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[];
82 /* I don't remember why XCPU ...
83 * This is used to wake the asender,
84 * and to interrupt sending the sending task
87 #define DRBD_SIG SIGXCPU
89 /* This is used to stop/restart our threads.
90 * Cannot use SIGTERM nor SIGKILL, since these
91 * are sent out by init on runlevel changes
92 * I choose SIGHUP for now.
94 #define DRBD_SIGKILL SIGHUP
96 /* All EEs on the free list should have ID_VACANT (== 0)
97 * freshly allocated EEs get !ID_VACANT (== 1)
98 * so if it says "cannot dereference null pointer at address 0x00000001",
99 * it is most likely one of these :( */
101 #define ID_IN_SYNC (4711ULL)
102 #define ID_OUT_OF_SYNC (4712ULL)
104 #define ID_SYNCER (-1ULL)
106 #define is_syncer_block_id(id) ((id) == ID_SYNCER)
111 /* to shorten dev_warn(DEV, "msg"); and relatives statements */
112 #define DEV (disk_to_dev(mdev->vdisk))
114 #define D_ASSERT(exp) if (!(exp)) \
115 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
117 #define ERR_IF(exp) if (({ \
118 int _b = (exp) != 0; \
119 if (_b) dev_err(DEV, "ASSERT FAILED: %s: (%s) in %s:%d\n", \
120 __func__, #exp, __FILE__, __LINE__); \
124 /* Defines to control fault insertion */
126 DRBD_FAULT_MD_WR = 0, /* meta data write */
127 DRBD_FAULT_MD_RD = 1, /* read */
128 DRBD_FAULT_RS_WR = 2, /* resync */
129 DRBD_FAULT_RS_RD = 3,
130 DRBD_FAULT_DT_WR = 4, /* data */
131 DRBD_FAULT_DT_RD = 5,
132 DRBD_FAULT_DT_RA = 6, /* data read ahead */
133 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */
134 DRBD_FAULT_AL_EE = 8, /* alloc ee */
135 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
141 _drbd_insert_fault(struct drbd_conf *mdev, unsigned int type);
144 drbd_insert_fault(struct drbd_conf *mdev, unsigned int type) {
145 #ifdef CONFIG_DRBD_FAULT_INJECTION
147 (enable_faults & (1<<type)) &&
148 _drbd_insert_fault(mdev, type);
154 /* integer division, round _UP_ to the next integer */
155 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
156 /* usual integer division */
157 #define div_floor(A, B) ((A)/(B))
159 /* drbd_meta-data.c (still in drbd_main.c) */
160 /* 4th incarnation of the disk layout. */
161 #define DRBD_MD_MAGIC (DRBD_MAGIC+4)
163 extern struct drbd_conf **minor_table;
164 extern struct ratelimit_state drbd_ratelimit_state;
168 /* receiver (data socket) */
170 P_DATA_REPLY = 0x01, /* Response to P_DATA_REQUEST */
171 P_RS_DATA_REPLY = 0x02, /* Response to P_RS_DATA_REQUEST */
174 P_BECOME_SYNC_TARGET = 0x05,
175 P_BECOME_SYNC_SOURCE = 0x06,
176 P_UNPLUG_REMOTE = 0x07, /* Used at various times to hint the peer */
177 P_DATA_REQUEST = 0x08, /* Used to ask for a data block */
178 P_RS_DATA_REQUEST = 0x09, /* Used to ask for a data block for resync */
185 P_AUTH_CHALLENGE = 0x10,
186 P_AUTH_RESPONSE = 0x11,
187 P_STATE_CHG_REQ = 0x12,
189 /* asender (meta socket */
192 P_RECV_ACK = 0x15, /* Used in protocol B */
193 P_WRITE_ACK = 0x16, /* Used in protocol C */
194 P_RS_WRITE_ACK = 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
195 P_DISCARD_ACK = 0x18, /* Used in proto C, two-primaries conflict detection */
196 P_NEG_ACK = 0x19, /* Sent if local disk is unusable */
197 P_NEG_DREPLY = 0x1a, /* Local disk is broken... */
198 P_NEG_RS_DREPLY = 0x1b, /* Local disk is broken... */
199 P_BARRIER_ACK = 0x1c,
200 P_STATE_CHG_REPLY = 0x1d,
202 /* "new" commands, no longer fitting into the ordering scheme above */
204 P_OV_REQUEST = 0x1e, /* data socket */
206 P_OV_RESULT = 0x20, /* meta socket */
207 P_CSUM_RS_REQUEST = 0x21, /* data socket */
208 P_RS_IS_IN_SYNC = 0x22, /* meta socket */
209 P_SYNC_PARAM89 = 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
210 P_COMPRESSED_BITMAP = 0x24, /* compressed or otherwise encoded bitmap transfer */
211 /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */
212 /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */
213 P_DELAY_PROBE = 0x27, /* is used on BOTH sockets */
214 P_OUT_OF_SYNC = 0x28, /* Mark as out of sync (Outrunning), data socket */
217 P_MAY_IGNORE = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
218 P_MAX_OPT_CMD = 0x101,
220 /* special command ids for handshake */
222 P_HAND_SHAKE_M = 0xfff1, /* First Packet on the MetaSock */
223 P_HAND_SHAKE_S = 0xfff2, /* First Packet on the Socket */
225 P_HAND_SHAKE = 0xfffe /* FIXED for the next century! */
228 static inline const char *cmdname(enum drbd_packets cmd)
230 /* THINK may need to become several global tables
231 * when we want to support more than
233 static const char *cmdnames[] = {
235 [P_DATA_REPLY] = "DataReply",
236 [P_RS_DATA_REPLY] = "RSDataReply",
237 [P_BARRIER] = "Barrier",
238 [P_BITMAP] = "ReportBitMap",
239 [P_BECOME_SYNC_TARGET] = "BecomeSyncTarget",
240 [P_BECOME_SYNC_SOURCE] = "BecomeSyncSource",
241 [P_UNPLUG_REMOTE] = "UnplugRemote",
242 [P_DATA_REQUEST] = "DataRequest",
243 [P_RS_DATA_REQUEST] = "RSDataRequest",
244 [P_SYNC_PARAM] = "SyncParam",
245 [P_SYNC_PARAM89] = "SyncParam89",
246 [P_PROTOCOL] = "ReportProtocol",
247 [P_UUIDS] = "ReportUUIDs",
248 [P_SIZES] = "ReportSizes",
249 [P_STATE] = "ReportState",
250 [P_SYNC_UUID] = "ReportSyncUUID",
251 [P_AUTH_CHALLENGE] = "AuthChallenge",
252 [P_AUTH_RESPONSE] = "AuthResponse",
254 [P_PING_ACK] = "PingAck",
255 [P_RECV_ACK] = "RecvAck",
256 [P_WRITE_ACK] = "WriteAck",
257 [P_RS_WRITE_ACK] = "RSWriteAck",
258 [P_DISCARD_ACK] = "DiscardAck",
259 [P_NEG_ACK] = "NegAck",
260 [P_NEG_DREPLY] = "NegDReply",
261 [P_NEG_RS_DREPLY] = "NegRSDReply",
262 [P_BARRIER_ACK] = "BarrierAck",
263 [P_STATE_CHG_REQ] = "StateChgRequest",
264 [P_STATE_CHG_REPLY] = "StateChgReply",
265 [P_OV_REQUEST] = "OVRequest",
266 [P_OV_REPLY] = "OVReply",
267 [P_OV_RESULT] = "OVResult",
268 [P_CSUM_RS_REQUEST] = "CsumRSRequest",
269 [P_RS_IS_IN_SYNC] = "CsumRSIsInSync",
270 [P_COMPRESSED_BITMAP] = "CBitmap",
271 [P_DELAY_PROBE] = "DelayProbe",
272 [P_OUT_OF_SYNC] = "OutOfSync",
276 if (cmd == P_HAND_SHAKE_M)
278 if (cmd == P_HAND_SHAKE_S)
280 if (cmd == P_HAND_SHAKE)
282 if (cmd >= P_MAX_CMD)
284 return cmdnames[cmd];
287 /* for sending/receiving the bitmap,
288 * possibly in some encoding scheme */
291 * stores total bits and long words
292 * of the bitmap, so we don't need to
293 * call the accessor functions over and again. */
294 unsigned long bm_bits;
295 unsigned long bm_words;
296 /* during xfer, current position within the bitmap */
297 unsigned long bit_offset;
298 unsigned long word_offset;
300 /* statistics; index: (h->command == P_BITMAP) */
305 extern void INFO_bm_xfer_stats(struct drbd_conf *mdev,
306 const char *direction, struct bm_xfer_ctx *c);
308 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
310 /* word_offset counts "native long words" (32 or 64 bit),
312 * Encoded packet may end at an unaligned bit offset.
