1 /* 2 drbd_int.h 3 4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg. 5 6 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. 7 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. 8 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. 9 10 drbd is free software; you can redistribute it and/or modify 11 it under the terms of the GNU General Public License as published by 12 the Free Software Foundation; either version 2, or (at your option) 13 any later version. 14 15 drbd is distributed in the hope that it will be useful, 16 but WITHOUT ANY WARRANTY; without even the implied warranty of 17 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 18 GNU General Public License for more details. 19 20 You should have received a copy of the GNU General Public License 21 along with drbd; see the file COPYING. If not, write to 22 the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. 23 24 */ 25 26 #ifndef _DRBD_INT_H 27 #define _DRBD_INT_H 28 29 #include <linux/compiler.h> 30 #include <linux/types.h> 31 #include <linux/list.h> 32 #include <linux/sched.h> 33 #include <linux/bitops.h> 34 #include <linux/slab.h> 35 #include <linux/crypto.h> 36 #include <linux/ratelimit.h> 37 #include <linux/tcp.h> 38 #include <linux/mutex.h> 39 #include <linux/major.h> 40 #include <linux/blkdev.h> 41 #include <linux/genhd.h> 42 #include <net/tcp.h> 43 #include <linux/lru_cache.h> 44 #include <linux/prefetch.h> 45 46 #ifdef __CHECKER__ 47 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr"))) 48 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read"))) 49 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write"))) 50 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call"))) 51 #else 52 # define __protected_by(x) 53 # define __protected_read_by(x) 54 # define __protected_write_by(x) 55 # define __must_hold(x) 56 #endif 57 58 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0) 59 60 /* module parameter, defined in drbd_main.c */ 61 extern unsigned int minor_count; 62 extern bool disable_sendpage; 63 extern bool allow_oos; 64 extern unsigned int cn_idx; 65 66 #ifdef CONFIG_DRBD_FAULT_INJECTION 67 extern int enable_faults; 68 extern int fault_rate; 69 extern int fault_devs; 70 #endif 71 72 extern char usermode_helper[]; 73 74 75 /* I don't remember why XCPU ... 76 * This is used to wake the asender, 77 * and to interrupt sending the sending task 78 * on disconnect. 79 */ 80 #define DRBD_SIG SIGXCPU 81 82 /* This is used to stop/restart our threads. 83 * Cannot use SIGTERM nor SIGKILL, since these 84 * are sent out by init on runlevel changes 85 * I choose SIGHUP for now. 86 */ 87 #define DRBD_SIGKILL SIGHUP 88 89 /* All EEs on the free list should have ID_VACANT (== 0) 90 * freshly allocated EEs get !ID_VACANT (== 1) 91 * so if it says "cannot dereference null pointer at address 0x00000001", 92 * it is most likely one of these :( */ 93 94 #define ID_IN_SYNC (4711ULL) 95 #define ID_OUT_OF_SYNC (4712ULL) 96 97 #define ID_SYNCER (-1ULL) 98 #define ID_VACANT 0 99 #define is_syncer_block_id(id) ((id) == ID_SYNCER) 100 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL) 101 102 struct drbd_conf; 103 104 105 /* to shorten dev_warn(DEV, "msg"); and relatives statements */ 106 #define DEV (disk_to_dev(mdev->vdisk)) 107 108 #define D_ASSERT(exp) if (!(exp)) \ 109 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__) 110 111 #define ERR_IF(exp) if (({ \ 112 int _b = (exp) != 0; \ 113 if (_b) dev_err(DEV, "ASSERT FAILED: %s: (%s) in %s:%d\n", \ 114 __func__, #exp, __FILE__, __LINE__); \ 115 _b; \ 116 })) 117 118 /* Defines to control fault insertion */ 119 enum { 120 DRBD_FAULT_MD_WR = 0, /* meta data write */ 121 DRBD_FAULT_MD_RD = 1, /* read */ 122 DRBD_FAULT_RS_WR = 2, /* resync */ 123 DRBD_FAULT_RS_RD = 3, 124 DRBD_FAULT_DT_WR = 4, /* data */ 125 DRBD_FAULT_DT_RD = 5, 126 DRBD_FAULT_DT_RA = 6, /* data read ahead */ 127 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */ 128 DRBD_FAULT_AL_EE = 8, /* alloc ee */ 129 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */ 130 131 DRBD_FAULT_MAX, 132 }; 133 134 extern unsigned int 135 _drbd_insert_fault(struct drbd_conf *mdev, unsigned int type); 136 137 static inline int 138 drbd_insert_fault(struct drbd_conf *mdev, unsigned int type) { 139 #ifdef CONFIG_DRBD_FAULT_INJECTION 140 return fault_rate && 141 (enable_faults & (1<<type)) && 142 _drbd_insert_fault(mdev, type); 143 #else 144 return 0; 145 #endif 146 } 147 148 /* integer division, round _UP_ to the next integer */ 149 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0)) 150 /* usual integer division */ 151 #define div_floor(A, B) ((A)/(B)) 152 153 /* drbd_meta-data.c (still in drbd_main.c) */ 154 /* 4th incarnation of the disk layout. */ 155 #define DRBD_MD_MAGIC (DRBD_MAGIC+4) 156 157 extern struct drbd_conf **minor_table; 158 extern struct ratelimit_state drbd_ratelimit_state; 159 160 /* on the wire */ 161 enum drbd_packets { 162 /* receiver (data socket) */ 163 P_DATA = 0x00, 164 P_DATA_REPLY = 0x01, /* Response to P_DATA_REQUEST */ 165 P_RS_DATA_REPLY = 0x02, /* Response to P_RS_DATA_REQUEST */ 166 P_BARRIER = 0x03, 167 P_BITMAP = 0x04, 168 P_BECOME_SYNC_TARGET = 0x05, 169 P_BECOME_SYNC_SOURCE = 0x06, 170 P_UNPLUG_REMOTE = 0x07, /* Used at various times to hint the peer */ 171 P_DATA_REQUEST = 0x08, /* Used to ask for a data block */ 172 P_RS_DATA_REQUEST = 0x09, /* Used to ask for a data block for resync */ 173 P_SYNC_PARAM = 0x0a, 174 P_PROTOCOL = 0x0b, 175 P_UUIDS = 0x0c, 176 P_SIZES = 0x0d, 177 P_STATE = 0x0e, 178 P_SYNC_UUID = 0x0f, 179 P_AUTH_CHALLENGE = 0x10, 180 P_AUTH_RESPONSE = 0x11, 181 P_STATE_CHG_REQ = 0x12, 182 183 /* asender (meta socket */ 184 P_PING = 0x13, 185 P_PING_ACK = 0x14, 186 P_RECV_ACK = 0x15, /* Used in protocol B */ 187 P_WRITE_ACK = 0x16, /* Used in protocol C */ 188 P_RS_WRITE_ACK = 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */ 189 P_DISCARD_ACK = 0x18, /* Used in proto C, two-primaries conflict detection */ 190 P_NEG_ACK = 0x19, /* Sent if local disk is unusable */ 191 P_NEG_DREPLY = 0x1a, /* Local disk is broken... */ 192 P_NEG_RS_DREPLY = 0x1b, /* Local disk is broken... */ 193 P_BARRIER_ACK = 0x1c, 194 P_STATE_CHG_REPLY = 0x1d, 195 196 /* "new" commands, no longer fitting into the ordering scheme above */ 197 198 P_OV_REQUEST = 0x1e, /* data socket */ 199 P_OV_REPLY = 0x1f, 200 P_OV_RESULT = 0x20, /* meta socket */ 201 P_CSUM_RS_REQUEST = 0x21, /* data socket */ 202 P_RS_IS_IN_SYNC = 0x22, /* meta socket */ 203 P_SYNC_PARAM89 = 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */ 204 P_COMPRESSED_BITMAP = 0x24, /* compressed or otherwise encoded bitmap transfer */ 205 /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */ 206 /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */ 207 P_DELAY_PROBE = 0x27, /* is used on BOTH sockets */ 208 P_OUT_OF_SYNC = 0x28, /* Mark as out of sync (Outrunning), data socket */ 209 P_RS_CANCEL = 0x29, /* meta: Used to cancel RS_DATA_REQUEST packet by SyncSource */ 210 211 P_MAX_CMD = 0x2A, 212 P_MAY_IGNORE = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */ 213 P_MAX_OPT_CMD = 0x101, 214 215 /* special command ids for handshake */ 216 217 P_HAND_SHAKE_M = 0xfff1, /* First Packet on the MetaSock */ 218 P_HAND_SHAKE_S = 0xfff2, /* First Packet on the Socket */ 219 220 P_HAND_SHAKE = 0xfffe /* FIXED for the next century! */ 221 }; 222 223 static inline const char *cmdname(enum drbd_packets cmd) 224 { 225 /* THINK may need to become several global tables 226 * when we want to support more than 227 * one PRO_VERSION */ 228 static const char *cmdnames[] = { 229 [P_DATA] = "Data", 230 [P_DATA_REPLY] = "DataReply", 231 [P_RS_DATA_REPLY] = "RSDataReply", 232 [P_BARRIER] = "Barrier", 233 [P_BITMAP] = "ReportBitMap", 234 [P_BECOME_SYNC_TARGET] = "BecomeSyncTarget", 235 [P_BECOME_SYNC_SOURCE] = "BecomeSyncSource", 236 [P_UNPLUG_REMOTE] = "UnplugRemote", 237 [P_DATA_REQUEST] = "DataRequest", 238 [P_RS_DATA_REQUEST] = "RSDataRequest", 239 [P_SYNC_PARAM] = "SyncParam", 240 [P_SYNC_PARAM89] = "SyncParam89", 241 [P_PROTOCOL] = "ReportProtocol", 242 [P_UUIDS] = "ReportUUIDs", 243 [P_SIZES] = "ReportSizes", 244 [P_STATE] = "ReportState", 245 [P_SYNC_UUID] = "ReportSyncUUID", 246 [P_AUTH_CHALLENGE] = "AuthChallenge", 247 [P_AUTH_RESPONSE] = "AuthResponse", 248 [P_PING] = "Ping", 249 [P_PING_ACK] = "PingAck", 250 [P_RECV_ACK] = "RecvAck", 251 [P_WRITE_ACK] = "WriteAck", 252 [P_RS_WRITE_ACK] = "RSWriteAck", 253 [P_DISCARD_ACK] = "DiscardAck", 254 [P_NEG_ACK] = "NegAck", 255 [P_NEG_DREPLY] = "NegDReply", 256 [P_NEG_RS_DREPLY] = "NegRSDReply", 257 [P_BARRIER_ACK] = "BarrierAck", 258 [P_STATE_CHG_REQ] = "StateChgRequest", 259 [P_STATE_CHG_REPLY] = "StateChgReply", 260 [P_OV_REQUEST] = "OVRequest", 261 [P_OV_REPLY] = "OVReply", 262 [P_OV_RESULT] = "OVResult", 263 [P_CSUM_RS_REQUEST] = "CsumRSRequest", 264 [P_RS_IS_IN_SYNC] = "CsumRSIsInSync", 265 [P_COMPRESSED_BITMAP] = "CBitmap", 266 [P_DELAY_PROBE] = "DelayProbe", 267 [P_OUT_OF_SYNC] = "OutOfSync", 268 [P_MAX_CMD] = NULL, 269 }; 270 271 if (cmd == P_HAND_SHAKE_M) 272 return "HandShakeM"; 273 if (cmd == P_HAND_SHAKE_S) 274 return "HandShakeS"; 275 if (cmd == P_HAND_SHAKE) 276 return "HandShake"; 277 if (cmd >= P_MAX_CMD) 278 return "Unknown"; 279 return cmdnames[cmd]; 280 } 281 282 /* for sending/receiving the bitmap, 283 * possibly in some encoding scheme */ 284 struct bm_xfer_ctx { 285 /* "const" 286 * stores total bits and long words 287 * of the bitmap, so we don't need to 288 * call the accessor functions over and again. */ 289 unsigned long bm_bits; 290 unsigned long bm_words; 291 /* during xfer, current position within the bitmap */ 292 unsigned long bit_offset; 293 unsigned long word_offset; 294 295 /* statistics; index: (h->command == P_BITMAP) */ 296 unsigned packets[2]; 297 unsigned bytes[2]; 298 }; 299 300 extern void INFO_bm_xfer_stats(struct drbd_conf *mdev, 301 const char *direction, struct bm_xfer_ctx *c); 302 303 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c) 304 { 305 /* word_offset counts "native long words" (32 or 64 bit), 306 * aligned at 64 bit. 307 * Encoded packet may end at an unaligned bit offset. 308 * In case a fallback clear text packet is transmitted in 309 * between, we adjust this offset back to the last 64bit 310 * aligned "native long word", which makes coding and decoding 311 * the plain text bitmap much more convenient. */ 312 #if BITS_PER_LONG == 64 313 c->word_offset = c->bit_offset >> 6; 314 #elif BITS_PER_LONG == 32 315 c->word_offset = c->bit_offset >> 5; 316 c->word_offset &= ~(1UL); 317 #else 318 # error "unsupported BITS_PER_LONG" 319 #endif 320 } 321 322 #ifndef __packed 323 #define __packed __attribute__((packed)) 324 #endif 325 326 /* This is the layout for a packet on the wire. 327 * The byteorder is the network byte order. 328 * (except block_id and barrier fields. 329 * these are pointers to local structs 330 * and have no relevance for the partner, 331 * which just echoes them as received.) 