested, or |
| 351 | * our lower level cannot handle trim, |
| 352 | * and we want to fall back to zeroout instead */ |
| 353 | __EE_ZEROOUT, |
| 354 | |
| 355 | /* In case a barrier failed, |
| 356 | * we need to resubmit without the barrier flag. */ |
| 357 | __EE_RESUBMITTED, |
| 358 | |
| 359 | /* we may have several bios per peer request. |
| 360 | * if any of those fail, we set this flag atomically |
| 361 | * from the endio callback */ |
| 362 | __EE_WAS_ERROR, |
| 363 | |
| 364 | /* This ee has a pointer to a digest instead of a block id */ |
| 365 | __EE_HAS_DIGEST, |
| 366 | |
| 367 | /* Conflicting local requests need to be restarted after this request */ |
| 368 | __EE_RESTART_REQUESTS, |
| 369 | |
| 370 | /* The peer wants a write ACK for this (wire proto C) */ |
| 371 | __EE_SEND_WRITE_ACK, |
| 372 | |
| 373 | /* Is set when net_conf had two_primaries set while creating this peer_req */ |
| 374 | __EE_IN_INTERVAL_TREE, |
| 375 | |
| 376 | /* for debugfs: */ |
| 377 | /* has this been submitted, or does it still wait for something else? */ |
| 378 | __EE_SUBMITTED, |
| 379 | |
| 380 | /* this is/was a write request */ |
| 381 | __EE_WRITE, |
| 382 | |
| 383 | /* this is/was a write same request */ |
| 384 | __EE_WRITE_SAME, |
| 385 | |
| 386 | /* this originates from application on peer |
| 387 | * (not some resync or verify or other DRBD internal request) */ |
| 388 | __EE_APPLICATION, |
| 389 | |
| 390 | /* If it contains only 0 bytes, send back P_RS_DEALLOCATED */ |
| 391 | __EE_RS_THIN_REQ, |
| 392 | }; |
| 393 | #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO) |
| 394 | #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC) |
| 395 | #define EE_TRIM (1<<__EE_TRIM) |
| 396 | #define EE_ZEROOUT (1<<__EE_ZEROOUT) |
| 397 | #define EE_RESUBMITTED (1<<__EE_RESUBMITTED) |
| 398 | #define EE_WAS_ERROR (1<<__EE_WAS_ERROR) |
| 399 | #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST) |
| 400 | #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS) |
| 401 | #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK) |
| 402 | #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE) |
| 403 | #define EE_SUBMITTED (1<<__EE_SUBMITTED) |
| 404 | #define EE_WRITE (1<<__EE_WRITE) |
| 405 | #define EE_WRITE_SAME (1<<__EE_WRITE_SAME) |
| 406 | #define EE_APPLICATION (1<<__EE_APPLICATION) |
| 407 | #define EE_RS_THIN_REQ (1<<__EE_RS_THIN_REQ) |
| 408 | |
| 409 | /* flag bits per device */ |
| 410 | enum { |
| 411 | UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */ |
| 412 | MD_DIRTY, /* current uuids and flags not yet on disk */ |
| 413 | USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */ |
| 414 | CL_ST_CHG_SUCCESS, |
| 415 | CL_ST_CHG_FAIL, |
| 416 | CRASHED_PRIMARY, /* This node was a crashed primary. |
| 417 | * Gets cleared when the state.conn |
| 418 | * goes into C_CONNECTED state. */ |
| 419 | CONSIDER_RESYNC, |
| 420 | |
| 421 | MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */ |
| 422 | |
| 423 | BITMAP_IO, /* suspend application io; |
| 424 | once no more io in flight, start bitmap io */ |
| 425 | BITMAP_IO_QUEUED, /* Started bitmap IO */ |
| 426 | WAS_IO_ERROR, /* Local disk failed, returned IO error */ |
| 427 | WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */ |
| 428 | FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */ |
| 429 | RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */ |
| 430 | RESIZE_PENDING, /* Size change detected locally, waiting for the response from |
| 431 | * the peer, if it changed there as well. */ |
| 432 | NEW_CUR_UUID, /* Create new current UUID when thawing IO */ |
| 433 | AL_SUSPENDED, /* Activity logging is currently suspended. */ |
| 434 | AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */ |
| 435 | B_RS_H_DONE, /* Before resync handler done (already executed) */ |
| 436 | DISCARD_MY_DATA, /* discard_my_data flag per volume */ |
| 437 | READ_BALANCE_RR, |
| 438 | |
| 439 | FLUSH_PENDING, /* if set, device->flush_jif is when we submitted that flush |
| 440 | * from drbd_flush_after_epoch() */ |
| 441 | |
| 442 | /* cleared only after backing device related structures have been destroyed. */ |
| 443 | GOING_DISKLESS, /* Disk is being detached, because of io-error, or admin request. */ |
| 444 | |
| 445 | /* to be used in drbd_device_post_work() */ |
| 446 | GO_DISKLESS, /* tell worker to schedule cleanup before detach */ |
| 447 | DESTROY_DISK, /* tell worker to close backing devices and destroy related structures. */ |
| 448 | MD_SYNC, /* tell worker to call drbd_md_sync() */ |
| 449 | RS_START, /* tell worker to start resync/OV */ |
| 450 | RS_PROGRESS, /* tell worker that resync made significant progress */ |
| 451 | RS_DONE, /* tell worker that resync is done */ |
| 452 | }; |
| 453 | |
| 454 | struct drbd_bitmap; /* opaque for drbd_device */ |
| 455 | |
| 456 | /* definition of bits in bm_flags to be used in drbd_bm_lock |
| 457 | * and drbd_bitmap_io and friends. */ |
| 458 | enum bm_flag { |
| 459 | /* currently locked for bulk operation */ |
| 460 | BM_LOCKED_MASK = 0xf, |
| 461 | |
| 462 | /* in detail, that is: */ |
| 463 | BM_DONT_CLEAR = 0x1, |
| 464 | BM_DONT_SET = 0x2, |
| 465 | BM_DONT_TEST = 0x4, |
| 466 | |
| 467 | /* so we can mark it locked for bulk operation, |
| 468 | * and still allow all non-bulk operations */ |
| 469 | BM_IS_LOCKED = 0x8, |
| 470 | |
| 471 | /* (test bit, count bit) allowed (common case) */ |
| 472 | BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED, |
| 473 | |
| 474 | /* testing bits, as well as setting new bits allowed, but clearing bits |
| 475 | * would be unexpected. Used during bitmap receive. Setting new bits |
| 476 | * requires sending of "out-of-sync" information, though. */ |
| 477 | BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED, |
| 478 | |
| 479 | /* for drbd_bm_write_copy_pages, everything is allowed, |
| 480 | * only concurrent bulk operations are locked out. */ |
| 481 | BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED, |
| 482 | }; |
| 483 | |
| 484 | struct drbd_work_queue { |
| 485 | struct list_head q; |
| 486 | spinlock_t q_lock; /* to protect the list. */ |
| 487 | wait_queue_head_t q_wait; |
| 488 | }; |
| 489 | |
| 490 | struct drbd_socket { |
| 491 | struct mutex mutex; |
| 492 | struct socket *socket; |
| 493 | /* this way we get our |
| 494 | * send/receive buffers off the stack */ |
| 495 | void *sbuf; |
| 496 | void *rbuf; |
| 497 | }; |
| 498 | |
| 499 | struct drbd_md { |
| 500 | u64 md_offset; /* sector offset to 'super' block */ |
| 501 | |
| 502 | u64 la_size_sect; /* last agreed size, unit sectors */ |
| 503 | spinlock_t uuid_lock; |
| 504 | u64 uuid[UI_SIZE]; |
| 505 | u64 device_uuid; |
| 506 | u32 flags; |
| 507 | u32 md_size_sect; |
| 508 | |
| 509 | s32 al_offset; /* signed relative sector offset to activity log */ |
| 510 | s32 bm_offset; /* signed relative sector offset to bitmap */ |
| 511 | |
| 512 | /* cached value of bdev->disk_conf->meta_dev_idx (see below) */ |
| 513 | s32 meta_dev_idx; |
| 514 | |
| 515 | /* see al_tr_number_to_on_disk_sector() */ |
| 516 | u32 al_stripes; |
| 517 | u32 al_stripe_size_4k; |
| 518 | u32 al_size_4k; /* cached product of the above */ |
| 519 | }; |
| 520 | |
| 521 | struct drbd_backing_dev { |
| 522 | struct block_device *backing_bdev; |
| 523 | struct file *backing_bdev_file; |
| 524 | struct block_device *md_bdev; |
| 525 | struct file *f_md_bdev; |
| 526 | struct drbd_md md; |
| 527 | struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */ |
| 528 | sector_t known_size; /* last known size of that backing device */ |
| 529 | }; |
| 530 | |
| 531 | struct drbd_md_io { |
| 532 | struct page *page; |
| 533 | unsigned long start_jif; /* last call to drbd_md_get_buffer */ |
| 534 | unsigned long submit_jif; /* last _drbd_md_sync_page_io() submit */ |
| 535 | const char *current_use; |
| 536 | atomic_t in_use; |
| 537 | unsigned int done; |
| 538 | int error; |
| 539 | }; |
| 540 | |
| 541 | struct bm_io_work { |
| 542 | struct drbd_work w; |
| 543 | struct drbd_peer_device *peer_device; |
| 544 | char *why; |
| 545 | enum bm_flag flags; |
| 546 | int (*io_fn)(struct drbd_device *device, struct drbd_peer_device *peer_device); |
| 547 | void (*done)(struct drbd_device *device, int rv); |
| 548 | }; |
| 549 | |
| 550 | struct fifo_buffer { |
| 551 | unsigned int head_index; |
| 552 | unsigned int size; |
| 553 | int total; /* sum of all values */ |
| 554 | int values[] __counted_by(size); |
| 555 | }; |
| 556 | extern struct fifo_buffer *fifo_alloc(unsigned int fifo_size); |
| 557 | |
| 558 | /* flag bits per connection */ |
| 559 | enum { |
| 560 | NET_CONGESTED, /* The data socket is congested */ |
| 561 | RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */ |
| 562 | SEND_PING, |
| 563 | GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */ |
| 564 | CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */ |
| 565 | CONN_WD_ST_CHG_OKAY, |
| 566 | CONN_WD_ST_CHG_FAIL, |
| 567 | CONN_DRY_RUN, /* Expect disconnect after resync handshake. */ |
| 568 | CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */ |
| 569 | STATE_SENT, /* Do not change state/UUIDs while this is set */ |
| 570 | CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC) |
| 571 | * pending, from drbd worker context. |
| 572 | */ |
| 573 | DISCONNECT_SENT, |
| 574 | |
| 575 | DEVICE_WORK_PENDING, /* tell worker that some device has pending work */ |
| 576 | }; |
| 577 | |
| 578 | enum which_state { NOW, OLD = NOW, NEW }; |
| 579 | |
| 580 | struct drbd_resource { |
| 581 | char *name; |
| 582 | #ifdef CONFIG_DEBUG_FS |
| 583 | struct dentry *debugfs_res; |
| 584 | struct dentry *debugfs_res_volumes; |
| 585 | struct dentry *debugfs_res_connections; |
| 586 | struct dentry *debugfs_res_in_flight_summary; |
| 587 | #endif |
| 588 | struct kref kref; |
| 589 | struct idr devices; /* volume number to device mapping */ |
| 590 | struct list_head connections; |
| 591 | struct list_head resources; |
| 592 | struct res_opts res_opts; |
| 593 | struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */ |
| 594 | struct mutex adm_mutex; /* mutex to serialize administrative requests */ |
| 595 | spinlock_t req_lock; |
| 596 | |
| 597 | unsigned susp:1; /* IO suspended by user */ |
| 598 | unsigned susp_nod:1; /* IO suspended because no data */ |
| 599 | unsigned susp_fen:1; /* IO suspended because fence peer handler runs */ |
| 600 | |
| 601 | enum write_ordering_e write_ordering; |
