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 int seq_num; 716 unsigned long start_time; 717 }; 718 719 struct drbd_tl_epoch { 720 struct drbd_work w; 721 struct list_head requests; /* requests before */ 722 struct drbd_tl_epoch *next; /* pointer to the next barrier */ 723 unsigned int br_number; /* the barriers identifier. */ 724 int n_writes; /* number of requests attached before this barrier */ 725 }; 726 727 struct drbd_request; 728 729 /* These Tl_epoch_entries may be in one of 6 lists: 730 active_ee .. data packet being written 731 sync_ee .. syncer block being written 732 done_ee .. block written, need to send P_WRITE_ACK 733 read_ee .. [RS]P_DATA_REQUEST being read 734 */ 735 736 struct drbd_epoch { 737 struct list_head list; 738 unsigned int barrier_nr; 739 atomic_t epoch_size; /* increased on every request added. */ 740 atomic_t active; /* increased on every req. added, and dec on every finished. */ 741 unsigned long flags; 742 }; 743 744 /* drbd_epoch flag bits */ 745 enum { 746 DE_HAVE_BARRIER_NUMBER, 747 }; 748 749 enum epoch_event { 750 EV_PUT, 751 EV_GOT_BARRIER_NR, 752 EV_BECAME_LAST, 753 EV_CLEANUP = 32, /* used as flag */ 754 }; 755 756 struct drbd_wq_barrier { 757 struct drbd_work w; 758 struct completion done; 759 }; 760 761 struct digest_info { 762 int digest_size; 763 void *digest; 764 }; 765 766 struct drbd_epoch_entry { 767 struct drbd_work w; 768 struct hlist_node collision; 769 struct drbd_epoch *epoch; /* for writes */ 770 struct drbd_conf *mdev; 771 struct page *pages; 772 atomic_t pending_bios; 773 unsigned int size; 774 /* see comments on ee flag bits below */ 775 unsigned long flags; 776 sector_t sector; 777 union { 778 u64 block_id; 779 struct digest_info *digest; 780 }; 781 }; 782 783 /* ee flag bits. 784 * While corresponding bios are in flight, the only modification will be 785 * set_bit WAS_ERROR, which has to be atomic. 786 * If no bios are in flight yet, or all have been completed, 787 * non-atomic modification to ee->flags is ok. 788 */ 789 enum { 790 __EE_CALL_AL_COMPLETE_IO, 791 __EE_MAY_SET_IN_SYNC, 792 793 /* In case a barrier failed, 794 * we need to resubmit without the barrier flag. */ 795 __EE_RESUBMITTED, 796 797 /* we may have several bios per epoch entry. 798 * if any of those fail, we set this flag atomically 799 * from the endio callback */ 800 __EE_WAS_ERROR, 801 802 /* This ee has a pointer to a digest instead of a block id */ 803 __EE_HAS_DIGEST, 804 }; 805 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO) 806 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC) 807 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED) 808 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR) 809 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST) 810 811 /* global flag bits */ 812 enum { 813 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */ 814 SIGNAL_ASENDER, /* whether asender wants to be interrupted */ 815 SEND_PING, /* whether asender should send a ping asap */ 816 817 UNPLUG_QUEUED, /* only relevant with kernel 2.4 */ 818 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */ 819 MD_DIRTY, /* current uuids and flags not yet on disk */ 820 DISCARD_CONCURRENT, /* Set on one node, cleared on the peer! */ 821 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */ 822 CLUSTER_ST_CHANGE, /* Cluster wide state change going on... */ 823 CL_ST_CHG_SUCCESS, 824 CL_ST_CHG_FAIL, 825 CRASHED_PRIMARY, /* This node was a crashed primary. 826 * Gets cleared when the state.conn 827 * goes into C_CONNECTED state. */ 828 NO_BARRIER_SUPP, /* underlying block device doesn't implement barriers */ 829 CONSIDER_RESYNC, 830 831 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */ 832 SUSPEND_IO, /* suspend application io */ 833 BITMAP_IO, /* suspend application io; 834 once no more io in flight, start bitmap io */ 835 BITMAP_IO_QUEUED, /* Started bitmap IO */ 836 GO_DISKLESS, /* Disk is being detached, on io-error or admin request. */ 837 WAS_IO_ERROR, /* Local disk failed returned IO error */ 838 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */ 839 NET_CONGESTED, /* The data socket is congested */ 840 841 CONFIG_PENDING, /* serialization of (re)configuration requests. 842 * if set, also prevents the device from dying */ 843 DEVICE_DYING, /* device became unconfigured, 844 * but worker thread is still handling the cleanup. 845 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed, 846 * while this is set. */ 847 RESIZE_PENDING, /* Size change detected locally, waiting for the response from 848 * the peer, if it changed there as well. */ 849 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */ 850 GOT_PING_ACK, /* set when we receive a ping_ack packet, misc wait gets woken */ 851 NEW_CUR_UUID, /* Create new current UUID when thawing IO */ 852 AL_SUSPENDED, /* Activity logging is currently suspended. */ 853 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */ 854 }; 855 856 struct drbd_bitmap; /* opaque for drbd_conf */ 857 858 /* definition of bits in bm_flags to be used in drbd_bm_lock 859 * and drbd_bitmap_io and friends. */ 860 enum bm_flag { 861 /* do we need to kfree, or vfree bm_pages? */ 862 BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */ 863 864 /* currently locked for bulk operation */ 865 BM_LOCKED_MASK = 0x7, 866 867 /* in detail, that is: */ 868 BM_DONT_CLEAR = 0x1, 869 BM_DONT_SET = 0x2, 870 BM_DONT_TEST = 0x4, 871 872 /* (test bit, count bit) allowed (common case) */ 873 BM_LOCKED_TEST_ALLOWED = 0x3, 874 875 /* testing bits, as well as setting new bits allowed, but clearing bits 876 * would be unexpected. Used during bitmap receive. Setting new bits 877 * requires sending of "out-of-sync" information, though. */ 878 BM_LOCKED_SET_ALLOWED = 0x1, 879 880 /* clear is not expected while bitmap is locked for bulk operation */ 881 }; 882 883 884 /* TODO sort members for performance 885 * MAYBE group them further */ 886 887 /* THINK maybe we actually want to use the default "event/%s" worker threads 888 * or similar in linux 2.6, which uses per cpu data and threads. 889 */ 890 struct drbd_work_queue { 891 struct list_head q; 892 struct semaphore s; /* producers up it, worker down()s it */ 893 spinlock_t q_lock; /* to protect the list. */ 894 }; 895 896 struct drbd_socket { 897 struct drbd_work_queue work; 898 struct mutex mutex; 899 struct socket *socket; 900 /* this way we get our 901 * send/receive buffers off the stack */ 902 union p_polymorph sbuf; 903 union p_polymorph rbuf; 904 }; 905 906 struct drbd_md { 907 u64 md_offset; /* sector offset to 'super' block */ 908 909 u64 la_size_sect; /* last agreed size, unit sectors */ 910 u64 uuid[UI_SIZE]; 911 u64 device_uuid; 912 u32 flags; 913 u32 md_size_sect; 914 915 s32 al_offset; /* signed relative sector offset to al area */ 916 s32 bm_offset; /* signed relative sector offset to bitmap */ 917 918 /* u32 al_nr_extents; important for restoring the AL 919 * is stored into sync_conf.al_extents, which in turn 920 * gets applied to act_log->nr_elements 921 */ 922 }; 923 924 /* for sync_conf and other types... */ 925 #define NL_PACKET(name, number, fields) struct name { fields }; 926 #define NL_INTEGER(pn,pr,member) int member; 927 #define NL_INT64(pn,pr,member) __u64 member; 928 #define NL_BIT(pn,pr,member) unsigned member:1; 929 #define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len; 930 #include <linux/drbd_nl.h> 931 932 struct drbd_backing_dev { 933 struct block_device *backing_bdev; 934 struct block_device *md_bdev; 935 struct drbd_md md; 936 struct disk_conf dc; /* The user provided config... */ 937 sector_t known_size; /* last known size of that backing device */ 938 }; 939 940 struct drbd_md_io { 941 struct drbd_conf *mdev; 942 struct completion event; 943 int error; 944 }; 945 946 struct bm_io_work { 947 struct drbd_work w; 948 char *why; 949 enum bm_flag flags; 950 int (*io_fn)(struct drbd_conf *mdev); 951 void (*done)(struct drbd_conf *mdev, int rv); 952 }; 953 954 enum write_ordering_e { 955 WO_none, 956 WO_drain_io, 957 WO_bdev_flush, 958 }; 959 960 struct fifo_buffer { 961 int *values; 962 unsigned int head_index; 963 unsigned int size; 964 }; 965 966 struct drbd_conf { 967 /* things that are stored as / read from meta data on disk */ 968 unsigned long flags; 969 970 /* configured by drbdsetup */ 971 struct net_conf *net_conf; /* protected by get_net_conf() and put_net_conf() */ 972 struct syncer_conf sync_conf; 973 struct drbd_backing_dev *ldev __protected_by(local); 974 975 sector_t p_size; /* partner's disk size */ 976 struct request_queue *rq_queue; 977 struct block_device *this_bdev; 978 struct gendisk *vdisk; 979 980 struct drbd_socket data; /* data/barrier/cstate/parameter packets */ 981 struct drbd_socket meta; /* ping/ack (metadata) packets */ 982 int agreed_pro_version; /* actually used protocol version */ 983 unsigned long last_received; /* in jiffies, either socket */ 984 unsigned int ko_count; 985 struct drbd_work resync_work, 986 unplug_work, 987 go_diskless, 988 md_sync_work, 989 start_resync_work; 990 struct timer_list resync_timer; 991 struct timer_list md_sync_timer; 992 struct timer_list start_resync_timer; 993 struct timer_list request_timer; 994 #ifdef DRBD_DEBUG_MD_SYNC 995 struct { 996 unsigned int line; 997 const char* func; 998 } last_md_mark_dirty; 999 #endif 1000 1001 /* Used after attach while negotiating new disk state. */ 1002 union drbd_state new_state_tmp; 1003 1004 union drbd_state state; 1005 wait_queue_head_t misc_wait; 1006 wait_queue_head_t state_wait; /* upon each state change. */ 1007 wait_queue_head_t net_cnt_wait; 1008 unsigned int send_cnt; 1009 unsigned int recv_cnt; 1010 unsigned int read_cnt; 1011 unsigned int writ_cnt; 1012 unsigned int al_writ_cnt; 1013 unsigned int bm_writ_cnt; 1014 atomic_t ap_bio_cnt; /* Requests we need to complete */ 1015 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */ 1016 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */ 1017 atomic_t unacked_cnt; /* Need to send replys for */ 1018 atomic_t local_cnt; /* Waiting for local completion */ 1019 atomic_t net_cnt; /* Users of net_conf */ 1020 spinlock_t req_lock; 1021 struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */ 1022 struct drbd_tl_epoch *newest_tle; 1023 struct drbd_tl_epoch *oldest_tle; 1024 struct list_head out_of_sequence_requests; 1025 struct hlist_head *tl_hash; 1026 unsigned int tl_hash_s; 1027 1028 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */ 1029 unsigned long rs_total; 1030 /* number of resync blocks that failed in this run */ 1031 unsigned long rs_failed; 1032 /* Syncer's start time [unit jiffies] */ 1033 unsigned long rs_start; 1034 /* cumulated time in PausedSyncX state [unit jiffies] */ 1035 unsigned long rs_paused; 1036 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */ 1037 unsigned long rs_same_csum; 1038 #define DRBD_SYNC_MARKS 8 1039 #define DRBD_SYNC_MARK_STEP (3*HZ) 1040 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */ 1041 unsigned long rs_mark_left[DRBD_SYNC_MARKS]; 1042 /* marks's time [unit jiffies] */ 1043 unsigned long rs_mark_time[DRBD_SYNC_MARKS]; 1044 /* current index into rs_mark_{left,time} */ 1045 int rs_last_mark; 1046 1047 /* where does the admin want us to start? (sector) */ 1048 sector_t ov_start_sector; 1049 /* where are we now? (sector) */ 1050 sector_t ov_position; 1051 /* Start sector of out of sync range (to merge printk reporting). */ 1052 sector_t ov_last_oos_start; 1053 /* size of out-of-sync range in sectors. */ 1054 sector_t ov_last_oos_size; 1055 unsigned long ov_left; /* in bits */ 1056 struct crypto_hash *csums_tfm; 1057 struct crypto_hash *verify_tfm; 1058 1059 struct drbd_thread receiver; 1060 struct drbd_thread worker; 1061 struct drbd_thread asender; 1062 struct drbd_bitmap *bitmap; 1063 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */ 1064 1065 /* Used to track operations of resync... */ 1066 struct lru_cache *resync; 1067 /* Number of locked elements in resync LRU */ 1068 unsigned int resync_locked; 1069 /* resync extent number waiting for application requests */ 1070 unsigned int resync_wenr; 1071 1072 int open_cnt; 1073 u64 *p_uuid; 1074 struct drbd_epoch *current_epoch; 1075 spinlock_t epoch_lock; 1076 unsigned int epochs; 1077 enum write_ordering_e write_ordering; 1078 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */ 1079 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */ 1080 struct list_head done_ee; /* send ack */ 1081 struct list_head read_ee; /* IO in progress (any read) */ 1082 struct list_head net_ee; /* zero-copy network send in progress */ 1083 struct hlist_head *ee_hash; /* is proteced by req_lock! */ 1084 unsigned int ee_hash_s; 1085 1086 /* this one is protected by ee_lock, single thread */ 1087 struct drbd_epoch_entry *last_write_w_barrier; 1088 1089 int next_barrier_nr; 1090 struct hlist_head *app_reads_hash; /* is proteced by req_lock */ 1091 struct list_head resync_reads; 1092 atomic_t pp_in_use; /* allocated from page pool */ 1093 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */ 1094 wait_queue_head_t ee_wait; 1095 struct page *md_io_page; /* one page buffer for md_io */ 1096 struct page *md_io_tmpp; /* for logical_block_size != 512 */ 1097 struct mutex md_io_mutex; /* protects the md_io_buffer */ 1098 spinlock_t al_lock; 1099 wait_queue_head_t al_wait; 1100 struct lru_cache *act_log; /* activity log */ 1101 unsigned int al_tr_number; 1102 int al_tr_cycle; 1103 int al_tr_pos; /* position of the next transaction in the journal */ 1104 struct crypto_hash *cram_hmac_tfm; 1105 struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */ 1106 struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */ 1107 void *int_dig_out; 1108 void *int_dig_in; 1109 void *int_dig_vv; 1110 wait_queue_head_t seq_wait; 1111 atomic_t packet_seq; 1112 unsigned int peer_seq; 1113 spinlock_t peer_seq_lock; 1114 unsigned int minor; 1115 unsigned long comm_bm_set; /* communicated number of set bits. */ 1116 cpumask_var_t cpu_mask; 1117 struct bm_io_work bm_io_work; 1118 u64 ed_uuid; /* UUID of the exposed data */ 1119 struct mutex state_mutex; 1120 char congestion_reason; /* Why we where congested... */ 1121 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */ 1122 atomic_t rs_sect_ev; /* for submitted resync data rate, both */ 1123 int rs_last_sect_ev; /* counter to compare with */ 1124 int rs_last_events; /* counter of read or write "events" (unit sectors) 1125 * on the lower level device when we last looked. */ 1126 int c_sync_rate; /* current resync rate after syncer throttle magic */ 1127 struct fifo_buffer rs_plan_s; /* correction values of resync planer */ 1128 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */ 1129 int rs_planed; /* resync sectors already planned */ 1130 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */ 1131 int peer_max_bio_size; 1132 int local_max_bio_size; 1133 }; 1134 1135 static inline struct drbd_conf *minor_to_mdev(unsigned int minor) 1136 { 1137 struct drbd_conf *mdev; 1138 1139 mdev = minor < minor_count ? minor_table[minor] : NULL; 1140 1141 return mdev; 1142 } 1143 1144 static inline unsigned int mdev_to_minor(struct drbd_conf *mdev) 1145 { 1146 return mdev->minor; 1147 } 1148 1149 /* returns 1 if it was successful, 1150 * returns 0 if there was no data socket. 