332 * 333 * NOTE that the payload starts at a long aligned offset, 334 * regardless of 32 or 64 bit arch! 335 */ 336 struct p_header80 { 337 u32 magic; 338 u16 command; 339 u16 length; /* bytes of data after this header */ 340 u8 payload[0]; 341 } __packed; 342 343 /* Header for big packets, Used for data packets exceeding 64kB */ 344 struct p_header95 { 345 u16 magic; /* use DRBD_MAGIC_BIG here */ 346 u16 command; 347 u32 length; /* Use only 24 bits of that. Ignore the highest 8 bit. */ 348 u8 payload[0]; 349 } __packed; 350 351 union p_header { 352 struct p_header80 h80; 353 struct p_header95 h95; 354 }; 355 356 /* 357 * short commands, packets without payload, plain p_header: 358 * P_PING 359 * P_PING_ACK 360 * P_BECOME_SYNC_TARGET 361 * P_BECOME_SYNC_SOURCE 362 * P_UNPLUG_REMOTE 363 */ 364 365 /* 366 * commands with out-of-struct payload: 367 * P_BITMAP (no additional fields) 368 * P_DATA, P_DATA_REPLY (see p_data) 369 * P_COMPRESSED_BITMAP (see receive_compressed_bitmap) 370 */ 371 372 /* these defines must not be changed without changing the protocol version */ 373 #define DP_HARDBARRIER 1 /* depricated */ 374 #define DP_RW_SYNC 2 /* equals REQ_SYNC */ 375 #define DP_MAY_SET_IN_SYNC 4 376 #define DP_UNPLUG 8 /* not used anymore */ 377 #define DP_FUA 16 /* equals REQ_FUA */ 378 #define DP_FLUSH 32 /* equals REQ_FLUSH */ 379 #define DP_DISCARD 64 /* equals REQ_DISCARD */ 380 381 struct p_data { 382 union p_header head; 383 u64 sector; /* 64 bits sector number */ 384 u64 block_id; /* to identify the request in protocol B&C */ 385 u32 seq_num; 386 u32 dp_flags; 387 } __packed; 388 389 /* 390 * commands which share a struct: 391 * p_block_ack: 392 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C), 393 * P_DISCARD_ACK (proto C, two-primaries conflict detection) 394 * p_block_req: 395 * P_DATA_REQUEST, P_RS_DATA_REQUEST 396 */ 397 struct p_block_ack { 398 struct p_header80 head; 399 u64 sector; 400 u64 block_id; 401 u32 blksize; 402 u32 seq_num; 403 } __packed; 404 405 406 struct p_block_req { 407 struct p_header80 head; 408 u64 sector; 409 u64 block_id; 410 u32 blksize; 411 u32 pad; /* to multiple of 8 Byte */ 412 } __packed; 413 414 /* 415 * commands with their own struct for additional fields: 416 * P_HAND_SHAKE 417 * P_BARRIER 418 * P_BARRIER_ACK 419 * P_SYNC_PARAM 420 * ReportParams 421 */ 422 423 struct p_handshake { 424 struct p_header80 head; /* 8 bytes */ 425 u32 protocol_min; 426 u32 feature_flags; 427 u32 protocol_max; 428 429 /* should be more than enough for future enhancements 430 * for now, feature_flags and the reserverd array shall be zero. 431 */ 432 433 u32 _pad; 434 u64 reserverd[7]; 435 } __packed; 436 /* 80 bytes, FIXED for the next century */ 437 438 struct p_barrier { 439 struct p_header80 head; 440 u32 barrier; /* barrier number _handle_ only */ 441 u32 pad; /* to multiple of 8 Byte */ 442 } __packed; 443 444 struct p_barrier_ack { 445 struct p_header80 head; 446 u32 barrier; 447 u32 set_size; 448 } __packed; 449 450 struct p_rs_param { 451 struct p_header80 head; 452 u32 rate; 453 454 /* Since protocol version 88 and higher. */ 455 char verify_alg[0]; 456 } __packed; 457 458 struct p_rs_param_89 { 459 struct p_header80 head; 460 u32 rate; 461 /* protocol version 89: */ 462 char verify_alg[SHARED_SECRET_MAX]; 463 char csums_alg[SHARED_SECRET_MAX]; 464 } __packed; 465 466 struct p_rs_param_95 { 467 struct p_header80 head; 468 u32 rate; 469 char verify_alg[SHARED_SECRET_MAX]; 470 char csums_alg[SHARED_SECRET_MAX]; 471 u32 c_plan_ahead; 472 u32 c_delay_target; 473 u32 c_fill_target; 474 u32 c_max_rate; 475 } __packed; 476 477 enum drbd_conn_flags { 478 CF_WANT_LOSE = 1, 479 CF_DRY_RUN = 2, 480 }; 481 482 struct p_protocol { 483 struct p_header80 head; 484 u32 protocol; 485 u32 after_sb_0p; 486 u32 after_sb_1p; 487 u32 after_sb_2p; 488 u32 conn_flags; 489 u32 two_primaries; 490 491 /* Since protocol version 87 and higher. */ 492 char integrity_alg[0]; 493 494 } __packed; 495 496 struct p_uuids { 497 struct p_header80 head; 498 u64 uuid[UI_EXTENDED_SIZE]; 499 } __packed; 500 501 struct p_rs_uuid { 502 struct p_header80 head; 503 u64 uuid; 504 } __packed; 505 506 struct p_sizes { 507 struct p_header80 head; 508 u64 d_size; /* size of disk */ 509 u64 u_size; /* user requested size */ 510 u64 c_size; /* current exported size */ 511 u32 max_bio_size; /* Maximal size of a BIO */ 512 u16 queue_order_type; /* not yet implemented in DRBD*/ 513 u16 dds_flags; /* use enum dds_flags here. */ 514 } __packed; 515 516 struct p_state { 517 struct p_header80 head; 518 u32 state; 519 } __packed; 520 521 struct p_req_state { 522 struct p_header80 head; 523 u32 mask; 524 u32 val; 525 } __packed; 526 527 struct p_req_state_reply { 528 struct p_header80 head; 529 u32 retcode; 530 } __packed; 531 532 struct p_drbd06_param { 533 u64 size; 534 u32 state; 535 u32 blksize; 536 u32 protocol; 537 u32 version; 538 u32 gen_cnt[5]; 539 u32 bit_map_gen[5]; 540 } __packed; 541 542 struct p_discard { 543 struct p_header80 head; 544 u64 block_id; 545 u32 seq_num; 546 u32 pad; 547 } __packed; 548 549 struct p_block_desc { 550 struct p_header80 head; 551 u64 sector; 552 u32 blksize; 553 u32 pad; /* to multiple of 8 Byte */ 554 } __packed; 555 556 /* Valid values for the encoding field. 557 * Bump proto version when changing this. */ 558 enum drbd_bitmap_code { 559 /* RLE_VLI_Bytes = 0, 560 * and other bit variants had been defined during 561 * algorithm evaluation. */ 562 RLE_VLI_Bits = 2, 563 }; 564 565 struct p_compressed_bm { 566 struct p_header80 head; 567 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code 568 * (encoding & 0x80): polarity (set/unset) of first runlength 569 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits 570 * used to pad up to head.length bytes 571 */ 572 u8 encoding; 573 574 u8 code[0]; 575 } __packed; 576 577 struct p_delay_probe93 { 578 struct p_header80 head; 579 u32 seq_num; /* sequence number to match the two probe packets */ 580 u32 offset; /* usecs the probe got sent after the reference time point */ 581 } __packed; 582 583 /* DCBP: Drbd Compressed Bitmap Packet ... */ 584 static inline enum drbd_bitmap_code 585 DCBP_get_code(struct p_compressed_bm *p) 586 { 587 return (enum drbd_bitmap_code)(p->encoding & 0x0f); 588 } 589 590 static inline void 591 DCBP_set_code(struct p_compressed_bm *p, enum drbd_bitmap_code code) 592 { 593 BUG_ON(code & ~0xf); 594 p->encoding = (p->encoding & ~0xf) | code; 595 } 596 597 static inline int 598 DCBP_get_start(struct p_compressed_bm *p) 599 { 600 return (p->encoding & 0x80) != 0; 601 } 602 603 static inline void 604 DCBP_set_start(struct p_compressed_bm *p, int set) 605 { 606 p->encoding = (p->encoding & ~0x80) | (set ? 0x80 : 0); 607 } 608 609 static inline int 610 DCBP_get_pad_bits(struct p_compressed_bm *p) 611 { 612 return (p->encoding >> 4) & 0x7; 613 } 614 615 static inline void 616 DCBP_set_pad_bits(struct p_compressed_bm *p, int n) 617 { 618 BUG_ON(n & ~0x7); 619 p->encoding = (p->encoding & (~0x7 << 4)) | (n << 4); 620 } 621 622 /* one bitmap packet, including the p_header, 623 * should fit within one _architecture independend_ page. 624 * so we need to use the fixed size 4KiB page size 625 * most architectures have used for a long time. 626 */ 627 #define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header80)) 628 #define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long)) 629 #define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm)) 630 #if (PAGE_SIZE < 4096) 631 /* drbd_send_bitmap / receive_bitmap would break horribly */ 632 #error "PAGE_SIZE too small" 633 #endif 634 635 union p_polymorph { 636 union p_header header; 637 struct p_handshake handshake; 638 struct p_data data; 639 struct p_block_ack block_ack; 640 struct p_barrier barrier; 641 struct p_barrier_ack barrier_ack; 642 struct p_rs_param_89 rs_param_89; 643 struct p_rs_param_95 rs_param_95; 644 struct p_protocol protocol; 645 struct p_sizes sizes; 646 struct p_uuids uuids; 647 struct p_state state; 648 struct p_req_state req_state; 649 struct p_req_state_reply req_state_reply; 650 struct p_block_req block_req; 651 struct p_delay_probe93 delay_probe93; 652 struct p_rs_uuid rs_uuid; 653 struct p_block_desc block_desc; 654 } __packed; 655 656 /**********************************************************************/ 657 enum drbd_thread_state { 658 None, 659 Running, 660 Exiting, 661 Restarting 662 }; 663 664 struct drbd_thread { 665 spinlock_t t_lock; 666 struct task_struct *task; 667 struct completion stop; 668 enum drbd_thread_state t_state; 669 int (*function) (struct drbd_thread *); 670 struct drbd_conf *mdev; 671 int reset_cpu_mask; 672 }; 673 674 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi) 675 { 676 /* THINK testing the t_state seems to be uncritical in all cases 677 * (but thread_{start,stop}), so we can read it *without* the lock. 678 * --lge */ 679 680 smp_rmb(); 681 return thi->t_state; 682 } 683 684 struct drbd_work; 685 typedef int (*drbd_work_cb)(struct drbd_conf *, struct drbd_work *, int cancel); 686 struct drbd_work { 687 struct list_head list; 688 drbd_work_cb cb; 689 }; 690 691 struct drbd_tl_epoch; 692 struct drbd_request { 693 struct drbd_work w; 694 struct drbd_conf *mdev; 695 696 /* if local IO is not allowed, will be NULL. 697 * if local IO _is_ allowed, holds the locally submitted bio clone, 698 * or, after local IO completion, the ERR_PTR(error). 699 * see drbd_endio_pri(). */ 700 struct bio *private_bio; 701 702 struct hlist_node collision; 703 sector_t sector; 704 unsigned int size; 705 unsigned int epoch; /* barrier_nr */ 706 707 /* barrier_nr: used to check on "completion" whether this req was in 708 * the current epoch, and we therefore have to close it, 709 * starting a new epoch... 710 */ 711 712 struct list_head tl_requests; /* ring list in the transfer log */ 713 struct bio *master_bio; /* master bio pointer */ 714 unsigned long rq_state; /* see comments above _req_mod() */ 715 unsigned long start_time; 716 }; 717 718 struct drbd_tl_epoch { 719 struct drbd_work w; 720 struct list_head requests; /* requests before */ 721 struct drbd_tl_epoch *next; /* pointer to the next barrier */ 722 unsigned int br_number; /* the barriers identifier. */ 723 int n_writes; /* number of requests attached before this barrier */ 724 }; 725 726 struct drbd_request; 727 728 /* These Tl_epoch_entries may be in one of 6 lists: 729 active_ee .. data packet being written 730 sync_ee .. syncer block being written 731 done_ee .. block written, need to send P_WRITE_ACK 732 read_ee .. [RS]P_DATA_REQUEST being read 733 */ 734 735 struct drbd_epoch { 736 struct list_head list; 737 unsigned int barrier_nr; 738 atomic_t epoch_size; /* increased on every request added. */ 739 atomic_t active; /* increased on every req. added, and dec on every finished. */ 740 unsigned long flags; 741 }; 742 743 /* drbd_epoch flag bits */ 744 enum { 745 DE_HAVE_BARRIER_NUMBER, 746 }; 747 748 enum epoch_event { 749 EV_PUT, 750 EV_GOT_BARRIER_NR, 751 EV_BECAME_LAST, 752 EV_CLEANUP = 32, /* used as flag */ 753 }; 754 755 struct drbd_wq_barrier { 756 struct drbd_work w; 757 struct completion done; 758 }; 759 760 struct digest_info { 761 int digest_size; 762 void *digest; 763 }; 764 765 struct drbd_epoch_entry { 766 struct drbd_work w; 767 struct hlist_node collision; 768 struct drbd_epoch *epoch; /* for writes */ 769 struct drbd_conf *mdev; 770 struct page *pages; 771 atomic_t pending_bios; 772 unsigned int size; 773 /* see comments on ee flag bits below */ 774 unsigned long flags; 775 sector_t sector; 776 union { 777 u64 block_id; 778 struct digest_info *digest; 779 }; 780 }; 781 782 /* ee flag bits. 783 * While corresponding bios are in flight, the only modification will be 784 * set_bit WAS_ERROR, which has to be atomic. 