4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
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>.
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)
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.
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.
29 #include <linux/compiler.h>
30 #include <linux/types.h>
31 #include <linux/version.h>
32 #include <linux/list.h>
33 #include <linux/sched.h>
34 #include <linux/bitops.h>
35 #include <linux/slab.h>
36 #include <linux/crypto.h>
37 #include <linux/ratelimit.h>
38 #include <linux/tcp.h>
39 #include <linux/mutex.h>
40 #include <linux/major.h>
41 #include <linux/blkdev.h>
42 #include <linux/genhd.h>
44 #include <linux/lru_cache.h>
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")))
52 # define __protected_by(x)
53 # define __protected_read_by(x)
54 # define __protected_write_by(x)
55 # define __must_hold(x)
58 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
60 /* module parameter, defined in drbd_main.c */
61 extern unsigned int minor_count;
62 extern int disable_sendpage;
64 extern unsigned int cn_idx;
66 #ifdef CONFIG_DRBD_FAULT_INJECTION
67 extern int enable_faults;
68 extern int fault_rate;
69 extern int fault_devs;
72 extern char usermode_helper[];
75 /* I don't remember why XCPU ...
76 * This is used to wake the asender,
77 * and to interrupt sending the sending task
80 #define DRBD_SIG SIGXCPU
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.
87 #define DRBD_SIGKILL SIGHUP
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 :( */
94 #define ID_IN_SYNC (4711ULL)
95 #define ID_OUT_OF_SYNC (4712ULL)
97 #define ID_SYNCER (-1ULL)
99 #define is_syncer_block_id(id) ((id) == ID_SYNCER)
104 /* to shorten dev_warn(DEV, "msg"); and relatives statements */
105 #define DEV (disk_to_dev(mdev->vdisk))
107 #define D_ASSERT(exp) if (!(exp)) \
108 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
110 #define ERR_IF(exp) if (({ \
111 int _b = (exp) != 0; \
112 if (_b) dev_err(DEV, "ASSERT FAILED: %s: (%s) in %s:%d\n", \
113 __func__, #exp, __FILE__, __LINE__); \
117 /* Defines to control fault insertion */
119 DRBD_FAULT_MD_WR = 0, /* meta data write */
120 DRBD_FAULT_MD_RD = 1, /* read */
121 DRBD_FAULT_RS_WR = 2, /* resync */
122 DRBD_FAULT_RS_RD = 3,
123 DRBD_FAULT_DT_WR = 4, /* data */
124 DRBD_FAULT_DT_RD = 5,
125 DRBD_FAULT_DT_RA = 6, /* data read ahead */
126 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */
127 DRBD_FAULT_AL_EE = 8, /* alloc ee */
128 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
134 _drbd_insert_fault(struct drbd_conf *mdev, unsigned int type);
137 drbd_insert_fault(struct drbd_conf *mdev, unsigned int type) {
138 #ifdef CONFIG_DRBD_FAULT_INJECTION
140 (enable_faults & (1<<type)) &&
141 _drbd_insert_fault(mdev, type);
147 /* integer division, round _UP_ to the next integer */
148 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
149 /* usual integer division */
150 #define div_floor(A, B) ((A)/(B))
152 /* drbd_meta-data.c (still in drbd_main.c) */
153 /* 4th incarnation of the disk layout. */
154 #define DRBD_MD_MAGIC (DRBD_MAGIC+4)
156 extern struct drbd_conf **minor_table;
157 extern struct ratelimit_state drbd_ratelimit_state;
161 /* receiver (data socket) */
163 P_DATA_REPLY = 0x01, /* Response to P_DATA_REQUEST */
164 P_RS_DATA_REPLY = 0x02, /* Response to P_RS_DATA_REQUEST */
167 P_BECOME_SYNC_TARGET = 0x05,
168 P_BECOME_SYNC_SOURCE = 0x06,
169 P_UNPLUG_REMOTE = 0x07, /* Used at various times to hint the peer */
170 P_DATA_REQUEST = 0x08, /* Used to ask for a data block */
171 P_RS_DATA_REQUEST = 0x09, /* Used to ask for a data block for resync */
178 P_AUTH_CHALLENGE = 0x10,
179 P_AUTH_RESPONSE = 0x11,
180 P_STATE_CHG_REQ = 0x12,
182 /* asender (meta socket */
185 P_RECV_ACK = 0x15, /* Used in protocol B */
186 P_WRITE_ACK = 0x16, /* Used in protocol C */
187 P_RS_WRITE_ACK = 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
188 P_DISCARD_ACK = 0x18, /* Used in proto C, two-primaries conflict detection */
189 P_NEG_ACK = 0x19, /* Sent if local disk is unusable */
190 P_NEG_DREPLY = 0x1a, /* Local disk is broken... */
191 P_NEG_RS_DREPLY = 0x1b, /* Local disk is broken... */
192 P_BARRIER_ACK = 0x1c,
193 P_STATE_CHG_REPLY = 0x1d,
195 /* "new" commands, no longer fitting into the ordering scheme above */
197 P_OV_REQUEST = 0x1e, /* data socket */
199 P_OV_RESULT = 0x20, /* meta socket */
200 P_CSUM_RS_REQUEST = 0x21, /* data socket */
201 P_RS_IS_IN_SYNC = 0x22, /* meta socket */
202 P_SYNC_PARAM89 = 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
203 P_COMPRESSED_BITMAP = 0x24, /* compressed or otherwise encoded bitmap transfer */
204 /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */
205 /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */
206 P_DELAY_PROBE = 0x27, /* is used on BOTH sockets */
207 P_OUT_OF_SYNC = 0x28, /* Mark as out of sync (Outrunning), data socket */
210 P_MAY_IGNORE = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
211 P_MAX_OPT_CMD = 0x101,
213 /* special command ids for handshake */
215 P_HAND_SHAKE_M = 0xfff1, /* First Packet on the MetaSock */
216 P_HAND_SHAKE_S = 0xfff2, /* First Packet on the Socket */
218 P_HAND_SHAKE = 0xfffe /* FIXED for the next century! */
221 static inline const char *cmdname(enum drbd_packets cmd)
223 /* THINK may need to become several global tables
224 * when we want to support more than
226 static const char *cmdnames[] = {
228 [P_DATA_REPLY] = "DataReply",
229 [P_RS_DATA_REPLY] = "RSDataReply",
230 [P_BARRIER] = "Barrier",
231 [P_BITMAP] = "ReportBitMap",
232 [P_BECOME_SYNC_TARGET] = "BecomeSyncTarget",
233 [P_BECOME_SYNC_SOURCE] = "BecomeSyncSource",
234 [P_UNPLUG_REMOTE] = "UnplugRemote",
235 [P_DATA_REQUEST] = "DataRequest",
236 [P_RS_DATA_REQUEST] = "RSDataRequest",
237 [P_SYNC_PARAM] = "SyncParam",
238 [P_SYNC_PARAM89] = "SyncParam89",
239 [P_PROTOCOL] = "ReportProtocol",
240 [P_UUIDS] = "ReportUUIDs",
241 [P_SIZES] = "ReportSizes",
242 [P_STATE] = "ReportState",
243 [P_SYNC_UUID] = "ReportSyncUUID",
244 [P_AUTH_CHALLENGE] = "AuthChallenge",
245 [P_AUTH_RESPONSE] = "AuthResponse",
247 [P_PING_ACK] = "PingAck",
248 [P_RECV_ACK] = "RecvAck",
249 [P_WRITE_ACK] = "WriteAck",
250 [P_RS_WRITE_ACK] = "RSWriteAck",
251 [P_DISCARD_ACK] = "DiscardAck",
252 [P_NEG_ACK] = "NegAck",
253 [P_NEG_DREPLY] = "NegDReply",
254 [P_NEG_RS_DREPLY] = "NegRSDReply",
255 [P_BARRIER_ACK] = "BarrierAck",
256 [P_STATE_CHG_REQ] = "StateChgRequest",
257 [P_STATE_CHG_REPLY] = "StateChgReply",
258 [P_OV_REQUEST] = "OVRequest",
259 [P_OV_REPLY] = "OVReply",
260 [P_OV_RESULT] = "OVResult",
261 [P_CSUM_RS_REQUEST] = "CsumRSRequest",
262 [P_RS_IS_IN_SYNC] = "CsumRSIsInSync",
263 [P_COMPRESSED_BITMAP] = "CBitmap",
264 [P_DELAY_PROBE] = "DelayProbe",
265 [P_OUT_OF_SYNC] = "OutOfSync",
269 if (cmd == P_HAND_SHAKE_M)
271 if (cmd == P_HAND_SHAKE_S)
273 if (cmd == P_HAND_SHAKE)
275 if (cmd >= P_MAX_CMD)
277 return cmdnames[cmd];
280 /* for sending/receiving the bitmap,
281 * possibly in some encoding scheme */
284 * stores total bits and long words
285 * of the bitmap, so we don't need to
286 * call the accessor functions over and again. */
287 unsigned long bm_bits;
288 unsigned long bm_words;
289 /* during xfer, current position within the bitmap */
290 unsigned long bit_offset;
291 unsigned long word_offset;
293 /* statistics; index: (h->command == P_BITMAP) */
298 extern void INFO_bm_xfer_stats(struct drbd_conf *mdev,
299 const char *direction, struct bm_xfer_ctx *c);
301 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
303 /* word_offset counts "native long words" (32 or 64 bit),
305 * Encoded packet may end at an unaligned bit offset.