1151 * so wherever you are going to use the data.socket, e.g. do 1152 * if (!drbd_get_data_sock(mdev)) 1153 * return 0; 1154 * CODE(); 1155 * drbd_put_data_sock(mdev); 1156 */ 1157 static inline int drbd_get_data_sock(struct drbd_conf *mdev) 1158 { 1159 mutex_lock(&mdev->data.mutex); 1160 /* drbd_disconnect() could have called drbd_free_sock() 1161 * while we were waiting in down()... */ 1162 if (unlikely(mdev->data.socket == NULL)) { 1163 mutex_unlock(&mdev->data.mutex); 1164 return 0; 1165 } 1166 return 1; 1167 } 1168 1169 static inline void drbd_put_data_sock(struct drbd_conf *mdev) 1170 { 1171 mutex_unlock(&mdev->data.mutex); 1172 } 1173 1174 /* 1175 * function declarations 1176 *************************/ 1177 1178 /* drbd_main.c */ 1179 1180 enum chg_state_flags { 1181 CS_HARD = 1, 1182 CS_VERBOSE = 2, 1183 CS_WAIT_COMPLETE = 4, 1184 CS_SERIALIZE = 8, 1185 CS_ORDERED = CS_WAIT_COMPLETE + CS_SERIALIZE, 1186 }; 1187 1188 enum dds_flags { 1189 DDSF_FORCED = 1, 1190 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */ 1191 }; 1192 1193 extern void drbd_init_set_defaults(struct drbd_conf *mdev); 1194 extern enum drbd_state_rv drbd_change_state(struct drbd_conf *mdev, 1195 enum chg_state_flags f, 1196 union drbd_state mask, 1197 union drbd_state val); 1198 extern void drbd_force_state(struct drbd_conf *, union drbd_state, 1199 union drbd_state); 1200 extern enum drbd_state_rv _drbd_request_state(struct drbd_conf *, 1201 union drbd_state, 1202 union drbd_state, 1203 enum chg_state_flags); 1204 extern enum drbd_state_rv __drbd_set_state(struct drbd_conf *, union drbd_state, 1205 enum chg_state_flags, 1206 struct completion *done); 1207 extern void print_st_err(struct drbd_conf *, union drbd_state, 1208 union drbd_state, int); 1209 extern int drbd_thread_start(struct drbd_thread *thi); 1210 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait); 1211 #ifdef CONFIG_SMP 1212 extern void drbd_thread_current_set_cpu(struct drbd_conf *mdev); 1213 extern void drbd_calc_cpu_mask(struct drbd_conf *mdev); 1214 #else 1215 #define drbd_thread_current_set_cpu(A) ({}) 1216 #define drbd_calc_cpu_mask(A) ({}) 1217 #endif 1218 extern void drbd_free_resources(struct drbd_conf *mdev); 1219 extern void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr, 1220 unsigned int set_size); 1221 extern void tl_clear(struct drbd_conf *mdev); 1222 extern void _tl_add_barrier(struct drbd_conf *, struct drbd_tl_epoch *); 1223 extern void drbd_free_sock(struct drbd_conf *mdev); 1224 extern int drbd_send(struct drbd_conf *mdev, struct socket *sock, 1225 void *buf, size_t size, unsigned msg_flags); 1226 extern int drbd_send_protocol(struct drbd_conf *mdev); 1227 extern int drbd_send_uuids(struct drbd_conf *mdev); 1228 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev); 1229 extern int drbd_gen_and_send_sync_uuid(struct drbd_conf *mdev); 1230 extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags); 1231 extern int _drbd_send_state(struct drbd_conf *mdev); 1232 extern int drbd_send_state(struct drbd_conf *mdev); 1233 extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock, 1234 enum drbd_packets cmd, struct p_header80 *h, 1235 size_t size, unsigned msg_flags); 1236 #define USE_DATA_SOCKET 1 1237 #define USE_META_SOCKET 0 1238 extern int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket, 1239 enum drbd_packets cmd, struct p_header80 *h, 1240 size_t size); 1241 extern int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd, 1242 char *data, size_t size); 1243 extern int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc); 1244 extern int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr, 1245 u32 set_size); 1246 extern int drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd, 1247 struct drbd_epoch_entry *e); 1248 extern int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packets cmd, 1249 struct p_block_req *rp); 1250 extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd, 1251 struct p_data *dp, int data_size); 1252 extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd, 1253 sector_t sector, int blksize, u64 block_id); 1254 extern int drbd_send_oos(struct drbd_conf *mdev, struct drbd_request *req); 1255 extern int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd, 1256 struct drbd_epoch_entry *e); 1257 extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req); 1258 extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd, 1259 sector_t sector, int size, u64 block_id); 1260 extern int drbd_send_drequest_csum(struct drbd_conf *mdev, 1261 sector_t sector,int size, 1262 void *digest, int digest_size, 1263 enum drbd_packets cmd); 1264 extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size); 1265 1266 extern int drbd_send_bitmap(struct drbd_conf *mdev); 1267 extern int _drbd_send_bitmap(struct drbd_conf *mdev); 1268 extern int drbd_send_sr_reply(struct drbd_conf *mdev, enum drbd_state_rv retcode); 1269 extern void drbd_free_bc(struct drbd_backing_dev *ldev); 1270 extern void drbd_mdev_cleanup(struct drbd_conf *mdev); 1271 void drbd_print_uuids(struct drbd_conf *mdev, const char *text); 1272 1273 extern void drbd_md_sync(struct drbd_conf *mdev); 1274 extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev); 1275 extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local); 1276 extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local); 1277 extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local); 1278 extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local); 1279 extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local); 1280 extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local); 1281 extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local); 1282 extern int drbd_md_test_flag(struct drbd_backing_dev *, int); 1283 #ifndef DRBD_DEBUG_MD_SYNC 1284 extern void drbd_md_mark_dirty(struct drbd_conf *mdev); 1285 #else 1286 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ ) 1287 extern void drbd_md_mark_dirty_(struct drbd_conf *mdev, 1288 unsigned int line, const char *func); 1289 #endif 1290 extern void drbd_queue_bitmap_io(struct drbd_conf *mdev, 1291 int (*io_fn)(struct drbd_conf *), 1292 void (*done)(struct drbd_conf *, int), 1293 char *why, enum bm_flag flags); 1294 extern int drbd_bitmap_io(struct drbd_conf *mdev, 1295 int (*io_fn)(struct drbd_conf *), 1296 char *why, enum bm_flag flags); 1297 extern int drbd_bmio_set_n_write(struct drbd_conf *mdev); 1298 extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev); 1299 extern void drbd_go_diskless(struct drbd_conf *mdev); 1300 extern void drbd_ldev_destroy(struct drbd_conf *mdev); 1301 1302 1303 /* Meta data layout 1304 We reserve a 128MB Block (4k aligned) 1305 * either at the end of the backing device 1306 * or on a separate meta data device. */ 1307 1308 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */ 1309 /* The following numbers are sectors */ 1310 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */ 1311 #define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */ 1312 /* Allows up to about 3.8TB */ 1313 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE) 1314 1315 /* Since the smalles IO unit is usually 512 byte */ 1316 #define MD_SECTOR_SHIFT 9 1317 #define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT) 1318 1319 /* activity log */ 1320 #define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */ 1321 #define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */ 1322 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT) 1323 1324 #if BITS_PER_LONG == 32 1325 #define LN2_BPL 5 1326 #define cpu_to_lel(A) cpu_to_le32(A) 1327 #define lel_to_cpu(A) le32_to_cpu(A) 1328 #elif BITS_PER_LONG == 64 1329 #define LN2_BPL 6 1330 #define cpu_to_lel(A) cpu_to_le64(A) 1331 #define lel_to_cpu(A) le64_to_cpu(A) 1332 #else 1333 #error "LN2 of BITS_PER_LONG unknown!" 1334 #endif 1335 1336 /* resync bitmap */ 1337 /* 16MB sized 'bitmap extent' to track syncer usage */ 1338 struct bm_extent { 1339 int rs_left; /* number of bits set (out of sync) in this extent. */ 1340 int rs_failed; /* number of failed resync requests in this extent. */ 1341 unsigned long flags; 1342 struct lc_element lce; 1343 }; 1344 1345 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */ 1346 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */ 1347 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */ 1348 1349 /* drbd_bitmap.c */ 1350 /* 1351 * We need to store one bit for a block. 1352 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap. 1353 * Bit 0 ==> local node thinks this block is binary identical on both nodes 1354 * Bit 1 ==> local node thinks this block needs to be synced. 