313 * In case a fallback clear text packet is transmitted in
314 * between, we adjust this offset back to the last 64bit
315 * aligned "native long word", which makes coding and decoding
316 * the plain text bitmap much more convenient. */
317 #if BITS_PER_LONG == 64
318 c->word_offset = c->bit_offset >> 6;
319 #elif BITS_PER_LONG == 32
320 c->word_offset = c->bit_offset >> 5;
321 c->word_offset &= ~(1UL);
323 # error "unsupported BITS_PER_LONG"
328 #define __packed __attribute__((packed))
331 /* This is the layout for a packet on the wire.
332 * The byteorder is the network byte order.
333 * (except block_id and barrier fields.
334 * these are pointers to local structs
335 * and have no relevance for the partner,
336 * which just echoes them as received.)
338 * NOTE that the payload starts at a long aligned offset,
339 * regardless of 32 or 64 bit arch!
344 u16 length; /* bytes of data after this header */
348 /* Header for big packets, Used for data packets exceeding 64kB */
350 u16 magic; /* use DRBD_MAGIC_BIG here */
352 u32 length; /* Use only 24 bits of that. Ignore the highest 8 bit. */
357 struct p_header80 h80;
358 struct p_header95 h95;
362 * short commands, packets without payload, plain p_header:
365 * P_BECOME_SYNC_TARGET
366 * P_BECOME_SYNC_SOURCE
371 * commands with out-of-struct payload:
372 * P_BITMAP (no additional fields)
373 * P_DATA, P_DATA_REPLY (see p_data)
374 * P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
377 /* these defines must not be changed without changing the protocol version */
378 #define DP_HARDBARRIER 1 /* depricated */
379 #define DP_RW_SYNC 2 /* equals REQ_SYNC */
380 #define DP_MAY_SET_IN_SYNC 4
381 #define DP_UNPLUG 8 /* not used anymore */
382 #define DP_FUA 16 /* equals REQ_FUA */
383 #define DP_FLUSH 32 /* equals REQ_FLUSH */
384 #define DP_DISCARD 64 /* equals REQ_DISCARD */
388 u64 sector; /* 64 bits sector number */
389 u64 block_id; /* to identify the request in protocol B&C */
395 * commands which share a struct:
397 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
398 * P_DISCARD_ACK (proto C, two-primaries conflict detection)
400 * P_DATA_REQUEST, P_RS_DATA_REQUEST
403 struct p_header80 head;
412 struct p_header80 head;
416 u32 pad; /* to multiple of 8 Byte */
420 * commands with their own struct for additional fields:
429 struct p_header80 head; /* 8 bytes */
434 /* should be more than enough for future enhancements
435 * for now, feature_flags and the reserverd array shall be zero.
441 /* 80 bytes, FIXED for the next century */
444 struct p_header80 head;
445 u32 barrier; /* barrier number _handle_ only */
446 u32 pad; /* to multiple of 8 Byte */
449 struct p_barrier_ack {
450 struct p_header80 head;
456 struct p_header80 head;
459 /* Since protocol version 88 and higher. */
463 struct p_rs_param_89 {
464 struct p_header80 head;
466 /* protocol version 89: */
467 char verify_alg[SHARED_SECRET_MAX];
468 char csums_alg[SHARED_SECRET_MAX];
471 struct p_rs_param_95 {
472 struct p_header80 head;
474 char verify_alg[SHARED_SECRET_MAX];
475 char csums_alg[SHARED_SECRET_MAX];
482 enum drbd_conn_flags {
488 struct p_header80 head;
496 /* Since protocol version 87 and higher. */
497 char integrity_alg[0];
502 struct p_header80 head;
503 u64 uuid[UI_EXTENDED_SIZE];
507 struct p_header80 head;
512 struct p_header80 head;
513 u64 d_size; /* size of disk */
514 u64 u_size; /* user requested size */
515 u64 c_size; /* current exported size */
516 u32 max_bio_size; /* Maximal size of a BIO */
517 u16 queue_order_type; /* not yet implemented in DRBD*/
518 u16 dds_flags; /* use enum dds_flags here. */
522 struct p_header80 head;
527 struct p_header80 head;
532 struct p_req_state_reply {
533 struct p_header80 head;
537 struct p_drbd06_param {
548 struct p_header80 head;
554 struct p_block_desc {
555 struct p_header80 head;
558 u32 pad; /* to multiple of 8 Byte */
561 /* Valid values for the encoding field.
562 * Bump proto version when changing this. */
563 enum drbd_bitmap_code {
564 /* RLE_VLI_Bytes = 0,
565 * and other bit variants had been defined during
566 * algorithm evaluation. */
570 struct p_compressed_bm {
571 struct p_header80 head;
572 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
573 * (encoding & 0x80): polarity (set/unset) of first runlength
574 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
575 * used to pad up to head.length bytes
582 struct p_delay_probe93 {
583 struct p_header80 head;
584 u32 seq_num; /* sequence number to match the two probe packets */
585 u32 offset; /* usecs the probe got sent after the reference time point */
588 /* DCBP: Drbd Compressed Bitmap Packet ... */
589 static inline enum drbd_bitmap_code
590 DCBP_get_code(struct p_compressed_bm *p)
592 return (enum drbd_bitmap_code)(p->encoding & 0x0f);
596 DCBP_set_code(struct p_compressed_bm *p, enum drbd_bitmap_code code)
599 p->encoding = (p->encoding & ~0xf) | code;
603 DCBP_get_start(struct p_compressed_bm *p)
605 return (p->encoding & 0x80) != 0;
609 DCBP_set_start(struct p_compressed_bm *p, int set)
611 p->encoding = (p->encoding & ~0x80) | (set ? 0x80 : 0);
615 DCBP_get_pad_bits(struct p_compressed_bm *p)
617 return (p->encoding >> 4) & 0x7;
621 DCBP_set_pad_bits(struct p_compressed_bm *p, int n)
624 p->encoding = (p->encoding & (~0x7 << 4)) | (n << 4);
627 /* one bitmap packet, including the p_header,
628 * should fit within one _architecture independend_ page.