306 * In case a fallback clear text packet is transmitted in
307 * between, we adjust this offset back to the last 64bit
308 * aligned "native long word", which makes coding and decoding
309 * the plain text bitmap much more convenient. */
310 #if BITS_PER_LONG == 64
311 c->word_offset = c->bit_offset >> 6;
312 #elif BITS_PER_LONG == 32
313 c->word_offset = c->bit_offset >> 5;
314 c->word_offset &= ~(1UL);
316 # error "unsupported BITS_PER_LONG"
321 #define __packed __attribute__((packed))
324 /* This is the layout for a packet on the wire.
325 * The byteorder is the network byte order.
326 * (except block_id and barrier fields.
327 * these are pointers to local structs
328 * and have no relevance for the partner,
329 * which just echoes them as received.)
331 * NOTE that the payload starts at a long aligned offset,
332 * regardless of 32 or 64 bit arch!
337 u16 length; /* bytes of data after this header */
341 /* Header for big packets, Used for data packets exceeding 64kB */
343 u16 magic; /* use DRBD_MAGIC_BIG here */
345 u32 length; /* Use only 24 bits of that. Ignore the highest 8 bit. */
350 struct p_header80 h80;
351 struct p_header95 h95;
355 * short commands, packets without payload, plain p_header:
358 * P_BECOME_SYNC_TARGET
359 * P_BECOME_SYNC_SOURCE
364 * commands with out-of-struct payload:
365 * P_BITMAP (no additional fields)
366 * P_DATA, P_DATA_REPLY (see p_data)
367 * P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
370 /* these defines must not be changed without changing the protocol version */
371 #define DP_HARDBARRIER 1 /* depricated */
372 #define DP_RW_SYNC 2 /* equals REQ_SYNC */
373 #define DP_MAY_SET_IN_SYNC 4
374 #define DP_UNPLUG 8 /* not used anymore */
375 #define DP_FUA 16 /* equals REQ_FUA */
376 #define DP_FLUSH 32 /* equals REQ_FLUSH */
377 #define DP_DISCARD 64 /* equals REQ_DISCARD */
381 u64 sector; /* 64 bits sector number */
382 u64 block_id; /* to identify the request in protocol B&C */
388 * commands which share a struct:
390 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
391 * P_DISCARD_ACK (proto C, two-primaries conflict detection)
393 * P_DATA_REQUEST, P_RS_DATA_REQUEST
396 struct p_header80 head;
405 struct p_header80 head;
409 u32 pad; /* to multiple of 8 Byte */
413 * commands with their own struct for additional fields:
422 struct p_header80 head; /* 8 bytes */
427 /* should be more than enough for future enhancements
428 * for now, feature_flags and the reserverd array shall be zero.
434 /* 80 bytes, FIXED for the next century */
437 struct p_header80 head;
438 u32 barrier; /* barrier number _handle_ only */
439 u32 pad; /* to multiple of 8 Byte */
442 struct p_barrier_ack {
443 struct p_header80 head;
449 struct p_header80 head;
452 /* Since protocol version 88 and higher. */
456 struct p_rs_param_89 {
457 struct p_header80 head;
459 /* protocol version 89: */
460 char verify_alg[SHARED_SECRET_MAX];
461 char csums_alg[SHARED_SECRET_MAX];
464 struct p_rs_param_95 {
465 struct p_header80 head;
467 char verify_alg[SHARED_SECRET_MAX];
468 char csums_alg[SHARED_SECRET_MAX];
475 enum drbd_conn_flags {
481 struct p_header80 head;
489 /* Since protocol version 87 and higher. */
490 char integrity_alg[0];
495 struct p_header80 head;
496 u64 uuid[UI_EXTENDED_SIZE];
500 struct p_header80 head;
505 struct p_header80 head;
506 u64 d_size; /* size of disk */
507 u64 u_size; /* user requested size */
508 u64 c_size; /* current exported size */
509 u32 max_bio_size; /* Maximal size of a BIO */
510 u16 queue_order_type; /* not yet implemented in DRBD*/
511 u16 dds_flags; /* use enum dds_flags here. */
515 struct p_header80 head;
520 struct p_header80 head;
525 struct p_req_state_reply {
526 struct p_header80 head;
530 struct p_drbd06_param {
541 struct p_header80 head;
547 struct p_block_desc {
548 struct p_header80 head;
551 u32 pad; /* to multiple of 8 Byte */
554 /* Valid values for the encoding field.
555 * Bump proto version when changing this. */
556 enum drbd_bitmap_code {
557 /* RLE_VLI_Bytes = 0,
558 * and other bit variants had been defined during
559 * algorithm evaluation. */
563 struct p_compressed_bm {
564 struct p_header80 head;
565 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
566 * (encoding & 0x80): polarity (set/unset) of first runlength
567 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
568 * used to pad up to head.length bytes
575 struct p_delay_probe93 {
576 struct p_header80 head;
577 u32 seq_num; /* sequence number to match the two probe packets */
578 u32 offset; /* usecs the probe got sent after the reference time point */
581 /* DCBP: Drbd Compressed Bitmap Packet ... */
582 static inline enum drbd_bitmap_code
583 DCBP_get_code(struct p_compressed_bm *p)
585 return (enum drbd_bitmap_code)(p->encoding & 0x0f);
589 DCBP_set_code(struct p_compressed_bm *p, enum drbd_bitmap_code code)
592 p->encoding = (p->encoding & ~0xf) | code;
596 DCBP_get_start(struct p_compressed_bm *p)
598 return (p->encoding & 0x80) != 0;
602 DCBP_set_start(struct p_compressed_bm *p, int set)
604 p->encoding = (p->encoding & ~0x80) | (set ? 0x80 : 0);
608 DCBP_get_pad_bits(struct p_compressed_bm *p)
610 return (p->encoding >> 4) & 0x7;
614 DCBP_set_pad_bits(struct p_compressed_bm *p, int n)
617 p->encoding = (p->encoding & (~0x7 << 4)) | (n << 4);
620 /* one bitmap packet, including the p_header,
621 * should fit within one _architecture independend_ page.
622 * so we need to use the fixed size 4KiB page size
623 * most architechtures have used for a long time.
625 #define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header80))
626 #define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
627 #define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm))
628 #if (PAGE_SIZE < 4096)
629 /* drbd_send_bitmap / receive_bitmap would break horribly */
630 #error "PAGE_SIZE too small"
634 union p_header header;
635 struct p_handshake handshake;
637 struct p_block_ack block_ack;
638 struct p_barrier barrier;
639 struct p_barrier_ack barrier_ack;
640 struct p_rs_param_89 rs_param_89;
641 struct p_rs_param_95 rs_param_95;
642 struct p_protocol protocol;
643 struct p_sizes sizes;
644 struct p_uuids uuids;
645 struct p_state state;
646 struct p_req_state req_state;
647 struct p_req_state_reply req_state_reply;
648 struct p_block_req block_req;
649 struct p_delay_probe93 delay_probe93;
650 struct p_rs_uuid rs_uuid;
651 struct p_block_desc block_desc;
654 /**********************************************************************/
655 enum drbd_thread_state {
664 struct task_struct *task;
665 struct completion stop;
666 enum drbd_thread_state t_state;
667 int (*function) (struct drbd_thread *);
668 struct drbd_conf *mdev;
672 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
674 /* THINK testing the t_state seems to be uncritical in all cases
675 * (but thread_{start,stop}), so we can read it *without* the lock.
684 * Having this as the first member of a struct provides sort of "inheritance".
685 * "derived" structs can be "drbd_queue_work()"ed.
686 * The callback should know and cast back to the descendant struct.
687 * drbd_request and drbd_epoch_entry are descendants of drbd_work.
690 typedef int (*drbd_work_cb)(struct drbd_conf *, struct drbd_work *, int cancel);
692 struct list_head list;
696 struct drbd_tl_epoch;
697 struct drbd_request {
699 struct drbd_conf *mdev;
701 /* if local IO is not allowed, will be NULL.
702 * if local IO _is_ allowed, holds the locally submitted bio clone,
703 * or, after local IO completion, the ERR_PTR(error).