785 * If no bios are in flight yet, or all have been completed, 786 * non-atomic modification to ee->flags is ok. 787 */ 788 enum { 789 __EE_CALL_AL_COMPLETE_IO, 790 __EE_MAY_SET_IN_SYNC, 791 792 /* In case a barrier failed, 793 * we need to resubmit without the barrier flag. */ 794 __EE_RESUBMITTED, 795 796 /* we may have several bios per epoch entry. 797 * if any of those fail, we set this flag atomically 798 * from the endio callback */ 799 __EE_WAS_ERROR, 800 801 /* This ee has a pointer to a digest instead of a block id */ 802 __EE_HAS_DIGEST, 803 }; 804 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO) 805 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC) 806 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED) 807 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR) 808 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST) 809 810 /* global flag bits */ 811 enum { 812 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */ 813 SIGNAL_ASENDER, /* whether asender wants to be interrupted */ 814 SEND_PING, /* whether asender should send a ping asap */ 815 816 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */ 817 MD_DIRTY, /* current uuids and flags not yet on disk */ 818 DISCARD_CONCURRENT, /* Set on one node, cleared on the peer! */ 819 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */ 820 CLUSTER_ST_CHANGE, /* Cluster wide state change going on... */ 821 CL_ST_CHG_SUCCESS, 822 CL_ST_CHG_FAIL, 823 CRASHED_PRIMARY, /* This node was a crashed primary. 824 * Gets cleared when the state.conn 825 * goes into C_CONNECTED state. */ 826 CONSIDER_RESYNC, 827 828 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */ 829 SUSPEND_IO, /* suspend application io */ 830 BITMAP_IO, /* suspend application io; 831 once no more io in flight, start bitmap io */ 832 BITMAP_IO_QUEUED, /* Started bitmap IO */ 833 GO_DISKLESS, /* Disk is being detached, on io-error or admin request. */ 834 WAS_IO_ERROR, /* Local disk failed returned IO error */ 835 FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */ 836 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */ 837 NET_CONGESTED, /* The data socket is congested */ 838 839 CONFIG_PENDING, /* serialization of (re)configuration requests. 840 * if set, also prevents the device from dying */ 841 DEVICE_DYING, /* device became unconfigured, 842 * but worker thread is still handling the cleanup. 843 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed, 844 * while this is set. */ 845 RESIZE_PENDING, /* Size change detected locally, waiting for the response from 846 * the peer, if it changed there as well. */ 847 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */ 848 GOT_PING_ACK, /* set when we receive a ping_ack packet, misc wait gets woken */ 849 NEW_CUR_UUID, /* Create new current UUID when thawing IO */ 850 AL_SUSPENDED, /* Activity logging is currently suspended. */ 851 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */ 852 STATE_SENT, /* Do not change state/UUIDs while this is set */ 853 854 CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC) 855 * pending, from drbd worker context. 856 * If set, bdi_write_congested() returns true, 857 * so shrink_page_list() would not recurse into, 858 * and potentially deadlock on, this drbd worker. 859 */ 860 }; 861 862 struct drbd_bitmap; /* opaque for drbd_conf */ 863 864 /* definition of bits in bm_flags to be used in drbd_bm_lock 865 * and drbd_bitmap_io and friends. */ 866 enum bm_flag { 867 /* do we need to kfree, or vfree bm_pages? */ 868 BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */ 869 870 /* currently locked for bulk operation */ 871 BM_LOCKED_MASK = 0xf, 872 873 /* in detail, that is: */ 874 BM_DONT_CLEAR = 0x1, 875 BM_DONT_SET = 0x2, 876 BM_DONT_TEST = 0x4, 877 878 /* so we can mark it locked for bulk operation, 879 * and still allow all non-bulk operations */ 880 BM_IS_LOCKED = 0x8, 881 882 /* (test bit, count bit) allowed (common case) */ 883 BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED, 884 885 /* testing bits, as well as setting new bits allowed, but clearing bits 886 * would be unexpected. Used during bitmap receive. Setting new bits 887 * requires sending of "out-of-sync" information, though. */ 888 BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED, 889 890 /* for drbd_bm_write_copy_pages, everything is allowed, 891 * only concurrent bulk operations are locked out. */ 892 BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED, 893 }; 894 895 struct drbd_work_queue { 896 struct list_head q; 897 struct semaphore s; /* producers up it, worker down()s it */ 898 spinlock_t q_lock; /* to protect the list. */ 899 }; 900 901 struct drbd_socket { 902 struct drbd_work_queue work; 903 struct mutex mutex; 904 struct socket *socket; 905 /* this way we get our 906 * send/receive buffers off the stack */ 907 union p_polymorph sbuf; 908 union p_polymorph rbuf; 909 }; 910 911 struct drbd_md { 912 u64 md_offset; /* sector offset to 'super' block */ 913 914 u64 la_size_sect; /* last agreed size, unit sectors */ 915 u64 uuid[UI_SIZE]; 916 u64 device_uuid; 917 u32 flags; 918 u32 md_size_sect; 919 920 s32 al_offset; /* signed relative sector offset to al area */ 921 s32 bm_offset; /* signed relative sector offset to bitmap */ 922 923 /* u32 al_nr_extents; important for restoring the AL 924 * is stored into sync_conf.al_extents, which in turn 925 * gets applied to act_log->nr_elements 926 */ 927 }; 928 929 /* for sync_conf and other types... */ 930 #define NL_PACKET(name, number, fields) struct name { fields }; 931 #define NL_INTEGER(pn,pr,member) int member; 932 #define NL_INT64(pn,pr,member) __u64 member; 933 #define NL_BIT(pn,pr,member) unsigned member:1; 934 #define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len; 935 #include <linux/drbd_nl.h> 936 937 struct drbd_backing_dev { 938 struct block_device *backing_bdev; 939 struct block_device *md_bdev; 940 struct drbd_md md; 941 struct disk_conf dc; /* The user provided config... */ 942 sector_t known_size; /* last known size of that backing device */ 943 }; 944 945 struct drbd_md_io { 946 unsigned int done; 947 int error; 948 }; 949 950 struct bm_io_work { 951 struct drbd_work w; 952 char *why; 953 enum bm_flag flags; 954 int (*io_fn)(struct drbd_conf *mdev); 955 void (*done)(struct drbd_conf *mdev, int rv); 956 }; 957 958 enum write_ordering_e { 959 WO_none, 960 WO_drain_io, 961 WO_bdev_flush, 962 }; 963 964 struct fifo_buffer { 965 int *values; 966 unsigned int head_index; 967 unsigned int size; 968 }; 969 970 struct drbd_conf { 971 /* things that are stored as / read from meta data on disk */ 972 unsigned long flags; 973 974 /* configured by drbdsetup */ 975 struct net_conf *net_conf; /* protected by get_net_conf() and put_net_conf() */ 976 struct syncer_conf sync_conf; 977 struct drbd_backing_dev *ldev __protected_by(local); 978 979 sector_t p_size; /* partner's disk size */ 980 struct request_queue *rq_queue; 981 struct block_device *this_bdev; 982 struct gendisk *vdisk; 983 984 struct drbd_socket data; /* data/barrier/cstate/parameter packets */ 985 struct drbd_socket meta; /* ping/ack (metadata) packets */ 986 int agreed_pro_version; /* actually used protocol version */ 987 unsigned long last_received; /* in jiffies, either socket */ 988 unsigned int ko_count; 989 struct drbd_work resync_work, 990 unplug_work, 991 go_diskless, 992 md_sync_work, 993 start_resync_work; 994 struct timer_list resync_timer; 995 struct timer_list md_sync_timer; 996 struct timer_list start_resync_timer; 997 struct timer_list request_timer; 998 #ifdef DRBD_DEBUG_MD_SYNC 999 struct { 1000 unsigned int line; 1001 const char* func; 1002 } last_md_mark_dirty; 1003 #endif 1004 1005 /* Used after attach while negotiating new disk state. */ 1006 union drbd_state new_state_tmp; 1007 1008 union drbd_state state; 1009 wait_queue_head_t misc_wait; 1010 wait_queue_head_t state_wait; /* upon each state change. */ 1011 wait_queue_head_t net_cnt_wait; 1012 unsigned int send_cnt; 1013 unsigned int recv_cnt; 1014 unsigned int read_cnt; 1015 unsigned int writ_cnt; 1016 unsigned int al_writ_cnt; 1017 unsigned int bm_writ_cnt; 1018 atomic_t ap_bio_cnt; /* Requests we need to complete */ 1019 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */ 1020 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */ 1021 atomic_t unacked_cnt; /* Need to send replys for */ 1022 atomic_t local_cnt; /* Waiting for local completion */ 1023 atomic_t net_cnt; /* Users of net_conf */ 1024 spinlock_t req_lock; 1025 struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */ 1026 struct drbd_tl_epoch *newest_tle; 1027 struct drbd_tl_epoch *oldest_tle; 1028 struct list_head out_of_sequence_requests; 1029 struct list_head barrier_acked_requests; 1030 struct hlist_head *tl_hash; 1031 unsigned int tl_hash_s; 1032 1033 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */ 1034 unsigned long rs_total; 1035 /* number of resync blocks that failed in this run */ 1036 unsigned long rs_failed; 1037 /* Syncer's start time [unit jiffies] */ 1038 unsigned long rs_start; 1039 /* cumulated time in PausedSyncX state [unit jiffies] */ 1040 unsigned long rs_paused; 1041 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */ 1042 unsigned long rs_same_csum; 1043 #define DRBD_SYNC_MARKS 8 1044 #define DRBD_SYNC_MARK_STEP (3*HZ) 1045 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */ 1046 unsigned long rs_mark_left[DRBD_SYNC_MARKS]; 1047 /* marks's time [unit jiffies] */ 1048 unsigned long rs_mark_time[DRBD_SYNC_MARKS]; 1049 /* current index into rs_mark_{left,time} */ 1050 int rs_last_mark; 1051 1052 /* where does the admin want us to start? (sector) */ 1053 sector_t ov_start_sector; 1054 /* where are we now? (sector) */ 1055 sector_t ov_position; 1056 /* Start sector of out of sync range (to merge printk reporting). */ 1057 sector_t ov_last_oos_start; 1058 /* size of out-of-sync range in sectors. */ 1059 sector_t ov_last_oos_size; 1060 unsigned long ov_left; /* in bits */ 1061 struct crypto_hash *csums_tfm; 1062 struct crypto_hash *verify_tfm; 1063 1064 