629 * so we need to use the fixed size 4KiB page size
630 * most architechtures have used for a long time.
632 #define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header80))
633 #define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
634 #define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm))
635 #if (PAGE_SIZE < 4096)
636 /* drbd_send_bitmap / receive_bitmap would break horribly */
637 #error "PAGE_SIZE too small"
641 union p_header header;
642 struct p_handshake handshake;
644 struct p_block_ack block_ack;
645 struct p_barrier barrier;
646 struct p_barrier_ack barrier_ack;
647 struct p_rs_param_89 rs_param_89;
648 struct p_rs_param_95 rs_param_95;
649 struct p_protocol protocol;
650 struct p_sizes sizes;
651 struct p_uuids uuids;
652 struct p_state state;
653 struct p_req_state req_state;
654 struct p_req_state_reply req_state_reply;
655 struct p_block_req block_req;
656 struct p_delay_probe93 delay_probe93;
657 struct p_rs_uuid rs_uuid;
658 struct p_block_desc block_desc;
661 /**********************************************************************/
662 enum drbd_thread_state {
671 struct task_struct *task;
672 struct completion stop;
673 enum drbd_thread_state t_state;
674 int (*function) (struct drbd_thread *);
675 struct drbd_conf *mdev;
679 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
681 /* THINK testing the t_state seems to be uncritical in all cases
682 * (but thread_{start,stop}), so we can read it *without* the lock.
691 * Having this as the first member of a struct provides sort of "inheritance".
692 * "derived" structs can be "drbd_queue_work()"ed.
693 * The callback should know and cast back to the descendant struct.
694 * drbd_request and drbd_epoch_entry are descendants of drbd_work.
697 typedef int (*drbd_work_cb)(struct drbd_conf *, struct drbd_work *, int cancel);
699 struct list_head list;
703 struct drbd_tl_epoch;
704 struct drbd_request {
706 struct drbd_conf *mdev;
708 /* if local IO is not allowed, will be NULL.
709 * if local IO _is_ allowed, holds the locally submitted bio clone,
710 * or, after local IO completion, the ERR_PTR(error).
711 * see drbd_endio_pri(). */
712 struct bio *private_bio;
714 struct hlist_node colision;
717 unsigned int epoch; /* barrier_nr */
719 /* barrier_nr: used to check on "completion" whether this req was in
720 * the current epoch, and we therefore have to close it,
721 * starting a new epoch...
724 /* up to here, the struct layout is identical to drbd_epoch_entry;
725 * we might be able to use that to our advantage... */
727 struct list_head tl_requests; /* ring list in the transfer log */
728 struct bio *master_bio; /* master bio pointer */
729 unsigned long rq_state; /* see comments above _req_mod() */
731 unsigned long start_time;
734 struct drbd_tl_epoch {
736 struct list_head requests; /* requests before */
737 struct drbd_tl_epoch *next; /* pointer to the next barrier */
738 unsigned int br_number; /* the barriers identifier. */
739 int n_writes; /* number of requests attached before this barrier */
744 /* These Tl_epoch_entries may be in one of 6 lists:
745 active_ee .. data packet being written
746 sync_ee .. syncer block being written
747 done_ee .. block written, need to send P_WRITE_ACK
748 read_ee .. [RS]P_DATA_REQUEST being read
752 struct list_head list;
753 unsigned int barrier_nr;
754 atomic_t epoch_size; /* increased on every request added. */
755 atomic_t active; /* increased on every req. added, and dec on every finished. */
759 /* drbd_epoch flag bits */
761 DE_HAVE_BARRIER_NUMBER,
768 EV_CLEANUP = 32, /* used as flag */
771 struct drbd_wq_barrier {
773 struct completion done;
781 struct drbd_epoch_entry {
783 struct hlist_node colision;
784 struct drbd_epoch *epoch; /* for writes */
785 struct drbd_conf *mdev;
787 atomic_t pending_bios;
789 /* see comments on ee flag bits below */
794 struct digest_info *digest;
799 * While corresponding bios are in flight, the only modification will be
800 * set_bit WAS_ERROR, which has to be atomic.
801 * If no bios are in flight yet, or all have been completed,
802 * non-atomic modification to ee->flags is ok.
805 __EE_CALL_AL_COMPLETE_IO,
806 __EE_MAY_SET_IN_SYNC,
808 /* In case a barrier failed,
809 * we need to resubmit without the barrier flag. */
812 /* we may have several bios per epoch entry.
813 * if any of those fail, we set this flag atomically
814 * from the endio callback */
817 /* This ee has a pointer to a digest instead of a block id */
820 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
821 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
822 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
823 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
824 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
826 /* global flag bits */
828 CREATE_BARRIER, /* next P_DATA is preceeded by a P_BARRIER */
829 SIGNAL_ASENDER, /* whether asender wants to be interrupted */
830 SEND_PING, /* whether asender should send a ping asap */
832 UNPLUG_QUEUED, /* only relevant with kernel 2.4 */
833 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
834 MD_DIRTY, /* current uuids and flags not yet on disk */
835 DISCARD_CONCURRENT, /* Set on one node, cleared on the peer! */
836 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
837 CLUSTER_ST_CHANGE, /* Cluster wide state change going on... */
840 CRASHED_PRIMARY, /* This node was a crashed primary.
841 * Gets cleared when the state.conn
842 * goes into C_CONNECTED state. */
843 WRITE_BM_AFTER_RESYNC, /* A kmalloc() during resync failed */
846 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */
847 SUSPEND_IO, /* suspend application io */
848 BITMAP_IO, /* suspend application io;
849 once no more io in flight, start bitmap io */
850 BITMAP_IO_QUEUED, /* Started bitmap IO */
851 GO_DISKLESS, /* Disk is being detached, on io-error or admin request. */
852 WAS_IO_ERROR, /* Local disk failed returned IO error */
853 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
854 NET_CONGESTED, /* The data socket is congested */
856 CONFIG_PENDING, /* serialization of (re)configuration requests.
857 * if set, also prevents the device from dying */
858 DEVICE_DYING, /* device became unconfigured,
859 * but worker thread is still handling the cleanup.