704 * see drbd_endio_pri(). */
705 struct bio *private_bio;
707 struct hlist_node colision;
710 unsigned int epoch; /* barrier_nr */
712 /* barrier_nr: used to check on "completion" whether this req was in
713 * the current epoch, and we therefore have to close it,
714 * starting a new epoch...
717 /* up to here, the struct layout is identical to drbd_epoch_entry;
718 * we might be able to use that to our advantage... */
720 struct list_head tl_requests; /* ring list in the transfer log */
721 struct bio *master_bio; /* master bio pointer */
722 unsigned long rq_state; /* see comments above _req_mod() */
724 unsigned long start_time;
727 struct drbd_tl_epoch {
729 struct list_head requests; /* requests before */
730 struct drbd_tl_epoch *next; /* pointer to the next barrier */
731 unsigned int br_number; /* the barriers identifier. */
732 int n_writes; /* number of requests attached before this barrier */
737 /* These Tl_epoch_entries may be in one of 6 lists:
738 active_ee .. data packet being written
739 sync_ee .. syncer block being written
740 done_ee .. block written, need to send P_WRITE_ACK
741 read_ee .. [RS]P_DATA_REQUEST being read
745 struct list_head list;
746 unsigned int barrier_nr;
747 atomic_t epoch_size; /* increased on every request added. */
748 atomic_t active; /* increased on every req. added, and dec on every finished. */
752 /* drbd_epoch flag bits */
754 DE_HAVE_BARRIER_NUMBER,
761 EV_CLEANUP = 32, /* used as flag */
764 struct drbd_wq_barrier {
766 struct completion done;
774 struct drbd_epoch_entry {
776 struct hlist_node colision;
777 struct drbd_epoch *epoch; /* for writes */
778 struct drbd_conf *mdev;
780 atomic_t pending_bios;
782 /* see comments on ee flag bits below */
787 struct digest_info *digest;
792 * While corresponding bios are in flight, the only modification will be
793 * set_bit WAS_ERROR, which has to be atomic.
794 * If no bios are in flight yet, or all have been completed,
795 * non-atomic modification to ee->flags is ok.
798 __EE_CALL_AL_COMPLETE_IO,
799 __EE_MAY_SET_IN_SYNC,
801 /* In case a barrier failed,
802 * we need to resubmit without the barrier flag. */
805 /* we may have several bios per epoch entry.
806 * if any of those fail, we set this flag atomically
807 * from the endio callback */
810 /* This ee has a pointer to a digest instead of a block id */
813 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
814 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
815 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
816 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
817 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
819 /* global flag bits */
821 CREATE_BARRIER, /* next P_DATA is preceeded by a P_BARRIER */
822 SIGNAL_ASENDER, /* whether asender wants to be interrupted */
823 SEND_PING, /* whether asender should send a ping asap */
825 UNPLUG_QUEUED, /* only relevant with kernel 2.4 */
826 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
827 MD_DIRTY, /* current uuids and flags not yet on disk */
828 DISCARD_CONCURRENT, /* Set on one node, cleared on the peer! */
829 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
830 CLUSTER_ST_CHANGE, /* Cluster wide state change going on... */
833 CRASHED_PRIMARY, /* This node was a crashed primary.
834 * Gets cleared when the state.conn
835 * goes into C_CONNECTED state. */
836 NO_BARRIER_SUPP, /* underlying block device doesn't implement barriers */
839 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */
840 SUSPEND_IO, /* suspend application io */
841 BITMAP_IO, /* suspend application io;
842 once no more io in flight, start bitmap io */
843 BITMAP_IO_QUEUED, /* Started bitmap IO */
844 GO_DISKLESS, /* Disk is being detached, on io-error or admin request. */
845 WAS_IO_ERROR, /* Local disk failed returned IO error */
846 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
847 NET_CONGESTED, /* The data socket is congested */
849 CONFIG_PENDING, /* serialization of (re)configuration requests.
850 * if set, also prevents the device from dying */
851 DEVICE_DYING, /* device became unconfigured,
852 * but worker thread is still handling the cleanup.
853 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
854 * while this is set. */
855 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
856 * the peer, if it changed there as well. */
857 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
858 GOT_PING_ACK, /* set when we receive a ping_ack packet, misc wait gets woken */
859 NEW_CUR_UUID, /* Create new current UUID when thawing IO */
860 AL_SUSPENDED, /* Activity logging is currently suspended. */
863 struct drbd_bitmap; /* opaque for drbd_conf */
865 /* TODO sort members for performance
866 * MAYBE group them further */
868 /* THINK maybe we actually want to use the default "event/%s" worker threads
869 * or similar in linux 2.6, which uses per cpu data and threads.
871 struct drbd_work_queue {
873 struct semaphore s; /* producers up it, worker down()s it */
874 spinlock_t q_lock; /* to protect the list. */
878 struct drbd_work_queue work;
880 struct socket *socket;
881 /* this way we get our
882 * send/receive buffers off the stack */
883 union p_polymorph sbuf;
884 union p_polymorph rbuf;
888 u64 md_offset; /* sector offset to 'super' block */
890 u64 la_size_sect; /* last agreed size, unit sectors */
896 s32 al_offset; /* signed relative sector offset to al area */
897 s32 bm_offset; /* signed relative sector offset to bitmap */
899 /* u32 al_nr_extents; important for restoring the AL
900 * is stored into sync_conf.al_extents, which in turn
901 * gets applied to act_log->nr_elements
905 /* for sync_conf and other types... */
906 #define NL_PACKET(name, number, fields) struct name { fields };
907 #define NL_INTEGER(pn,pr,member) int member;
908 #define NL_INT64(pn,pr,member) __u64 member;
909 #define NL_BIT(pn,pr,member) unsigned member:1;
910 #define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
911 #include "linux/drbd_nl.h"
913 struct drbd_backing_dev {
914 struct block_device *backing_bdev;
915 struct block_device *md_bdev;
917 struct disk_conf dc; /* The user provided config... */
918 sector_t known_size; /* last known size of that backing device */
922 struct drbd_conf *mdev;
923 struct completion event;
930 int (*io_fn)(struct drbd_conf *mdev);
931 void (*done)(struct drbd_conf *mdev, int rv);
934 enum write_ordering_e {
942 unsigned int head_index;
947 /* things that are stored as / read from meta data on disk */
950 /* configured by drbdsetup */
951 struct net_conf *net_conf; /* protected by get_net_conf() and put_net_conf() */
952 struct syncer_conf sync_conf;
953 struct drbd_backing_dev *ldev __protected_by(local);
955 sector_t p_size; /* partner's disk size */
956 struct request_queue *rq_queue;
957 struct block_device *this_bdev;
958 struct gendisk *vdisk;
960 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
961 struct drbd_socket meta; /* ping/ack (metadata) packets */
962 int agreed_pro_version; /* actually used protocol version */
963 unsigned long last_received; /* in jiffies, either socket */
964 unsigned int ko_count;
965 struct drbd_work resync_work,
970 struct timer_list resync_timer;
971 struct timer_list md_sync_timer;
972 #ifdef DRBD_DEBUG_MD_SYNC
976 } last_md_mark_dirty;
979 /* Used after attach while negotiating new disk state. */
980 union drbd_state new_state_tmp;
982 union drbd_state state;
983 wait_queue_head_t misc_wait;
984 wait_queue_head_t state_wait; /* upon each state change. */
985 wait_queue_head_t net_cnt_wait;
986 unsigned int send_cnt;
987 unsigned int recv_cnt;
988 unsigned int read_cnt;
989 unsigned int writ_cnt;
990 unsigned int al_writ_cnt;
991 unsigned int bm_writ_cnt;
992 atomic_t ap_bio_cnt; /* Requests we need to complete */
993 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
994 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
995 atomic_t unacked_cnt; /* Need to send replys for */
996 atomic_t local_cnt; /* Waiting for local completion */
997 atomic_t net_cnt; /* Users of net_conf */
999 struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */
1000 struct drbd_tl_epoch *newest_tle;
1001 struct drbd_tl_epoch *oldest_tle;
1002 struct list_head out_of_sequence_requests;
1003 struct hlist_head *tl_hash;
1004 unsigned int tl_hash_s;
1006 /* blocks to sync in this run [unit BM_BLOCK_SIZE] */
1007 unsigned long rs_total;
1008 /* number of sync IOs that failed in this run */
1009 unsigned long rs_failed;
1010 /* Syncer's start time [unit jiffies] */
1011 unsigned long rs_start;
1012 /* cumulated time in PausedSyncX state [unit jiffies] */
1013 unsigned long rs_paused;
1014 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
1015 unsigned long rs_same_csum;
1016 #define DRBD_SYNC_MARKS 8
1017 #define DRBD_SYNC_MARK_STEP (3*HZ)
1018 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
1019 unsigned long rs_mark_left[DRBD_SYNC_MARKS];
1020 /* marks's time [unit jiffies] */
1021 unsigned long rs_mark_time[DRBD_SYNC_MARKS];
1022 /* current index into rs_mark_{left,time} */
1025 /* where does the admin want us to start? (sector) */
1026 sector_t ov_start_sector;
1027 /* where are we now? (sector) */
1028 sector_t ov_position;
1029 /* Start sector of out of sync range (to merge printk reporting). */
1030 sector_t ov_last_oos_start;
1031 /* size of out-of-sync range in sectors. */
1032 sector_t ov_last_oos_size;
1033 unsigned long ov_left; /* in bits */
1034 struct crypto_hash *csums_tfm;
1035 struct crypto_hash *verify_tfm;
1037 struct drbd_thread receiver;
1038 struct drbd_thread worker;
1039 struct drbd_thread asender;
1040 struct drbd_bitmap *bitmap;
1041 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
1043 /* Used to track operations of resync... */
1044 struct lru_cache *resync;
1045 /* Number of locked elements in resync LRU */
1046 unsigned int resync_locked;
1047 /* resync extent number waiting for application requests */
1048 unsigned int resync_wenr;
1052 struct drbd_epoch *current_epoch;