unsigned long last_reattach_jif; 1065 unsigned long last_reconnect_jif; 1066 struct drbd_thread receiver; 1067 struct drbd_thread worker; 1068 struct drbd_thread asender; 1069 struct drbd_bitmap *bitmap; 1070 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */ 1071 1072 /* Used to track operations of resync... */ 1073 struct lru_cache *resync; 1074 /* Number of locked elements in resync LRU */ 1075 unsigned int resync_locked; 1076 /* resync extent number waiting for application requests */ 1077 unsigned int resync_wenr; 1078 1079 int open_cnt; 1080 u64 *p_uuid; 1081 struct drbd_epoch *current_epoch; 1082 spinlock_t epoch_lock; 1083 unsigned int epochs; 1084 enum write_ordering_e write_ordering; 1085 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */ 1086 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */ 1087 struct list_head done_ee; /* send ack */ 1088 struct list_head read_ee; /* IO in progress (any read) */ 1089 struct list_head net_ee; /* zero-copy network send in progress */ 1090 struct hlist_head *ee_hash; /* is proteced by req_lock! */ 1091 unsigned int ee_hash_s; 1092 1093 /* this one is protected by ee_lock, single thread */ 1094 struct drbd_epoch_entry *last_write_w_barrier; 1095 1096 int next_barrier_nr; 1097 struct hlist_head *app_reads_hash; /* is proteced by req_lock */ 1098 struct list_head resync_reads; 1099 atomic_t pp_in_use; /* allocated from page pool */ 1100 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */ 1101 wait_queue_head_t ee_wait; 1102 struct page *md_io_page; /* one page buffer for md_io */ 1103 struct page *md_io_tmpp; /* for logical_block_size != 512 */ 1104 struct drbd_md_io md_io; 1105 atomic_t md_io_in_use; /* protects the md_io, md_io_page and md_io_tmpp */ 1106 spinlock_t al_lock; 1107 wait_queue_head_t al_wait; 1108 struct lru_cache *act_log; /* activity log */ 1109 unsigned int al_tr_number; 1110 int al_tr_cycle; 1111 int al_tr_pos; /* position of the next transaction in the journal */ 1112 struct crypto_hash *cram_hmac_tfm; 1113 struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */ 1114 struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */ 1115 void *int_dig_out; 1116 void *int_dig_in; 1117 void *int_dig_vv; 1118 wait_queue_head_t seq_wait; 1119 atomic_t packet_seq; 1120 unsigned int peer_seq; 1121 spinlock_t peer_seq_lock; 1122 unsigned int minor; 1123 unsigned long comm_bm_set; /* communicated number of set bits. */ 1124 cpumask_var_t cpu_mask; 1125 struct bm_io_work bm_io_work; 1126 u64 ed_uuid; /* UUID of the exposed data */ 1127 struct mutex state_mutex; 1128 char congestion_reason; /* Why we where congested... */ 1129 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */ 1130 atomic_t rs_sect_ev; /* for submitted resync data rate, both */ 1131 int rs_last_sect_ev; /* counter to compare with */ 1132 int rs_last_events; /* counter of read or write "events" (unit sectors) 1133 * on the lower level device when we last looked. */ 1134 int c_sync_rate; /* current resync rate after syncer throttle magic */ 1135 struct fifo_buffer rs_plan_s; /* correction values of resync planer */ 1136 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */ 1137 int rs_planed; /* resync sectors already planned */ 1138 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */ 1139 unsigned int peer_max_bio_size; 1140 unsigned int local_max_bio_size; 1141 }; 1142 1143 static inline struct drbd_conf *minor_to_mdev(unsigned int minor) 1144 { 1145 struct drbd_conf *mdev; 1146 1147 mdev = minor < minor_count ? minor_table[minor] : NULL; 1148 1149 return mdev; 1150 } 1151 1152 static inline unsigned int mdev_to_minor(struct drbd_conf *mdev) 1153 { 1154 return mdev->minor; 1155 } 1156 1157 /* returns 1 if it was successful, 1158 * returns 0 if there was no data socket. 1159 * so wherever you are going to use the data.socket, e.g. do 1160 * if (!drbd_get_data_sock(mdev)) 1161 * return 0; 1162 * CODE(); 1163 * drbd_put_data_sock(mdev); 1164 */ 1165 static inline int drbd_get_data_sock(struct drbd_conf *mdev) 1166 { 1167 mutex_lock(&mdev->data.mutex); 1168 /* drbd_disconnect() could have called drbd_free_sock() 1169 * while we were waiting in down()... */ 1170 if (unlikely(mdev->data.socket == NULL)) { 1171 mutex_unlock(&mdev->data.mutex); 1172 return 0; 1173 } 1174 return 1; 1175 } 1176 1177 static inline void drbd_put_data_sock(struct drbd_conf *mdev) 1178 { 1179 mutex_unlock(&mdev->data.mutex); 1180 } 1181 1182 /* 1183 * function declarations 1184 *************************/ 1185 1186 /* drbd_main.c */ 1187 1188 enum chg_state_flags { 1189 CS_HARD = 1, 1190 CS_VERBOSE = 2, 1191 CS_WAIT_COMPLETE = 4, 1192 CS_SERIALIZE = 8, 1193 CS_ORDERED = CS_WAIT_COMPLETE + CS_SERIALIZE, 1194 }; 1195 1196 enum dds_flags { 1197 DDSF_FORCED = 1, 1198 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */ 1199 }; 1200 1201 extern void drbd_init_set_defaults(struct drbd_conf *mdev); 1202 extern enum drbd_state_rv drbd_change_state(struct drbd_conf *mdev, 1203 enum chg_state_flags f, 1204 union drbd_state mask, 1205 union drbd_state val); 1206 extern void drbd_force_state(struct drbd_conf *, union drbd_state, 1207 union drbd_state); 1208 extern enum drbd_state_rv _drbd_request_state(struct drbd_conf *, 1209 union drbd_state, 1210 union drbd_state, 1211 enum chg_state_flags); 1212 extern enum drbd_state_rv __drbd_set_state(struct drbd_conf *, union drbd_state, 1213 enum chg_state_flags, 1214 struct completion *done); 1215 extern void print_st_err(struct drbd_conf *, union drbd_state, 1216 union drbd_state, int); 1217 extern int drbd_thread_start(struct drbd_thread *thi); 1218 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait); 1219 #ifdef CONFIG_SMP 1220 extern void drbd_thread_current_set_cpu(struct drbd_conf *mdev); 1221 extern void drbd_calc_cpu_mask(struct drbd_conf *mdev); 1222 #else 1223 #define drbd_thread_current_set_cpu(A) ({}) 1224 #define drbd_calc_cpu_mask(A) ({}) 1225 #endif 1226 extern void drbd_free_resources(struct drbd_conf *mdev); 1227 extern void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr, 1228 unsigned int set_size); 1229 extern void tl_clear(struct drbd_conf *mdev); 1230 extern void _tl_add_barrier(struct drbd_conf *, struct drbd_tl_epoch *); 1231 extern void drbd_free_sock(struct drbd_conf *mdev); 1232 extern int drbd_send(struct drbd_conf *mdev, struct socket *sock, 1233 void *buf, size_t size, unsigned msg_flags); 1234 extern int drbd_send_protocol(struct drbd_conf *mdev); 1235 extern int drbd_send_uuids(struct drbd_conf *mdev); 1236 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev); 1237 extern int drbd_gen_and_send_sync_uuid(struct drbd_conf *mdev); 1238 extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags); 1239 extern int drbd_send_state(struct drbd_conf *mdev, union drbd_state s); 1240 extern int drbd_send_current_state(struct drbd_conf *mdev); 1241 extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock, 1242 enum drbd_packets cmd, struct p_header80 *h, 1243 size_t size, unsigned msg_flags); 1244 #define USE_DATA_SOCKET 1 1245 #define USE_META_SOCKET 0 1246 extern int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket, 1247 enum drbd_packets cmd, struct p_header80 *h, 1248 size_t size); 1249 extern int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd, 1250 char *data, size_t size); 1251 extern int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc); 1252 extern int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr, 1253 u32 set_size); 1254 extern int drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd, 1255 struct drbd_epoch_entry *e); 1256 extern int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packets cmd, 1257 struct p_block_req *rp); 1258 extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd, 1259 struct p_data *dp, int data_size); 1260 extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd, 1261 sector_t sector, int blksize, u64 block_id); 1262 extern int drbd_send_oos(struct drbd_conf *mdev, struct drbd_request *req); 1263 extern int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd, 1264 struct drbd_epoch_entry *e); 1265 extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req); 1266 extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd, 1267 sector_t sector, int size, u64 block_id); 1268 extern int drbd_send_drequest_csum(struct drbd_conf *mdev, 1269 sector_t sector,int size, 1270 void *digest, int digest_size, 1271 enum drbd_packets cmd); 1272 extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size); 1273 1274 extern int drbd_send_bitmap(struct drbd_conf *mdev); 1275 extern int _drbd_send_bitmap(struct drbd_conf *mdev); 1276 extern int drbd_send_sr_reply(struct drbd_conf *mdev, enum drbd_state_rv retcode); 1277 extern void drbd_free_bc(struct drbd_backing_dev *ldev); 1278 extern void drbd_mdev_cleanup(struct drbd_conf *mdev); 1279 void drbd_print_uuids(struct drbd_conf *mdev, const char *text); 1280 1281 extern void drbd_md_sync(struct drbd_conf *mdev); 1282 extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev); 1283 extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local); 1284 extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local); 1285 extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local); 1286 extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local); 1287 extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local); 1288 extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local); 1289 extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local); 1290 extern int drbd_md_test_flag(struct drbd_backing_dev *, int); 1291 #ifndef DRBD_DEBUG_MD_SYNC 1292 extern void drbd_md_mark_dirty(struct drbd_conf *mdev); 1293 #else 1294 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ ) 1295 extern void drbd_md_mark_dirty_(struct drbd_conf *mdev, 1296 unsigned int line, const char *func); 1297 #endif 1298 extern void drbd_queue_bitmap_io(struct drbd_conf *mdev, 1299 int (*io_fn)(struct drbd_conf *), 1300 void (*done)(struct drbd_conf *, int), 1301 char *why, enum bm_flag flags); 1302 extern int drbd_bitmap_io(struct drbd_conf *mdev, 1303 int (*io_fn)(struct drbd_conf *), 1304 char *why, enum bm_flag flags); 1305 extern int drbd_bmio_set_n_write(struct drbd_conf *mdev); 1306 extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev); 1307 extern void drbd_go_diskless(struct drbd_conf *mdev); 1308 extern void drbd_ldev_destroy(struct drbd_conf *mdev); 1309 1310 1311 /* Meta data layout 1312 We reserve a 128MB Block (4k aligned) 1313 * either at the end of the backing device 1314 * or on a separate meta data device. */ 1315 1316 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */ 1317 /* The following numbers are sectors */ 1318 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */ 1319 #define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */ 1320 /* Allows up to about 3.8TB */ 1321 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE) 1322 1323 /* Since the smalles IO unit is usually 512 byte */ 1324 #define MD_SECTOR_SHIFT 9 1325 #define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT) 1326 1327 /* activity log */ 1328 #define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */ 1329 #define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */ 1330 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT) 1331 1332 #if BITS_PER_LONG == 32 1333 #define LN2_BPL 5 1334 #define cpu_to_lel(A) cpu_to_le32(A) 1335 #define lel_to_cpu(A) le32_to_cpu(A) 1336 #elif BITS_PER_LONG == 64 1337 #define LN2_BPL 6 1338 #define cpu_to_lel(A) cpu_to_le64(A) 1339 #define lel_to_cpu(A) le64_to_cpu(A) 1340 #else 1341 #error "LN2 of BITS_PER_LONG unknown!" 