860 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
861 * while this is set. */
862 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
863 * the peer, if it changed there as well. */
864 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
865 GOT_PING_ACK, /* set when we receive a ping_ack packet, misc wait gets woken */
866 NEW_CUR_UUID, /* Create new current UUID when thawing IO */
867 AL_SUSPENDED, /* Activity logging is currently suspended. */
870 struct drbd_bitmap; /* opaque for drbd_conf */
872 /* TODO sort members for performance
873 * MAYBE group them further */
875 /* THINK maybe we actually want to use the default "event/%s" worker threads
876 * or similar in linux 2.6, which uses per cpu data and threads.
878 struct drbd_work_queue {
880 struct semaphore s; /* producers up it, worker down()s it */
881 spinlock_t q_lock; /* to protect the list. */
885 struct drbd_work_queue work;
887 struct socket *socket;
888 /* this way we get our
889 * send/receive buffers off the stack */
890 union p_polymorph sbuf;
891 union p_polymorph rbuf;
895 u64 md_offset; /* sector offset to 'super' block */
897 u64 la_size_sect; /* last agreed size, unit sectors */
903 s32 al_offset; /* signed relative sector offset to al area */
904 s32 bm_offset; /* signed relative sector offset to bitmap */
906 /* u32 al_nr_extents; important for restoring the AL
907 * is stored into sync_conf.al_extents, which in turn
908 * gets applied to act_log->nr_elements
912 /* for sync_conf and other types... */
913 #define NL_PACKET(name, number, fields) struct name { fields };
914 #define NL_INTEGER(pn,pr,member) int member;
915 #define NL_INT64(pn,pr,member) __u64 member;
916 #define NL_BIT(pn,pr,member) unsigned member:1;
917 #define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
918 #include "linux/drbd_nl.h"
920 struct drbd_backing_dev {
921 struct block_device *backing_bdev;
922 struct block_device *md_bdev;
924 struct disk_conf dc; /* The user provided config... */
925 sector_t known_size; /* last known size of that backing device */
929 struct drbd_conf *mdev;
930 struct completion event;
937 int (*io_fn)(struct drbd_conf *mdev);
938 void (*done)(struct drbd_conf *mdev, int rv);
941 enum write_ordering_e {
949 unsigned int head_index;
954 /* things that are stored as / read from meta data on disk */
957 /* configured by drbdsetup */
958 struct net_conf *net_conf; /* protected by get_net_conf() and put_net_conf() */
959 struct syncer_conf sync_conf;
960 struct drbd_backing_dev *ldev __protected_by(local);
962 sector_t p_size; /* partner's disk size */
963 struct request_queue *rq_queue;
964 struct block_device *this_bdev;
965 struct gendisk *vdisk;
967 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
968 struct drbd_socket meta; /* ping/ack (metadata) packets */
969 int agreed_pro_version; /* actually used protocol version */
970 unsigned long last_received; /* in jiffies, either socket */
971 unsigned int ko_count;
972 struct drbd_work resync_work,
977 struct timer_list resync_timer;
978 struct timer_list md_sync_timer;
979 #ifdef DRBD_DEBUG_MD_SYNC
983 } last_md_mark_dirty;
986 /* Used after attach while negotiating new disk state. */
987 union drbd_state new_state_tmp;
989 union drbd_state state;
990 wait_queue_head_t misc_wait;
991 wait_queue_head_t state_wait; /* upon each state change. */
992 wait_queue_head_t net_cnt_wait;
993 unsigned int send_cnt;
994 unsigned int recv_cnt;
995 unsigned int read_cnt;
996 unsigned int writ_cnt;
997 unsigned int al_writ_cnt;
998 unsigned int bm_writ_cnt;
999 atomic_t ap_bio_cnt; /* Requests we need to complete */
1000 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
1001 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
1002 atomic_t unacked_cnt; /* Need to send replys for */
1003 atomic_t local_cnt; /* Waiting for local completion */
1004 atomic_t net_cnt; /* Users of net_conf */
1005 spinlock_t req_lock;
1006 struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */
1007 struct drbd_tl_epoch *newest_tle;
1008 struct drbd_tl_epoch *oldest_tle;
1009 struct list_head out_of_sequence_requests;
1010 struct hlist_head *tl_hash;
1011 unsigned int tl_hash_s;
1013 /* blocks to sync in this run [unit BM_BLOCK_SIZE] */
1014 unsigned long rs_total;
1015 /* number of sync IOs that failed in this run */
1016 unsigned long rs_failed;
1017 /* Syncer's start time [unit jiffies] */
1018 unsigned long rs_start;
1019 /* cumulated time in PausedSyncX state [unit jiffies] */
1020 unsigned long rs_paused;
1021 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
1022 unsigned long rs_same_csum;
1023 #define DRBD_SYNC_MARKS 8
1024 #define DRBD_SYNC_MARK_STEP (3*HZ)
1025 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
1026 unsigned long rs_mark_left[DRBD_SYNC_MARKS];
1027 /* marks's time [unit jiffies] */
1028 unsigned long rs_mark_time[DRBD_SYNC_MARKS];
1029 /* current index into rs_mark_{left,time} */
1032 /* where does the admin want us to start? (sector) */
1033 sector_t ov_start_sector;
1034 /* where are we now? (sector) */
1035 sector_t ov_position;
1036 /* Start sector of out of sync range (to merge printk reporting). */
1037 sector_t ov_last_oos_start;
1038 /* size of out-of-sync range in sectors. */
1039 sector_t ov_last_oos_size;
1040 unsigned long ov_left; /* in bits */
1041 struct crypto_hash *csums_tfm;
1042 struct crypto_hash *verify_tfm;
1044 struct drbd_thread receiver;
1045 struct drbd_thread worker;
1046 struct drbd_thread asender;
1047 struct drbd_bitmap *bitmap;
1048 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
1050 /* Used to track operations of resync... */
1051 struct lru_cache *resync;
1052 /* Number of locked elements in resync LRU */
1053 unsigned int resync_locked;
1054 /* resync extent number waiting for application requests */
1055 unsigned int resync_wenr;
1059 struct drbd_epoch *current_epoch;
1060 spinlock_t epoch_lock;
1061 unsigned int epochs;
1062 enum write_ordering_e write_ordering;
1063 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
1064 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
1065 struct list_head done_ee; /* send ack */
1066 struct list_head read_ee; /* IO in progress (any read) */
1067 struct list_head net_ee; /* zero-copy network send in progress */
1068 struct hlist_head *ee_hash; /* is proteced by req_lock! */
1069 unsigned int ee_hash_s;
1071 /* this one is protected by ee_lock, single thread */
1072 struct drbd_epoch_entry *last_write_w_barrier;
1074 int next_barrier_nr;
1075 struct hlist_head *app_reads_hash; /* is proteced by req_lock */
1076 struct list_head resync_reads;
1077 atomic_t pp_in_use; /* allocated from page pool */
1078 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */
1079 wait_queue_head_t ee_wait;
1080 struct page *md_io_page; /* one page buffer for md_io */
1081 struct page *md_io_tmpp; /* for logical_block_size != 512 */
1082 struct mutex md_io_mutex; /* protects the md_io_buffer */
1084 wait_queue_head_t al_wait;
1085 struct lru_cache *act_log; /* activity log */
1086 unsigned int al_tr_number;
1088 int al_tr_pos; /* position of the next transaction in the journal */
1089 struct crypto_hash *cram_hmac_tfm;
1090 struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */
1091 struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */
1095 wait_queue_head_t seq_wait;
1096 atomic_t packet_seq;
1097 unsigned int peer_seq;
1098 spinlock_t peer_seq_lock;
1100 unsigned long comm_bm_set; /* communicated number of set bits. */
1101 cpumask_var_t cpu_mask;
1102 struct bm_io_work bm_io_work;
1103 u64 ed_uuid; /* UUID of the exposed data */
1104 struct mutex state_mutex;
1105 char congestion_reason; /* Why we where congested... */
1106 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
1107 atomic_t rs_sect_ev; /* for submitted resync data rate, both */
1108 int rs_last_sect_ev; /* counter to compare with */
1109 int rs_last_events; /* counter of read or write "events" (unit sectors)
1110 * on the lower level device when we last looked. */
1111 int c_sync_rate; /* current resync rate after syncer throttle magic */
1112 struct fifo_buffer rs_plan_s; /* correction values of resync planer */
1113 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
1114 int rs_planed; /* resync sectors already planed */
1115 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
1118 static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
1120 struct drbd_conf *mdev;
1122 mdev = minor < minor_count ? minor_table[minor] : NULL;
1127 static inline unsigned int mdev_to_minor(struct drbd_conf *mdev)
1132 /* returns 1 if it was successfull,
1133 * returns 0 if there was no data socket.