1053 spinlock_t epoch_lock;
1054 unsigned int epochs;
1055 enum write_ordering_e write_ordering;
1056 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
1057 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
1058 struct list_head done_ee; /* send ack */
1059 struct list_head read_ee; /* IO in progress (any read) */
1060 struct list_head net_ee; /* zero-copy network send in progress */
1061 struct hlist_head *ee_hash; /* is proteced by req_lock! */
1062 unsigned int ee_hash_s;
1064 /* this one is protected by ee_lock, single thread */
1065 struct drbd_epoch_entry *last_write_w_barrier;
1067 int next_barrier_nr;
1068 struct hlist_head *app_reads_hash; /* is proteced by req_lock */
1069 struct list_head resync_reads;
1070 atomic_t pp_in_use; /* allocated from page pool */
1071 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */
1072 wait_queue_head_t ee_wait;
1073 struct page *md_io_page; /* one page buffer for md_io */
1074 struct page *md_io_tmpp; /* for logical_block_size != 512 */
1075 struct mutex md_io_mutex; /* protects the md_io_buffer */
1077 wait_queue_head_t al_wait;
1078 struct lru_cache *act_log; /* activity log */
1079 unsigned int al_tr_number;
1081 int al_tr_pos; /* position of the next transaction in the journal */
1082 struct crypto_hash *cram_hmac_tfm;
1083 struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */
1084 struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */
1088 wait_queue_head_t seq_wait;
1089 atomic_t packet_seq;
1090 unsigned int peer_seq;
1091 spinlock_t peer_seq_lock;
1093 unsigned long comm_bm_set; /* communicated number of set bits. */
1094 cpumask_var_t cpu_mask;
1095 struct bm_io_work bm_io_work;
1096 u64 ed_uuid; /* UUID of the exposed data */
1097 struct mutex state_mutex;
1098 char congestion_reason; /* Why we where congested... */
1099 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
1100 atomic_t rs_sect_ev; /* for submitted resync data rate, both */
1101 int rs_last_sect_ev; /* counter to compare with */
1102 int rs_last_events; /* counter of read or write "events" (unit sectors)
1103 * on the lower level device when we last looked. */
1104 int c_sync_rate; /* current resync rate after syncer throttle magic */
1105 struct fifo_buffer rs_plan_s; /* correction values of resync planer */
1106 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
1107 int rs_planed; /* resync sectors already planed */
1108 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
1111 static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
1113 struct drbd_conf *mdev;
1115 mdev = minor < minor_count ? minor_table[minor] : NULL;
1120 static inline unsigned int mdev_to_minor(struct drbd_conf *mdev)
1125 /* returns 1 if it was successfull,
1126 * returns 0 if there was no data socket.
1127 * so wherever you are going to use the data.socket, e.g. do
1128 * if (!drbd_get_data_sock(mdev))
1131 * drbd_put_data_sock(mdev);
1133 static inline int drbd_get_data_sock(struct drbd_conf *mdev)
1135 mutex_lock(&mdev->data.mutex);
1136 /* drbd_disconnect() could have called drbd_free_sock()
1137 * while we were waiting in down()... */
1138 if (unlikely(mdev->data.socket == NULL)) {
1139 mutex_unlock(&mdev->data.mutex);
1145 static inline void drbd_put_data_sock(struct drbd_conf *mdev)
1147 mutex_unlock(&mdev->data.mutex);
1151 * function declarations
1152 *************************/
1156 enum chg_state_flags {
1159 CS_WAIT_COMPLETE = 4,
1161 CS_ORDERED = CS_WAIT_COMPLETE + CS_SERIALIZE,
1166 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1169 extern void drbd_init_set_defaults(struct drbd_conf *mdev);
1170 extern enum drbd_state_rv drbd_change_state(struct drbd_conf *mdev,
1171 enum chg_state_flags f,
1172 union drbd_state mask,
1173 union drbd_state val);
1174 extern void drbd_force_state(struct drbd_conf *, union drbd_state,
1176 extern enum drbd_state_rv _drbd_request_state(struct drbd_conf *,
1179 enum chg_state_flags);
1180 extern enum drbd_state_rv __drbd_set_state(struct drbd_conf *, union drbd_state,
1181 enum chg_state_flags,
1182 struct completion *done);
1183 extern void print_st_err(struct drbd_conf *, union drbd_state,
1184 union drbd_state, int);
1185 extern int drbd_thread_start(struct drbd_thread *thi);
1186 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1188 extern void drbd_thread_current_set_cpu(struct drbd_conf *mdev);
1189 extern void drbd_calc_cpu_mask(struct drbd_conf *mdev);
1191 #define drbd_thread_current_set_cpu(A) ({})
1192 #define drbd_calc_cpu_mask(A) ({})
1194 extern void drbd_free_resources(struct drbd_conf *mdev);
1195 extern void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr,
1196 unsigned int set_size);
1197 extern void tl_clear(struct drbd_conf *mdev);
1198 enum drbd_req_event;
1199 extern void tl_restart(struct drbd_conf *mdev, enum drbd_req_event what);
1200 extern void _tl_add_barrier(struct drbd_conf *, struct drbd_tl_epoch *);
1201 extern void drbd_free_sock(struct drbd_conf *mdev);
1202 extern int drbd_send(struct drbd_conf *mdev, struct socket *sock,
1203 void *buf, size_t size, unsigned msg_flags);
1204 extern int drbd_send_protocol(struct drbd_conf *mdev);
1205 extern int drbd_send_uuids(struct drbd_conf *mdev);
1206 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
1207 extern int drbd_send_sync_uuid(struct drbd_conf *mdev, u64 val);
1208 extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags);
1209 extern int _drbd_send_state(struct drbd_conf *mdev);
1210 extern int drbd_send_state(struct drbd_conf *mdev);
1211 extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock,
1212 enum drbd_packets cmd, struct p_header80 *h,
1213 size_t size, unsigned msg_flags);
1214 #define USE_DATA_SOCKET 1
1215 #define USE_META_SOCKET 0
1216 extern int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket,
1217 enum drbd_packets cmd, struct p_header80 *h,
1219 extern int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd,
1220 char *data, size_t size);
1221 extern int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc);
1222 extern int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr,
1224 extern int drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd,
1225 struct drbd_epoch_entry *e);
1226 extern int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packets cmd,
1227 struct p_block_req *rp);
1228 extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd,
1229 struct p_data *dp, int data_size);
1230 extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd,
1231 sector_t sector, int blksize, u64 block_id);
1232 extern int drbd_send_oos(struct drbd_conf *mdev, struct drbd_request *req);
1233 extern int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd,
1234 struct drbd_epoch_entry *e);
1235 extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
1236 extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1237 sector_t sector, int size, u64 block_id);
1238 extern int drbd_send_drequest_csum(struct drbd_conf *mdev,
1239 sector_t sector,int size,
1240 void *digest, int digest_size,
1241 enum drbd_packets cmd);
1242 extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size);
1244 extern int drbd_send_bitmap(struct drbd_conf *mdev);
1245 extern int _drbd_send_bitmap(struct drbd_conf *mdev);
1246 extern int drbd_send_sr_reply(struct drbd_conf *mdev, enum drbd_state_rv retcode);
1247 extern void drbd_free_bc(struct drbd_backing_dev *ldev);
1248 extern void drbd_mdev_cleanup(struct drbd_conf *mdev);
1250 /* drbd_meta-data.c (still in drbd_main.c) */
1251 extern void drbd_md_sync(struct drbd_conf *mdev);
1252 extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
1253 /* maybe define them below as inline? */
1254 extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1255 extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1256 extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1257 extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1258 extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
1259 extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
1260 extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
1261 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1262 #ifndef DRBD_DEBUG_MD_SYNC
1263 extern void drbd_md_mark_dirty(struct drbd_conf *mdev);
1265 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1266 extern void drbd_md_mark_dirty_(struct drbd_conf *mdev,
1267 unsigned int line, const char *func);
1269 extern void drbd_queue_bitmap_io(struct drbd_conf *mdev,
1270 int (*io_fn)(struct drbd_conf *),
1271 void (*done)(struct drbd_conf *, int),
1273 extern int drbd_bmio_set_n_write(struct drbd_conf *mdev);
1274 extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev);
1275 extern int drbd_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), char *why);
1276 extern void drbd_go_diskless(struct drbd_conf *mdev);
1277 extern void drbd_ldev_destroy(struct drbd_conf *mdev);
1281 We reserve a 128MB Block (4k aligned)
1282 * either at the end of the backing device
1283 * or on a separate meta data device. */
1285 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1286 /* The following numbers are sectors */
1287 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1288 #define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */
1289 /* Allows up to about 3.8TB */
1290 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1292 /* Since the smalles IO unit is usually 512 byte */
1293 #define MD_SECTOR_SHIFT 9
1294 #define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT)
1297 #define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1298 #define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */
1299 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1301 #if BITS_PER_LONG == 32
1303 #define cpu_to_lel(A) cpu_to_le32(A)
1304 #define lel_to_cpu(A) le32_to_cpu(A)
1305 #elif BITS_PER_LONG == 64
1307 #define cpu_to_lel(A) cpu_to_le64(A)
1308 #define lel_to_cpu(A) le64_to_cpu(A)
1310 #error "LN2 of BITS_PER_LONG unknown!"