| 602 | |
| 603 | cpumask_var_t cpu_mask; |
| 604 | }; |
| 605 | |
| 606 | struct drbd_thread_timing_details |
| 607 | { |
| 608 | unsigned long start_jif; |
| 609 | void *cb_addr; |
| 610 | const char *caller_fn; |
| 611 | unsigned int line; |
| 612 | unsigned int cb_nr; |
| 613 | }; |
| 614 | |
| 615 | struct drbd_connection { |
| 616 | struct list_head connections; |
| 617 | struct drbd_resource *resource; |
| 618 | #ifdef CONFIG_DEBUG_FS |
| 619 | struct dentry *debugfs_conn; |
| 620 | struct dentry *debugfs_conn_callback_history; |
| 621 | struct dentry *debugfs_conn_oldest_requests; |
| 622 | #endif |
| 623 | struct kref kref; |
| 624 | struct idr peer_devices; /* volume number to peer device mapping */ |
| 625 | enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */ |
| 626 | struct mutex cstate_mutex; /* Protects graceful disconnects */ |
| 627 | unsigned int connect_cnt; /* Inc each time a connection is established */ |
| 628 | |
| 629 | unsigned long flags; |
| 630 | struct net_conf *net_conf; /* content protected by rcu */ |
| 631 | wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */ |
| 632 | |
| 633 | struct sockaddr_storage my_addr; |
| 634 | int my_addr_len; |
| 635 | struct sockaddr_storage peer_addr; |
| 636 | int peer_addr_len; |
| 637 | |
| 638 | struct drbd_socket data; /* data/barrier/cstate/parameter packets */ |
| 639 | struct drbd_socket meta; /* ping/ack (metadata) packets */ |
| 640 | int agreed_pro_version; /* actually used protocol version */ |
| 641 | u32 agreed_features; |
| 642 | unsigned long last_received; /* in jiffies, either socket */ |
| 643 | unsigned int ko_count; |
| 644 | |
| 645 | struct list_head transfer_log; /* all requests not yet fully processed */ |
| 646 | |
| 647 | struct crypto_shash *cram_hmac_tfm; |
| 648 | struct crypto_shash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */ |
| 649 | struct crypto_shash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */ |
| 650 | struct crypto_shash *csums_tfm; |
| 651 | struct crypto_shash *verify_tfm; |
| 652 | void *int_dig_in; |
| 653 | void *int_dig_vv; |
| 654 | |
| 655 | /* receiver side */ |
| 656 | struct drbd_epoch *current_epoch; |
| 657 | spinlock_t epoch_lock; |
| 658 | unsigned int epochs; |
| 659 | atomic_t current_tle_nr; /* transfer log epoch number */ |
| 660 | unsigned current_tle_writes; /* writes seen within this tl epoch */ |
| 661 | |
| 662 | unsigned long last_reconnect_jif; |
| 663 | /* empty member on older kernels without blk_start_plug() */ |
| 664 | struct blk_plug receiver_plug; |
| 665 | struct drbd_thread receiver; |
| 666 | struct drbd_thread worker; |
| 667 | struct drbd_thread ack_receiver; |
| 668 | struct workqueue_struct *ack_sender; |
| 669 | |
| 670 | /* cached pointers, |
| 671 | * so we can look up the oldest pending requests more quickly. |
| 672 | * protected by resource->req_lock */ |
| 673 | struct drbd_request *req_next; /* DRBD 9: todo.req_next */ |
| 674 | struct drbd_request *req_ack_pending; |
| 675 | struct drbd_request *req_not_net_done; |
| 676 | |
| 677 | /* sender side */ |
| 678 | struct drbd_work_queue sender_work; |
| 679 | |
| 680 | #define DRBD_THREAD_DETAILS_HIST 16 |
| 681 | unsigned int w_cb_nr; /* keeps counting up */ |
| 682 | unsigned int r_cb_nr; /* keeps counting up */ |
| 683 | struct drbd_thread_timing_details w_timing_details[DRBD_THREAD_DETAILS_HIST]; |
| 684 | struct drbd_thread_timing_details r_timing_details[DRBD_THREAD_DETAILS_HIST]; |
| 685 | |
| 686 | struct { |
| 687 | unsigned long last_sent_barrier_jif; |
| 688 | |
| 689 | /* whether this sender thread |
| 690 | * has processed a single write yet. */ |
| 691 | bool seen_any_write_yet; |
| 692 | |
| 693 | /* Which barrier number to send with the next P_BARRIER */ |
| 694 | int current_epoch_nr; |
| 695 | |
| 696 | /* how many write requests have been sent |
| 697 | * with req->epoch == current_epoch_nr. |
| 698 | * If none, no P_BARRIER will be sent. */ |
| 699 | unsigned current_epoch_writes; |
| 700 | } send; |
| 701 | }; |
| 702 | |
| 703 | static inline bool has_net_conf(struct drbd_connection *connection) |
| 704 | { |
| 705 | bool has_net_conf; |
| 706 | |
| 707 | rcu_read_lock(); |
| 708 | has_net_conf = rcu_dereference(connection->net_conf); |
| 709 | rcu_read_unlock(); |
| 710 | |
| 711 | return has_net_conf; |
| 712 | } |
| 713 | |
| 714 | void __update_timing_details( |
| 715 | struct drbd_thread_timing_details *tdp, |
| 716 | unsigned int *cb_nr, |
| 717 | void *cb, |
| 718 | const char *fn, const unsigned int line); |
| 719 | |
| 720 | #define update_worker_timing_details(c, cb) \ |
| 721 | __update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ ) |
| 722 | #define update_receiver_timing_details(c, cb) \ |
| 723 | __update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ ) |
| 724 | |
| 725 | struct submit_worker { |
| 726 | struct workqueue_struct *wq; |
| 727 | struct work_struct worker; |
| 728 | |
| 729 | /* protected by ..->resource->req_lock */ |
| 730 | struct list_head writes; |
| 731 | }; |
| 732 | |
| 733 | struct drbd_peer_device { |
| 734 | struct list_head peer_devices; |
| 735 | struct drbd_device *device; |
| 736 | struct drbd_connection *connection; |
| 737 | struct work_struct send_acks_work; |
| 738 | #ifdef CONFIG_DEBUG_FS |
| 739 | struct dentry *debugfs_peer_dev; |
| 740 | #endif |
| 741 | }; |
| 742 | |
| 743 | struct drbd_device { |
| 744 | struct drbd_resource *resource; |
| 745 | struct list_head peer_devices; |
| 746 | struct list_head pending_bitmap_io; |
| 747 | |
| 748 | unsigned long flush_jif; |
| 749 | #ifdef CONFIG_DEBUG_FS |
| 750 | struct dentry *debugfs_minor; |
| 751 | struct dentry *debugfs_vol; |
| 752 | struct dentry *debugfs_vol_oldest_requests; |
| 753 | struct dentry *debugfs_vol_act_log_extents; |
| 754 | struct dentry *debugfs_vol_resync_extents; |
| 755 | struct dentry *debugfs_vol_data_gen_id; |
| 756 | struct dentry *debugfs_vol_ed_gen_id; |
| 757 | #endif |
| 758 | |
| 759 | unsigned int vnr; /* volume number within the connection */ |
| 760 | unsigned int minor; /* device minor number */ |
| 761 | |
| 762 | struct kref kref; |
| 763 | |
| 764 | /* things that are stored as / read from meta data on disk */ |
| 765 | unsigned long flags; |
| 766 | |
| 767 | /* configured by drbdsetup */ |
| 768 | struct drbd_backing_dev *ldev; |
| 769 | |
| 770 | sector_t p_size; /* partner's disk size */ |
| 771 | struct request_queue *rq_queue; |
| 772 | struct gendisk *vdisk; |
| 773 | |
| 774 | unsigned long last_reattach_jif; |
| 775 | struct drbd_work resync_work; |
| 776 | struct drbd_work unplug_work; |
| 777 | struct timer_list resync_timer; |
| 778 | struct timer_list md_sync_timer; |
| 779 | struct timer_list start_resync_timer; |
| 780 | struct timer_list request_timer; |
| 781 | |
| 782 | /* Used after attach while negotiating new disk state. */ |
| 783 | union drbd_state new_state_tmp; |
| 784 | |
| 785 | union drbd_dev_state state; |
| 786 | wait_queue_head_t misc_wait; |
| 787 | wait_queue_head_t state_wait; /* upon each state change. */ |
| 788 | unsigned int send_cnt; |
| 789 | unsigned int recv_cnt; |
| 790 | unsigned int read_cnt; |
| 791 | unsigned int writ_cnt; |
| 792 | unsigned int al_writ_cnt; |
| 793 | unsigned int bm_writ_cnt; |
| 794 | atomic_t ap_bio_cnt; /* Requests we need to complete */ |
| 795 | atomic_t ap_actlog_cnt; /* Requests waiting for activity log */ |
| 796 | atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */ |
| 797 | atomic_t rs_pending_cnt; /* RS request/data packets on the wire */ |
| 798 | atomic_t unacked_cnt; /* Need to send replies for */ |
| 799 | atomic_t local_cnt; /* Waiting for local completion */ |
| 800 | atomic_t suspend_cnt; |
| 801 | |
| 802 | /* Interval tree of pending local requests */ |
| 803 | struct rb_root read_requests; |
| 804 | struct rb_root write_requests; |
| 805 | |
| 806 | /* for statistics and timeouts */ |
| 807 | /* [0] read, [1] write */ |
| 808 | struct list_head pending_master_completion[2]; |
| 809 | struct list_head pending_completion[2]; |
| 810 | |
| 811 | /* use checksums for *this* resync */ |
| 812 | bool use_csums; |
| 813 | /* blocks to resync in this run [unit BM_BLOCK_SIZE] */ |
| 814 | unsigned long rs_total; |
| 815 | /* number of resync blocks that failed in this run */ |
| 816 | unsigned long rs_failed; |
| 817 | /* Syncer's start time [unit jiffies] */ |
| 818 | unsigned long rs_start; |
| 819 | /* cumulated time in PausedSyncX state [unit jiffies] */ |
| 820 | unsigned long rs_paused; |
| 821 | /* skipped because csum was equal [unit BM_BLOCK_SIZE] */ |
| 822 | unsigned long rs_same_csum; |
| 823 | #define DRBD_SYNC_MARKS 8 |
| 824 | #define DRBD_SYNC_MARK_STEP (3*HZ) |
| 825 | /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */ |
| 826 | unsigned long rs_mark_left[DRBD_SYNC_MARKS]; |
| 827 | /* marks's time [unit jiffies] */ |
| 828 | unsigned long rs_mark_time[DRBD_SYNC_MARKS]; |
| 829 | /* current index into rs_mark_{left,time} */ |
| 830 | int rs_last_mark; |
| 831 | unsigned long rs_last_bcast; /* [unit jiffies] */ |
| 832 | |
| 833 | /* where does the admin want us to start? (sector) */ |
| 834 | sector_t ov_start_sector; |
| 835 | sector_t ov_stop_sector; |
| 836 | /* where are we now? (sector) */ |
| 837 | sector_t ov_position; |
| 838 | /* Start sector of out of sync range (to merge printk reporting). */ |
| 839 | sector_t ov_last_oos_start; |
| 840 | /* size of out-of-sync range in sectors. */ |
| 841 | sector_t ov_last_oos_size; |
| 842 | unsigned long ov_left; /* in bits */ |
| 843 | |
| 844 | struct drbd_bitmap *bitmap; |
| 845 | unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */ |
| 846 | |
| 847 | /* Used to track operations of resync... */ |
| 848 | struct lru_cache *resync; |
| 849 | /* Number of locked elements in resync LRU */ |
| 850 | unsigned int resync_locked; |
| 851 | /* resync extent number waiting for application requests */ |
| 852 | unsigned int resync_wenr; |
| 853 | |
| 854 | int open_cnt; |
| 855 | u64 *p_uuid; |
| 856 | |
| 857 | struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */ |
| 858 | struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */ |
| 859 | struct list_head done_ee; /* need to send P_WRITE_ACK */ |
| 860 | struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */ |
| 861 | struct list_head net_ee; /* zero-copy network send in progress */ |
| 862 | |
| 863 | struct list_head resync_reads; |
| 864 | atomic_t pp_in_use; /* allocated from page pool */ |