1134 * so wherever you are going to use the data.socket, e.g. do
1135 * if (!drbd_get_data_sock(mdev))
1138 * drbd_put_data_sock(mdev);
1140 static inline int drbd_get_data_sock(struct drbd_conf *mdev)
1142 mutex_lock(&mdev->data.mutex);
1143 /* drbd_disconnect() could have called drbd_free_sock()
1144 * while we were waiting in down()... */
1145 if (unlikely(mdev->data.socket == NULL)) {
1146 mutex_unlock(&mdev->data.mutex);
1152 static inline void drbd_put_data_sock(struct drbd_conf *mdev)
1154 mutex_unlock(&mdev->data.mutex);
1158 * function declarations
1159 *************************/
1163 enum chg_state_flags {
1166 CS_WAIT_COMPLETE = 4,
1168 CS_ORDERED = CS_WAIT_COMPLETE + CS_SERIALIZE,
1173 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1176 extern void drbd_init_set_defaults(struct drbd_conf *mdev);
1177 extern int drbd_change_state(struct drbd_conf *mdev, enum chg_state_flags f,
1178 union drbd_state mask, union drbd_state val);
1179 extern void drbd_force_state(struct drbd_conf *, union drbd_state,
1181 extern int _drbd_request_state(struct drbd_conf *, union drbd_state,
1182 union drbd_state, enum chg_state_flags);
1183 extern int __drbd_set_state(struct drbd_conf *, union drbd_state,
1184 enum chg_state_flags, struct completion *done);
1185 extern void print_st_err(struct drbd_conf *, union drbd_state,
1186 union drbd_state, int);
1187 extern int drbd_thread_start(struct drbd_thread *thi);
1188 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1190 extern void drbd_thread_current_set_cpu(struct drbd_conf *mdev);
1191 extern void drbd_calc_cpu_mask(struct drbd_conf *mdev);
1193 #define drbd_thread_current_set_cpu(A) ({})
1194 #define drbd_calc_cpu_mask(A) ({})
1196 extern void drbd_free_resources(struct drbd_conf *mdev);
1197 extern void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr,
1198 unsigned int set_size);
1199 extern void tl_clear(struct drbd_conf *mdev);
1200 enum drbd_req_event;
1201 extern void tl_restart(struct drbd_conf *mdev, enum drbd_req_event what);
1202 extern void _tl_add_barrier(struct drbd_conf *, struct drbd_tl_epoch *);
1203 extern void drbd_free_sock(struct drbd_conf *mdev);
1204 extern int drbd_send(struct drbd_conf *mdev, struct socket *sock,
1205 void *buf, size_t size, unsigned msg_flags);
1206 extern int drbd_send_protocol(struct drbd_conf *mdev);
1207 extern int drbd_send_uuids(struct drbd_conf *mdev);
1208 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
1209 extern int drbd_send_sync_uuid(struct drbd_conf *mdev, u64 val);
1210 extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags);
1211 extern int _drbd_send_state(struct drbd_conf *mdev);
1212 extern int drbd_send_state(struct drbd_conf *mdev);
1213 extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock,
1214 enum drbd_packets cmd, struct p_header80 *h,
1215 size_t size, unsigned msg_flags);
1216 #define USE_DATA_SOCKET 1
1217 #define USE_META_SOCKET 0
1218 extern int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket,
1219 enum drbd_packets cmd, struct p_header80 *h,
1221 extern int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd,
1222 char *data, size_t size);
1223 extern int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc);
1224 extern int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr,
1226 extern int drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd,
1227 struct drbd_epoch_entry *e);
1228 extern int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packets cmd,
1229 struct p_block_req *rp);
1230 extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd,
1231 struct p_data *dp, int data_size);
1232 extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd,
1233 sector_t sector, int blksize, u64 block_id);
1234 extern int drbd_send_oos(struct drbd_conf *mdev, struct drbd_request *req);
1235 extern int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd,
1236 struct drbd_epoch_entry *e);
1237 extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
1238 extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1239 sector_t sector, int size, u64 block_id);
1240 extern int drbd_send_drequest_csum(struct drbd_conf *mdev,
1241 sector_t sector,int size,
1242 void *digest, int digest_size,
1243 enum drbd_packets cmd);
1244 extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size);
1246 extern int drbd_send_bitmap(struct drbd_conf *mdev);
1247 extern int _drbd_send_bitmap(struct drbd_conf *mdev);
1248 extern int drbd_send_sr_reply(struct drbd_conf *mdev, int retcode);
1249 extern void drbd_free_bc(struct drbd_backing_dev *ldev);
1250 extern void drbd_mdev_cleanup(struct drbd_conf *mdev);
1252 /* drbd_meta-data.c (still in drbd_main.c) */
1253 extern void drbd_md_sync(struct drbd_conf *mdev);
1254 extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
1255 /* maybe define them below as inline? */
1256 extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1257 extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1258 extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1259 extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1260 extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
1261 extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
1262 extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
1263 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1264 #ifndef DRBD_DEBUG_MD_SYNC
1265 extern void drbd_md_mark_dirty(struct drbd_conf *mdev);
1267 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1268 extern void drbd_md_mark_dirty_(struct drbd_conf *mdev,
1269 unsigned int line, const char *func);
1271 extern void drbd_queue_bitmap_io(struct drbd_conf *mdev,
1272 int (*io_fn)(struct drbd_conf *),
1273 void (*done)(struct drbd_conf *, int),
1275 extern int drbd_bmio_set_n_write(struct drbd_conf *mdev);
1276 extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev);
1277 extern int drbd_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), char *why);
1278 extern void drbd_go_diskless(struct drbd_conf *mdev);
1279 extern void drbd_ldev_destroy(struct drbd_conf *mdev);
1283 We reserve a 128MB Block (4k aligned)
1284 * either at the end of the backing device
1285 * or on a separate meta data device. */
1287 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1288 /* The following numbers are sectors */
1289 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1290 #define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */
1291 /* Allows up to about 3.8TB */
1292 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1294 /* Since the smalles IO unit is usually 512 byte */
1295 #define MD_SECTOR_SHIFT 9
1296 #define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT)
1299 #define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1300 #define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */
1301 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1303 #if BITS_PER_LONG == 32
1305 #define cpu_to_lel(A) cpu_to_le32(A)
1306 #define lel_to_cpu(A) le32_to_cpu(A)
1307 #elif BITS_PER_LONG == 64
1309 #define cpu_to_lel(A) cpu_to_le64(A)
1310 #define lel_to_cpu(A) le64_to_cpu(A)
1312 #error "LN2 of BITS_PER_LONG unknown!"