1314 /* 16MB sized 'bitmap extent' to track syncer usage */
1316 int rs_left; /* number of bits set (out of sync) in this extent. */
1317 int rs_failed; /* number of failed resync requests in this extent. */
1318 unsigned long flags;
1319 struct lc_element lce;
1322 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1323 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1324 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1328 * We need to store one bit for a block.
1329 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1330 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1331 * Bit 1 ==> local node thinks this block needs to be synced.
1334 #define SLEEP_TIME (HZ/10)
1336 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1337 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1338 /* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1339 * per sector of on disk bitmap */
1340 #define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */
1341 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1343 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1344 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1347 /* thus many _storage_ sectors are described by one bit */
1348 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1349 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1350 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1352 /* bit to represented kilo byte conversion */
1353 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1355 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1356 * _storage_ sector is located in */
1357 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1359 /* how much _storage_ sectors we have per bitmap sector */
1360 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1361 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1363 /* in one sector of the bitmap, we have this many activity_log extents. */
1364 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1365 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1367 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1368 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1370 /* the extent in "PER_EXTENT" below is an activity log extent
1371 * we need that many (long words/bytes) to store the bitmap
1372 * of one AL_EXTENT_SIZE chunk of storage.
1373 * we can store the bitmap for that many AL_EXTENTS within
1374 * one sector of the _on_disk_ bitmap:
1375 * bit 0 bit 37 bit 38 bit (512*8)-1
1376 * ...|........|........|.. // ..|........|
1377 * sect. 0 `296 `304 ^(512*8*8)-1
1379 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1380 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1381 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1384 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1385 #define DRBD_MAX_SECTORS_BM \
1386 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1387 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1388 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1389 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1390 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1391 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1392 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1394 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1395 /* 16 TB in units of sectors */
1396 #if BITS_PER_LONG == 32
1397 /* adjust by one page worth of bitmap,
1398 * so we won't wrap around in drbd_bm_find_next_bit.
1399 * you should use 64bit OS for that much storage, anyways. */
1400 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1402 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0x1LU << 32)
1406 /* Sector shift value for the "hash" functions of tl_hash and ee_hash tables.
1407 * With a value of 8 all IO in one 128K block make it to the same slot of the
1410 #define DRBD_MAX_BIO_SIZE (1U<<(9+HT_SHIFT))
1412 #define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */
1414 /* Number of elements in the app_reads_hash */
1415 #define APP_R_HSIZE 15
1417 extern int drbd_bm_init(struct drbd_conf *mdev);
1418 extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits);
1419 extern void drbd_bm_cleanup(struct drbd_conf *mdev);
1420 extern void drbd_bm_set_all(struct drbd_conf *mdev);
1421 extern void drbd_bm_clear_all(struct drbd_conf *mdev);
1422 extern int drbd_bm_set_bits(
1423 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1424 extern int drbd_bm_clear_bits(
1425 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1426 /* bm_set_bits variant for use while holding drbd_bm_lock */
1427 extern void _drbd_bm_set_bits(struct drbd_conf *mdev,
1428 const unsigned long s, const unsigned long e);
1429 extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr);
1430 extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr);
1431 extern int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local);
1432 extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
1433 extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
1434 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev,
1435 unsigned long al_enr);
1436 extern size_t drbd_bm_words(struct drbd_conf *mdev);
1437 extern unsigned long drbd_bm_bits(struct drbd_conf *mdev);
1438 extern sector_t drbd_bm_capacity(struct drbd_conf *mdev);
1439 extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1440 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1441 extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1442 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo);
1443 extern unsigned long _drbd_bm_total_weight(struct drbd_conf *mdev);
1444 extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev);
1445 extern int drbd_bm_rs_done(struct drbd_conf *mdev);
1446 /* for receive_bitmap */
1447 extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset,
1448 size_t number, unsigned long *buffer);
1449 /* for _drbd_send_bitmap */
1450 extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset,
1451 size_t number, unsigned long *buffer);
1453 extern void drbd_bm_lock(struct drbd_conf *mdev, char *why);
1454 extern void drbd_bm_unlock(struct drbd_conf *mdev);
1456 extern int drbd_bm_count_bits(struct drbd_conf *mdev, const unsigned long s, const unsigned long e);
1459 extern struct kmem_cache *drbd_request_cache;
1460 extern struct kmem_cache *drbd_ee_cache; /* epoch entries */
1461 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1462 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1463 extern mempool_t *drbd_request_mempool;
1464 extern mempool_t *drbd_ee_mempool;
1466 extern struct page *drbd_pp_pool; /* drbd's page pool */
1467 extern spinlock_t drbd_pp_lock;
1468 extern int drbd_pp_vacant;
1469 extern wait_queue_head_t drbd_pp_wait;
1471 extern rwlock_t global_state_lock;
1473 extern struct drbd_conf *drbd_new_device(unsigned int minor);
1474 extern void drbd_free_mdev(struct drbd_conf *mdev);
1476 extern int proc_details;
1479 extern int drbd_make_request(struct request_queue *q, struct bio *bio);
1480 extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req);
1481 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1482 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1486 extern void drbd_suspend_io(struct drbd_conf *mdev);
1487 extern void drbd_resume_io(struct drbd_conf *mdev);
1488 extern char *ppsize(char *buf, unsigned long long size);
1489 extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int);
1490 enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
1491 extern enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local);
1492 extern void resync_after_online_grow(struct drbd_conf *);
1493 extern void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int) __must_hold(local);
1494 extern enum drbd_state_rv drbd_set_role(struct drbd_conf *mdev,
1495 enum drbd_role new_role,
1497 extern enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev);
1498 extern void drbd_try_outdate_peer_async(struct drbd_conf *mdev);
1499 extern int drbd_khelper(struct drbd_conf *mdev, char *cmd);
1502 extern int drbd_worker(struct drbd_thread *thi);
1503 extern int drbd_alter_sa(struct drbd_conf *mdev, int na);
1504 extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side);
1505 extern void resume_next_sg(struct drbd_conf *mdev);
1506 extern void suspend_other_sg(struct drbd_conf *mdev);
1507 extern int drbd_resync_finished(struct drbd_conf *mdev);
1508 /* maybe rather drbd_main.c ? */
1509 extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
1510 struct drbd_backing_dev *bdev, sector_t sector, int rw);
1511 extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int);
1512 extern void drbd_rs_controller_reset(struct drbd_conf *mdev);
1514 static inline void ov_oos_print(struct drbd_conf *mdev)
1516 if (mdev->ov_last_oos_size) {
1517 dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n",
1518 (unsigned long long)mdev->ov_last_oos_start,
1519 (unsigned long)mdev->ov_last_oos_size);
1521 mdev->ov_last_oos_size=0;
1525 extern void drbd_csum_bio(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
1526 extern void drbd_csum_ee(struct drbd_conf *, struct crypto_hash *, struct drbd_epoch_entry *, void *);
1527 /* worker callbacks */
1528 extern int w_req_cancel_conflict(struct drbd_conf *, struct drbd_work *, int);