| 865 | atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */ |
| 866 | wait_queue_head_t ee_wait; |
| 867 | struct drbd_md_io md_io; |
| 868 | spinlock_t al_lock; |
| 869 | wait_queue_head_t al_wait; |
| 870 | struct lru_cache *act_log; /* activity log */ |
| 871 | unsigned int al_tr_number; |
| 872 | int al_tr_cycle; |
| 873 | wait_queue_head_t seq_wait; |
| 874 | atomic_t packet_seq; |
| 875 | unsigned int peer_seq; |
| 876 | spinlock_t peer_seq_lock; |
| 877 | unsigned long comm_bm_set; /* communicated number of set bits. */ |
| 878 | struct bm_io_work bm_io_work; |
| 879 | u64 ed_uuid; /* UUID of the exposed data */ |
| 880 | struct mutex own_state_mutex; |
| 881 | struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */ |
| 882 | char congestion_reason; /* Why we where congested... */ |
| 883 | atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */ |
| 884 | atomic_t rs_sect_ev; /* for submitted resync data rate, both */ |
| 885 | int rs_last_sect_ev; /* counter to compare with */ |
| 886 | int rs_last_events; /* counter of read or write "events" (unit sectors) |
| 887 | * on the lower level device when we last looked. */ |
| 888 | int c_sync_rate; /* current resync rate after syncer throttle magic */ |
| 889 | struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */ |
| 890 | int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */ |
| 891 | atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */ |
| 892 | unsigned int peer_max_bio_size; |
| 893 | unsigned int local_max_bio_size; |
| 894 | |
| 895 | /* any requests that would block in drbd_make_request() |
| 896 | * are deferred to this single-threaded work queue */ |
| 897 | struct submit_worker submit; |
| 898 | }; |
| 899 | |
| 900 | struct drbd_bm_aio_ctx { |
| 901 | struct drbd_device *device; |
| 902 | struct list_head list; /* on device->pending_bitmap_io */; |
| 903 | unsigned long start_jif; |
| 904 | atomic_t in_flight; |
| 905 | unsigned int done; |
| 906 | unsigned flags; |
| 907 | #define BM_AIO_COPY_PAGES 1 |
| 908 | #define BM_AIO_WRITE_HINTED 2 |
| 909 | #define BM_AIO_WRITE_ALL_PAGES 4 |
| 910 | #define BM_AIO_READ 8 |
| 911 | int error; |
| 912 | struct kref kref; |
| 913 | }; |
| 914 | |
| 915 | struct drbd_config_context { |
| 916 | /* assigned from drbd_genlmsghdr */ |
| 917 | unsigned int minor; |
| 918 | /* assigned from request attributes, if present */ |
| 919 | unsigned int volume; |
| 920 | #define VOLUME_UNSPECIFIED (-1U) |
| 921 | /* pointer into the request skb, |
| 922 | * limited lifetime! */ |
| 923 | char *resource_name; |
| 924 | struct nlattr *my_addr; |
| 925 | struct nlattr *peer_addr; |
| 926 | |
| 927 | /* reply buffer */ |
| 928 | struct sk_buff *reply_skb; |
| 929 | /* pointer into reply buffer */ |
| 930 | struct drbd_genlmsghdr *reply_dh; |
| 931 | /* resolved from attributes, if possible */ |
| 932 | struct drbd_device *device; |
| 933 | struct drbd_resource *resource; |
| 934 | struct drbd_connection *connection; |
| 935 | }; |
| 936 | |
| 937 | static inline struct drbd_device *minor_to_device(unsigned int minor) |
| 938 | { |
| 939 | return (struct drbd_device *)idr_find(&drbd_devices, minor); |
| 940 | } |
| 941 | |
| 942 | static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device) |
| 943 | { |
| 944 | return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices); |
| 945 | } |
| 946 | |
| 947 | static inline struct drbd_peer_device * |
| 948 | conn_peer_device(struct drbd_connection *connection, int volume_number) |
| 949 | { |
| 950 | return idr_find(&connection->peer_devices, volume_number); |
| 951 | } |
| 952 | |
| 953 | #define for_each_resource(resource, _resources) \ |
| 954 | list_for_each_entry(resource, _resources, resources) |
| 955 | |
| 956 | #define for_each_resource_rcu(resource, _resources) \ |
| 957 | list_for_each_entry_rcu(resource, _resources, resources) |
| 958 | |
| 959 | #define for_each_resource_safe(resource, tmp, _resources) \ |
| 960 | list_for_each_entry_safe(resource, tmp, _resources, resources) |
| 961 | |
| 962 | #define for_each_connection(connection, resource) \ |
| 963 | list_for_each_entry(connection, &resource->connections, connections) |
| 964 | |
| 965 | #define for_each_connection_rcu(connection, resource) \ |
| 966 | list_for_each_entry_rcu(connection, &resource->connections, connections) |
| 967 | |
| 968 | #define for_each_connection_safe(connection, tmp, resource) \ |
| 969 | list_for_each_entry_safe(connection, tmp, &resource->connections, connections) |
| 970 | |
| 971 | #define for_each_peer_device(peer_device, device) \ |
| 972 | list_for_each_entry(peer_device, &device->peer_devices, peer_devices) |
| 973 | |
| 974 | #define for_each_peer_device_rcu(peer_device, device) \ |
| 975 | list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices) |
| 976 | |
| 977 | #define for_each_peer_device_safe(peer_device, tmp, device) \ |
| 978 | list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices) |
| 979 | |
| 980 | static inline unsigned int device_to_minor(struct drbd_device *device) |
| 981 | { |
| 982 | return device->minor; |
| 983 | } |
| 984 | |
| 985 | /* |
| 986 | * function declarations |
| 987 | *************************/ |
| 988 | |
| 989 | /* drbd_main.c */ |
| 990 | |
| 991 | enum dds_flags { |
| 992 | DDSF_FORCED = 1, |
| 993 | DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */ |
| 994 | }; |
| 995 | |
| 996 | extern void drbd_init_set_defaults(struct drbd_device *device); |
| 997 | extern int drbd_thread_start(struct drbd_thread *thi); |
| 998 | extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait); |
| 999 | #ifdef CONFIG_SMP |
| 1000 | extern void drbd_thread_current_set_cpu(struct drbd_thread *thi); |
| 1001 | #else |
| 1002 | #define drbd_thread_current_set_cpu(A) ({}) |
| 1003 | #endif |
| 1004 | extern void tl_release(struct drbd_connection *, unsigned int barrier_nr, |
| 1005 | unsigned int set_size); |
| 1006 | extern void tl_clear(struct drbd_connection *); |
| 1007 | extern void drbd_free_sock(struct drbd_connection *connection); |
| 1008 | extern int drbd_send(struct drbd_connection *connection, struct socket *sock, |
| 1009 | void *buf, size_t size, unsigned msg_flags); |
| 1010 | extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t, |
| 1011 | unsigned); |
| 1012 | |
| 1013 | extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd); |
| 1014 | extern int drbd_send_protocol(struct drbd_connection *connection); |
| 1015 | extern int drbd_send_uuids(struct drbd_peer_device *); |
| 1016 | extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *); |
| 1017 | extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *); |
| 1018 | extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags); |
| 1019 | extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s); |
| 1020 | extern int drbd_send_current_state(struct drbd_peer_device *); |
| 1021 | extern int drbd_send_sync_param(struct drbd_peer_device *); |
| 1022 | extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr, |
| 1023 | u32 set_size); |
| 1024 | extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet, |
| 1025 | struct drbd_peer_request *); |
| 1026 | extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet, |
| 1027 | struct p_block_req *rp); |
| 1028 | extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet, |
| 1029 | struct p_data *dp, int data_size); |
| 1030 | extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet, |
| 1031 | sector_t sector, int blksize, u64 block_id); |
| 1032 | extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *); |
| 1033 | extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet, |
| 1034 | struct drbd_peer_request *); |
| 1035 | extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req); |
| 1036 | extern int drbd_send_drequest(struct drbd_peer_device *, int cmd, |
| 1037 | sector_t sector, int size, u64 block_id); |
| 1038 | extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector, |
| 1039 | int size, void *digest, int digest_size, |
| 1040 | enum drbd_packet cmd); |
| 1041 | extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size); |
| 1042 | |
| 1043 | extern int drbd_send_bitmap(struct drbd_device *device, struct drbd_peer_device *peer_device); |
| 1044 | extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode); |
| 1045 | extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode); |
| 1046 | extern int drbd_send_rs_deallocated(struct drbd_peer_device *, struct drbd_peer_request *); |
| 1047 | extern void drbd_backing_dev_free(struct drbd_device *device, struct drbd_backing_dev *ldev); |
| 1048 | extern void drbd_device_cleanup(struct drbd_device *device); |
| 1049 | extern void drbd_print_uuids(struct drbd_device *device, const char *text); |
| 1050 | extern void drbd_queue_unplug(struct drbd_device *device); |
| 1051 | |
| 1052 | extern void conn_md_sync(struct drbd_connection *connection); |
| 1053 | extern void drbd_md_write(struct drbd_device *device, void *buffer); |
| 1054 | extern void drbd_md_sync(struct drbd_device *device); |
| 1055 | extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev); |
| 1056 | extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); |
| 1057 | extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); |
| 1058 | extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local); |
| 1059 | extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local); |
| 1060 | extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local); |
| 1061 | extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); |
| 1062 | extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local); |
| 1063 | extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local); |
| 1064 | extern int drbd_md_test_flag(struct drbd_backing_dev *, int); |
| 1065 | extern void drbd_md_mark_dirty(struct drbd_device *device); |
| 1066 | extern void drbd_queue_bitmap_io(struct drbd_device *device, |
| 1067 | int (*io_fn)(struct drbd_device *, struct drbd_peer_device *), |
| 1068 | void (*done)(struct drbd_device *, int), |
| 1069 | char *why, enum bm_flag flags, |
| 1070 | struct drbd_peer_device *peer_device); |
| 1071 | extern int drbd_bitmap_io(struct drbd_device *device, |
| 1072 | int (*io_fn)(struct drbd_device *, struct drbd_peer_device *), |
| 1073 | char *why, enum bm_flag flags, |
| 1074 | struct drbd_peer_device *peer_device); |
| 1075 | extern int drbd_bitmap_io_from_worker(struct drbd_device *device, |