1355 */ 1356 1357 #define SLEEP_TIME (HZ/10) 1358 1359 #define BM_BLOCK_SHIFT 12 /* 4k per bit */ 1360 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT) 1361 /* (9+3) : 512 bytes @ 8 bits; representing 16M storage 1362 * per sector of on disk bitmap */ 1363 #define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */ 1364 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT) 1365 1366 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12) 1367 #error "HAVE YOU FIXED drbdmeta AS WELL??" 1368 #endif 1369 1370 /* thus many _storage_ sectors are described by one bit */ 1371 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9)) 1372 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9)) 1373 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1) 1374 1375 /* bit to represented kilo byte conversion */ 1376 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10)) 1377 1378 /* in which _bitmap_ extent (resp. sector) the bit for a certain 1379 * _storage_ sector is located in */ 1380 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9)) 1381 1382 /* how much _storage_ sectors we have per bitmap sector */ 1383 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9)) 1384 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1) 1385 1386 /* in one sector of the bitmap, we have this many activity_log extents. */ 1387 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT)) 1388 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL)) 1389 1390 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT) 1391 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1) 1392 1393 /* the extent in "PER_EXTENT" below is an activity log extent 1394 * we need that many (long words/bytes) to store the bitmap 1395 * of one AL_EXTENT_SIZE chunk of storage. 1396 * we can store the bitmap for that many AL_EXTENTS within 1397 * one sector of the _on_disk_ bitmap: 1398 * bit 0 bit 37 bit 38 bit (512*8)-1 1399 * ...|........|........|.. // ..|........| 1400 * sect. 0 `296 `304 ^(512*8*8)-1 1401 * 1402 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG ) 1403 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128 1404 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4 1405 */ 1406 1407 #define DRBD_MAX_SECTORS_32 (0xffffffffLU) 1408 #define DRBD_MAX_SECTORS_BM \ 1409 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9))) 1410 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32 1411 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM 1412 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM 1413 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32 1414 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32 1415 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32 1416 #else 1417 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM 1418 /* 16 TB in units of sectors */ 1419 #if BITS_PER_LONG == 32 1420 /* adjust by one page worth of bitmap, 1421 * so we won't wrap around in drbd_bm_find_next_bit. 1422 * you should use 64bit OS for that much storage, anyways. */ 1423 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff) 1424 #else 1425 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */ 1426 #define DRBD_MAX_SECTORS_FLEX (1UL << 51) 1427 /* corresponds to (1UL << 38) bits right now. */ 1428 #endif 1429 #endif 1430 1431 /* Sector shift value for the "hash" functions of tl_hash and ee_hash tables. 1432 * With a value of 8 all IO in one 128K block make it to the same slot of the 1433 * hash table. */ 1434 #define HT_SHIFT 8 1435 #define DRBD_MAX_BIO_SIZE (1U<<(9+HT_SHIFT)) 1436 #define DRBD_MAX_BIO_SIZE_SAFE (1 << 12) /* Works always = 4k */ 1437 1438 #define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */ 1439 1440 /* Number of elements in the app_reads_hash */ 1441 #define APP_R_HSIZE 15 1442 1443 extern int drbd_bm_init(struct drbd_conf *mdev); 1444 extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits); 1445 extern void drbd_bm_cleanup(struct drbd_conf *mdev); 1446 extern void drbd_bm_set_all(struct drbd_conf *mdev); 1447 extern void drbd_bm_clear_all(struct drbd_conf *mdev); 1448 /* set/clear/test only a few bits at a time */ 1449 extern int drbd_bm_set_bits( 1450 struct drbd_conf *mdev, unsigned long s, unsigned long e); 1451 extern int drbd_bm_clear_bits( 1452 struct drbd_conf *mdev, unsigned long s, unsigned long e); 1453 extern int drbd_bm_count_bits( 1454 struct drbd_conf *mdev, const unsigned long s, const unsigned long e); 1455 /* bm_set_bits variant for use while holding drbd_bm_lock, 1456 * may process the whole bitmap in one go */ 1457 extern void _drbd_bm_set_bits(struct drbd_conf *mdev, 1458 const unsigned long s, const unsigned long e); 1459 extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr); 1460 extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr); 1461 extern int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local); 1462 extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local); 1463 extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local); 1464 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev, 1465 unsigned long al_enr); 1466 extern size_t drbd_bm_words(struct drbd_conf *mdev); 1467 extern unsigned long drbd_bm_bits(struct drbd_conf *mdev); 1468 extern sector_t drbd_bm_capacity(struct drbd_conf *mdev); 1469 1470 #define DRBD_END_OF_BITMAP (~(unsigned long)0) 1471 extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo); 1472 /* bm_find_next variants for use while you hold drbd_bm_lock() */ 1473 extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo); 1474 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo); 1475 extern unsigned long _drbd_bm_total_weight(struct drbd_conf *mdev); 1476 extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev); 1477 extern int drbd_bm_rs_done(struct drbd_conf *mdev); 1478 /* for receive_bitmap */ 1479 extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset, 1480 size_t number, unsigned long *buffer); 1481 /* for _drbd_send_bitmap */ 1482 extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset, 1483 size_t number, unsigned long *buffer); 1484 1485 extern void drbd_bm_lock(struct drbd_conf *mdev, char *why, enum bm_flag flags); 1486 extern void drbd_bm_unlock(struct drbd_conf *mdev); 1487 /* drbd_main.c */ 1488 1489 extern struct kmem_cache *drbd_request_cache; 1490 extern struct kmem_cache *drbd_ee_cache; /* epoch entries */ 1491 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */ 1492 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */ 1493 extern mempool_t *drbd_request_mempool; 1494 extern mempool_t *drbd_ee_mempool; 1495 1496 extern struct page *drbd_pp_pool; /* drbd's page pool */ 1497 extern spinlock_t drbd_pp_lock; 1498 extern int drbd_pp_vacant; 1499 extern wait_queue_head_t drbd_pp_wait; 1500 1501 extern rwlock_t global_state_lock; 1502 1503 extern struct drbd_conf *drbd_new_device(unsigned int minor); 1504 extern void drbd_free_mdev(struct drbd_conf *mdev); 1505 1506 extern int proc_details; 1507 1508 /* drbd_req */ 1509 extern void drbd_make_request(struct request_queue *q, struct bio *bio); 1510 extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req); 1511 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec); 1512 extern int is_valid_ar_handle(struct drbd_request *, sector_t); 1513 1514 1515 /* drbd_nl.c */ 1516 extern void drbd_suspend_io(struct drbd_conf *mdev); 1517 extern void drbd_resume_io(struct drbd_conf *mdev); 1518 extern char *ppsize(char *buf, unsigned long long size); 1519 extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int); 1520 enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 }; 1521 extern enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local); 1522 extern void resync_after_online_grow(struct drbd_conf *); 1523 extern void drbd_reconsider_max_bio_size(struct drbd_conf *mdev); 1524 extern enum drbd_state_rv drbd_set_role(struct drbd_conf *mdev, 1525 enum drbd_role new_role, 1526 int force); 1527 extern enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev); 1528 extern void drbd_try_outdate_peer_async(struct drbd_conf *mdev); 1529 extern int drbd_khelper(struct drbd_conf *mdev, char *cmd); 1530 1531 /* drbd_worker.c */ 1532 extern int drbd_worker(struct drbd_thread *thi); 1533 extern int drbd_alter_sa(struct drbd_conf *mdev, int na); 1534 extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side); 