1529 extern int w_read_retry_remote(struct drbd_conf *, struct drbd_work *, int);
1530 extern int w_e_end_data_req(struct drbd_conf *, struct drbd_work *, int);
1531 extern int w_e_end_rsdata_req(struct drbd_conf *, struct drbd_work *, int);
1532 extern int w_e_end_csum_rs_req(struct drbd_conf *, struct drbd_work *, int);
1533 extern int w_e_end_ov_reply(struct drbd_conf *, struct drbd_work *, int);
1534 extern int w_e_end_ov_req(struct drbd_conf *, struct drbd_work *, int);
1535 extern int w_ov_finished(struct drbd_conf *, struct drbd_work *, int);
1536 extern int w_resync_inactive(struct drbd_conf *, struct drbd_work *, int);
1537 extern int w_resume_next_sg(struct drbd_conf *, struct drbd_work *, int);
1538 extern int w_send_write_hint(struct drbd_conf *, struct drbd_work *, int);
1539 extern int w_send_dblock(struct drbd_conf *, struct drbd_work *, int);
1540 extern int w_send_barrier(struct drbd_conf *, struct drbd_work *, int);
1541 extern int w_send_read_req(struct drbd_conf *, struct drbd_work *, int);
1542 extern int w_prev_work_done(struct drbd_conf *, struct drbd_work *, int);
1543 extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int);
1544 extern int w_restart_disk_io(struct drbd_conf *, struct drbd_work *, int);
1545 extern int w_send_oos(struct drbd_conf *, struct drbd_work *, int);
1546 extern int w_start_resync(struct drbd_conf *, struct drbd_work *, int);
1548 extern void resync_timer_fn(unsigned long data);
1550 /* drbd_receiver.c */
1551 extern int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector);
1552 extern int drbd_submit_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e,
1553 const unsigned rw, const int fault_type);
1554 extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list);
1555 extern struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev,
1558 unsigned int data_size,
1559 gfp_t gfp_mask) __must_hold(local);
1560 extern void drbd_free_some_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e,
1562 #define drbd_free_ee(m,e) drbd_free_some_ee(m, e, 0)
1563 #define drbd_free_net_ee(m,e) drbd_free_some_ee(m, e, 1)
1564 extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1565 struct list_head *head);
1566 extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1567 struct list_head *head);
1568 extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled);
1569 extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed);
1570 extern void drbd_flush_workqueue(struct drbd_conf *mdev);
1571 extern void drbd_free_tl_hash(struct drbd_conf *mdev);
1573 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1574 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1575 static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1576 char __user *optval, int optlen)
1579 if (level == SOL_SOCKET)
1580 err = sock_setsockopt(sock, level, optname, optval, optlen);
1582 err = sock->ops->setsockopt(sock, level, optname, optval,
1587 static inline void drbd_tcp_cork(struct socket *sock)
1590 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1591 (char __user *)&val, sizeof(val));
1594 static inline void drbd_tcp_uncork(struct socket *sock)
1597 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1598 (char __user *)&val, sizeof(val));
1601 static inline void drbd_tcp_nodelay(struct socket *sock)
1604 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1605 (char __user *)&val, sizeof(val));
1608 static inline void drbd_tcp_quickack(struct socket *sock)
1611 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1612 (char __user *)&val, sizeof(val));
1615 void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo);
1618 extern struct proc_dir_entry *drbd_proc;
1619 extern const struct file_operations drbd_proc_fops;
1620 extern const char *drbd_conn_str(enum drbd_conns s);
1621 extern const char *drbd_role_str(enum drbd_role s);
1624 extern void drbd_al_begin_io(struct drbd_conf *mdev, sector_t sector);
1625 extern void drbd_al_complete_io(struct drbd_conf *mdev, sector_t sector);
1626 extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector);
1627 extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1628 extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1629 extern void drbd_rs_cancel_all(struct drbd_conf *mdev);
1630 extern int drbd_rs_del_all(struct drbd_conf *mdev);
1631 extern void drbd_rs_failed_io(struct drbd_conf *mdev,
1632 sector_t sector, int size);
1633 extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *);
1634 extern void drbd_advance_rs_marks(struct drbd_conf *mdev, unsigned long still_to_go);
1635 extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
1636 int size, const char *file, const unsigned int line);
1637 #define drbd_set_in_sync(mdev, sector, size) \
1638 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1639 extern int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
1640 int size, const char *file, const unsigned int line);
1641 #define drbd_set_out_of_sync(mdev, sector, size) \
1642 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1643 extern void drbd_al_apply_to_bm(struct drbd_conf *mdev);
1644 extern void drbd_al_shrink(struct drbd_conf *mdev);
1649 void drbd_nl_cleanup(void);
1650 int __init drbd_nl_init(void);
1651 void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state);
1652 void drbd_bcast_sync_progress(struct drbd_conf *mdev);
1653 void drbd_bcast_ee(struct drbd_conf *mdev,
1654 const char *reason, const int dgs,
1655 const char* seen_hash, const char* calc_hash,
1656 const struct drbd_epoch_entry* e);
1660 * DOC: DRBD State macros
1662 * These macros are used to express state changes in easily readable form.
1664 * The NS macros expand to a mask and a value, that can be bit ored onto the
1665 * current state as soon as the spinlock (req_lock) was taken.
1667 * The _NS macros are used for state functions that get called with the
1668 * spinlock. These macros expand directly to the new state value.
1670 * Besides the basic forms NS() and _NS() additional _?NS[23] are defined
1671 * to express state changes that affect more than one aspect of the state.
1673 * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY)
1674 * Means that the network connection was established and that the peer
1675 * is in secondary role.
1677 #define role_MASK R_MASK
1678 #define peer_MASK R_MASK
1679 #define disk_MASK D_MASK
1680 #define pdsk_MASK D_MASK
1681 #define conn_MASK C_MASK
1683 #define user_isp_MASK 1
1684 #define aftr_isp_MASK 1
1685 #define susp_nod_MASK 1
1686 #define susp_fen_MASK 1
1689 ({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \
1690 ({ union drbd_state val; val.i = 0; val.T = (S); val; })
1691 #define NS2(T1, S1, T2, S2) \
1692 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1693 mask.T2 = T2##_MASK; mask; }), \
1694 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1695 val.T2 = (S2); val; })
1696 #define NS3(T1, S1, T2, S2, T3, S3) \
1697 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1698 mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \
1699 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1700 val.T2 = (S2); val.T3 = (S3); val; })
1702 #define _NS(D, T, S) \
1703 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; })
1704 #define _NS2(D, T1, S1, T2, S2) \
1705 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1706 __ns.T2 = (S2); __ns; })
1707 #define _NS3(D, T1, S1, T2, S2, T3, S3) \
1708 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1709 __ns.T2 = (S2); __ns.T3 = (S3); __ns; })
1712 * inline helper functions
1713 *************************/
1715 /* see also page_chain_add and friends in drbd_receiver.c */
1716 static inline struct page *page_chain_next(struct page *page)
1718 return (struct page *)page_private(page);
1720 #define page_chain_for_each(page) \
1721 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1722 page = page_chain_next(page))
1723 #define page_chain_for_each_safe(page, n) \
1724 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1726 static inline int drbd_bio_has_active_page(struct bio *bio)
1728 struct bio_vec *bvec;
1731 __bio_for_each_segment(bvec, bio, i, 0) {
1732 if (page_count(bvec->bv_page) > 1)
1739 static inline int drbd_ee_has_active_page(struct drbd_epoch_entry *e)
1741 struct page *page = e->pages;
1742 page_chain_for_each(page) {
1743 if (page_count(page) > 1)
1750 static inline void drbd_state_lock(struct drbd_conf *mdev)
1752 wait_event(mdev->misc_wait,
1753 !test_and_set_bit(CLUSTER_ST_CHANGE, &mdev->flags));
1756 static inline void drbd_state_unlock(struct drbd_conf *mdev)
1758 clear_bit(CLUSTER_ST_CHANGE, &mdev->flags);
1759 wake_up(&mdev->misc_wait);
1762 static inline enum drbd_state_rv
1763 _drbd_set_state(struct drbd_conf *mdev, union drbd_state ns,
1764 enum chg_state_flags flags, struct completion *done)
1766 enum drbd_state_rv rv;
1768 read_lock(&global_state_lock);
1769 rv = __drbd_set_state(mdev, ns, flags, done);
1770 read_unlock(&global_state_lock);
1776 * drbd_request_state() - Reqest a state change
1777 * @mdev: DRBD device.
1778 * @mask: mask of state bits to change.
1779 * @val: value of new state bits.
1781 * This is the most graceful way of requesting a state change. It is verbose
1782 * quite verbose in case the state change is not possible, and all those
1783 * state changes are globally serialized.