1342 #endif 1343 1344 /* resync bitmap */ 1345 /* 16MB sized 'bitmap extent' to track syncer usage */ 1346 struct bm_extent { 1347 int rs_left; /* number of bits set (out of sync) in this extent. */ 1348 int rs_failed; /* number of failed resync requests in this extent. */ 1349 unsigned long flags; 1350 struct lc_element lce; 1351 }; 1352 1353 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */ 1354 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */ 1355 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */ 1356 1357 /* drbd_bitmap.c */ 1358 /* 1359 * We need to store one bit for a block. 1360 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap. 1361 * Bit 0 ==> local node thinks this block is binary identical on both nodes 1362 * Bit 1 ==> local node thinks this block needs to be synced. 1363 */ 1364 1365 #define SLEEP_TIME (HZ/10) 1366 1367 #define BM_BLOCK_SHIFT 12 /* 4k per bit */ 1368 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT) 1369 /* (9+3) : 512 bytes @ 8 bits; representing 16M storage 1370 * per sector of on disk bitmap */ 1371 #define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */ 1372 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT) 1373 1374 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12) 1375 #error "HAVE YOU FIXED drbdmeta AS WELL??" 1376 #endif 1377 1378 /* thus many _storage_ sectors are described by one bit */ 1379 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9)) 1380 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9)) 1381 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1) 1382 1383 /* bit to represented kilo byte conversion */ 1384 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10)) 1385 1386 /* in which _bitmap_ extent (resp. sector) the bit for a certain 1387 * _storage_ sector is located in */ 1388 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9)) 1389 1390 /* how much _storage_ sectors we have per bitmap sector */ 1391 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9)) 1392 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1) 1393 1394 /* in one sector of the bitmap, we have this many activity_log extents. */ 1395 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT)) 1396 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL)) 1397 1398 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT) 1399 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1) 1400 1401 /* the extent in "PER_EXTENT" below is an activity log extent 1402 * we need that many (long words/bytes) to store the bitmap 1403 * of one AL_EXTENT_SIZE chunk of storage. 1404 * we can store the bitmap for that many AL_EXTENTS within 1405 * one sector of the _on_disk_ bitmap: 1406 * bit 0 bit 37 bit 38 bit (512*8)-1 1407 * ...|........|........|.. // ..|........| 1408 * sect. 0 `296 `304 ^(512*8*8)-1 1409 * 1410 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG ) 1411 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128 1412 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4 1413 */ 1414 1415 #define DRBD_MAX_SECTORS_32 (0xffffffffLU) 1416 #define DRBD_MAX_SECTORS_BM \ 1417 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9))) 1418 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32 1419 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM 1420 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM 1421 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32 1422 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32 1423 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32 1424 #else 1425 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM 1426 /* 16 TB in units of sectors */ 1427 #if BITS_PER_LONG == 32 1428 /* adjust by one page worth of bitmap, 1429 * so we won't wrap around in drbd_bm_find_next_bit. 1430 * you should use 64bit OS for that much storage, anyways. */ 1431 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff) 1432 #else 1433 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */ 1434 #define DRBD_MAX_SECTORS_FLEX (1UL << 51) 1435 /* corresponds to (1UL << 38) bits right now. */ 1436 #endif 1437 #endif 1438 1439 /* Sector shift value for the "hash" functions of tl_hash and ee_hash tables. 1440 * With a value of 8 all IO in one 128K block make it to the same slot of the 1441 * hash table. */ 1442 #define HT_SHIFT 8 1443 #define DRBD_MAX_BIO_SIZE (1U<<(9+HT_SHIFT)) 1444 #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */ 1445 1446 #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* The old header only allows packets up to 32Kib data */ 1447 1448 /* Number of elements in the app_reads_hash */ 1449 #define APP_R_HSIZE 15 1450 1451 extern int drbd_bm_init(struct drbd_conf *mdev); 1452 extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits); 1453 extern void drbd_bm_cleanup(struct drbd_conf *mdev); 1454 extern void drbd_bm_set_all(struct drbd_conf *mdev); 1455 extern void drbd_bm_clear_all(struct drbd_conf *mdev); 1456 /* set/clear/test only a few bits at a time */ 1457 extern int drbd_bm_set_bits( 1458 struct drbd_conf *mdev, unsigned long s, unsigned long e); 1459 extern int drbd_bm_clear_bits( 1460 struct drbd_conf *mdev, unsigned long s, unsigned long e); 1461 extern int drbd_bm_count_bits( 1462 struct drbd_conf *mdev, const unsigned long s, const unsigned long e); 1463 /* bm_set_bits variant for use while holding drbd_bm_lock, 1464 * may process the whole bitmap in one go */ 1465 extern void _drbd_bm_set_bits(struct drbd_conf *mdev, 1466 const unsigned long s, const unsigned long e); 1467 extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr); 1468 extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr); 1469 extern int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local); 1470 extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local); 1471 extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local); 1472 extern int drbd_bm_write_all(struct drbd_conf *mdev) __must_hold(local); 1473 extern int drbd_bm_write_copy_pages(struct drbd_conf *mdev) __must_hold(local); 1474 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev, 1475 unsigned long al_enr); 1476 extern size_t drbd_bm_words(struct drbd_conf *mdev); 1477 extern unsigned long drbd_bm_bits(struct drbd_conf *mdev); 1478 extern sector_t drbd_bm_capacity(struct drbd_conf *mdev); 1479 1480 #define DRBD_END_OF_BITMAP (~(unsigned long)0) 1481 extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo); 1482 /* bm_find_next variants for use while you hold drbd_bm_lock() */ 1483 extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo); 1484 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo); 1485 extern unsigned long _drbd_bm_total_weight(struct drbd_conf *mdev); 1486 extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev); 1487 extern int drbd_bm_rs_done(struct drbd_conf *mdev); 1488 /* for receive_bitmap */ 1489 extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset, 1490 size_t number, unsigned long *buffer); 1491 /* for _drbd_send_bitmap */ 1492 extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset, 1493 size_t number, unsigned long *buffer); 1494 1495 extern void drbd_bm_lock(struct drbd_conf *mdev, char *why, enum bm_flag flags); 1496 extern void drbd_bm_unlock(struct drbd_conf *mdev); 1497 /* drbd_main.c */ 1498 1499 extern struct kmem_cache *drbd_request_cache; 1500 extern struct kmem_cache *drbd_ee_cache; /* epoch entries */ 1501 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */ 1502 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */ 1503 extern mempool_t *drbd_request_mempool; 1504 extern mempool_t *drbd_ee_mempool; 1505 1506 /* drbd's page pool, used to buffer data received from the peer, 1507 * or data requested by the peer. 1508 * 1509 * This does not have an emergency reserve. 1510 * 1511 * When allocating from this pool, it first takes pages from the pool. 1512 * Only if the pool is depleted will try to allocate from the system. 1513 * 1514 * The assumption is that pages taken from this pool will be processed, 1515 * and given back, "quickly", and then can be recycled, so we can avoid 1516 * frequent calls to alloc_page(), and still will be able to make progress even 1517 * under memory pressure. 1518 */ 1519 extern struct page *drbd_pp_pool; 1520 extern spinlock_t drbd_pp_lock; 1521 extern int drbd_pp_vacant; 1522 extern wait_queue_head_t drbd_pp_wait; 1523 1524 /* We also need a standard (emergency-reserve backed) page pool 1525 * for meta data IO (activity log, bitmap). 1526 * We can keep it global, as long as it is used as "N pages at a time". 1527 * 128 should be plenty, currently we probably can get away with as few as 1. 