1535 extern void resume_next_sg(struct drbd_conf *mdev); 1536 extern void suspend_other_sg(struct drbd_conf *mdev); 1537 extern int drbd_resync_finished(struct drbd_conf *mdev); 1538 /* maybe rather drbd_main.c ? */ 1539 extern int drbd_md_sync_page_io(struct drbd_conf *mdev, 1540 struct drbd_backing_dev *bdev, sector_t sector, int rw); 1541 extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int); 1542 extern void drbd_rs_controller_reset(struct drbd_conf *mdev); 1543 1544 static inline void ov_oos_print(struct drbd_conf *mdev) 1545 { 1546 if (mdev->ov_last_oos_size) { 1547 dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n", 1548 (unsigned long long)mdev->ov_last_oos_start, 1549 (unsigned long)mdev->ov_last_oos_size); 1550 } 1551 mdev->ov_last_oos_size=0; 1552 } 1553 1554 1555 extern void drbd_csum_bio(struct drbd_conf *, struct crypto_hash *, struct bio *, void *); 1556 extern void drbd_csum_ee(struct drbd_conf *, struct crypto_hash *, struct drbd_epoch_entry *, void *); 1557 /* worker callbacks */ 1558 extern int w_req_cancel_conflict(struct drbd_conf *, struct drbd_work *, int); 1559 extern int w_read_retry_remote(struct drbd_conf *, struct drbd_work *, int); 1560 extern int w_e_end_data_req(struct drbd_conf *, struct drbd_work *, int); 1561 extern int w_e_end_rsdata_req(struct drbd_conf *, struct drbd_work *, int); 1562 extern int w_e_end_csum_rs_req(struct drbd_conf *, struct drbd_work *, int); 1563 extern int w_e_end_ov_reply(struct drbd_conf *, struct drbd_work *, int); 1564 extern int w_e_end_ov_req(struct drbd_conf *, struct drbd_work *, int); 1565 extern int w_ov_finished(struct drbd_conf *, struct drbd_work *, int); 1566 extern int w_resync_timer(struct drbd_conf *, struct drbd_work *, int); 1567 extern int w_resume_next_sg(struct drbd_conf *, struct drbd_work *, int); 1568 extern int w_send_write_hint(struct drbd_conf *, struct drbd_work *, int); 1569 extern int w_send_dblock(struct drbd_conf *, struct drbd_work *, int); 1570 extern int w_send_barrier(struct drbd_conf *, struct drbd_work *, int); 1571 extern int w_send_read_req(struct drbd_conf *, struct drbd_work *, int); 1572 extern int w_prev_work_done(struct drbd_conf *, struct drbd_work *, int); 1573 extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int); 1574 extern int w_restart_disk_io(struct drbd_conf *, struct drbd_work *, int); 1575 extern int w_send_oos(struct drbd_conf *, struct drbd_work *, int); 1576 extern int w_start_resync(struct drbd_conf *, struct drbd_work *, int); 1577 1578 extern void resync_timer_fn(unsigned long data); 1579 extern void start_resync_timer_fn(unsigned long data); 1580 1581 /* drbd_receiver.c */ 1582 extern int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector); 1583 extern int drbd_submit_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e, 1584 const unsigned rw, const int fault_type); 1585 extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list); 1586 extern struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev, 1587 u64 id, 1588 sector_t sector, 1589 unsigned int data_size, 1590 gfp_t gfp_mask) __must_hold(local); 1591 extern void drbd_free_some_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e, 1592 int is_net); 1593 #define drbd_free_ee(m,e) drbd_free_some_ee(m, e, 0) 1594 #define drbd_free_net_ee(m,e) drbd_free_some_ee(m, e, 1) 1595 extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev, 1596 struct list_head *head); 1597 extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev, 1598 struct list_head *head); 1599 extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled); 1600 extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed); 1601 extern void drbd_flush_workqueue(struct drbd_conf *mdev); 1602 extern void drbd_free_tl_hash(struct drbd_conf *mdev); 1603 1604 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to 1605 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */ 1606 static inline int drbd_setsockopt(struct socket *sock, int level, int optname, 1607 char __user *optval, int optlen) 1608 { 1609 int err; 1610 if (level == SOL_SOCKET) 1611 err = sock_setsockopt(sock, level, optname, optval, optlen); 1612 else 1613 err = sock->ops->setsockopt(sock, level, optname, optval, 1614 optlen); 1615 return err; 1616 } 1617 1618 static inline void drbd_tcp_cork(struct socket *sock) 1619 { 1620 int __user val = 1; 1621 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK, 1622 (char __user *)&val, sizeof(val)); 1623 } 1624 1625 static inline void drbd_tcp_uncork(struct socket *sock) 1626 { 1627 int __user val = 0; 1628 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK, 1629 (char __user *)&val, sizeof(val)); 1630 } 1631 1632 static inline void drbd_tcp_nodelay(struct socket *sock) 1633 { 1634 int __user val = 1; 1635 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY, 1636 (char __user *)&val, sizeof(val)); 1637 } 1638 1639 static inline void drbd_tcp_quickack(struct socket *sock) 1640 { 1641 int __user val = 2; 1642 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK, 1643 (char __user *)&val, sizeof(val)); 1644 } 1645 1646 void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo); 1647 1648 /* drbd_proc.c */ 1649 extern struct proc_dir_entry *drbd_proc; 1650 extern const struct file_operations drbd_proc_fops; 1651 extern const char *drbd_conn_str(enum drbd_conns s); 1652 extern const char *drbd_role_str(enum drbd_role s); 1653 1654 /* drbd_actlog.c */ 1655 extern void drbd_al_begin_io(struct drbd_conf *mdev, sector_t sector); 1656 extern void drbd_al_complete_io(struct drbd_conf *mdev, sector_t sector); 1657 extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector); 1658 extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector); 1659 extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector); 1660 extern void drbd_rs_cancel_all(struct drbd_conf *mdev); 1661 extern int drbd_rs_del_all(struct drbd_conf *mdev); 1662 extern void drbd_rs_failed_io(struct drbd_conf *mdev, 1663 sector_t sector, int size); 1664 extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *); 1665 extern void drbd_advance_rs_marks(struct drbd_conf *mdev, unsigned long still_to_go); 1666 extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector, 1667 int size, const char *file, const unsigned int line); 1668 #define drbd_set_in_sync(mdev, sector, size) \ 1669 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__) 1670 extern int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector, 1671 int size, const char *file, const unsigned int line); 1672 #define drbd_set_out_of_sync(mdev, sector, size) \ 1673 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__) 1674 extern void drbd_al_apply_to_bm(struct drbd_conf *mdev); 1675 extern void drbd_al_shrink(struct drbd_conf *mdev); 1676 1677 1678 /* drbd_nl.c */ 1679 1680 void drbd_nl_cleanup(void); 1681 int __init drbd_nl_init(void); 1682 void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state); 1683 void drbd_bcast_sync_progress(struct drbd_conf *mdev); 1684 void drbd_bcast_ee(struct drbd_conf *mdev, 1685 const char *reason, const int dgs, 1686 const char* seen_hash, const char* calc_hash, 1687 const struct drbd_epoch_entry* e); 1688 1689 1690 /** 1691 * DOC: DRBD State macros 1692 * 1693 * These macros are used to express state changes in easily readable form. 1694 * 1695 * The NS macros expand to a mask and a value, that can be bit ored onto the 1696 * current state as soon as the spinlock (req_lock) was taken. 1697 * 1698 * The _NS macros are used for state functions that get called with the 1699 * spinlock. These macros expand directly to the new state value. 1700 * 1701 * Besides the basic forms NS() and _NS() additional _?NS[23] are defined 1702 * to express state changes that affect more than one aspect of the state. 