1785 static inline int drbd_request_state(struct drbd_conf *mdev,
1786 union drbd_state mask,
1787 union drbd_state val)
1789 return _drbd_request_state(mdev, mask, val, CS_VERBOSE + CS_ORDERED);
1792 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1793 static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach, const char *where)
1795 switch (mdev->ldev->dc.on_io_error) {
1798 if (__ratelimit(&drbd_ratelimit_state))
1799 dev_err(DEV, "Local IO failed in %s.\n", where);
1802 /* NOTE fall through to detach case if forcedetach set */
1804 case EP_CALL_HELPER:
1805 set_bit(WAS_IO_ERROR, &mdev->flags);
1806 if (mdev->state.disk > D_FAILED) {
1807 _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL);
1809 "Local IO failed in %s. Detaching...\n", where);
1816 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1817 * @mdev: DRBD device.
1818 * @error: Error code passed to the IO completion callback
1819 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1821 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1823 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1824 static inline void drbd_chk_io_error_(struct drbd_conf *mdev,
1825 int error, int forcedetach, const char *where)
1828 unsigned long flags;
1829 spin_lock_irqsave(&mdev->req_lock, flags);
1830 __drbd_chk_io_error_(mdev, forcedetach, where);
1831 spin_unlock_irqrestore(&mdev->req_lock, flags);
1837 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1838 * @bdev: Meta data block device.
1840 * BTW, for internal meta data, this happens to be the maximum capacity
1841 * we could agree upon with our peer node.
1843 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1845 switch (bdev->dc.meta_dev_idx) {
1846 case DRBD_MD_INDEX_INTERNAL:
1847 case DRBD_MD_INDEX_FLEX_INT:
1848 return bdev->md.md_offset + bdev->md.bm_offset;
1849 case DRBD_MD_INDEX_FLEX_EXT:
1851 return bdev->md.md_offset;
1856 * drbd_md_last_sector() - Return the last sector number of the meta data area
1857 * @bdev: Meta data block device.
1859 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1861 switch (bdev->dc.meta_dev_idx) {
1862 case DRBD_MD_INDEX_INTERNAL:
1863 case DRBD_MD_INDEX_FLEX_INT:
1864 return bdev->md.md_offset + MD_AL_OFFSET - 1;
1865 case DRBD_MD_INDEX_FLEX_EXT:
1867 return bdev->md.md_offset + bdev->md.md_size_sect;
1871 /* Returns the number of 512 byte sectors of the device */
1872 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1874 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1875 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0;
1879 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1880 * @bdev: Meta data block device.
1882 * returns the capacity we announce to out peer. we clip ourselves at the
1883 * various MAX_SECTORS, because if we don't, current implementation will
1884 * oops sooner or later
1886 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1889 switch (bdev->dc.meta_dev_idx) {
1890 case DRBD_MD_INDEX_INTERNAL:
1891 case DRBD_MD_INDEX_FLEX_INT:
1892 s = drbd_get_capacity(bdev->backing_bdev)
1893 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1894 drbd_md_first_sector(bdev))
1897 case DRBD_MD_INDEX_FLEX_EXT:
1898 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1899 drbd_get_capacity(bdev->backing_bdev));
1900 /* clip at maximum size the meta device can support */
1901 s = min_t(sector_t, s,
1902 BM_EXT_TO_SECT(bdev->md.md_size_sect
1903 - bdev->md.bm_offset));
1906 s = min_t(sector_t, DRBD_MAX_SECTORS,
1907 drbd_get_capacity(bdev->backing_bdev));
1913 * drbd_md_ss__() - Return the sector number of our meta data super block
1914 * @mdev: DRBD device.
1915 * @bdev: Meta data block device.
1917 static inline sector_t drbd_md_ss__(struct drbd_conf *mdev,
1918 struct drbd_backing_dev *bdev)
1920 switch (bdev->dc.meta_dev_idx) {
1921 default: /* external, some index */
1922 return MD_RESERVED_SECT * bdev->dc.meta_dev_idx;
1923 case DRBD_MD_INDEX_INTERNAL:
1924 /* with drbd08, internal meta data is always "flexible" */
1925 case DRBD_MD_INDEX_FLEX_INT:
1926 /* sizeof(struct md_on_disk_07) == 4k
1927 * position: last 4k aligned block of 4k size */
1928 if (!bdev->backing_bdev) {
1929 if (__ratelimit(&drbd_ratelimit_state)) {
1930 dev_err(DEV, "bdev->backing_bdev==NULL\n");
1935 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL)
1937 case DRBD_MD_INDEX_FLEX_EXT:
1943 drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w)
1945 unsigned long flags;
1946 spin_lock_irqsave(&q->q_lock, flags);
1947 list_add(&w->list, &q->q);
1948 up(&q->s); /* within the spinlock,
1949 see comment near end of drbd_worker() */
1950 spin_unlock_irqrestore(&q->q_lock, flags);
1954 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1956 unsigned long flags;
1957 spin_lock_irqsave(&q->q_lock, flags);
1958 list_add_tail(&w->list, &q->q);
1959 up(&q->s); /* within the spinlock,
1960 see comment near end of drbd_worker() */
1961 spin_unlock_irqrestore(&q->q_lock, flags);
1964 static inline void wake_asender(struct drbd_conf *mdev)
1966 if (test_bit(SIGNAL_ASENDER, &mdev->flags))
1967 force_sig(DRBD_SIG, mdev->asender.task);
1970 static inline void request_ping(struct drbd_conf *mdev)
1972 set_bit(SEND_PING, &mdev->flags);
1976 static inline int drbd_send_short_cmd(struct drbd_conf *mdev,
1977 enum drbd_packets cmd)
1979 struct p_header80 h;
1980 return drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, &h, sizeof(h));
1983 static inline int drbd_send_ping(struct drbd_conf *mdev)
1985 struct p_header80 h;
1986 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING, &h, sizeof(h));
1989 static inline int drbd_send_ping_ack(struct drbd_conf *mdev)
1991 struct p_header80 h;
1992 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING_ACK, &h, sizeof(h));
1995 static inline void drbd_thread_stop(struct drbd_thread *thi)
1997 _drbd_thread_stop(thi, false, true);
2000 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
2002 _drbd_thread_stop(thi, false, false);
2005 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
2007 _drbd_thread_stop(thi, true, false);
2010 /* counts how many answer packets packets we expect from our peer,
2011 * for either explicit application requests,
2012 * or implicit barrier packets as necessary.
2015 * _req_mod(req, queue_for_net_write or queue_for_net_read);
2016 * it is much easier and equally valid to count what we queue for the
2017 * worker, even before it actually was queued or send.
2018 * (drbd_make_request_common; recovery path on read io-error)
2020 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
2021 * _req_mod(req, data_received)
2022 * [from receive_DataReply]
2023 * _req_mod(req, write_acked_by_peer or recv_acked_by_peer or neg_acked)
2024 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
2025 * for some reason it is NOT decreased in got_NegAck,
2026 * but in the resulting cleanup code from report_params.
2027 * we should try to remember the reason for that...
2028 * _req_mod(req, send_failed or send_canceled)
2029 * _req_mod(req, connection_lost_while_pending)
2030 * [from tl_clear_barrier]
2032 static inline void inc_ap_pending(struct drbd_conf *mdev)
2034 atomic_inc(&mdev->ap_pending_cnt);
2037 #define ERR_IF_CNT_IS_NEGATIVE(which) \
2038 if (atomic_read(&mdev->which) < 0) \
2039 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
2040 __func__ , __LINE__ , \
2041 atomic_read(&mdev->which))
2043 #define dec_ap_pending(mdev) do { \
2044 typecheck(struct drbd_conf *, mdev); \
2045 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \
2046 wake_up(&mdev->misc_wait); \
2047 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
2049 /* counts how many resync-related answers we still expect from the peer
2051 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
2052 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK whith ID_SYNCER)
2053 * (or P_NEG_ACK with ID_SYNCER)
2055 static inline void inc_rs_pending(struct drbd_conf *mdev)
2057 atomic_inc(&mdev->rs_pending_cnt);
2060 #define dec_rs_pending(mdev) do { \
2061 typecheck(struct drbd_conf *, mdev); \
2062 atomic_dec(&mdev->rs_pending_cnt); \
2063 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
2065 /* counts how many answers we still need to send to the peer.