| 1076 | int (*io_fn)(struct drbd_device *, struct drbd_peer_device *), |
| 1077 | char *why, enum bm_flag flags, |
| 1078 | struct drbd_peer_device *peer_device); |
| 1079 | extern int drbd_bmio_set_n_write(struct drbd_device *device, |
| 1080 | struct drbd_peer_device *peer_device) __must_hold(local); |
| 1081 | extern int drbd_bmio_clear_n_write(struct drbd_device *device, |
| 1082 | struct drbd_peer_device *peer_device) __must_hold(local); |
| 1083 | |
| 1084 | /* Meta data layout |
| 1085 | * |
| 1086 | * We currently have two possible layouts. |
| 1087 | * Offsets in (512 byte) sectors. |
| 1088 | * external: |
| 1089 | * |----------- md_size_sect ------------------| |
| 1090 | * [ 4k superblock ][ activity log ][ Bitmap ] |
| 1091 | * | al_offset == 8 | |
| 1092 | * | bm_offset = al_offset + X | |
| 1093 | * ==> bitmap sectors = md_size_sect - bm_offset |
| 1094 | * |
| 1095 | * Variants: |
| 1096 | * old, indexed fixed size meta data: |
| 1097 | * |
| 1098 | * internal: |
| 1099 | * |----------- md_size_sect ------------------| |
| 1100 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*] |
| 1101 | * | al_offset < 0 | |
| 1102 | * | bm_offset = al_offset - Y | |
| 1103 | * ==> bitmap sectors = Y = al_offset - bm_offset |
| 1104 | * |
| 1105 | * [padding*] are zero or up to 7 unused 512 Byte sectors to the |
| 1106 | * end of the device, so that the [4k superblock] will be 4k aligned. |
| 1107 | * |
| 1108 | * The activity log consists of 4k transaction blocks, |
| 1109 | * which are written in a ring-buffer, or striped ring-buffer like fashion, |
| 1110 | * which are writtensize used to be fixed 32kB, |
| 1111 | * but is about to become configurable. |
| 1112 | */ |
| 1113 | |
| 1114 | /* Our old fixed size meta data layout |
| 1115 | * allows up to about 3.8TB, so if you want more, |
| 1116 | * you need to use the "flexible" meta data format. */ |
| 1117 | #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */ |
| 1118 | #define MD_4kB_SECT 8 |
| 1119 | #define MD_32kB_SECT 64 |
| 1120 | |
| 1121 | /* One activity log extent represents 4M of storage */ |
| 1122 | #define AL_EXTENT_SHIFT 22 |
| 1123 | #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT) |
| 1124 | |
| 1125 | /* We could make these currently hardcoded constants configurable |
| 1126 | * variables at create-md time (or even re-configurable at runtime?). |
| 1127 | * Which will require some more changes to the DRBD "super block" |
| 1128 | * and attach code. |
| 1129 | * |
| 1130 | * updates per transaction: |
| 1131 | * This many changes to the active set can be logged with one transaction. |
| 1132 | * This number is arbitrary. |
| 1133 | * context per transaction: |
| 1134 | * This many context extent numbers are logged with each transaction. |
| 1135 | * This number is resulting from the transaction block size (4k), the layout |
| 1136 | * of the transaction header, and the number of updates per transaction. |
| 1137 | * See drbd_actlog.c:struct al_transaction_on_disk |
| 1138 | * */ |
| 1139 | #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary |
| 1140 | #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4 |
| 1141 | |
| 1142 | #if BITS_PER_LONG == 32 |
| 1143 | #define LN2_BPL 5 |
| 1144 | #define cpu_to_lel(A) cpu_to_le32(A) |
| 1145 | #define lel_to_cpu(A) le32_to_cpu(A) |
| 1146 | #elif BITS_PER_LONG == 64 |
| 1147 | #define LN2_BPL 6 |
| 1148 | #define cpu_to_lel(A) cpu_to_le64(A) |
| 1149 | #define lel_to_cpu(A) le64_to_cpu(A) |
| 1150 | #else |
| 1151 | #error "LN2 of BITS_PER_LONG unknown!" |
| 1152 | #endif |
| 1153 | |
| 1154 | /* resync bitmap */ |
| 1155 | /* 16MB sized 'bitmap extent' to track syncer usage */ |
| 1156 | struct bm_extent { |
| 1157 | int rs_left; /* number of bits set (out of sync) in this extent. */ |
| 1158 | int rs_failed; /* number of failed resync requests in this extent. */ |
| 1159 | unsigned long flags; |
| 1160 | struct lc_element lce; |
| 1161 | }; |
| 1162 | |
| 1163 | #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */ |
| 1164 | #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */ |
| 1165 | #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */ |
| 1166 | |
| 1167 | /* drbd_bitmap.c */ |
| 1168 | /* |
| 1169 | * We need to store one bit for a block. |
| 1170 | * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap. |
| 1171 | * Bit 0 ==> local node thinks this block is binary identical on both nodes |
| 1172 | * Bit 1 ==> local node thinks this block needs to be synced. |
| 1173 | */ |
| 1174 | |
| 1175 | #define SLEEP_TIME (HZ/10) |
| 1176 | |
| 1177 | /* We do bitmap IO in units of 4k blocks. |
| 1178 | * We also still have a hardcoded 4k per bit relation. */ |
| 1179 | #define BM_BLOCK_SHIFT 12 /* 4k per bit */ |
| 1180 | #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT) |
| 1181 | /* mostly arbitrarily set the represented size of one bitmap extent, |
| 1182 | * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap |
| 1183 | * at 4k per bit resolution) */ |
| 1184 | #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */ |
| 1185 | #define BM_EXT_SIZE (1<<BM_EXT_SHIFT) |
| 1186 | |
| 1187 | #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12) |
| 1188 | #error "HAVE YOU FIXED drbdmeta AS WELL??" |
| 1189 | #endif |
| 1190 | |
| 1191 | /* thus many _storage_ sectors are described by one bit */ |
| 1192 | #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9)) |
| 1193 | #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9)) |
| 1194 | #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1) |
| 1195 | |
| 1196 | /* bit to represented kilo byte conversion */ |
| 1197 | #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10)) |
| 1198 | |
| 1199 | /* in which _bitmap_ extent (resp. sector) the bit for a certain |
| 1200 | * _storage_ sector is located in */ |
| 1201 | #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9)) |
| 1202 | #define BM_BIT_TO_EXT(x) ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT)) |
| 1203 | |
| 1204 | /* first storage sector a bitmap extent corresponds to */ |
| 1205 | #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9)) |
| 1206 | /* how much _storage_ sectors we have per bitmap extent */ |
| 1207 | #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1) |
| 1208 | /* how many bits are covered by one bitmap extent (resync extent) */ |
| 1209 | #define BM_BITS_PER_EXT (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT)) |
| 1210 | |
| 1211 | #define BM_BLOCKS_PER_BM_EXT_MASK (BM_BITS_PER_EXT - 1) |
| 1212 | |
| 1213 | |
| 1214 | /* in one sector of the bitmap, we have this many activity_log extents. */ |
| 1215 | #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT)) |
| 1216 | |
| 1217 | /* the extent in "PER_EXTENT" below is an activity log extent |
| 1218 | * we need that many (long words/bytes) to store the bitmap |
| 1219 | * of one AL_EXTENT_SIZE chunk of storage. |
| 1220 | * we can store the bitmap for that many AL_EXTENTS within |
| 1221 | * one sector of the _on_disk_ bitmap: |
| 1222 | * bit 0 bit 37 bit 38 bit (512*8)-1 |
| 1223 | * ...|........|........|.. // ..|........| |
| 1224 | * sect. 0 `296 `304 ^(512*8*8)-1 |
| 1225 | * |
| 1226 | #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG ) |
| 1227 | #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128 |
| 1228 | #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4 |
| 1229 | */ |
| 1230 | |
| 1231 | #define DRBD_MAX_SECTORS_32 (0xffffffffLU) |
| 1232 | /* we have a certain meta data variant that has a fixed on-disk size of 128 |
| 1233 | * MiB, of which 4k are our "superblock", and 32k are the fixed size activity |
| 1234 | * log, leaving this many sectors for the bitmap. |
| 1235 | */ |
| 1236 | |
| 1237 | #define DRBD_MAX_SECTORS_FIXED_BM \ |
| 1238 | ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9))) |
| 1239 | #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM |
| 1240 | /* 16 TB in units of sectors */ |
| 1241 | #if BITS_PER_LONG == 32 |
| 1242 | /* adjust by one page worth of bitmap, |
| 1243 | * so we won't wrap around in drbd_bm_find_next_bit. |
| 1244 | * you should use 64bit OS for that much storage, anyways. */ |
| 1245 | #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff) |
| 1246 | #else |
| 1247 | /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */ |
| 1248 | #define DRBD_MAX_SECTORS_FLEX (1UL << 51) |
| 1249 | /* corresponds to (1UL << 38) bits right now. */ |
| 1250 | #endif |
| 1251 | |
| 1252 | /* Estimate max bio size as 256 * PAGE_SIZE, |
| 1253 | * so for typical PAGE_SIZE of 4k, that is (1<<20) Byte. |
| 1254 | * Since we may live in a mixed-platform cluster, |
| 1255 | * we limit us to a platform agnostic constant here for now. |
| 1256 | * A followup commit may allow even bigger BIO sizes, |
| 1257 | * once we thought that through. */ |
| 1258 | #define DRBD_MAX_BIO_SIZE (1U << 20) |
| 1259 | #if DRBD_MAX_BIO_SIZE > (BIO_MAX_VECS << PAGE_SHIFT) |
| 1260 | #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE |
| 1261 | #endif |
| 1262 | #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */ |
| 1263 | |
| 1264 | #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */ |
| 1265 | #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */ |
| 1266 | |
| 1267 | /* For now, don't allow more than half of what we can "activate" in one |
| 1268 | * activity log transaction to be discarded in one go. We may need to rework |
| 1269 | * drbd_al_begin_io() to allow for even larger discard ranges */ |
| 1270 | #define DRBD_MAX_BATCH_BIO_SIZE (AL_UPDATES_PER_TRANSACTION/2*AL_EXTENT_SIZE) |
| 1271 | #define DRBD_MAX_BBIO_SECTORS (DRBD_MAX_BATCH_BIO_SIZE >> 9) |
| 1272 | |
| 1273 | extern int drbd_bm_init(struct drbd_device *device); |
| 1274 | extern int drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits); |
| 1275 | extern void drbd_bm_cleanup(struct drbd_device *device); |
| 1276 | extern void drbd_bm_set_all(struct drbd_device *device); |
| 1277 | extern void drbd_bm_clear_all(struct drbd_device *device); |
| 1278 | /* set/clear/test only a few bits at a time */ |
| 1279 | extern int drbd_bm_set_bits( |
| 1280 | struct drbd_device *device, unsigned long s, unsigned long e); |
| 1281 | extern int drbd_bm_clear_bits( |
| 1282 | struct drbd_device *device, unsigned long s, unsigned long e); |
| 1283 | extern int drbd_bm_count_bits( |
| 1284 | struct drbd_device *device, const unsigned long s, const unsigned long e); |