1703 * 1704 * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY) 1705 * Means that the network connection was established and that the peer 1706 * is in secondary role. 1707 */ 1708 #define role_MASK R_MASK 1709 #define peer_MASK R_MASK 1710 #define disk_MASK D_MASK 1711 #define pdsk_MASK D_MASK 1712 #define conn_MASK C_MASK 1713 #define susp_MASK 1 1714 #define user_isp_MASK 1 1715 #define aftr_isp_MASK 1 1716 #define susp_nod_MASK 1 1717 #define susp_fen_MASK 1 1718 1719 #define NS(T, S) \ 1720 ({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \ 1721 ({ union drbd_state val; val.i = 0; val.T = (S); val; }) 1722 #define NS2(T1, S1, T2, S2) \ 1723 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \ 1724 mask.T2 = T2##_MASK; mask; }), \ 1725 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \ 1726 val.T2 = (S2); val; }) 1727 #define NS3(T1, S1, T2, S2, T3, S3) \ 1728 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \ 1729 mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \ 1730 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \ 1731 val.T2 = (S2); val.T3 = (S3); val; }) 1732 1733 #define _NS(D, T, S) \ 1734 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; }) 1735 #define _NS2(D, T1, S1, T2, S2) \ 1736 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \ 1737 __ns.T2 = (S2); __ns; }) 1738 #define _NS3(D, T1, S1, T2, S2, T3, S3) \ 1739 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \ 1740 __ns.T2 = (S2); __ns.T3 = (S3); __ns; }) 1741 1742 /* 1743 * inline helper functions 1744 *************************/ 1745 1746 /* see also page_chain_add and friends in drbd_receiver.c */ 1747 static inline struct page *page_chain_next(struct page *page) 1748 { 1749 return (struct page *)page_private(page); 1750 } 1751 #define page_chain_for_each(page) \ 1752 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \ 1753 page = page_chain_next(page)) 1754 #define page_chain_for_each_safe(page, n) \ 1755 for (; page && ({ n = page_chain_next(page); 1; }); page = n) 1756 1757 static inline int drbd_bio_has_active_page(struct bio *bio) 1758 { 1759 struct bio_vec *bvec; 1760 int i; 1761 1762 __bio_for_each_segment(bvec, bio, i, 0) { 1763 if (page_count(bvec->bv_page) > 1) 1764 return 1; 1765 } 1766 1767 return 0; 1768 } 1769 1770 static inline int drbd_ee_has_active_page(struct drbd_epoch_entry *e) 1771 { 1772 struct page *page = e->pages; 1773 page_chain_for_each(page) { 1774 if (page_count(page) > 1) 1775 return 1; 1776 } 1777 return 0; 1778 } 1779 1780 1781 static inline void drbd_state_lock(struct drbd_conf *mdev) 1782 { 1783 wait_event(mdev->misc_wait, 1784 !test_and_set_bit(CLUSTER_ST_CHANGE, &mdev->flags)); 1785 } 1786 1787 static inline void drbd_state_unlock(struct drbd_conf *mdev) 1788 { 1789 clear_bit(CLUSTER_ST_CHANGE, &mdev->flags); 1790 wake_up(&mdev->misc_wait); 1791 } 1792 1793 static inline enum drbd_state_rv 1794 _drbd_set_state(struct drbd_conf *mdev, union drbd_state ns, 1795 enum chg_state_flags flags, struct completion *done) 1796 { 1797 enum drbd_state_rv rv; 1798 1799 read_lock(&global_state_lock); 1800 rv = __drbd_set_state(mdev, ns, flags, done); 1801 read_unlock(&global_state_lock); 1802 1803 return rv; 1804 } 1805 1806 /** 1807 * drbd_request_state() - Reqest a state change 1808 * @mdev: DRBD device. 1809 * @mask: mask of state bits to change. 1810 * @val: value of new state bits. 1811 * 1812 * This is the most graceful way of requesting a state change. It is verbose 1813 * quite verbose in case the state change is not possible, and all those 1814 * state changes are globally serialized. 1815 */ 1816 static inline int drbd_request_state(struct drbd_conf *mdev, 1817 union drbd_state mask, 1818 union drbd_state val) 1819 { 1820 return _drbd_request_state(mdev, mask, val, CS_VERBOSE + CS_ORDERED); 1821 } 1822 1823 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__) 1824 static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach, const char *where) 1825 { 1826 switch (mdev->ldev->dc.on_io_error) { 1827 case EP_PASS_ON: 1828 if (!forcedetach) { 1829 if (__ratelimit(&drbd_ratelimit_state)) 1830 dev_err(DEV, "Local IO failed in %s.\n", where); 1831 if (mdev->state.disk > D_INCONSISTENT) 1832 _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_HARD, NULL); 1833 break; 1834 } 1835 /* NOTE fall through to detach case if forcedetach set */ 1836 case EP_DETACH: 1837 case EP_CALL_HELPER: 1838 set_bit(WAS_IO_ERROR, &mdev->flags); 1839 if (mdev->state.disk > D_FAILED) { 1840 _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL); 1841 dev_err(DEV, 1842 "Local IO failed in %s. Detaching...\n", where); 1843 } 1844 break; 1845 } 1846 } 1847 1848 /** 1849 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers 1850 * @mdev: DRBD device. 1851 * @error: Error code passed to the IO completion callback 1852 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data 1853 * 1854 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED) 1855 */ 1856 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__) 1857 static inline void drbd_chk_io_error_(struct drbd_conf *mdev, 1858 int error, int forcedetach, const char *where) 1859 { 1860 if (error) { 1861 unsigned long flags; 1862 spin_lock_irqsave(&mdev->req_lock, flags); 1863 __drbd_chk_io_error_(mdev, forcedetach, where); 1864 spin_unlock_irqrestore(&mdev->req_lock, flags); 1865 } 1866 } 1867 1868 1869 /** 1870 * drbd_md_first_sector() - Returns the first sector number of the meta data area 1871 * @bdev: Meta data block device. 1872 * 1873 * BTW, for internal meta data, this happens to be the maximum capacity 1874 * we could agree upon with our peer node. 1875 */ 1876 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev) 1877 { 1878 switch (bdev->dc.meta_dev_idx) { 1879 case DRBD_MD_INDEX_INTERNAL: 1880 case DRBD_MD_INDEX_FLEX_INT: 1881 return bdev->md.md_offset + bdev->md.bm_offset; 1882 case DRBD_MD_INDEX_FLEX_EXT: 1883 default: 1884 return bdev->md.md_offset; 1885 } 1886 } 1887 1888 /** 1889 * drbd_md_last_sector() - Return the last sector number of the meta data area 1890 * @bdev: Meta data block device. 1891 */ 1892 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev) 1893 { 1894 switch (bdev->dc.meta_dev_idx) { 1895 case DRBD_MD_INDEX_INTERNAL: 1896 case DRBD_MD_INDEX_FLEX_INT: 1897 return bdev->md.md_offset + MD_AL_OFFSET - 1; 1898 case DRBD_MD_INDEX_FLEX_EXT: 1899 default: 1900 return bdev->md.md_offset + bdev->md.md_size_sect; 1901 } 1902 } 1903 1904 /* Returns the number of 512 byte sectors of the device */ 1905 static inline sector_t drbd_get_capacity(struct block_device *bdev) 1906 { 1907 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */ 1908 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0; 1909 } 1910 1911 /** 1912 * drbd_get_max_capacity() - Returns the capacity we announce to out peer 1913 * @bdev: Meta data block device. 1914 * 1915 * returns the capacity we announce to out peer. we clip ourselves at the 1916 * various MAX_SECTORS, because if we don't, current implementation will 1917 * oops sooner or later 1918 */ 1919 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev) 1920 { 1921 sector_t s; 1922 switch (bdev->dc.meta_dev_idx) { 1923 case DRBD_MD_INDEX_INTERNAL: 1924 case DRBD_MD_INDEX_FLEX_INT: 1925 s = drbd_get_capacity(bdev->backing_bdev) 1926 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX, 1927 drbd_md_first_sector(bdev)) 1928 : 0; 1929 break; 1930 case DRBD_MD_INDEX_FLEX_EXT: 1931 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX, 1932 drbd_get_capacity(bdev->backing_bdev)); 1933 /* clip at maximum size the meta device can support */ 1934 s = min_t(sector_t, s, 1935 BM_EXT_TO_SECT(bdev->md.md_size_sect 1936 - bdev->md.bm_offset)); 1937 break; 1938 default: 1939 s = min_t(sector_t, DRBD_MAX_SECTORS, 1940 drbd_get_capacity(bdev->backing_bdev)); 1941 } 1942 return s; 1943 } 1944 1945 /** 1946 * drbd_md_ss__() - Return the sector number of our meta data super block 1947 * @mdev: DRBD device. 1948 * @bdev: Meta data block device. 