1316 /* 16MB sized 'bitmap extent' to track syncer usage */
1318 int rs_left; /* number of bits set (out of sync) in this extent. */
1319 int rs_failed; /* number of failed resync requests in this extent. */
1320 unsigned long flags;
1321 struct lc_element lce;
1324 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1325 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1326 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1330 * We need to store one bit for a block.
1331 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1332 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1333 * Bit 1 ==> local node thinks this block needs to be synced.
1336 #define SLEEP_TIME (HZ/10)
1338 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1339 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1340 /* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1341 * per sector of on disk bitmap */
1342 #define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */
1343 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1345 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1346 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1349 /* thus many _storage_ sectors are described by one bit */
1350 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1351 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1352 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1354 /* bit to represented kilo byte conversion */
1355 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1357 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1358 * _storage_ sector is located in */
1359 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1361 /* how much _storage_ sectors we have per bitmap sector */
1362 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1363 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1365 /* in one sector of the bitmap, we have this many activity_log extents. */
1366 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1367 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1369 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1370 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1372 /* the extent in "PER_EXTENT" below is an activity log extent
1373 * we need that many (long words/bytes) to store the bitmap
1374 * of one AL_EXTENT_SIZE chunk of storage.
1375 * we can store the bitmap for that many AL_EXTENTS within
1376 * one sector of the _on_disk_ bitmap:
1377 * bit 0 bit 37 bit 38 bit (512*8)-1
1378 * ...|........|........|.. // ..|........|
1379 * sect. 0 `296 `304 ^(512*8*8)-1
1381 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1382 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1383 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1386 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1387 #define DRBD_MAX_SECTORS_BM \
1388 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1389 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1390 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1391 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1392 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1393 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1394 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1396 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1397 /* 16 TB in units of sectors */
1398 #if BITS_PER_LONG == 32
1399 /* adjust by one page worth of bitmap,
1400 * so we won't wrap around in drbd_bm_find_next_bit.
1401 * you should use 64bit OS for that much storage, anyways. */
1402 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1404 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0x1LU << 32)
1408 /* Sector shift value for the "hash" functions of tl_hash and ee_hash tables.
1409 * With a value of 8 all IO in one 128K block make it to the same slot of the
1412 #define DRBD_MAX_BIO_SIZE (1U<<(9+HT_SHIFT))
1414 #define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */
1416 /* Number of elements in the app_reads_hash */
1417 #define APP_R_HSIZE 15
1419 extern int drbd_bm_init(struct drbd_conf *mdev);
1420 extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits);
1421 extern void drbd_bm_cleanup(struct drbd_conf *mdev);
1422 extern void drbd_bm_set_all(struct drbd_conf *mdev);
1423 extern void drbd_bm_clear_all(struct drbd_conf *mdev);
1424 extern int drbd_bm_set_bits(
1425 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1426 extern int drbd_bm_clear_bits(
1427 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1428 /* bm_set_bits variant for use while holding drbd_bm_lock */
1429 extern void _drbd_bm_set_bits(struct drbd_conf *mdev,
1430 const unsigned long s, const unsigned long e);
1431 extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr);
1432 extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr);
1433 extern int drbd_bm_write_sect(struct drbd_conf *mdev, unsigned long enr) __must_hold(local);
1434 extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
1435 extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
1436 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev,
1437 unsigned long al_enr);
1438 extern size_t drbd_bm_words(struct drbd_conf *mdev);
1439 extern unsigned long drbd_bm_bits(struct drbd_conf *mdev);
1440 extern sector_t drbd_bm_capacity(struct drbd_conf *mdev);
1441 extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1442 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1443 extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1444 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo);
1445 extern unsigned long _drbd_bm_total_weight(struct drbd_conf *mdev);
1446 extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev);
1447 extern int drbd_bm_rs_done(struct drbd_conf *mdev);
1448 /* for receive_bitmap */
1449 extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset,
1450 size_t number, unsigned long *buffer);
1451 /* for _drbd_send_bitmap and drbd_bm_write_sect */
1452 extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset,
1453 size_t number, unsigned long *buffer);
1455 extern void drbd_bm_lock(struct drbd_conf *mdev, char *why);
1456 extern void drbd_bm_unlock(struct drbd_conf *mdev);
1458 extern int drbd_bm_count_bits(struct drbd_conf *mdev, const unsigned long s, const unsigned long e);
1461 extern struct kmem_cache *drbd_request_cache;
1462 extern struct kmem_cache *drbd_ee_cache; /* epoch entries */
1463 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1464 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1465 extern mempool_t *drbd_request_mempool;
1466 extern mempool_t *drbd_ee_mempool;
1468 extern struct page *drbd_pp_pool; /* drbd's page pool */
1469 extern spinlock_t drbd_pp_lock;
1470 extern int drbd_pp_vacant;
1471 extern wait_queue_head_t drbd_pp_wait;
1473 extern rwlock_t global_state_lock;
1475 extern struct drbd_conf *drbd_new_device(unsigned int minor);
1476 extern void drbd_free_mdev(struct drbd_conf *mdev);
1478 extern int proc_details;
1481 extern int drbd_make_request(struct request_queue *q, struct bio *bio);
1482 extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req);
1483 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1484 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1488 extern void drbd_suspend_io(struct drbd_conf *mdev);
1489 extern void drbd_resume_io(struct drbd_conf *mdev);
1490 extern char *ppsize(char *buf, unsigned long long size);
1491 extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int);
1492 enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
1493 extern enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local);
1494 extern void resync_after_online_grow(struct drbd_conf *);
1495 extern void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int) __must_hold(local);
1496 extern int drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role,