1528 */ 1529 #define DRBD_MIN_POOL_PAGES 128 1530 extern mempool_t *drbd_md_io_page_pool; 1531 1532 /* We also need to make sure we get a bio 1533 * when we need it for housekeeping purposes */ 1534 extern struct bio_set *drbd_md_io_bio_set; 1535 /* to allocate from that set */ 1536 extern struct bio *bio_alloc_drbd(gfp_t gfp_mask); 1537 1538 extern rwlock_t global_state_lock; 1539 1540 extern struct drbd_conf *drbd_new_device(unsigned int minor); 1541 extern void drbd_free_mdev(struct drbd_conf *mdev); 1542 1543 extern int proc_details; 1544 1545 /* drbd_req */ 1546 extern void drbd_make_request(struct request_queue *q, struct bio *bio); 1547 extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req); 1548 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec); 1549 extern int is_valid_ar_handle(struct drbd_request *, sector_t); 1550 1551 1552 /* drbd_nl.c */ 1553 extern void drbd_suspend_io(struct drbd_conf *mdev); 1554 extern void drbd_resume_io(struct drbd_conf *mdev); 1555 extern char *ppsize(char *buf, unsigned long long size); 1556 extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int); 1557 enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 }; 1558 extern enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local); 1559 extern void resync_after_online_grow(struct drbd_conf *); 1560 extern void drbd_reconsider_max_bio_size(struct drbd_conf *mdev); 1561 extern enum drbd_state_rv drbd_set_role(struct drbd_conf *mdev, 1562 enum drbd_role new_role, 1563 int force); 1564 extern enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev); 1565 extern void drbd_try_outdate_peer_async(struct drbd_conf *mdev); 1566 extern int drbd_khelper(struct drbd_conf *mdev, char *cmd); 1567 1568 /* drbd_worker.c */ 1569 extern int drbd_worker(struct drbd_thread *thi); 1570 extern int drbd_alter_sa(struct drbd_conf *mdev, int na); 1571 extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side); 1572 extern void resume_next_sg(struct drbd_conf *mdev); 1573 extern void suspend_other_sg(struct drbd_conf *mdev); 1574 extern int drbd_resync_finished(struct drbd_conf *mdev); 1575 /* maybe rather drbd_main.c ? */ 1576 extern void *drbd_md_get_buffer(struct drbd_conf *mdev); 1577 extern void drbd_md_put_buffer(struct drbd_conf *mdev); 1578 extern int drbd_md_sync_page_io(struct drbd_conf *mdev, 1579 struct drbd_backing_dev *bdev, sector_t sector, int rw); 1580 extern void wait_until_done_or_disk_failure(struct drbd_conf *mdev, struct drbd_backing_dev *bdev, 1581 unsigned int *done); 1582 extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int); 1583 extern void drbd_rs_controller_reset(struct drbd_conf *mdev); 1584 1585 static inline void ov_oos_print(struct drbd_conf *mdev) 1586 { 1587 if (mdev->ov_last_oos_size) { 1588 dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n", 1589 (unsigned long long)mdev->ov_last_oos_start, 1590 (unsigned long)mdev->ov_last_oos_size); 1591 } 1592 mdev->ov_last_oos_size=0; 1593 } 1594 1595 1596 extern void drbd_csum_bio(struct drbd_conf *, struct crypto_hash *, struct bio *, void *); 1597 extern void drbd_csum_ee(struct drbd_conf *, struct crypto_hash *, struct drbd_epoch_entry *, void *); 1598 /* worker callbacks */ 1599 extern int w_req_cancel_conflict(struct drbd_conf *, struct drbd_work *, int); 1600 extern int w_read_retry_remote(struct drbd_conf *, struct drbd_work *, int); 1601 extern int w_e_end_data_req(struct drbd_conf *, struct drbd_work *, int); 1602 extern int w_e_end_rsdata_req(struct drbd_conf *, struct drbd_work *, int); 1603 extern int w_e_end_csum_rs_req(struct drbd_conf *, struct drbd_work *, int); 1604 extern int w_e_end_ov_reply(struct drbd_conf *, struct drbd_work *, int); 1605 extern int w_e_end_ov_req(struct drbd_conf *, struct drbd_work *, int); 1606 extern int w_ov_finished(struct drbd_conf *, struct drbd_work *, int); 1607 extern int w_resync_timer(struct drbd_conf *, struct drbd_work *, int); 1608 extern int w_resume_next_sg(struct drbd_conf *, struct drbd_work *, int); 1609 extern int w_send_write_hint(struct drbd_conf *, struct drbd_work *, int); 1610 extern int w_send_dblock(struct drbd_conf *, struct drbd_work *, int); 1611 extern int w_send_barrier(struct drbd_conf *, struct drbd_work *, int); 1612 extern int w_send_read_req(struct drbd_conf *, struct drbd_work *, int); 1613 extern int w_prev_work_done(struct drbd_conf *, struct drbd_work *, int); 1614 extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int); 1615 extern int w_restart_disk_io(struct drbd_conf *, struct drbd_work *, int); 1616 extern int w_send_oos(struct drbd_conf *, struct drbd_work *, int); 1617 extern int w_start_resync(struct drbd_conf *, struct drbd_work *, int); 1618 1619 extern void resync_timer_fn(unsigned long data); 1620 extern void start_resync_timer_fn(unsigned long data); 1621 1622 /* drbd_receiver.c */ 1623 extern int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector); 1624 extern int drbd_submit_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e, 1625 const unsigned rw, const int fault_type); 1626 extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list); 1627 extern struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev, 1628 u64 id, 1629 sector_t sector, 1630 unsigned int data_size, 1631 gfp_t gfp_mask) __must_hold(local); 1632 extern void drbd_free_some_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e, 1633 int is_net); 1634 #define drbd_free_ee(m,e) drbd_free_some_ee(m, e, 0) 1635 #define drbd_free_net_ee(m,e) drbd_free_some_ee(m, e, 1) 1636 extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev, 1637 struct list_head *head); 1638 extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev, 1639 struct list_head *head); 1640 extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled); 1641 extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed); 1642 extern void drbd_flush_workqueue(struct drbd_conf *mdev); 1643 extern void drbd_free_tl_hash(struct drbd_conf *mdev); 1644 1645 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to 1646 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */ 1647 static inline int drbd_setsockopt(struct socket *sock, int level, int optname, 1648 char __user *optval, int optlen) 1649 { 1650 int err; 1651 if (level == SOL_SOCKET) 1652 err = sock_setsockopt(sock, level, optname, optval, optlen); 1653 else 1654 err = sock->ops->setsockopt(sock, level, optname, optval, 1655 optlen); 1656 return err; 1657 } 1658 1659 static inline void drbd_tcp_cork(struct socket *sock) 1660 { 1661 int __user val = 1; 1662 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK, 1663 (char __user *)&val, sizeof(val)); 1664 } 1665 1666 static inline void drbd_tcp_uncork(struct socket *sock) 1667 { 1668 int __user val = 0; 1669 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK, 1670 (char __user *)&val, sizeof(val)); 1671 } 1672 1673 static inline void drbd_tcp_nodelay(struct socket *sock) 1674 { 1675 int __user val = 1; 1676 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY, 1677 (char __user *)&val, sizeof(val)); 1678 } 1679 1680 static inline void drbd_tcp_quickack(struct socket *sock) 1681 { 1682 int __user val = 2; 1683 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK, 1684 (char __user *)&val, sizeof(val)); 1685 } 1686 1687 void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo); 1688 1689 /* drbd_proc.c */ 1690 extern struct proc_dir_entry *drbd_proc; 1691 extern const struct file_operations drbd_proc_fops; 1692 extern const char *drbd_conn_str(enum drbd_conns s); 1693 extern const char *drbd_role_str(enum drbd_role s); 1694 1695 /* drbd_actlog.c */ 1696 extern void drbd_al_begin_io(struct drbd_conf *mdev, sector_t sector); 1697 extern void drbd_al_complete_io(struct drbd_conf *mdev, sector_t sector); 1698 extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector); 1699 extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector); 1700 extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector); 1701 extern void drbd_rs_cancel_all(struct drbd_conf *mdev); 1702 extern int drbd_rs_del_all(struct drbd_conf *mdev); 1703 extern void drbd_rs_failed_io(struct drbd_conf *mdev, 1704 sector_t sector, int size); 1705 extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *); 1706 extern void drbd_advance_rs_marks(struct drbd_conf *mdev, unsigned long still_to_go); 1707 extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector, 1708 int size, const char *file, const unsigned int line); 1709 #define drbd_set_in_sync(mdev, sector, size) \ 1710 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__) 1711 extern int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector, 1712 int size, const char *file, const unsigned int line); 1713 #define drbd_set_out_of_sync(mdev, sector, size) \ 1714 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__) 1715 extern void drbd_al_apply_to_bm(struct drbd_conf *mdev); 1716 extern void drbd_al_shrink(struct drbd_conf *mdev); 1717 1718 1719 /* drbd_nl.c */ 1720 1721 void drbd_nl_cleanup(void); 1722 int __init drbd_nl_init(void); 1723 void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state); 1724 void drbd_bcast_sync_progress(struct drbd_conf *mdev); 1725 void drbd_bcast_ee(struct drbd_conf *mdev, 1726 const char *reason, const int dgs, 1727 const char* seen_hash, const char* calc_hash, 1728 const struct drbd_epoch_entry* e); 1729 1730 1731 /** 1732 * DOC: DRBD State macros 1733 * 1734 * These macros are used to express state changes in easily readable form. 1735 * 1736 * The NS macros expand to a mask and a value, that can be bit ored onto the 1737 * current state as soon as the spinlock (req_lock) was taken. 1738 * 1739 * The _NS macros are used for state functions that get called with the 1740 * spinlock. These macros expand directly to the new state value. 1741 * 1742 * Besides the basic forms NS() and _NS() additional _?NS[23] are defined 1743 * to express state changes that affect more than one aspect of the state. 1744 * 1745 * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY) 1746 * Means that the network connection was established and that the peer 1747 * is in secondary role. 1748 */ 1749 #define role_MASK R_MASK 1750 #define peer_MASK R_MASK 1751 #define disk_MASK D_MASK 1752 #define pdsk_MASK D_MASK 1753 #define conn_MASK C_MASK 1754 #define susp_MASK 1 1755 #define user_isp_MASK 1 1756 #define aftr_isp_MASK 1 1757 #define susp_nod_MASK 1 1758 #define susp_fen_MASK 1 1759 1760 #define NS(T, S) \ 1761 ({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \ 1762 ({ union drbd_state val; val.i = 0; val.T = (S); val; }) 1763 #define NS2(T1, S1, T2, S2) \ 1764 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \ 1765 mask.T2 = T2##_MASK; mask; }), \ 1766 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \ 1767 val.T2 = (S2); val; }) 1768 #define NS3(T1, S1, T2, S2, T3, S3) \ 1769 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \ 1770 mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \ 1771 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \ 1772 val.T2 = (S2); val.T3 = (S3); val; }) 1773 1774 #define _NS(D, T, S) \ 1775 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; }) 1776 #define _NS2(D, T1, S1, T2, S2) \ 1777 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \ 1778 __ns.T2 = (S2); __ns; }) 1779 #define _NS3(D, T1, S1, T2, S2, T3, S3) \ 1780 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \ 1781 __ns.T2 = (S2); __ns.T3 = (S3); __ns; }) 1782 1783 /* 1784 * inline helper functions 1785 *************************/ 1786 1787 /* see also page_chain_add and friends in drbd_receiver.c */ 1788 static inline struct page *page_chain_next(struct page *page) 1789 { 1790 return (struct page *)page_private(page); 1791 } 1792 #define page_chain_for_each(page) \ 1793 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \ 1794 page = page_chain_next(page)) 1795 #define page_chain_for_each_safe(page, n) \ 1796 for (; page && ({ n = page_chain_next(page); 1; }); page = n) 1797 1798 static inline int drbd_ee_has_active_page(struct drbd_epoch_entry *e) 1799 { 1800 struct page *page = e->pages; 1801 page_chain_for_each(page) { 1802 if (page_count(page) > 1) 1803 return 1; 1804 } 1805 return 0; 1806 } 1807 1808 static inline void drbd_state_lock(struct drbd_conf *mdev) 1809 { 1810 wait_event(mdev->misc_wait, 1811 !test_and_set_bit(CLUSTER_ST_CHANGE, &mdev->flags)); 1812 } 1813 1814 static inline void drbd_state_unlock(struct drbd_conf *mdev) 1815 { 1816 clear_bit(CLUSTER_ST_CHANGE, &mdev->flags); 1817 wake_up(&mdev->misc_wait); 1818 } 1819 1820 static inline enum drbd_state_rv 1821 _drbd_set_state(struct drbd_conf *mdev, union drbd_state ns, 1822 enum chg_state_flags flags, struct completion *done) 1823 { 1824 enum drbd_state_rv rv; 1825 1826 read_lock(&global_state_lock); 1827 rv = __drbd_set_state(mdev, ns, flags, done); 1828 read_unlock(&global_state_lock); 1829 1830 return rv; 1831 } 1832 1833 /** 1834 * drbd_request_state() - Reqest a state change 1835 * @mdev: DRBD device. 