| 1285 | /* bm_set_bits variant for use while holding drbd_bm_lock, |
| 1286 | * may process the whole bitmap in one go */ |
| 1287 | extern void _drbd_bm_set_bits(struct drbd_device *device, |
| 1288 | const unsigned long s, const unsigned long e); |
| 1289 | extern int drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr); |
| 1290 | extern int drbd_bm_e_weight(struct drbd_device *device, unsigned long enr); |
| 1291 | extern int drbd_bm_read(struct drbd_device *device, |
| 1292 | struct drbd_peer_device *peer_device) __must_hold(local); |
| 1293 | extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr); |
| 1294 | extern int drbd_bm_write(struct drbd_device *device, |
| 1295 | struct drbd_peer_device *peer_device) __must_hold(local); |
| 1296 | extern void drbd_bm_reset_al_hints(struct drbd_device *device) __must_hold(local); |
| 1297 | extern int drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local); |
| 1298 | extern int drbd_bm_write_lazy(struct drbd_device *device, unsigned upper_idx) __must_hold(local); |
| 1299 | extern int drbd_bm_write_all(struct drbd_device *device, |
| 1300 | struct drbd_peer_device *peer_device) __must_hold(local); |
| 1301 | extern int drbd_bm_write_copy_pages(struct drbd_device *device, |
| 1302 | struct drbd_peer_device *peer_device) __must_hold(local); |
| 1303 | extern size_t drbd_bm_words(struct drbd_device *device); |
| 1304 | extern unsigned long drbd_bm_bits(struct drbd_device *device); |
| 1305 | extern sector_t drbd_bm_capacity(struct drbd_device *device); |
| 1306 | |
| 1307 | #define DRBD_END_OF_BITMAP (~(unsigned long)0) |
| 1308 | extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); |
| 1309 | /* bm_find_next variants for use while you hold drbd_bm_lock() */ |
| 1310 | extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); |
| 1311 | extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo); |
| 1312 | extern unsigned long _drbd_bm_total_weight(struct drbd_device *device); |
| 1313 | extern unsigned long drbd_bm_total_weight(struct drbd_device *device); |
| 1314 | /* for receive_bitmap */ |
| 1315 | extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset, |
| 1316 | size_t number, unsigned long *buffer); |
| 1317 | /* for _drbd_send_bitmap */ |
| 1318 | extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset, |
| 1319 | size_t number, unsigned long *buffer); |
| 1320 | |
| 1321 | extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags); |
| 1322 | extern void drbd_bm_unlock(struct drbd_device *device); |
| 1323 | /* drbd_main.c */ |
| 1324 | |
| 1325 | extern struct kmem_cache *drbd_request_cache; |
| 1326 | extern struct kmem_cache *drbd_ee_cache; /* peer requests */ |
| 1327 | extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */ |
| 1328 | extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */ |
| 1329 | extern mempool_t drbd_request_mempool; |
| 1330 | extern mempool_t drbd_ee_mempool; |
| 1331 | |
| 1332 | /* drbd's page pool, used to buffer data received from the peer, |
| 1333 | * or data requested by the peer. |
| 1334 | * |
| 1335 | * This does not have an emergency reserve. |
| 1336 | * |
| 1337 | * When allocating from this pool, it first takes pages from the pool. |
| 1338 | * Only if the pool is depleted will try to allocate from the system. |
| 1339 | * |
| 1340 | * The assumption is that pages taken from this pool will be processed, |
| 1341 | * and given back, "quickly", and then can be recycled, so we can avoid |
| 1342 | * frequent calls to alloc_page(), and still will be able to make progress even |
| 1343 | * under memory pressure. |
| 1344 | */ |
| 1345 | extern struct page *drbd_pp_pool; |
| 1346 | extern spinlock_t drbd_pp_lock; |
| 1347 | extern int drbd_pp_vacant; |
| 1348 | extern wait_queue_head_t drbd_pp_wait; |
| 1349 | |
| 1350 | /* We also need a standard (emergency-reserve backed) page pool |
| 1351 | * for meta data IO (activity log, bitmap). |
| 1352 | * We can keep it global, as long as it is used as "N pages at a time". |
| 1353 | * 128 should be plenty, currently we probably can get away with as few as 1. |
| 1354 | */ |
| 1355 | #define DRBD_MIN_POOL_PAGES 128 |
| 1356 | extern mempool_t drbd_md_io_page_pool; |
| 1357 | |
| 1358 | /* We also need to make sure we get a bio |
| 1359 | * when we need it for housekeeping purposes */ |
| 1360 | extern struct bio_set drbd_md_io_bio_set; |
| 1361 | |
| 1362 | /* And a bio_set for cloning */ |
| 1363 | extern struct bio_set drbd_io_bio_set; |
| 1364 | |
| 1365 | extern struct mutex resources_mutex; |
| 1366 | |
| 1367 | extern int conn_lowest_minor(struct drbd_connection *connection); |
| 1368 | extern enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor); |
| 1369 | extern void drbd_destroy_device(struct kref *kref); |
| 1370 | extern void drbd_delete_device(struct drbd_device *device); |
| 1371 | |
| 1372 | extern struct drbd_resource *drbd_create_resource(const char *name); |
| 1373 | extern void drbd_free_resource(struct drbd_resource *resource); |
| 1374 | |
| 1375 | extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts); |
| 1376 | extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts); |
| 1377 | extern void drbd_destroy_connection(struct kref *kref); |
| 1378 | extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len, |
| 1379 | void *peer_addr, int peer_addr_len); |
| 1380 | extern struct drbd_resource *drbd_find_resource(const char *name); |
| 1381 | extern void drbd_destroy_resource(struct kref *kref); |
| 1382 | extern void conn_free_crypto(struct drbd_connection *connection); |
| 1383 | |
| 1384 | /* drbd_req */ |
| 1385 | extern void do_submit(struct work_struct *ws); |
| 1386 | extern void __drbd_make_request(struct drbd_device *, struct bio *); |
| 1387 | void drbd_submit_bio(struct bio *bio); |
| 1388 | |
| 1389 | /* drbd_nl.c */ |
| 1390 | |
| 1391 | extern struct mutex notification_mutex; |
| 1392 | |
| 1393 | extern void drbd_suspend_io(struct drbd_device *device); |
| 1394 | extern void drbd_resume_io(struct drbd_device *device); |
| 1395 | extern char *ppsize(char *buf, unsigned long long size); |
| 1396 | extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int); |
| 1397 | enum determine_dev_size { |
| 1398 | DS_ERROR_SHRINK = -3, |
| 1399 | DS_ERROR_SPACE_MD = -2, |
| 1400 | DS_ERROR = -1, |
| 1401 | DS_UNCHANGED = 0, |
| 1402 | DS_SHRUNK = 1, |
| 1403 | DS_GREW = 2, |
| 1404 | DS_GREW_FROM_ZERO = 3, |
| 1405 | }; |
| 1406 | extern enum determine_dev_size |
| 1407 | drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local); |
| 1408 | extern void resync_after_online_grow(struct drbd_device *); |
| 1409 | extern void drbd_reconsider_queue_parameters(struct drbd_device *device, |
| 1410 | struct drbd_backing_dev *bdev, struct o_qlim *o); |
| 1411 | extern enum drbd_state_rv drbd_set_role(struct drbd_device *device, |
| 1412 | enum drbd_role new_role, |
| 1413 | int force); |
| 1414 | extern bool conn_try_outdate_peer(struct drbd_connection *connection); |
| 1415 | extern void conn_try_outdate_peer_async(struct drbd_connection *connection); |
| 1416 | extern enum drbd_peer_state conn_khelper(struct drbd_connection *connection, char *cmd); |
| 1417 | extern int drbd_khelper(struct drbd_device *device, char *cmd); |
| 1418 | |
| 1419 | /* drbd_worker.c */ |
| 1420 | /* bi_end_io handlers */ |
| 1421 | extern void drbd_md_endio(struct bio *bio); |
| 1422 | extern void drbd_peer_request_endio(struct bio *bio); |
| 1423 | extern void drbd_request_endio(struct bio *bio); |
| 1424 | extern int drbd_worker(struct drbd_thread *thi); |
| 1425 | enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor); |
| 1426 | void drbd_resync_after_changed(struct drbd_device *device); |
| 1427 | extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side); |
| 1428 | extern void resume_next_sg(struct drbd_device *device); |
| 1429 | extern void suspend_other_sg(struct drbd_device *device); |
| 1430 | extern int drbd_resync_finished(struct drbd_peer_device *peer_device); |
| 1431 | /* maybe rather drbd_main.c ? */ |
| 1432 | extern void *drbd_md_get_buffer(struct drbd_device *device, const char *intent); |
| 1433 | extern void drbd_md_put_buffer(struct drbd_device *device); |
| 1434 | extern int drbd_md_sync_page_io(struct drbd_device *device, |
| 1435 | struct drbd_backing_dev *bdev, sector_t sector, enum req_op op); |
| 1436 | extern void drbd_ov_out_of_sync_found(struct drbd_peer_device *peer_device, |
| 1437 | sector_t sector, int size); |
| 1438 | extern void wait_until_done_or_force_detached(struct drbd_device *device, |
| 1439 | struct drbd_backing_dev *bdev, unsigned int *done); |
| 1440 | extern void drbd_rs_controller_reset(struct drbd_peer_device *peer_device); |
| 1441 | |
| 1442 | static inline void ov_out_of_sync_print(struct drbd_peer_device *peer_device) |
| 1443 | { |
| 1444 | struct drbd_device *device = peer_device->device; |
| 1445 | |
| 1446 | if (device->ov_last_oos_size) { |
| 1447 | drbd_err(peer_device, "Out of sync: start=%llu, size=%lu (sectors)\n", |
| 1448 | (unsigned long long)device->ov_last_oos_start, |
| 1449 | (unsigned long)device->ov_last_oos_size); |
| 1450 | } |
| 1451 | device->ov_last_oos_size = 0; |
| 1452 | } |
| 1453 | |
| 1454 | |
| 1455 | extern void drbd_csum_bio(struct crypto_shash *, struct bio *, void *); |
| 1456 | extern void drbd_csum_ee(struct crypto_shash *, struct drbd_peer_request *, |
| 1457 | void *); |
| 1458 | /* worker callbacks */ |
| 1459 | extern int w_e_end_data_req(struct drbd_work *, int); |
| 1460 | extern int w_e_end_rsdata_req(struct drbd_work *, int); |
| 1461 | extern int w_e_end_csum_rs_req(struct drbd_work *, int); |
| 1462 | extern int w_e_end_ov_reply(struct drbd_work *, int); |
| 1463 | extern int w_e_end_ov_req(struct drbd_work *, int); |
| 1464 | extern int w_ov_finished(struct drbd_work *, int); |
| 1465 | extern int w_resync_timer(struct drbd_work *, int); |
| 1466 | extern int w_send_write_hint(struct drbd_work *, int); |
| 1467 | extern int w_send_dblock(struct drbd_work *, int); |
| 1468 | extern int w_send_read_req(struct drbd_work *, int); |
| 1469 | extern int w_restart_disk_io(struct drbd_work *, int); |
| 1470 | extern int w_send_out_of_sync(struct drbd_work *, int); |
| 1471 | |
| 1472 | extern void resync_timer_fn(struct timer_list *t); |
| 1473 | extern void start_resync_timer_fn(struct timer_list *t); |
| 1474 | |
| 1475 | extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req); |
| 1476 | |
| 1477 | /* drbd_receiver.c */ |