1498 extern enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev);
1499 extern void drbd_try_outdate_peer_async(struct drbd_conf *mdev);
1500 extern int drbd_khelper(struct drbd_conf *mdev, char *cmd);
1503 extern int drbd_worker(struct drbd_thread *thi);
1504 extern int drbd_alter_sa(struct drbd_conf *mdev, int na);
1505 extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side);
1506 extern void resume_next_sg(struct drbd_conf *mdev);
1507 extern void suspend_other_sg(struct drbd_conf *mdev);
1508 extern int drbd_resync_finished(struct drbd_conf *mdev);
1509 /* maybe rather drbd_main.c ? */
1510 extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
1511 struct drbd_backing_dev *bdev, sector_t sector, int rw);
1512 extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int);
1513 extern void drbd_rs_controller_reset(struct drbd_conf *mdev);
1515 static inline void ov_oos_print(struct drbd_conf *mdev)
1517 if (mdev->ov_last_oos_size) {
1518 dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n",
1519 (unsigned long long)mdev->ov_last_oos_start,
1520 (unsigned long)mdev->ov_last_oos_size);
1522 mdev->ov_last_oos_size=0;
1526 extern void drbd_csum_bio(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
1527 extern void drbd_csum_ee(struct drbd_conf *, struct crypto_hash *, struct drbd_epoch_entry *, void *);
1528 /* worker callbacks */
1529 extern int w_req_cancel_conflict(struct drbd_conf *, struct drbd_work *, int);
1530 extern int w_read_retry_remote(struct drbd_conf *, struct drbd_work *, int);
1531 extern int w_e_end_data_req(struct drbd_conf *, struct drbd_work *, int);
1532 extern int w_e_end_rsdata_req(struct drbd_conf *, struct drbd_work *, int);
1533 extern int w_e_end_csum_rs_req(struct drbd_conf *, struct drbd_work *, int);
1534 extern int w_e_end_ov_reply(struct drbd_conf *, struct drbd_work *, int);
1535 extern int w_e_end_ov_req(struct drbd_conf *, struct drbd_work *, int);
1536 extern int w_ov_finished(struct drbd_conf *, struct drbd_work *, int);
1537 extern int w_resync_inactive(struct drbd_conf *, struct drbd_work *, int);
1538 extern int w_resume_next_sg(struct drbd_conf *, struct drbd_work *, int);
1539 extern int w_send_write_hint(struct drbd_conf *, struct drbd_work *, int);
1540 extern int w_send_dblock(struct drbd_conf *, struct drbd_work *, int);
1541 extern int w_send_barrier(struct drbd_conf *, struct drbd_work *, int);
1542 extern int w_send_read_req(struct drbd_conf *, struct drbd_work *, int);
1543 extern int w_prev_work_done(struct drbd_conf *, struct drbd_work *, int);
1544 extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int);
1545 extern int w_restart_disk_io(struct drbd_conf *, struct drbd_work *, int);
1546 extern int w_send_oos(struct drbd_conf *, struct drbd_work *, int);
1547 extern int w_start_resync(struct drbd_conf *, struct drbd_work *, int);
1549 extern void resync_timer_fn(unsigned long data);
1551 /* drbd_receiver.c */
1552 extern int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector);
1553 extern int drbd_submit_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e,
1554 const unsigned rw, const int fault_type);
1555 extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list);
1556 extern struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev,
1559 unsigned int data_size,
1560 gfp_t gfp_mask) __must_hold(local);
1561 extern void drbd_free_some_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e,
1563 #define drbd_free_ee(m,e) drbd_free_some_ee(m, e, 0)
1564 #define drbd_free_net_ee(m,e) drbd_free_some_ee(m, e, 1)
1565 extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1566 struct list_head *head);
1567 extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1568 struct list_head *head);
1569 extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled);
1570 extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed);
1571 extern void drbd_flush_workqueue(struct drbd_conf *mdev);
1572 extern void drbd_free_tl_hash(struct drbd_conf *mdev);
1574 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1575 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1576 static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1577 char __user *optval, int optlen)
1580 if (level == SOL_SOCKET)
1581 err = sock_setsockopt(sock, level, optname, optval, optlen);
1583 err = sock->ops->setsockopt(sock, level, optname, optval,
1588 static inline void drbd_tcp_cork(struct socket *sock)
1591 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1592 (char __user *)&val, sizeof(val));
1595 static inline void drbd_tcp_uncork(struct socket *sock)
1598 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1599 (char __user *)&val, sizeof(val));
1602 static inline void drbd_tcp_nodelay(struct socket *sock)
1605 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1606 (char __user *)&val, sizeof(val));
1609 static inline void drbd_tcp_quickack(struct socket *sock)
1612 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1613 (char __user *)&val, sizeof(val));
1616 void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo);
1619 extern struct proc_dir_entry *drbd_proc;
1620 extern const struct file_operations drbd_proc_fops;
1621 extern const char *drbd_conn_str(enum drbd_conns s);
1622 extern const char *drbd_role_str(enum drbd_role s);
1625 extern void drbd_al_begin_io(struct drbd_conf *mdev, sector_t sector);
1626 extern void drbd_al_complete_io(struct drbd_conf *mdev, sector_t sector);
1627 extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector);
1628 extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1629 extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1630 extern void drbd_rs_cancel_all(struct drbd_conf *mdev);
1631 extern int drbd_rs_del_all(struct drbd_conf *mdev);
1632 extern void drbd_rs_failed_io(struct drbd_conf *mdev,
1633 sector_t sector, int size);
1634 extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *);
1635 extern void drbd_advance_rs_marks(struct drbd_conf *mdev, unsigned long still_to_go);
1636 extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
1637 int size, const char *file, const unsigned int line);
1638 #define drbd_set_in_sync(mdev, sector, size) \
1639 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1640 extern int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
1641 int size, const char *file, const unsigned int line);
1642 #define drbd_set_out_of_sync(mdev, sector, size) \
1643 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1644 extern void drbd_al_apply_to_bm(struct drbd_conf *mdev);
1645 extern void drbd_al_to_on_disk_bm(struct drbd_conf *mdev);
1646 extern void drbd_al_shrink(struct drbd_conf *mdev);
1651 void drbd_nl_cleanup(void);
1652 int __init drbd_nl_init(void);
1653 void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state);
1654 void drbd_bcast_sync_progress(struct drbd_conf *mdev);
1655 void drbd_bcast_ee(struct drbd_conf *mdev,
1656 const char *reason, const int dgs,
1657 const char* seen_hash, const char* calc_hash,
1658 const struct drbd_epoch_entry* e);
1662 * DOC: DRBD State macros
1664 * These macros are used to express state changes in easily readable form.
1666 * The NS macros expand to a mask and a value, that can be bit ored onto the
1667 * current state as soon as the spinlock (req_lock) was taken.
1669 * The _NS macros are used for state functions that get called with the
1670 * spinlock. These macros expand directly to the new state value.
1672 * Besides the basic forms NS() and _NS() additional _?NS[23] are defined
1673 * to express state changes that affect more than one aspect of the state.
1675 * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY)
1676 * Means that the network connection was established and that the peer
1677 * is in secondary role.