1836 * @mask: mask of state bits to change. 1837 * @val: value of new state bits. 1838 * 1839 * This is the most graceful way of requesting a state change. It is verbose 1840 * quite verbose in case the state change is not possible, and all those 1841 * state changes are globally serialized. 1842 */ 1843 static inline int drbd_request_state(struct drbd_conf *mdev, 1844 union drbd_state mask, 1845 union drbd_state val) 1846 { 1847 return _drbd_request_state(mdev, mask, val, CS_VERBOSE + CS_ORDERED); 1848 } 1849 1850 enum drbd_force_detach_flags { 1851 DRBD_IO_ERROR, 1852 DRBD_META_IO_ERROR, 1853 DRBD_FORCE_DETACH, 1854 }; 1855 1856 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__) 1857 static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, 1858 enum drbd_force_detach_flags forcedetach, 1859 const char *where) 1860 { 1861 switch (mdev->ldev->dc.on_io_error) { 1862 case EP_PASS_ON: 1863 if (forcedetach == DRBD_IO_ERROR) { 1864 if (__ratelimit(&drbd_ratelimit_state)) 1865 dev_err(DEV, "Local IO failed in %s.\n", where); 1866 if (mdev->state.disk > D_INCONSISTENT) 1867 _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_HARD, NULL); 1868 break; 1869 } 1870 /* NOTE fall through to detach case if forcedetach set */ 1871 case EP_DETACH: 1872 case EP_CALL_HELPER: 1873 set_bit(WAS_IO_ERROR, &mdev->flags); 1874 if (forcedetach == DRBD_FORCE_DETACH) 1875 set_bit(FORCE_DETACH, &mdev->flags); 1876 if (mdev->state.disk > D_FAILED) { 1877 _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL); 1878 dev_err(DEV, 1879 "Local IO failed in %s. Detaching...\n", where); 1880 } 1881 break; 1882 } 1883 } 1884 1885 /** 1886 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers 1887 * @mdev: DRBD device. 1888 * @error: Error code passed to the IO completion callback 1889 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data 1890 * 1891 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED) 1892 */ 1893 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__) 1894 static inline void drbd_chk_io_error_(struct drbd_conf *mdev, 1895 int error, enum drbd_force_detach_flags forcedetach, const char *where) 1896 { 1897 if (error) { 1898 unsigned long flags; 1899 spin_lock_irqsave(&mdev->req_lock, flags); 1900 __drbd_chk_io_error_(mdev, forcedetach, where); 1901 spin_unlock_irqrestore(&mdev->req_lock, flags); 1902 } 1903 } 1904 1905 1906 /** 1907 * drbd_md_first_sector() - Returns the first sector number of the meta data area 1908 * @bdev: Meta data block device. 1909 * 1910 * BTW, for internal meta data, this happens to be the maximum capacity 1911 * we could agree upon with our peer node. 1912 */ 1913 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev) 1914 { 1915 switch (bdev->dc.meta_dev_idx) { 1916 case DRBD_MD_INDEX_INTERNAL: 1917 case DRBD_MD_INDEX_FLEX_INT: 1918 return bdev->md.md_offset + bdev->md.bm_offset; 1919 case DRBD_MD_INDEX_FLEX_EXT: 1920 default: 1921 return bdev->md.md_offset; 1922 } 1923 } 1924 1925 /** 1926 * drbd_md_last_sector() - Return the last sector number of the meta data area 1927 * @bdev: Meta data block device. 1928 */ 1929 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev) 1930 { 1931 switch (bdev->dc.meta_dev_idx) { 1932 case DRBD_MD_INDEX_INTERNAL: 1933 case DRBD_MD_INDEX_FLEX_INT: 1934 return bdev->md.md_offset + MD_AL_OFFSET - 1; 1935 case DRBD_MD_INDEX_FLEX_EXT: 1936 default: 1937 return bdev->md.md_offset + bdev->md.md_size_sect; 1938 } 1939 } 1940 1941 /* Returns the number of 512 byte sectors of the device */ 1942 static inline sector_t drbd_get_capacity(struct block_device *bdev) 1943 { 1944 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */ 1945 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0; 1946 } 1947 1948 /** 1949 * drbd_get_max_capacity() - Returns the capacity we announce to out peer 1950 * @bdev: Meta data block device. 1951 * 1952 * returns the capacity we announce to out peer. we clip ourselves at the 1953 * various MAX_SECTORS, because if we don't, current implementation will 1954 * oops sooner or later 1955 */ 1956 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev) 1957 { 1958 sector_t s; 1959 switch (bdev->dc.meta_dev_idx) { 1960 case DRBD_MD_INDEX_INTERNAL: 1961 case DRBD_MD_INDEX_FLEX_INT: 1962 s = drbd_get_capacity(bdev->backing_bdev) 1963 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX, 1964 drbd_md_first_sector(bdev)) 1965 : 0; 1966 break; 1967 case DRBD_MD_INDEX_FLEX_EXT: 1968 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX, 1969 drbd_get_capacity(bdev->backing_bdev)); 1970 /* clip at maximum size the meta device can support */ 1971 s = min_t(sector_t, s, 1972 BM_EXT_TO_SECT(bdev->md.md_size_sect 1973 - bdev->md.bm_offset)); 1974 break; 1975 default: 1976 s = min_t(sector_t, DRBD_MAX_SECTORS, 1977 drbd_get_capacity(bdev->backing_bdev)); 1978 } 1979 return s; 1980 } 1981 1982 /** 1983 * drbd_md_ss__() - Return the sector number of our meta data super block 1984 * @mdev: DRBD device. 1985 * @bdev: Meta data block device. 1986 */ 1987 static inline sector_t drbd_md_ss__(struct drbd_conf *mdev, 1988 struct drbd_backing_dev *bdev) 1989 { 1990 switch (bdev->dc.meta_dev_idx) { 1991 default: /* external, some index */ 1992 return MD_RESERVED_SECT * bdev->dc.meta_dev_idx; 1993 case DRBD_MD_INDEX_INTERNAL: 1994 /* with drbd08, internal meta data is always "flexible" */ 1995 case DRBD_MD_INDEX_FLEX_INT: 1996 /* sizeof(struct md_on_disk_07) == 4k 1997 * position: last 4k aligned block of 4k size */ 1998 if (!bdev->backing_bdev) { 1999 if (__ratelimit(&drbd_ratelimit_state)) { 2000 dev_err(DEV, "bdev->backing_bdev==NULL\n"); 2001 dump_stack(); 2002 } 2003 return 0; 2004 } 2005 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) 2006 - MD_AL_OFFSET; 2007 case DRBD_MD_INDEX_FLEX_EXT: 2008 return 0; 2009 } 2010 } 2011 2012 static inline void 2013 drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w) 2014 { 2015 unsigned long flags; 2016 spin_lock_irqsave(&q->q_lock, flags); 2017 list_add(&w->list, &q->q); 2018 up(&q->s); /* within the spinlock, 2019 see comment near end of drbd_worker() */ 2020 spin_unlock_irqrestore(&q->q_lock, flags); 2021 } 2022 2023 static inline void 2024 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w) 2025 { 2026 unsigned long flags; 2027 spin_lock_irqsave(&q->q_lock, flags); 2028 list_add_tail(&w->list, &q->q); 2029 up(&q->s); /* within the spinlock, 2030 see comment near end of drbd_worker() */ 2031 spin_unlock_irqrestore(&q->q_lock, flags); 2032 } 2033 2034 static inline void wake_asender(struct drbd_conf *mdev) 2035 { 2036 if (test_bit(SIGNAL_ASENDER, &mdev->flags)) 2037 force_sig(DRBD_SIG, mdev->asender.task); 2038 } 2039 2040 static inline void request_ping(struct drbd_conf *mdev) 2041 { 2042 set_bit(SEND_PING, &mdev->flags); 2043 wake_asender(mdev); 2044 } 2045 2046 static inline int drbd_send_short_cmd(struct drbd_conf *mdev, 2047 enum drbd_packets cmd) 2048 { 2049 struct p_header80 h; 2050 return drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, &h, sizeof(h)); 2051 } 2052 2053 static inline int drbd_send_ping(struct drbd_conf *mdev) 2054 { 2055 struct p_header80 h; 2056 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING, &h, sizeof(h)); 2057 } 2058 2059 static inline int drbd_send_ping_ack(struct drbd_conf *mdev) 2060 { 2061 struct p_header80 h; 2062 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING_ACK, &h, sizeof(h)); 2063 } 2064 2065 static inline void drbd_thread_stop(struct drbd_thread *thi) 2066 { 2067 _drbd_thread_stop(thi, false, true); 2068 } 2069 2070 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi) 2071 { 2072 _drbd_thread_stop(thi, false, false); 2073 } 2074 2075 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi) 2076 { 2077 _drbd_thread_stop(thi, true, false); 2078 } 2079 2080 /* counts how many answer packets packets we expect from our peer, 2081 * for either explicit application requests, 2082 * or implicit barrier packets as necessary. 2083 * increased: 2084 * w_send_barrier 2085 * _req_mod(req, queue_for_net_write or queue_for_net_read); 2086 * it is much easier and equally valid to count what we queue for the 2087 * worker, even before it actually was queued or send. 2088 * (drbd_make_request_common; recovery path on read io-error) 2089 * decreased: 2090 * got_BarrierAck (respective tl_clear, tl_clear_barrier) 2091 * _req_mod(req, data_received) 2092 * [from receive_DataReply] 2093 * _req_mod(req, write_acked_by_peer or recv_acked_by_peer or neg_acked) 2094 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)] 2095 * for some reason it is NOT decreased in got_NegAck, 2096 * but in the resulting cleanup code from report_params. 2097 * we should try to remember the reason for that... 2098 * _req_mod(req, send_failed or send_canceled) 2099 * _req_mod(req, connection_lost_while_pending) 2100 * [from tl_clear_barrier] 2101 */ 2102 static inline void inc_ap_pending(struct drbd_conf *mdev) 2103 { 2104 atomic_inc(&mdev->ap_pending_cnt); 2105 } 2106 2107 #define ERR_IF_CNT_IS_NEGATIVE(which) \ 2108 if (atomic_read(&mdev->which) < 0) \ 2109 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \ 2110 __func__ , __LINE__ , \ 2111 atomic_read(&mdev->which)) 2112 2113 #define dec_ap_pending(mdev) do { \ 2114 typecheck(struct drbd_conf *, mdev); \ 2115 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \ 2116 wake_up(&mdev->misc_wait); \ 2117 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0) 2118 2119 /* counts how many resync-related answers we still expect from the peer 2120 * increase decrease 2121 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY) 2122 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER) 2123 * (or P_NEG_ACK with ID_SYNCER) 2124 */ 2125 static inline void inc_rs_pending(struct drbd_conf *mdev) 2126 { 2127 atomic_inc(&mdev->rs_pending_cnt); 2128 } 2129 2130 #define dec_rs_pending(mdev) do { \ 2131 typecheck(struct drbd_conf *, mdev); \ 2132 atomic_dec(&mdev->rs_pending_cnt); \ 2133 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0) 2134 2135 /* counts how many answers we still need to send to the peer. 2136 * increased on 2137 * receive_Data unless protocol A; 2138 * we need to send a P_RECV_ACK (proto B) 2139 * or P_WRITE_ACK (proto C) 2140 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK 2141 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA 2142 * receive_Barrier_* we need to send a P_BARRIER_ACK 2143 */ 2144 static inline void inc_unacked(struct drbd_conf *mdev) 2145 { 2146 atomic_inc(&mdev->unacked_cnt); 2147 } 2148 2149 #define dec_unacked(mdev) do { \ 2150 typecheck(struct drbd_conf *, mdev); \ 2151 atomic_dec(&mdev->unacked_cnt); \ 2152 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0) 2153 2154 #define sub_unacked(mdev, n) do { \ 2155 typecheck(struct drbd_conf *, mdev); \ 2156 atomic_sub(n, &mdev->unacked_cnt); \ 2157 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0) 2158 2159 2160 static inline void put_net_conf(struct drbd_conf *mdev) 2161 { 2162 if (atomic_dec_and_test(&mdev->net_cnt)) 2163 wake_up(&mdev->net_cnt_wait); 2164 } 2165 2166 /** 2167 * get_net_conf() - Increase ref count on mdev->net_conf; Returns 0 if nothing there 2168 * @mdev: DRBD device. 