1949 */ 1950 static inline sector_t drbd_md_ss__(struct drbd_conf *mdev, 1951 struct drbd_backing_dev *bdev) 1952 { 1953 switch (bdev->dc.meta_dev_idx) { 1954 default: /* external, some index */ 1955 return MD_RESERVED_SECT * bdev->dc.meta_dev_idx; 1956 case DRBD_MD_INDEX_INTERNAL: 1957 /* with drbd08, internal meta data is always "flexible" */ 1958 case DRBD_MD_INDEX_FLEX_INT: 1959 /* sizeof(struct md_on_disk_07) == 4k 1960 * position: last 4k aligned block of 4k size */ 1961 if (!bdev->backing_bdev) { 1962 if (__ratelimit(&drbd_ratelimit_state)) { 1963 dev_err(DEV, "bdev->backing_bdev==NULL\n"); 1964 dump_stack(); 1965 } 1966 return 0; 1967 } 1968 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) 1969 - MD_AL_OFFSET; 1970 case DRBD_MD_INDEX_FLEX_EXT: 1971 return 0; 1972 } 1973 } 1974 1975 static inline void 1976 drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w) 1977 { 1978 unsigned long flags; 1979 spin_lock_irqsave(&q->q_lock, flags); 1980 list_add(&w->list, &q->q); 1981 up(&q->s); /* within the spinlock, 1982 see comment near end of drbd_worker() */ 1983 spin_unlock_irqrestore(&q->q_lock, flags); 1984 } 1985 1986 static inline void 1987 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w) 1988 { 1989 unsigned long flags; 1990 spin_lock_irqsave(&q->q_lock, flags); 1991 list_add_tail(&w->list, &q->q); 1992 up(&q->s); /* within the spinlock, 1993 see comment near end of drbd_worker() */ 1994 spin_unlock_irqrestore(&q->q_lock, flags); 1995 } 1996 1997 static inline void wake_asender(struct drbd_conf *mdev) 1998 { 1999 if (test_bit(SIGNAL_ASENDER, &mdev->flags)) 2000 force_sig(DRBD_SIG, mdev->asender.task); 2001 } 2002 2003 static inline void request_ping(struct drbd_conf *mdev) 2004 { 2005 set_bit(SEND_PING, &mdev->flags); 2006 wake_asender(mdev); 2007 } 2008 2009 static inline int drbd_send_short_cmd(struct drbd_conf *mdev, 2010 enum drbd_packets cmd) 2011 { 2012 struct p_header80 h; 2013 return drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, &h, sizeof(h)); 2014 } 2015 2016 static inline int drbd_send_ping(struct drbd_conf *mdev) 2017 { 2018 struct p_header80 h; 2019 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING, &h, sizeof(h)); 2020 } 2021 2022 static inline int drbd_send_ping_ack(struct drbd_conf *mdev) 2023 { 2024 struct p_header80 h; 2025 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING_ACK, &h, sizeof(h)); 2026 } 2027 2028 static inline void drbd_thread_stop(struct drbd_thread *thi) 2029 { 2030 _drbd_thread_stop(thi, false, true); 2031 } 2032 2033 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi) 2034 { 2035 _drbd_thread_stop(thi, false, false); 2036 } 2037 2038 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi) 2039 { 2040 _drbd_thread_stop(thi, true, false); 2041 } 2042 2043 /* counts how many answer packets packets we expect from our peer, 2044 * for either explicit application requests, 2045 * or implicit barrier packets as necessary. 2046 * increased: 2047 * w_send_barrier 2048 * _req_mod(req, queue_for_net_write or queue_for_net_read); 2049 * it is much easier and equally valid to count what we queue for the 2050 * worker, even before it actually was queued or send. 2051 * (drbd_make_request_common; recovery path on read io-error) 2052 * decreased: 2053 * got_BarrierAck (respective tl_clear, tl_clear_barrier) 2054 * _req_mod(req, data_received) 2055 * [from receive_DataReply] 2056 * _req_mod(req, write_acked_by_peer or recv_acked_by_peer or neg_acked) 2057 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)] 2058 * for some reason it is NOT decreased in got_NegAck, 2059 * but in the resulting cleanup code from report_params. 2060 * we should try to remember the reason for that... 2061 * _req_mod(req, send_failed or send_canceled) 2062 * _req_mod(req, connection_lost_while_pending) 2063 * [from tl_clear_barrier] 2064 */ 2065 static inline void inc_ap_pending(struct drbd_conf *mdev) 2066 { 2067 atomic_inc(&mdev->ap_pending_cnt); 2068 } 2069 2070 #define ERR_IF_CNT_IS_NEGATIVE(which) \ 2071 if (atomic_read(&mdev->which) < 0) \ 2072 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \ 2073 __func__ , __LINE__ , \ 2074 atomic_read(&mdev->which)) 2075 2076 #define dec_ap_pending(mdev) do { \ 2077 typecheck(struct drbd_conf *, mdev); \ 2078 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \ 2079 wake_up(&mdev->misc_wait); \ 2080 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0) 2081 2082 /* counts how many resync-related answers we still expect from the peer 2083 * increase decrease 2084 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY) 2085 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER) 2086 * (or P_NEG_ACK with ID_SYNCER) 2087 */ 2088 static inline void inc_rs_pending(struct drbd_conf *mdev) 2089 { 2090 atomic_inc(&mdev->rs_pending_cnt); 2091 } 2092 2093 #define dec_rs_pending(mdev) do { \ 2094 typecheck(struct drbd_conf *, mdev); \ 2095 atomic_dec(&mdev->rs_pending_cnt); \ 2096 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0) 2097 2098 /* counts how many answers we still need to send to the peer. 2099 * increased on 2100 * receive_Data unless protocol A; 2101 * we need to send a P_RECV_ACK (proto B) 2102 * or P_WRITE_ACK (proto C) 2103 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK 2104 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA 2105 * receive_Barrier_* we need to send a P_BARRIER_ACK 2106 */ 2107 static inline void inc_unacked(struct drbd_conf *mdev) 2108 { 2109 atomic_inc(&mdev->unacked_cnt); 2110 } 2111 2112 #define dec_unacked(mdev) do { \ 2113 typecheck(struct drbd_conf *, mdev); \ 2114 atomic_dec(&mdev->unacked_cnt); \ 2115 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0) 2116 2117 #define sub_unacked(mdev, n) do { \ 2118 typecheck(struct drbd_conf *, mdev); \ 2119 atomic_sub(n, &mdev->unacked_cnt); \ 2120 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0) 2121 2122 2123 static inline void put_net_conf(struct drbd_conf *mdev) 2124 { 2125 if (atomic_dec_and_test(&mdev->net_cnt)) 2126 wake_up(&mdev->net_cnt_wait); 2127 } 2128 2129 /** 2130 * get_net_conf() - Increase ref count on mdev->net_conf; Returns 0 if nothing there 2131 * @mdev: DRBD device. 2132 * 2133 * You have to call put_net_conf() when finished working with mdev->net_conf. 2134 */ 2135 static inline int get_net_conf(struct drbd_conf *mdev) 2136 { 2137 int have_net_conf; 2138 2139 atomic_inc(&mdev->net_cnt); 2140 have_net_conf = mdev->state.conn >= C_UNCONNECTED; 2141 if (!have_net_conf) 2142 put_net_conf(mdev); 2143 return have_net_conf; 2144 } 2145 2146 /** 2147 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev 2148 * @M: DRBD device. 2149 * 2150 * You have to call put_ldev() when finished working with mdev->ldev. 2151 */ 2152 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT)) 2153 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS)) 2154 2155 static inline void put_ldev(struct drbd_conf *mdev) 2156 { 2157 int i = atomic_dec_return(&mdev->local_cnt); 2158 2159 /* This may be called from some endio handler, 2160 * so we must not sleep here. */ 2161 2162 __release(local); 2163 D_ASSERT(i >= 0); 2164 if (i == 0) { 2165 if (mdev->state.disk == D_DISKLESS) 2166 /* even internal references gone, safe to destroy */ 2167 drbd_ldev_destroy(mdev); 2168 if (mdev->state.disk == D_FAILED) 2169 /* all application IO references gone. */ 2170 drbd_go_diskless(mdev); 2171 wake_up(&mdev->misc_wait); 2172 } 2173 } 2174 2175 #ifndef __CHECKER__ 2176 static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins) 2177 { 2178 int io_allowed; 2179 2180 /* never get a reference while D_DISKLESS */ 2181 if (mdev->state.disk == D_DISKLESS) 2182 return 0; 2183 2184 atomic_inc(&mdev->local_cnt); 2185 io_allowed = (mdev->state.disk >= mins); 2186 if (!io_allowed) 2187 put_ldev(mdev); 2188 return io_allowed; 2189 } 2190 #else 2191 extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins); 2192 #endif 2193 2194 /* you must have an "get_ldev" reference */ 2195 static inline void drbd_get_syncer_progress(struct drbd_conf *mdev, 2196 unsigned long *bits_left, unsigned int *per_mil_done) 2197 { 2198 /* this is to break it at compile time when we change that, in case we 2199 * want to support more than (1<<32) bits on a 32bit arch. */ 2200 typecheck(unsigned long, mdev->rs_total); 2201 2202 /* note: both rs_total and rs_left are in bits, i.e. in 2203 * units of BM_BLOCK_SIZE. 2204 * for the percentage, we don't care. */ 2205 2206 if (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T) 2207 *bits_left = mdev->ov_left; 2208 else 2209 *bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed; 2210 /* >> 10 to prevent overflow, 2211 * +1 to prevent division by zero */ 2212 if (*bits_left > mdev->rs_total) { 2213 /* doh. maybe a logic bug somewhere. 