2067 * receive_Data unless protocol A;
2068 * we need to send a P_RECV_ACK (proto B)
2069 * or P_WRITE_ACK (proto C)
2070 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2071 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2072 * receive_Barrier_* we need to send a P_BARRIER_ACK
2074 static inline void inc_unacked(struct drbd_conf *mdev)
2076 atomic_inc(&mdev->unacked_cnt);
2079 #define dec_unacked(mdev) do { \
2080 typecheck(struct drbd_conf *, mdev); \
2081 atomic_dec(&mdev->unacked_cnt); \
2082 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2084 #define sub_unacked(mdev, n) do { \
2085 typecheck(struct drbd_conf *, mdev); \
2086 atomic_sub(n, &mdev->unacked_cnt); \
2087 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2090 static inline void put_net_conf(struct drbd_conf *mdev)
2092 if (atomic_dec_and_test(&mdev->net_cnt))
2093 wake_up(&mdev->net_cnt_wait);
2097 * get_net_conf() - Increase ref count on mdev->net_conf; Returns 0 if nothing there
2098 * @mdev: DRBD device.
2100 * You have to call put_net_conf() when finished working with mdev->net_conf.
2102 static inline int get_net_conf(struct drbd_conf *mdev)
2106 atomic_inc(&mdev->net_cnt);
2107 have_net_conf = mdev->state.conn >= C_UNCONNECTED;
2110 return have_net_conf;
2114 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2117 * You have to call put_ldev() when finished working with mdev->ldev.
2119 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2120 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2122 static inline void put_ldev(struct drbd_conf *mdev)
2124 int i = atomic_dec_return(&mdev->local_cnt);
2128 if (mdev->state.disk == D_DISKLESS)
2129 /* even internal references gone, safe to destroy */
2130 drbd_ldev_destroy(mdev);
2131 if (mdev->state.disk == D_FAILED)
2132 /* all application IO references gone. */
2133 drbd_go_diskless(mdev);
2134 wake_up(&mdev->misc_wait);
2139 static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
2143 /* never get a reference while D_DISKLESS */
2144 if (mdev->state.disk == D_DISKLESS)
2147 atomic_inc(&mdev->local_cnt);
2148 io_allowed = (mdev->state.disk >= mins);
2154 extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins);
2157 /* you must have an "get_ldev" reference */
2158 static inline void drbd_get_syncer_progress(struct drbd_conf *mdev,
2159 unsigned long *bits_left, unsigned int *per_mil_done)
2162 * this is to break it at compile time when we change that
2163 * (we may feel 4TB maximum storage per drbd is not enough)
2165 typecheck(unsigned long, mdev->rs_total);
2167 /* note: both rs_total and rs_left are in bits, i.e. in
2168 * units of BM_BLOCK_SIZE.
2169 * for the percentage, we don't care. */
2171 if (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T)
2172 *bits_left = mdev->ov_left;
2174 *bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
2175 /* >> 10 to prevent overflow,
2176 * +1 to prevent division by zero */
2177 if (*bits_left > mdev->rs_total) {
2178 /* doh. maybe a logic bug somewhere.
2179 * may also be just a race condition
2180 * between this and a disconnect during sync.
2181 * for now, just prevent in-kernel buffer overflow.
2184 dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2185 drbd_conn_str(mdev->state.conn),
2186 *bits_left, mdev->rs_total, mdev->rs_failed);
2189 /* make sure the calculation happens in long context */
2190 unsigned long tmp = 1000UL -
2191 (*bits_left >> 10)*1000UL
2192 / ((mdev->rs_total >> 10) + 1UL);
2193 *per_mil_done = tmp;
2198 /* this throttles on-the-fly application requests
2199 * according to max_buffers settings;
2200 * maybe re-implement using semaphores? */
2201 static inline int drbd_get_max_buffers(struct drbd_conf *mdev)
2203 int mxb = 1000000; /* arbitrary limit on open requests */
2204 if (get_net_conf(mdev)) {
2205 mxb = mdev->net_conf->max_buffers;
2211 static inline int drbd_state_is_stable(struct drbd_conf *mdev)
2213 union drbd_state s = mdev->state;
2215 /* DO NOT add a default clause, we want the compiler to warn us
2216 * for any newly introduced state we may have forgotten to add here */
2218 switch ((enum drbd_conns)s.conn) {
2219 /* new io only accepted when there is no connection, ... */
2221 case C_WF_CONNECTION:
2222 /* ... or there is a well established connection. */
2228 case C_PAUSED_SYNC_S:
2229 case C_PAUSED_SYNC_T:
2232 /* transitional states, IO allowed */
2233 case C_DISCONNECTING:
2237 case C_NETWORK_FAILURE:
2238 case C_PROTOCOL_ERROR:
2240 case C_WF_REPORT_PARAMS:
2241 case C_STARTING_SYNC_S:
2242 case C_STARTING_SYNC_T:
2245 /* Allow IO in BM exchange states with new protocols */
2247 if (mdev->agreed_pro_version < 96)
2251 /* no new io accepted in these states */
2253 case C_WF_SYNC_UUID:
2259 switch ((enum drbd_disk_state)s.disk) {
2261 case D_INCONSISTENT:
2265 /* disk state is stable as well. */
2268 /* no new io accepted during tansitional states */
2281 static inline int is_susp(union drbd_state s)
2283 return s.susp || s.susp_nod || s.susp_fen;
2286 static inline bool may_inc_ap_bio(struct drbd_conf *mdev)
2288 int mxb = drbd_get_max_buffers(mdev);
2290 if (is_susp(mdev->state))
2292 if (test_bit(SUSPEND_IO, &mdev->flags))
2295 /* to avoid potential deadlock or bitmap corruption,
2296 * in various places, we only allow new application io
2297 * to start during "stable" states. */
2299 /* no new io accepted when attaching or detaching the disk */
2300 if (!drbd_state_is_stable(mdev))
2303 /* since some older kernels don't have atomic_add_unless,
2304 * and we are within the spinlock anyways, we have this workaround. */
2305 if (atomic_read(&mdev->ap_bio_cnt) > mxb)
2307 if (test_bit(BITMAP_IO, &mdev->flags))
2312 static inline bool inc_ap_bio_cond(struct drbd_conf *mdev, int count)
2316 spin_lock_irq(&mdev->req_lock);
2317 rv = may_inc_ap_bio(mdev);
2319 atomic_add(count, &mdev->ap_bio_cnt);
2320 spin_unlock_irq(&mdev->req_lock);
2325 static inline void inc_ap_bio(struct drbd_conf *mdev, int count)
2328 * as long as the device is suspended
2329 * until the bitmap is no longer on the fly during connection
2330 * handshake as long as we would exeed the max_buffer limit.
2332 * to avoid races with the reconnect code,
2333 * we need to atomic_inc within the spinlock. */
2335 wait_event(mdev->misc_wait, inc_ap_bio_cond(mdev, count));
2338 static inline void dec_ap_bio(struct drbd_conf *mdev)
2340 int mxb = drbd_get_max_buffers(mdev);
2341 int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt);
2343 D_ASSERT(ap_bio >= 0);
2344 /* this currently does wake_up for every dec_ap_bio!
2345 * maybe rather introduce some type of hysteresis?
2346 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2348 wake_up(&mdev->misc_wait);
2349 if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) {
2350 if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags))
2351 drbd_queue_work(&mdev->data.work, &mdev->bm_io_work.w);
2355 static inline void drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val)
2357 mdev->ed_uuid = val;
2360 static inline int seq_cmp(u32 a, u32 b)
2362 /* we assume wrap around at 32bit.
2363 * for wrap around at 24bit (old atomic_t),
2367 return (s32)(a) - (s32)(b);
2369 #define seq_lt(a, b) (seq_cmp((a), (b)) < 0)
2370 #define seq_gt(a, b) (seq_cmp((a), (b)) > 0)
2371 #define seq_ge(a, b) (seq_cmp((a), (b)) >= 0)
2372 #define seq_le(a, b) (seq_cmp((a), (b)) <= 0)
2373 /* CAUTION: please no side effects in arguments! */
2374 #define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b)))
2376 static inline void update_peer_seq(struct drbd_conf *mdev, unsigned int new_seq)
2379 spin_lock(&mdev->peer_seq_lock);
2380 m = seq_max(mdev->peer_seq, new_seq);
2382 spin_unlock(&mdev->peer_seq_lock);
2384 wake_up(&mdev->seq_wait);
2387 static inline void drbd_update_congested(struct drbd_conf *mdev)
2389 struct sock *sk = mdev->data.socket->sk;
2390 if (sk->sk_wmem_queued > sk->sk_sndbuf * 4 / 5)
2391 set_bit(NET_CONGESTED, &mdev->flags);
2394 static inline int drbd_queue_order_type(struct drbd_conf *mdev)
2396 /* sorry, we currently have no working implementation
2397 * of distributed TCQ stuff */
2398 #ifndef QUEUE_ORDERED_NONE
2399 #define QUEUE_ORDERED_NONE 0
2401 return QUEUE_ORDERED_NONE;
2404 static inline void drbd_md_flush(struct drbd_conf *mdev)
2408 if (test_bit(MD_NO_FUA, &mdev->flags))
2411 r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL);
2413 set_bit(MD_NO_FUA, &mdev->flags);
2414 dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r);