2169 * 2170 * You have to call put_net_conf() when finished working with mdev->net_conf. 2171 */ 2172 static inline int get_net_conf(struct drbd_conf *mdev) 2173 { 2174 int have_net_conf; 2175 2176 atomic_inc(&mdev->net_cnt); 2177 have_net_conf = mdev->state.conn >= C_UNCONNECTED; 2178 if (!have_net_conf) 2179 put_net_conf(mdev); 2180 return have_net_conf; 2181 } 2182 2183 /** 2184 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev 2185 * @M: DRBD device. 2186 * 2187 * You have to call put_ldev() when finished working with mdev->ldev. 2188 */ 2189 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT)) 2190 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS)) 2191 2192 static inline void put_ldev(struct drbd_conf *mdev) 2193 { 2194 int i = atomic_dec_return(&mdev->local_cnt); 2195 2196 /* This may be called from some endio handler, 2197 * so we must not sleep here. */ 2198 2199 __release(local); 2200 D_ASSERT(i >= 0); 2201 if (i == 0) { 2202 if (mdev->state.disk == D_DISKLESS) 2203 /* even internal references gone, safe to destroy */ 2204 drbd_ldev_destroy(mdev); 2205 if (mdev->state.disk == D_FAILED) 2206 /* all application IO references gone. */ 2207 drbd_go_diskless(mdev); 2208 wake_up(&mdev->misc_wait); 2209 } 2210 } 2211 2212 #ifndef __CHECKER__ 2213 static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins) 2214 { 2215 int io_allowed; 2216 2217 /* never get a reference while D_DISKLESS */ 2218 if (mdev->state.disk == D_DISKLESS) 2219 return 0; 2220 2221 atomic_inc(&mdev->local_cnt); 2222 io_allowed = (mdev->state.disk >= mins); 2223 if (!io_allowed) 2224 put_ldev(mdev); 2225 return io_allowed; 2226 } 2227 #else 2228 extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins); 2229 #endif 2230 2231 /* you must have an "get_ldev" reference */ 2232 static inline void drbd_get_syncer_progress(struct drbd_conf *mdev, 2233 unsigned long *bits_left, unsigned int *per_mil_done) 2234 { 2235 /* this is to break it at compile time when we change that, in case we 2236 * want to support more than (1<<32) bits on a 32bit arch. */ 2237 typecheck(unsigned long, mdev->rs_total); 2238 2239 /* note: both rs_total and rs_left are in bits, i.e. in 2240 * units of BM_BLOCK_SIZE. 2241 * for the percentage, we don't care. */ 2242 2243 if (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T) 2244 *bits_left = mdev->ov_left; 2245 else 2246 *bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed; 2247 /* >> 10 to prevent overflow, 2248 * +1 to prevent division by zero */ 2249 if (*bits_left > mdev->rs_total) { 2250 /* doh. maybe a logic bug somewhere. 2251 * may also be just a race condition 2252 * between this and a disconnect during sync. 2253 * for now, just prevent in-kernel buffer overflow. 2254 */ 2255 smp_rmb(); 2256 dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n", 2257 drbd_conn_str(mdev->state.conn), 2258 *bits_left, mdev->rs_total, mdev->rs_failed); 2259 *per_mil_done = 0; 2260 } else { 2261 /* Make sure the division happens in long context. 2262 * We allow up to one petabyte storage right now, 2263 * at a granularity of 4k per bit that is 2**38 bits. 2264 * After shift right and multiplication by 1000, 2265 * this should still fit easily into a 32bit long, 2266 * so we don't need a 64bit division on 32bit arch. 2267 * Note: currently we don't support such large bitmaps on 32bit 2268 * arch anyways, but no harm done to be prepared for it here. 2269 */ 2270 unsigned int shift = mdev->rs_total > UINT_MAX ? 16 : 10; 2271 unsigned long left = *bits_left >> shift; 2272 unsigned long total = 1UL + (mdev->rs_total >> shift); 2273 unsigned long tmp = 1000UL - left * 1000UL/total; 2274 *per_mil_done = tmp; 2275 } 2276 } 2277 2278 2279 /* this throttles on-the-fly application requests 2280 * according to max_buffers settings; 2281 * maybe re-implement using semaphores? */ 2282 static inline int drbd_get_max_buffers(struct drbd_conf *mdev) 2283 { 2284 int mxb = 1000000; /* arbitrary limit on open requests */ 2285 if (get_net_conf(mdev)) { 2286 mxb = mdev->net_conf->max_buffers; 2287 put_net_conf(mdev); 2288 } 2289 return mxb; 2290 } 2291 2292 static inline int drbd_state_is_stable(struct drbd_conf *mdev) 2293 { 2294 union drbd_state s = mdev->state; 2295 2296 /* DO NOT add a default clause, we want the compiler to warn us 2297 * for any newly introduced state we may have forgotten to add here */ 2298 2299 switch ((enum drbd_conns)s.conn) { 2300 /* new io only accepted when there is no connection, ... */ 2301 case C_STANDALONE: 2302 case C_WF_CONNECTION: 2303 /* ... or there is a well established connection. */ 2304 case C_CONNECTED: 2305 case C_SYNC_SOURCE: 2306 case C_SYNC_TARGET: 2307 case C_VERIFY_S: 2308 case C_VERIFY_T: 2309 case C_PAUSED_SYNC_S: 2310 case C_PAUSED_SYNC_T: 2311 case C_AHEAD: 2312 case C_BEHIND: 2313 /* transitional states, IO allowed */ 2314 case C_DISCONNECTING: 2315 case C_UNCONNECTED: 2316 case C_TIMEOUT: 2317 case C_BROKEN_PIPE: 2318 case C_NETWORK_FAILURE: 2319 case C_PROTOCOL_ERROR: 2320 case C_TEAR_DOWN: 2321 case C_WF_REPORT_PARAMS: 2322 case C_STARTING_SYNC_S: 2323 case C_STARTING_SYNC_T: 2324 break; 2325 2326 /* Allow IO in BM exchange states with new protocols */ 2327 case C_WF_BITMAP_S: 2328 if (mdev->agreed_pro_version < 96) 2329 return 0; 2330 break; 2331 2332 /* no new io accepted in these states */ 2333 case C_WF_BITMAP_T: 2334 case C_WF_SYNC_UUID: 2335 case C_MASK: 2336 /* not "stable" */ 2337 return 0; 2338 } 2339 2340 switch ((enum drbd_disk_state)s.disk) { 2341 case D_DISKLESS: 2342 case D_INCONSISTENT: 2343 case D_OUTDATED: 2344 case D_CONSISTENT: 2345 case D_UP_TO_DATE: 2346 case D_FAILED: 2347 /* disk state is stable as well. */ 2348 break; 2349 2350 /* no new io accepted during tansitional states */ 2351 case D_ATTACHING: 2352 case D_NEGOTIATING: 2353 case D_UNKNOWN: 2354 case D_MASK: 2355 /* not "stable" */ 2356 return 0; 2357 } 2358 2359 return 1; 2360 } 2361 2362 static inline int is_susp(union drbd_state s) 2363 { 2364 return s.susp || s.susp_nod || s.susp_fen; 2365 } 2366 2367 static inline bool may_inc_ap_bio(struct drbd_conf *mdev) 2368 { 2369 int mxb = drbd_get_max_buffers(mdev); 2370 2371 if (is_susp(mdev->state)) 2372 return false; 2373 if (test_bit(SUSPEND_IO, &mdev->flags)) 2374 return false; 2375 2376 /* to avoid potential deadlock or bitmap corruption, 2377 * in various places, we only allow new application io 2378 * to start during "stable" states. */ 2379 2380 /* no new io accepted when attaching or detaching the disk */ 2381 if (!drbd_state_is_stable(mdev)) 2382 return false; 2383 2384 /* since some older kernels don't have atomic_add_unless, 2385 * and we are within the spinlock anyways, we have this workaround. */ 2386 if (atomic_read(&mdev->ap_bio_cnt) > mxb) 2387 return false; 2388 if (test_bit(BITMAP_IO, &mdev->flags)) 2389 return false; 2390 return true; 2391 } 2392 2393 static inline bool inc_ap_bio_cond(struct drbd_conf *mdev, int count) 2394 { 2395 bool rv = false; 2396 2397 spin_lock_irq(&mdev->req_lock); 2398 rv = may_inc_ap_bio(mdev); 2399 if (rv) 2400 atomic_add(count, &mdev->ap_bio_cnt); 2401 spin_unlock_irq(&mdev->req_lock); 2402 2403 return rv; 2404 } 2405 2406 static inline void inc_ap_bio(struct drbd_conf *mdev, int count) 2407 { 2408 /* we wait here 2409 * as long as the device is suspended 2410 * until the bitmap is no longer on the fly during connection 2411 * handshake as long as we would exeed the max_buffer limit. 2412 * 2413 * to avoid races with the reconnect code, 2414 * we need to atomic_inc within the spinlock. */ 2415 2416 wait_event(mdev->misc_wait, inc_ap_bio_cond(mdev, count)); 2417 } 2418 2419 static inline void dec_ap_bio(struct drbd_conf *mdev) 2420 { 2421 int mxb = drbd_get_max_buffers(mdev); 2422 int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt); 2423 2424 D_ASSERT(ap_bio >= 0); 2425 2426 if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) { 2427 if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags)) 2428 drbd_queue_work(&mdev->data.work, &mdev->bm_io_work.w); 2429 } 2430 2431 /* this currently does wake_up for every dec_ap_bio! 2432 * maybe rather introduce some type of hysteresis? 2433 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */ 2434 if (ap_bio < mxb) 2435 wake_up(&mdev->misc_wait); 2436 } 2437 2438 static inline int drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val) 2439 { 2440 int changed = mdev->ed_uuid != val; 2441 mdev->ed_uuid = val; 2442 return changed; 2443 } 2444 2445 static inline int seq_cmp(u32 a, u32 b) 2446 { 2447 /* we assume wrap around at 32bit. 2448 * for wrap around at 24bit (old atomic_t), 2449 * we'd have to 2450 * a <<= 8; b <<= 8; 2451 */ 2452 return (s32)(a) - (s32)(b); 2453 } 2454 #define seq_lt(a, b) (seq_cmp((a), (b)) < 0) 2455 #define seq_gt(a, b) (seq_cmp((a), (b)) > 0) 2456 #define seq_ge(a, b) (seq_cmp((a), (b)) >= 0) 2457 #define seq_le(a, b) (seq_cmp((a), (b)) <= 0) 2458 /* CAUTION: please no side effects in arguments! */ 2459 #define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b))) 2460 2461 static inline void update_peer_seq(struct drbd_conf *mdev, unsigned int new_seq) 2462 { 2463 unsigned int m; 2464 spin_lock(&mdev->peer_seq_lock); 2465 m = seq_max(mdev->peer_seq, new_seq); 2466 mdev->peer_seq = m; 2467 spin_unlock(&mdev->peer_seq_lock); 2468 if (m == new_seq) 2469 wake_up(&mdev->seq_wait); 2470 } 2471 2472 static inline void drbd_update_congested(struct drbd_conf *mdev) 2473 { 2474 struct sock *sk = mdev->data.socket->sk; 2475 if (sk->sk_wmem_queued > sk->sk_sndbuf * 4 / 5) 2476 set_bit(NET_CONGESTED, &mdev->flags); 2477 } 2478 2479 static inline int drbd_queue_order_type(struct drbd_conf *mdev) 2480 { 2481 /* sorry, we currently have no working implementation 2482 * of distributed TCQ stuff */ 2483 #ifndef QUEUE_ORDERED_NONE 2484 #define QUEUE_ORDERED_NONE 0 2485 #endif 2486 return QUEUE_ORDERED_NONE; 2487 } 2488 2489 static inline void drbd_md_flush(struct drbd_conf *mdev) 2490 { 2491 int r; 2492 2493 if (test_bit(MD_NO_FUA, &mdev->flags)) 2494 return; 2495 2496 r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL); 2497 if (r) { 2498 set_bit(MD_NO_FUA, &mdev->flags); 2499 dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r); 2500 } 2501 } 2502 2503 #endif 2504