| 1478 | extern int drbd_issue_discard_or_zero_out(struct drbd_device *device, |
| 1479 | sector_t start, unsigned int nr_sectors, int flags); |
| 1480 | extern int drbd_receiver(struct drbd_thread *thi); |
| 1481 | extern int drbd_ack_receiver(struct drbd_thread *thi); |
| 1482 | extern void drbd_send_acks_wf(struct work_struct *ws); |
| 1483 | extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device); |
| 1484 | extern bool drbd_rs_should_slow_down(struct drbd_peer_device *peer_device, sector_t sector, |
| 1485 | bool throttle_if_app_is_waiting); |
| 1486 | extern int drbd_submit_peer_request(struct drbd_peer_request *peer_req); |
| 1487 | extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *); |
| 1488 | extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64, |
| 1489 | sector_t, unsigned int, |
| 1490 | unsigned int, |
| 1491 | gfp_t) __must_hold(local); |
| 1492 | extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *, |
| 1493 | int); |
| 1494 | #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0) |
| 1495 | #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1) |
| 1496 | extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool); |
| 1497 | extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed); |
| 1498 | extern int drbd_connected(struct drbd_peer_device *); |
| 1499 | |
| 1500 | /* sets the number of 512 byte sectors of our virtual device */ |
| 1501 | void drbd_set_my_capacity(struct drbd_device *device, sector_t size); |
| 1502 | |
| 1503 | /* |
| 1504 | * used to submit our private bio |
| 1505 | */ |
| 1506 | static inline void drbd_submit_bio_noacct(struct drbd_device *device, |
| 1507 | int fault_type, struct bio *bio) |
| 1508 | { |
| 1509 | __release(local); |
| 1510 | if (!bio->bi_bdev) { |
| 1511 | drbd_err(device, "drbd_submit_bio_noacct: bio->bi_bdev == NULL\n"); |
| 1512 | bio->bi_status = BLK_STS_IOERR; |
| 1513 | bio_endio(bio); |
| 1514 | return; |
| 1515 | } |
| 1516 | |
| 1517 | if (drbd_insert_fault(device, fault_type)) |
| 1518 | bio_io_error(bio); |
| 1519 | else |
| 1520 | submit_bio_noacct(bio); |
| 1521 | } |
| 1522 | |
| 1523 | void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev, |
| 1524 | enum write_ordering_e wo); |
| 1525 | |
| 1526 | /* drbd_proc.c */ |
| 1527 | extern struct proc_dir_entry *drbd_proc; |
| 1528 | int drbd_seq_show(struct seq_file *seq, void *v); |
| 1529 | |
| 1530 | /* drbd_actlog.c */ |
| 1531 | extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i); |
| 1532 | extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i); |
| 1533 | extern void drbd_al_begin_io_commit(struct drbd_device *device); |
| 1534 | extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i); |
| 1535 | extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i); |
| 1536 | extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i); |
| 1537 | extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector); |
| 1538 | extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector); |
| 1539 | extern int drbd_try_rs_begin_io(struct drbd_peer_device *peer_device, sector_t sector); |
| 1540 | extern void drbd_rs_cancel_all(struct drbd_device *device); |
| 1541 | extern int drbd_rs_del_all(struct drbd_device *device); |
| 1542 | extern void drbd_rs_failed_io(struct drbd_peer_device *peer_device, |
| 1543 | sector_t sector, int size); |
| 1544 | extern void drbd_advance_rs_marks(struct drbd_peer_device *peer_device, unsigned long still_to_go); |
| 1545 | |
| 1546 | enum update_sync_bits_mode { RECORD_RS_FAILED, SET_OUT_OF_SYNC, SET_IN_SYNC }; |
| 1547 | extern int __drbd_change_sync(struct drbd_peer_device *peer_device, sector_t sector, int size, |
| 1548 | enum update_sync_bits_mode mode); |
| 1549 | #define drbd_set_in_sync(peer_device, sector, size) \ |
| 1550 | __drbd_change_sync(peer_device, sector, size, SET_IN_SYNC) |
| 1551 | #define drbd_set_out_of_sync(peer_device, sector, size) \ |
| 1552 | __drbd_change_sync(peer_device, sector, size, SET_OUT_OF_SYNC) |
| 1553 | #define drbd_rs_failed_io(peer_device, sector, size) \ |
| 1554 | __drbd_change_sync(peer_device, sector, size, RECORD_RS_FAILED) |
| 1555 | extern void drbd_al_shrink(struct drbd_device *device); |
| 1556 | extern int drbd_al_initialize(struct drbd_device *, void *); |
| 1557 | |
| 1558 | /* drbd_nl.c */ |
| 1559 | /* state info broadcast */ |
| 1560 | struct sib_info { |
| 1561 | enum drbd_state_info_bcast_reason sib_reason; |
| 1562 | union { |
| 1563 | struct { |
| 1564 | char *helper_name; |
| 1565 | unsigned helper_exit_code; |
| 1566 | }; |
| 1567 | struct { |
| 1568 | union drbd_state os; |
| 1569 | union drbd_state ns; |
| 1570 | }; |
| 1571 | }; |
| 1572 | }; |
| 1573 | void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib); |
| 1574 | |
| 1575 | extern int notify_resource_state(struct sk_buff *, |
| 1576 | unsigned int, |
| 1577 | struct drbd_resource *, |
| 1578 | struct resource_info *, |
| 1579 | enum drbd_notification_type); |
| 1580 | extern int notify_device_state(struct sk_buff *, |
| 1581 | unsigned int, |
| 1582 | struct drbd_device *, |
| 1583 | struct device_info *, |
| 1584 | enum drbd_notification_type); |
| 1585 | extern int notify_connection_state(struct sk_buff *, |
| 1586 | unsigned int, |
| 1587 | struct drbd_connection *, |
| 1588 | struct connection_info *, |
| 1589 | enum drbd_notification_type); |
| 1590 | extern int notify_peer_device_state(struct sk_buff *, |
| 1591 | unsigned int, |
| 1592 | struct drbd_peer_device *, |
| 1593 | struct peer_device_info *, |
| 1594 | enum drbd_notification_type); |
| 1595 | extern void notify_helper(enum drbd_notification_type, struct drbd_device *, |
| 1596 | struct drbd_connection *, const char *, int); |
| 1597 | |
| 1598 | /* |
| 1599 | * inline helper functions |
| 1600 | *************************/ |
| 1601 | |
| 1602 | /* see also page_chain_add and friends in drbd_receiver.c */ |
| 1603 | static inline struct page *page_chain_next(struct page *page) |
| 1604 | { |
| 1605 | return (struct page *)page_private(page); |
| 1606 | } |
| 1607 | #define page_chain_for_each(page) \ |
| 1608 | for (; page && ({ prefetch(page_chain_next(page)); 1; }); \ |
| 1609 | page = page_chain_next(page)) |
| 1610 | #define page_chain_for_each_safe(page, n) \ |
| 1611 | for (; page && ({ n = page_chain_next(page); 1; }); page = n) |
| 1612 | |
| 1613 | |
| 1614 | static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req) |
| 1615 | { |
| 1616 | struct page *page = peer_req->pages; |
| 1617 | page_chain_for_each(page) { |
| 1618 | if (page_count(page) > 1) |
| 1619 | return 1; |
| 1620 | } |
| 1621 | return 0; |
| 1622 | } |
| 1623 | |
| 1624 | static inline union drbd_state drbd_read_state(struct drbd_device *device) |
| 1625 | { |
| 1626 | struct drbd_resource *resource = device->resource; |
| 1627 | union drbd_state rv; |
| 1628 | |
| 1629 | rv.i = device->state.i; |
| 1630 | rv.susp = resource->susp; |
| 1631 | rv.susp_nod = resource->susp_nod; |
| 1632 | rv.susp_fen = resource->susp_fen; |
| 1633 | |
| 1634 | return rv; |
| 1635 | } |
| 1636 | |
| 1637 | enum drbd_force_detach_flags { |
| 1638 | DRBD_READ_ERROR, |
| 1639 | DRBD_WRITE_ERROR, |
| 1640 | DRBD_META_IO_ERROR, |
| 1641 | DRBD_FORCE_DETACH, |
| 1642 | }; |
| 1643 | |
| 1644 | #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__) |
| 1645 | static inline void __drbd_chk_io_error_(struct drbd_device *device, |
| 1646 | enum drbd_force_detach_flags df, |
| 1647 | const char *where) |
| 1648 | { |
| 1649 | enum drbd_io_error_p ep; |
| 1650 | |
| 1651 | rcu_read_lock(); |
| 1652 | ep = rcu_dereference(device->ldev->disk_conf)->on_io_error; |
| 1653 | rcu_read_unlock(); |
| 1654 | switch (ep) { |
| 1655 | case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */ |
| 1656 | if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) { |
| 1657 | if (drbd_ratelimit()) |
| 1658 | drbd_err(device, "Local IO failed in %s.\n", where); |
| 1659 | if (device->state.disk > D_INCONSISTENT) |
| 1660 | _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL); |
| 1661 | break; |
| 1662 | } |
| 1663 | fallthrough; /* for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */ |
| 1664 | case EP_DETACH: |
| 1665 | case EP_CALL_HELPER: |
| 1666 | /* Remember whether we saw a READ or WRITE error. |
| 1667 | * |
| 1668 | * Recovery of the affected area for WRITE failure is covered |
| 1669 | * by the activity log. |
| 1670 | * READ errors may fall outside that area though. Certain READ |
| 1671 | * errors can be "healed" by writing good data to the affected |
| 1672 | * blocks, which triggers block re-allocation in lower layers. |
| 1673 | * |
| 1674 | * If we can not write the bitmap after a READ error, |
| 1675 | * we may need to trigger a full sync (see w_go_diskless()). |
| 1676 | * |
| 1677 | * Force-detach is not really an IO error, but rather a |
| 1678 | * desperate measure to try to deal with a completely |
| 1679 | * unresponsive lower level IO stack. |
| 1680 | * Still it should be treated as a WRITE error. |
| 1681 | * |
| 1682 | * Meta IO error is always WRITE error: |
| 1683 | * we read meta data only once during attach, |
| 1684 | * which will fail in case of errors. |
| 1685 | */ |
| 1686 | set_bit(WAS_IO_ERROR, &device->flags); |
| 1687 | if (df == DRBD_READ_ERROR) |
| 1688 | set_bit(WAS_READ_ERROR, &device->flags); |
| 1689 | if (df == DRBD_FORCE_DETACH) |
| 1690 | set_bit(FORCE_DETACH, &device->flags); |
| 1691 | if (device->state.disk > D_FAILED) { |
| 1692 | _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL); |
| 1693 | drbd_err(device, |
| 1694 | "Local IO failed in %s. Detaching...\n", where); |
| 1695 | } |
| 1696 | break; |
| 1697 | } |
| 1698 | } |
| 1699 | |
| 1700 | /** |
| 1701 | * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers |
| 1702 | * @device: DRBD device. |
| 1703 | * @error: Error code passed to the IO completion callback |
| 1704 | * @forcedetach: Force detach. I.e. the error happened while accessing the meta data |
| 1705 | * |
| 1706 | * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED) |
| 1707 | */ |
| 1708 | #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__) |
| 1709 | static inline void drbd_chk_io_error_(struct drbd_device *device, |
| 1710 | int error, enum drbd_force_detach_flags forcedetach, const char *where) |
| 1711 | { |
| 1712 | if (error) { |
| 1713 | unsigned long flags; |
| 1714 | spin_lock_irqsave(&device->resource->req_lock, flags); |