2214 * may also be just a race condition 2215 * between this and a disconnect during sync. 2216 * for now, just prevent in-kernel buffer overflow. 2217 */ 2218 smp_rmb(); 2219 dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n", 2220 drbd_conn_str(mdev->state.conn), 2221 *bits_left, mdev->rs_total, mdev->rs_failed); 2222 *per_mil_done = 0; 2223 } else { 2224 /* Make sure the division happens in long context. 2225 * We allow up to one petabyte storage right now, 2226 * at a granularity of 4k per bit that is 2**38 bits. 2227 * After shift right and multiplication by 1000, 2228 * this should still fit easily into a 32bit long, 2229 * so we don't need a 64bit division on 32bit arch. 2230 * Note: currently we don't support such large bitmaps on 32bit 2231 * arch anyways, but no harm done to be prepared for it here. 2232 */ 2233 unsigned int shift = mdev->rs_total >= (1ULL << 32) ? 16 : 10; 2234 unsigned long left = *bits_left >> shift; 2235 unsigned long total = 1UL + (mdev->rs_total >> shift); 2236 unsigned long tmp = 1000UL - left * 1000UL/total; 2237 *per_mil_done = tmp; 2238 } 2239 } 2240 2241 2242 /* this throttles on-the-fly application requests 2243 * according to max_buffers settings; 2244 * maybe re-implement using semaphores? */ 2245 static inline int drbd_get_max_buffers(struct drbd_conf *mdev) 2246 { 2247 int mxb = 1000000; /* arbitrary limit on open requests */ 2248 if (get_net_conf(mdev)) { 2249 mxb = mdev->net_conf->max_buffers; 2250 put_net_conf(mdev); 2251 } 2252 return mxb; 2253 } 2254 2255 static inline int drbd_state_is_stable(struct drbd_conf *mdev) 2256 { 2257 union drbd_state s = mdev->state; 2258 2259 /* DO NOT add a default clause, we want the compiler to warn us 2260 * for any newly introduced state we may have forgotten to add here */ 2261 2262 switch ((enum drbd_conns)s.conn) { 2263 /* new io only accepted when there is no connection, ... */ 2264 case C_STANDALONE: 2265 case C_WF_CONNECTION: 2266 /* ... or there is a well established connection. */ 2267 case C_CONNECTED: 2268 case C_SYNC_SOURCE: 2269 case C_SYNC_TARGET: 2270 case C_VERIFY_S: 2271 case C_VERIFY_T: 2272 case C_PAUSED_SYNC_S: 2273 case C_PAUSED_SYNC_T: 2274 case C_AHEAD: 2275 case C_BEHIND: 2276 /* transitional states, IO allowed */ 2277 case C_DISCONNECTING: 2278 case C_UNCONNECTED: 2279 case C_TIMEOUT: 2280 case C_BROKEN_PIPE: 2281 case C_NETWORK_FAILURE: 2282 case C_PROTOCOL_ERROR: 2283 case C_TEAR_DOWN: 2284 case C_WF_REPORT_PARAMS: 2285 case C_STARTING_SYNC_S: 2286 case C_STARTING_SYNC_T: 2287 break; 2288 2289 /* Allow IO in BM exchange states with new protocols */ 2290 case C_WF_BITMAP_S: 2291 if (mdev->agreed_pro_version < 96) 2292 return 0; 2293 break; 2294 2295 /* no new io accepted in these states */ 2296 case C_WF_BITMAP_T: 2297 case C_WF_SYNC_UUID: 2298 case C_MASK: 2299 /* not "stable" */ 2300 return 0; 2301 } 2302 2303 switch ((enum drbd_disk_state)s.disk) { 2304 case D_DISKLESS: 2305 case D_INCONSISTENT: 2306 case D_OUTDATED: 2307 case D_CONSISTENT: 2308 case D_UP_TO_DATE: 2309 /* disk state is stable as well. */ 2310 break; 2311 2312 /* no new io accepted during tansitional states */ 2313 case D_ATTACHING: 2314 case D_FAILED: 2315 case D_NEGOTIATING: 2316 case D_UNKNOWN: 2317 case D_MASK: 2318 /* not "stable" */ 2319 return 0; 2320 } 2321 2322 return 1; 2323 } 2324 2325 static inline int is_susp(union drbd_state s) 2326 { 2327 return s.susp || s.susp_nod || s.susp_fen; 2328 } 2329 2330 static inline bool may_inc_ap_bio(struct drbd_conf *mdev) 2331 { 2332 int mxb = drbd_get_max_buffers(mdev); 2333 2334 if (is_susp(mdev->state)) 2335 return false; 2336 if (test_bit(SUSPEND_IO, &mdev->flags)) 2337 return false; 2338 2339 /* to avoid potential deadlock or bitmap corruption, 2340 * in various places, we only allow new application io 2341 * to start during "stable" states. */ 2342 2343 /* no new io accepted when attaching or detaching the disk */ 2344 if (!drbd_state_is_stable(mdev)) 2345 return false; 2346 2347 /* since some older kernels don't have atomic_add_unless, 2348 * and we are within the spinlock anyways, we have this workaround. */ 2349 if (atomic_read(&mdev->ap_bio_cnt) > mxb) 2350 return false; 2351 if (test_bit(BITMAP_IO, &mdev->flags)) 2352 return false; 2353 return true; 2354 } 2355 2356 static inline bool inc_ap_bio_cond(struct drbd_conf *mdev, int count) 2357 { 2358 bool rv = false; 2359 2360 spin_lock_irq(&mdev->req_lock); 2361 rv = may_inc_ap_bio(mdev); 2362 if (rv) 2363 atomic_add(count, &mdev->ap_bio_cnt); 2364 spin_unlock_irq(&mdev->req_lock); 2365 2366 return rv; 2367 } 2368 2369 static inline void inc_ap_bio(struct drbd_conf *mdev, int count) 2370 { 2371 /* we wait here 2372 * as long as the device is suspended 2373 * until the bitmap is no longer on the fly during connection 2374 * handshake as long as we would exeed the max_buffer limit. 2375 * 2376 * to avoid races with the reconnect code, 2377 * we need to atomic_inc within the spinlock. */ 2378 2379 wait_event(mdev->misc_wait, inc_ap_bio_cond(mdev, count)); 2380 } 2381 2382 static inline void dec_ap_bio(struct drbd_conf *mdev) 2383 { 2384 int mxb = drbd_get_max_buffers(mdev); 2385 int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt); 2386 2387 D_ASSERT(ap_bio >= 0); 2388 /* this currently does wake_up for every dec_ap_bio! 2389 * maybe rather introduce some type of hysteresis? 2390 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */ 2391 if (ap_bio < mxb) 2392 wake_up(&mdev->misc_wait); 2393 if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) { 2394 if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags)) 2395 drbd_queue_work(&mdev->data.work, &mdev->bm_io_work.w); 2396 } 2397 } 2398 2399 static inline int drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val) 2400 { 2401 int changed = mdev->ed_uuid != val; 2402 mdev->ed_uuid = val; 2403 return changed; 2404 } 2405 2406 static inline int seq_cmp(u32 a, u32 b) 2407 { 2408 /* we assume wrap around at 32bit. 2409 * for wrap around at 24bit (old atomic_t), 2410 * we'd have to 2411 * a <<= 8; b <<= 8; 2412 */ 2413 return (s32)(a) - (s32)(b); 2414 } 2415 #define seq_lt(a, b) (seq_cmp((a), (b)) < 0) 2416 #define seq_gt(a, b) (seq_cmp((a), (b)) > 0) 2417 #define seq_ge(a, b) (seq_cmp((a), (b)) >= 0) 2418 #define seq_le(a, b) (seq_cmp((a), (b)) <= 0) 2419 /* CAUTION: please no side effects in arguments! */ 2420 #define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b))) 2421 2422 static inline void update_peer_seq(struct drbd_conf *mdev, unsigned int new_seq) 2423 { 2424 unsigned int m; 2425 spin_lock(&mdev->peer_seq_lock); 2426 m = seq_max(mdev->peer_seq, new_seq); 2427 mdev->peer_seq = m; 2428 spin_unlock(&mdev->peer_seq_lock); 2429 if (m == new_seq) 2430 wake_up(&mdev->seq_wait); 2431 } 2432 2433 static inline void drbd_update_congested(struct drbd_conf *mdev) 2434 { 2435 struct sock *sk = mdev->data.socket->sk; 2436 if (sk->sk_wmem_queued > sk->sk_sndbuf * 4 / 5) 2437 set_bit(NET_CONGESTED, &mdev->flags); 2438 } 2439 2440 static inline int drbd_queue_order_type(struct drbd_conf *mdev) 2441 { 2442 /* sorry, we currently have no working implementation 2443 * of distributed TCQ stuff */ 2444 #ifndef QUEUE_ORDERED_NONE 2445 #define QUEUE_ORDERED_NONE 0 2446 #endif 2447 return QUEUE_ORDERED_NONE; 2448 } 2449 2450 static inline void drbd_md_flush(struct drbd_conf *mdev) 2451 { 2452 int r; 2453 2454 if (test_bit(MD_NO_FUA, &mdev->flags)) 2455 return; 2456 2457 r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL); 2458 if (r) { 2459 set_bit(MD_NO_FUA, &mdev->flags); 2460 dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r); 2461 } 2462 } 2463 2464 #endif 2465