| 1715 | __drbd_chk_io_error_(device, forcedetach, where); |
| 1716 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
| 1717 | } |
| 1718 | } |
| 1719 | |
| 1720 | |
| 1721 | /** |
| 1722 | * drbd_md_first_sector() - Returns the first sector number of the meta data area |
| 1723 | * @bdev: Meta data block device. |
| 1724 | * |
| 1725 | * BTW, for internal meta data, this happens to be the maximum capacity |
| 1726 | * we could agree upon with our peer node. |
| 1727 | */ |
| 1728 | static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev) |
| 1729 | { |
| 1730 | switch (bdev->md.meta_dev_idx) { |
| 1731 | case DRBD_MD_INDEX_INTERNAL: |
| 1732 | case DRBD_MD_INDEX_FLEX_INT: |
| 1733 | return bdev->md.md_offset + bdev->md.bm_offset; |
| 1734 | case DRBD_MD_INDEX_FLEX_EXT: |
| 1735 | default: |
| 1736 | return bdev->md.md_offset; |
| 1737 | } |
| 1738 | } |
| 1739 | |
| 1740 | /** |
| 1741 | * drbd_md_last_sector() - Return the last sector number of the meta data area |
| 1742 | * @bdev: Meta data block device. |
| 1743 | */ |
| 1744 | static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev) |
| 1745 | { |
| 1746 | switch (bdev->md.meta_dev_idx) { |
| 1747 | case DRBD_MD_INDEX_INTERNAL: |
| 1748 | case DRBD_MD_INDEX_FLEX_INT: |
| 1749 | return bdev->md.md_offset + MD_4kB_SECT -1; |
| 1750 | case DRBD_MD_INDEX_FLEX_EXT: |
| 1751 | default: |
| 1752 | return bdev->md.md_offset + bdev->md.md_size_sect -1; |
| 1753 | } |
| 1754 | } |
| 1755 | |
| 1756 | /* Returns the number of 512 byte sectors of the device */ |
| 1757 | static inline sector_t drbd_get_capacity(struct block_device *bdev) |
| 1758 | { |
| 1759 | return bdev ? bdev_nr_sectors(bdev) : 0; |
| 1760 | } |
| 1761 | |
| 1762 | /** |
| 1763 | * drbd_get_max_capacity() - Returns the capacity we announce to out peer |
| 1764 | * @bdev: Meta data block device. |
| 1765 | * |
| 1766 | * returns the capacity we announce to out peer. we clip ourselves at the |
| 1767 | * various MAX_SECTORS, because if we don't, current implementation will |
| 1768 | * oops sooner or later |
| 1769 | */ |
| 1770 | static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev) |
| 1771 | { |
| 1772 | sector_t s; |
| 1773 | |
| 1774 | switch (bdev->md.meta_dev_idx) { |
| 1775 | case DRBD_MD_INDEX_INTERNAL: |
| 1776 | case DRBD_MD_INDEX_FLEX_INT: |
| 1777 | s = drbd_get_capacity(bdev->backing_bdev) |
| 1778 | ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX, |
| 1779 | drbd_md_first_sector(bdev)) |
| 1780 | : 0; |
| 1781 | break; |
| 1782 | case DRBD_MD_INDEX_FLEX_EXT: |
| 1783 | s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX, |
| 1784 | drbd_get_capacity(bdev->backing_bdev)); |
| 1785 | /* clip at maximum size the meta device can support */ |
| 1786 | s = min_t(sector_t, s, |
| 1787 | BM_EXT_TO_SECT(bdev->md.md_size_sect |
| 1788 | - bdev->md.bm_offset)); |
| 1789 | break; |
| 1790 | default: |
| 1791 | s = min_t(sector_t, DRBD_MAX_SECTORS, |
| 1792 | drbd_get_capacity(bdev->backing_bdev)); |
| 1793 | } |
| 1794 | return s; |
| 1795 | } |
| 1796 | |
| 1797 | /** |
| 1798 | * drbd_md_ss() - Return the sector number of our meta data super block |
| 1799 | * @bdev: Meta data block device. |
| 1800 | */ |
| 1801 | static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev) |
| 1802 | { |
| 1803 | const int meta_dev_idx = bdev->md.meta_dev_idx; |
| 1804 | |
| 1805 | if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT) |
| 1806 | return 0; |
| 1807 | |
| 1808 | /* Since drbd08, internal meta data is always "flexible". |
| 1809 | * position: last 4k aligned block of 4k size */ |
| 1810 | if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL || |
| 1811 | meta_dev_idx == DRBD_MD_INDEX_FLEX_INT) |
| 1812 | return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8; |
| 1813 | |
| 1814 | /* external, some index; this is the old fixed size layout */ |
| 1815 | return MD_128MB_SECT * bdev->md.meta_dev_idx; |
| 1816 | } |
| 1817 | |
| 1818 | static inline void |
| 1819 | drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w) |
| 1820 | { |
| 1821 | unsigned long flags; |
| 1822 | spin_lock_irqsave(&q->q_lock, flags); |
| 1823 | list_add_tail(&w->list, &q->q); |
| 1824 | spin_unlock_irqrestore(&q->q_lock, flags); |
| 1825 | wake_up(&q->q_wait); |
| 1826 | } |
| 1827 | |
| 1828 | static inline void |
| 1829 | drbd_queue_work_if_unqueued(struct drbd_work_queue *q, struct drbd_work *w) |
| 1830 | { |
| 1831 | unsigned long flags; |
| 1832 | spin_lock_irqsave(&q->q_lock, flags); |
| 1833 | if (list_empty_careful(&w->list)) |
| 1834 | list_add_tail(&w->list, &q->q); |
| 1835 | spin_unlock_irqrestore(&q->q_lock, flags); |
| 1836 | wake_up(&q->q_wait); |
| 1837 | } |
| 1838 | |
| 1839 | static inline void |
| 1840 | drbd_device_post_work(struct drbd_device *device, int work_bit) |
| 1841 | { |
| 1842 | if (!test_and_set_bit(work_bit, &device->flags)) { |
| 1843 | struct drbd_connection *connection = |
| 1844 | first_peer_device(device)->connection; |
| 1845 | struct drbd_work_queue *q = &connection->sender_work; |
| 1846 | if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags)) |
| 1847 | wake_up(&q->q_wait); |
| 1848 | } |
| 1849 | } |
| 1850 | |
| 1851 | extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue); |
| 1852 | |
| 1853 | /* To get the ack_receiver out of the blocking network stack, |
| 1854 | * so it can change its sk_rcvtimeo from idle- to ping-timeout, |
| 1855 | * and send a ping, we need to send a signal. |
| 1856 | * Which signal we send is irrelevant. */ |
| 1857 | static inline void wake_ack_receiver(struct drbd_connection *connection) |
| 1858 | { |
| 1859 | struct task_struct *task = connection->ack_receiver.task; |
| 1860 | if (task && get_t_state(&connection->ack_receiver) == RUNNING) |
| 1861 | send_sig(SIGXCPU, task, 1); |
| 1862 | } |
| 1863 | |
| 1864 | static inline void request_ping(struct drbd_connection *connection) |
| 1865 | { |
| 1866 | set_bit(SEND_PING, &connection->flags); |
| 1867 | wake_ack_receiver(connection); |
| 1868 | } |
| 1869 | |
| 1870 | extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *); |
| 1871 | extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *); |
| 1872 | extern int conn_send_command(struct drbd_connection *, struct drbd_socket *, |
| 1873 | enum drbd_packet, unsigned int, void *, |
| 1874 | unsigned int); |
| 1875 | extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *, |
| 1876 | enum drbd_packet, unsigned int, void *, |
| 1877 | unsigned int); |
| 1878 | |
| 1879 | extern int drbd_send_ping(struct drbd_connection *connection); |
| 1880 | extern int drbd_send_ping_ack(struct drbd_connection *connection); |
| 1881 | extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state); |
| 1882 | extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state); |
| 1883 | |
| 1884 | static inline void drbd_thread_stop(struct drbd_thread *thi) |
| 1885 | { |
| 1886 | _drbd_thread_stop(thi, false, true); |
| 1887 | } |
| 1888 | |
| 1889 | static inline void drbd_thread_stop_nowait(struct drbd_thread *thi) |
| 1890 | { |
| 1891 | _drbd_thread_stop(thi, false, false); |
| 1892 | } |
| 1893 | |
| 1894 | static inline void drbd_thread_restart_nowait(struct drbd_thread *thi) |
| 1895 | { |
| 1896 | _drbd_thread_stop(thi, true, false); |
| 1897 | } |
| 1898 | |
| 1899 | /* counts how many answer packets packets we expect from our peer, |
| 1900 | * for either explicit application requests, |
| 1901 | * or implicit barrier packets as necessary. |
| 1902 | * increased: |
| 1903 | * w_send_barrier |
| 1904 | * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ); |
| 1905 | * it is much easier and equally valid to count what we queue for the |
| 1906 | * worker, even before it actually was queued or send. |
| 1907 | * (drbd_make_request_common; recovery path on read io-error) |
| 1908 | * decreased: |
| 1909 | * got_BarrierAck (respective tl_clear, tl_clear_barrier) |
| 1910 | * _req_mod(req, DATA_RECEIVED) |
| 1911 | * [from receive_DataReply] |
| 1912 | * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED) |
| 1913 | * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)] |
| 1914 | * for some reason it is NOT decreased in got_NegAck, |
| 1915 | * but in the resulting cleanup code from report_params. |
| 1916 | * we should try to remember the reason for that... |
| 1917 | * _req_mod(req, SEND_FAILED or SEND_CANCELED) |
| 1918 | * _req_mod(req, CONNECTION_LOST_WHILE_PENDING) |
| 1919 | * [from tl_clear_barrier] |
| 1920 | */ |
| 1921 | static inline void inc_ap_pending(struct drbd_device *device) |
| 1922 | { |
| 1923 | atomic_inc(&device->ap_pending_cnt); |
| 1924 | } |
| 1925 | |
| 1926 | #define dec_ap_pending(device) ((void)expect((device), __dec_ap_pending(device) >= 0)) |
| 1927 | static inline int __dec_ap_pending(struct drbd_device *device) |
| 1928 | { |
| 1929 | int ap_pending_cnt = atomic_dec_return(&device->ap_pending_cnt); |
| 1930 | |
| 1931 | if (ap_pending_cnt == 0) |
| 1932 | wake_up(&device->misc_wait); |
| 1933 | return ap_pending_cnt; |
| 1934 | } |
| 1935 | |
| 1936 | /* counts how many resync-related answers we still expect from the peer |
| 1937 | * increase decrease |
| 1938 | * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY) |
| 1939 | * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER) |
| 1940 | * (or P_NEG_ACK with ID_SYNCER) |
| 1941 | */ |
| 1942 | static inline void inc_rs_pending(struct drbd_peer_device *peer_device) |
| 1943 | { |
| 1944 | atomic_inc(&peer_device->device->rs_pending_cnt); |
| 1945 | } |
| 1946 | |
| 1947 | #define dec_rs_pending(peer_device) \ |
| 1948 | ((void)expect((peer_device), __dec_rs_pending(peer_device) >= 0)) |
| 1949 | static inline int __dec_rs_pending(struct drbd_peer_device *peer_device) |
| 1950 | { |
| 1951 | return atomic_dec_return(&peer_device->device->rs_pending_cnt); |
| 1952 | } |
| 1953 | |
| 1954 | /* counts how many answers we still need to send to the peer. |
| 1955 | * increased on |
| 1956 | * receive_Data unless protocol A; |
| 1957 | * we need to send a P_RECV_ACK (proto B) |
| 1958 | * or P_WRITE_ACK (proto C) |
| 1959 | * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK |
| 1960 | * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA |
| 1961 | * receive_Barrier_* we need to send a P_BARRIER_ACK |
| 1962 | */ |
| 1963 | static inline void inc_unacked(struct drbd_device *device) |
| 1964 | { |
| 1965 | atomic_inc(&device->unacked_cnt); |
| 1966 | } |
| 1967 | |
| 1968 | #define dec_unacked(device) ((void)expect(device, __dec_unacked(device) >= 0)) |
| 1969 | static inline int __dec_unacked(struct drbd_device *device) |
| 1970 | { |
| 1971 | return atomic_dec_return(&device->unacked_cnt); |
| 1972 | } |
| 1973 | |
| 1974 | #define sub_unacked(device, n) ((void)expect(device, __sub_unacked(device) >= 0)) |
| 1975 | static inline int __sub_unacked(struct drbd_device *device, int n) |
| 1976 | { |
| 1977 | return atomic_sub_return(n, &device->unacked_cnt); |
| 1978 | } |
| 1979 | |
| 1980 | static inline bool is_sync_target_state(enum drbd_conns connection_state) |
| 1981 | { |
| 1982 | return connection_state == C_SYNC_TARGET || |
| 1983 | connection_state == C_PAUSED_SYNC_T; |
| 1984 | } |
| 1985 | |
| 1986 | static inline bool is_sync_source_state(enum drbd_conns connection_state) |
| 1987 | { |
| 1988 | return connection_state == C_SYNC_SOURCE || |
| 1989 | connection_state == C_PAUSED_SYNC_S; |
| 1990 | } |
| 1991 | |
| 1992 | static inline bool is_sync_state(enum drbd_conns connection_state) |
| 1993 | { |
| 1994 | return is_sync_source_state(connection_state) || |
| 1995 | is_sync_target_state(connection_state); |
| 1996 | } |
| 1997 | |
| 1998 | /** |
| 1999 | * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev |
| 2000 | * @_device: DRBD device. |
| 2001 | * @_min_state: Minimum device state required for success. |
| 2002 | * |
| 2003 | * You have to call put_ldev() when finished working with device->ldev. |
| 2004 | */ |
| 2005 | #define get_ldev_if_state(_device, _min_state) \ |
| 2006 | (_get_ldev_if_state((_device), (_min_state)) ? \ |
| 2007 | ({ __acquire(x); true; }) : false) |
| 2008 | #define get_ldev(_device) get_ldev_if_state(_device, D_INCONSISTENT) |
| 2009 | |
| 2010 | static inline void put_ldev(struct drbd_device *device) |
| 2011 | { |
| 2012 | enum drbd_disk_state disk_state = device->state.disk; |
| 2013 | /* We must check the state *before* the atomic_dec becomes visible, |
| 2014 | * or we have a theoretical race where someone hitting zero, |
| 2015 | * while state still D_FAILED, will then see D_DISKLESS in the |
| 2016 | * condition below and calling into destroy, where he must not, yet. */ |
| 2017 | int i = atomic_dec_return(&device->local_cnt); |
| 2018 | |
| 2019 | /* This may be called from some endio handler, |
| 2020 | * so we must not sleep here. */ |
| 2021 | |
| 2022 | __release(local); |
| 2023 | D_ASSERT(device, i >= 0); |
| 2024 | if (i == 0) { |
| 2025 | if (disk_state == D_DISKLESS) |
| 2026 | /* even internal references gone, safe to destroy */ |
| 2027 | drbd_device_post_work(device, DESTROY_DISK); |
| 2028 | if (disk_state == D_FAILED) |
| 2029 | /* all application IO references gone. */ |
| 2030 | if (!test_and_set_bit(GOING_DISKLESS, &device->flags)) |
| 2031 | drbd_device_post_work(device, GO_DISKLESS); |
| 2032 | wake_up(&device->misc_wait); |
| 2033 | } |
| 2034 | } |
| 2035 | |
| 2036 | #ifndef __CHECKER__ |
| 2037 | static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins) |
| 2038 | { |
| 2039 | int io_allowed; |
| 2040 | |
| 2041 | /* never get a reference while D_DISKLESS */ |
| 2042 | if (device->state.disk == D_DISKLESS) |
| 2043 | return 0; |
| 2044 | |
| 2045 | atomic_inc(&device->local_cnt); |
| 2046 | io_allowed = (device->state.disk >= mins); |
| 2047 | if (!io_allowed) |
| 2048 | put_ldev(device); |
| 2049 | return io_allowed; |
| 2050 | } |
| 2051 | #else |
| 2052 | extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins); |
| 2053 | #endif |
| 2054 | |
| 2055 | /* this throttles on-the-fly application requests |
| 2056 | * according to max_buffers settings; |
| 2057 | * maybe re-implement using semaphores? */ |
| 2058 | static inline int drbd_get_max_buffers(struct drbd_device *device) |
| 2059 | { |
| 2060 | struct net_conf *nc; |
| 2061 | int mxb; |
| 2062 | |
| 2063 | rcu_read_lock(); |
| 2064 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
| 2065 | mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */ |
| 2066 | rcu_read_unlock(); |
| 2067 | |
| 2068 | return mxb; |
| 2069 | } |
| 2070 | |
| 2071 | static inline int drbd_state_is_stable(struct drbd_device *device) |
| 2072 | { |
| 2073 | union drbd_dev_state s = device->state; |
| 2074 | |
| 2075 | /* DO NOT add a default clause, we want the compiler to warn us |
| 2076 | * for any newly introduced state we may have forgotten to add here */ |
| 2077 | |
| 2078 | switch ((enum drbd_conns)s.conn) { |
| 2079 | /* new io only accepted when there is no connection, ... */ |
| 2080 | case C_STANDALONE: |
| 2081 | case C_WF_CONNECTION: |
| 2082 | /* ... or there is a well established connection. */ |
| 2083 | case C_CONNECTED: |
| 2084 | case C_SYNC_SOURCE: |
| 2085 | case C_SYNC_TARGET: |
| 2086 | case C_VERIFY_S: |
| 2087 | case C_VERIFY_T: |
| 2088 | case C_PAUSED_SYNC_S: |
| 2089 | case C_PAUSED_SYNC_T: |
| 2090 | case C_AHEAD: |
| 2091 | case C_BEHIND: |
| 2092 | /* transitional states, IO allowed */ |
| 2093 | case C_DISCONNECTING: |
| 2094 | case C_UNCONNECTED: |
| 2095 | case C_TIMEOUT: |
| 2096 | case C_BROKEN_PIPE: |
| 2097 | case C_NETWORK_FAILURE: |
| 2098 | case C_PROTOCOL_ERROR: |
| 2099 | case C_TEAR_DOWN: |
| 2100 | case C_WF_REPORT_PARAMS: |
| 2101 | case C_STARTING_SYNC_S: |
| 2102 | case C_STARTING_SYNC_T: |
| 2103 | break; |
| 2104 | |
| 2105 | /* Allow IO in BM exchange states with new protocols */ |
| 2106 | case C_WF_BITMAP_S: |
| 2107 | if (first_peer_device(device)->connection->agreed_pro_version < 96) |
| 2108 | return 0; |
| 2109 | break; |
| 2110 | |
| 2111 | /* no new io accepted in these states */ |
| 2112 | case C_WF_BITMAP_T: |
| 2113 | case C_WF_SYNC_UUID: |
| 2114 | case C_MASK: |
| 2115 | /* not "stable" */ |
| 2116 | return 0; |
| 2117 | } |
| 2118 | |
| 2119 | switch ((enum drbd_disk_state)s.disk) { |
| 2120 | case D_DISKLESS: |
| 2121 | case D_INCONSISTENT: |
| 2122 | case D_OUTDATED: |
| 2123 | case D_CONSISTENT: |
| 2124 | case D_UP_TO_DATE: |
| 2125 | case D_FAILED: |
| 2126 | /* disk state is stable as well. */ |
| 2127 | break; |
| 2128 | |
| 2129 | /* no new io accepted during transitional states */ |
| 2130 | case D_ATTACHING: |
| 2131 | case D_NEGOTIATING: |
| 2132 | case D_UNKNOWN: |
| 2133 | case D_MASK: |
| 2134 | /* not "stable" */ |
| 2135 | return 0; |
| 2136 | } |
| 2137 | |
| 2138 | return 1; |
| 2139 | } |
| 2140 | |
| 2141 | static inline int drbd_suspended(struct drbd_device *device) |
| 2142 | { |
| 2143 | struct drbd_resource *resource = device->resource; |
| 2144 | |
| 2145 | return resource->susp || resource->susp_fen || resource->susp_nod; |
| 2146 | } |
| 2147 | |
| 2148 | static inline bool may_inc_ap_bio(struct drbd_device *device) |
| 2149 | { |
| 2150 | int mxb = drbd_get_max_buffers(device); |
| 2151 | |
| 2152 | if (drbd_suspended(device)) |
| 2153 | return false; |
| 2154 | if (atomic_read(&device->suspend_cnt)) |
| 2155 | return false; |
| 2156 | |
| 2157 | /* to avoid potential deadlock or bitmap corruption, |
| 2158 | * in various places, we only allow new application io |
| 2159 | * to start during "stable" states. */ |
| 2160 | |
| 2161 | /* no new io accepted when attaching or detaching the disk */ |
| 2162 | if (!drbd_state_is_stable(device)) |
| 2163 | return false; |
| 2164 | |
| 2165 | /* since some older kernels don't have atomic_add_unless, |
| 2166 | * and we are within the spinlock anyways, we have this workaround. */ |
| 2167 | if (atomic_read(&device->ap_bio_cnt) > mxb) |
| 2168 | return false; |
| 2169 | if (test_bit(BITMAP_IO, &device->flags)) |
| 2170 | return false; |
| 2171 | return true; |
| 2172 | } |
| 2173 | |
| 2174 | static inline bool inc_ap_bio_cond(struct drbd_device *device) |
| 2175 | { |
| 2176 | bool rv = false; |
| 2177 | |
| 2178 | spin_lock_irq(&device->resource->req_lock); |
| 2179 | rv = may_inc_ap_bio(device); |
| 2180 | if (rv) |
| 2181 | atomic_inc(&device->ap_bio_cnt); |
| 2182 | spin_unlock_irq(&device->resource->req_lock); |
| 2183 | |
| 2184 | return rv; |
| 2185 | } |
| 2186 | |
| 2187 | static inline void inc_ap_bio(struct drbd_device *device) |
| 2188 | { |
| 2189 | /* we wait here |
| 2190 | * as long as the device is suspended |
| 2191 | * until the bitmap is no longer on the fly during connection |
| 2192 | * handshake as long as we would exceed the max_buffer limit. |
| 2193 | * |
| 2194 | * to avoid races with the reconnect code, |
| 2195 | * we need to atomic_inc within the spinlock. */ |
| 2196 | |
| 2197 | wait_event(device->misc_wait, inc_ap_bio_cond(device)); |
| 2198 | } |
| 2199 | |
| 2200 | static inline void dec_ap_bio(struct drbd_device *device) |
| 2201 | { |
| 2202 | int mxb = drbd_get_max_buffers(device); |
| 2203 | int ap_bio = atomic_dec_return(&device->ap_bio_cnt); |
| 2204 | |
| 2205 | D_ASSERT(device, ap_bio >= 0); |
| 2206 | |
| 2207 | if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) { |
| 2208 | if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags)) |
| 2209 | drbd_queue_work(&first_peer_device(device)-> |
| 2210 | connection->sender_work, |
| 2211 | &device->bm_io_work.w); |
| 2212 | } |
| 2213 | |
| 2214 | /* this currently does wake_up for every dec_ap_bio! |
| 2215 | * maybe rather introduce some type of hysteresis? |
| 2216 | * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */ |
| 2217 | if (ap_bio < mxb) |
| 2218 | wake_up(&device->misc_wait); |
| 2219 | } |
| 2220 | |
| 2221 | static inline bool verify_can_do_stop_sector(struct drbd_device *device) |
| 2222 | { |
| 2223 | return first_peer_device(device)->connection->agreed_pro_version >= 97 && |
| 2224 | first_peer_device(device)->connection->agreed_pro_version != 100; |
| 2225 | } |
| 2226 | |
| 2227 | static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val) |
| 2228 | { |
| 2229 | int changed = device->ed_uuid != val; |
| 2230 | device->ed_uuid = val; |
| 2231 | return changed; |
| 2232 | } |
| 2233 | |
| 2234 | static inline int drbd_queue_order_type(struct drbd_device *device) |
| 2235 | { |
| 2236 | /* sorry, we currently have no working implementation |
| 2237 | * of distributed TCQ stuff */ |
| 2238 | #ifndef QUEUE_ORDERED_NONE |
| 2239 | #define QUEUE_ORDERED_NONE 0 |
| 2240 | #endif |
| 2241 | return QUEUE_ORDERED_NONE; |
| 2242 | } |
| 2243 | |
| 2244 | static inline struct drbd_connection *first_connection(struct drbd_resource *resource) |
| 2245 | { |
| 2246 | return list_first_entry_or_null(&resource->connections, |
| 2247 | struct drbd_connection, connections); |
| 2248 | } |
| 2249 | |
| 2250 | #endif |