1679 #define role_MASK R_MASK
1680 #define peer_MASK R_MASK
1681 #define disk_MASK D_MASK
1682 #define pdsk_MASK D_MASK
1683 #define conn_MASK C_MASK
1685 #define user_isp_MASK 1
1686 #define aftr_isp_MASK 1
1687 #define susp_nod_MASK 1
1688 #define susp_fen_MASK 1
1691 ({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \
1692 ({ union drbd_state val; val.i = 0; val.T = (S); val; })
1693 #define NS2(T1, S1, T2, S2) \
1694 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1695 mask.T2 = T2##_MASK; mask; }), \
1696 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1697 val.T2 = (S2); val; })
1698 #define NS3(T1, S1, T2, S2, T3, S3) \
1699 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1700 mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \
1701 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1702 val.T2 = (S2); val.T3 = (S3); val; })
1704 #define _NS(D, T, S) \
1705 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; })
1706 #define _NS2(D, T1, S1, T2, S2) \
1707 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1708 __ns.T2 = (S2); __ns; })
1709 #define _NS3(D, T1, S1, T2, S2, T3, S3) \
1710 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1711 __ns.T2 = (S2); __ns.T3 = (S3); __ns; })
1714 * inline helper functions
1715 *************************/
1717 /* see also page_chain_add and friends in drbd_receiver.c */
1718 static inline struct page *page_chain_next(struct page *page)
1720 return (struct page *)page_private(page);
1722 #define page_chain_for_each(page) \
1723 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1724 page = page_chain_next(page))
1725 #define page_chain_for_each_safe(page, n) \
1726 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1728 static inline int drbd_bio_has_active_page(struct bio *bio)
1730 struct bio_vec *bvec;
1733 __bio_for_each_segment(bvec, bio, i, 0) {
1734 if (page_count(bvec->bv_page) > 1)
1741 static inline int drbd_ee_has_active_page(struct drbd_epoch_entry *e)
1743 struct page *page = e->pages;
1744 page_chain_for_each(page) {
1745 if (page_count(page) > 1)
1752 static inline void drbd_state_lock(struct drbd_conf *mdev)
1754 wait_event(mdev->misc_wait,
1755 !test_and_set_bit(CLUSTER_ST_CHANGE, &mdev->flags));
1758 static inline void drbd_state_unlock(struct drbd_conf *mdev)
1760 clear_bit(CLUSTER_ST_CHANGE, &mdev->flags);
1761 wake_up(&mdev->misc_wait);
1764 static inline int _drbd_set_state(struct drbd_conf *mdev,
1765 union drbd_state ns, enum chg_state_flags flags,
1766 struct completion *done)
1770 read_lock(&global_state_lock);
1771 rv = __drbd_set_state(mdev, ns, flags, done);
1772 read_unlock(&global_state_lock);
1778 * drbd_request_state() - Reqest a state change
1779 * @mdev: DRBD device.
1780 * @mask: mask of state bits to change.
1781 * @val: value of new state bits.
1783 * This is the most graceful way of requesting a state change. It is verbose
1784 * quite verbose in case the state change is not possible, and all those
1785 * state changes are globally serialized.
1787 static inline int drbd_request_state(struct drbd_conf *mdev,
1788 union drbd_state mask,
1789 union drbd_state val)
1791 return _drbd_request_state(mdev, mask, val, CS_VERBOSE + CS_ORDERED);
1794 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1795 static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach, const char *where)
1797 switch (mdev->ldev->dc.on_io_error) {
1800 if (__ratelimit(&drbd_ratelimit_state))
1801 dev_err(DEV, "Local IO failed in %s.\n", where);
1804 /* NOTE fall through to detach case if forcedetach set */
1806 case EP_CALL_HELPER:
1807 set_bit(WAS_IO_ERROR, &mdev->flags);
1808 if (mdev->state.disk > D_FAILED) {
1809 _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL);
1811 "Local IO failed in %s. Detaching...\n", where);
1818 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1819 * @mdev: DRBD device.
1820 * @error: Error code passed to the IO completion callback
1821 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1823 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1825 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1826 static inline void drbd_chk_io_error_(struct drbd_conf *mdev,
1827 int error, int forcedetach, const char *where)
1830 unsigned long flags;
1831 spin_lock_irqsave(&mdev->req_lock, flags);
1832 __drbd_chk_io_error_(mdev, forcedetach, where);
1833 spin_unlock_irqrestore(&mdev->req_lock, flags);
1839 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1840 * @bdev: Meta data block device.
1842 * BTW, for internal meta data, this happens to be the maximum capacity
1843 * we could agree upon with our peer node.
1845 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1847 switch (bdev->dc.meta_dev_idx) {
1848 case DRBD_MD_INDEX_INTERNAL:
1849 case DRBD_MD_INDEX_FLEX_INT:
1850 return bdev->md.md_offset + bdev->md.bm_offset;
1851 case DRBD_MD_INDEX_FLEX_EXT:
1853 return bdev->md.md_offset;
1858 * drbd_md_last_sector() - Return the last sector number of the meta data area
1859 * @bdev: Meta data block device.
1861 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1863 switch (bdev->dc.meta_dev_idx) {
1864 case DRBD_MD_INDEX_INTERNAL:
1865 case DRBD_MD_INDEX_FLEX_INT:
1866 return bdev->md.md_offset + MD_AL_OFFSET - 1;
1867 case DRBD_MD_INDEX_FLEX_EXT:
1869 return bdev->md.md_offset + bdev->md.md_size_sect;
1873 /* Returns the number of 512 byte sectors of the device */
1874 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1876 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1877 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0;
1881 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1882 * @bdev: Meta data block device.
1884 * returns the capacity we announce to out peer. we clip ourselves at the
1885 * various MAX_SECTORS, because if we don't, current implementation will
1886 * oops sooner or later
1888 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1891 switch (bdev->dc.meta_dev_idx) {
1892 case DRBD_MD_INDEX_INTERNAL:
1893 case DRBD_MD_INDEX_FLEX_INT:
1894 s = drbd_get_capacity(bdev->backing_bdev)
1895 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1896 drbd_md_first_sector(bdev))
1899 case DRBD_MD_INDEX_FLEX_EXT:
1900 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1901 drbd_get_capacity(bdev->backing_bdev));
1902 /* clip at maximum size the meta device can support */
1903 s = min_t(sector_t, s,
1904 BM_EXT_TO_SECT(bdev->md.md_size_sect
1905 - bdev->md.bm_offset));
1908 s = min_t(sector_t, DRBD_MAX_SECTORS,
1909 drbd_get_capacity(bdev->backing_bdev));
1915 * drbd_md_ss__() - Return the sector number of our meta data super block
1916 * @mdev: DRBD device.
1917 * @bdev: Meta data block device.
1919 static inline sector_t drbd_md_ss__(struct drbd_conf *mdev,
1920 struct drbd_backing_dev *bdev)
1922 switch (bdev->dc.meta_dev_idx) {
1923 default: /* external, some index */
1924 return MD_RESERVED_SECT * bdev->dc.meta_dev_idx;
1925 case DRBD_MD_INDEX_INTERNAL:
1926 /* with drbd08, internal meta data is always "flexible" */
1927 case DRBD_MD_INDEX_FLEX_INT:
1928 /* sizeof(struct md_on_disk_07) == 4k
1929 * position: last 4k aligned block of 4k size */
1930 if (!bdev->backing_bdev) {
1931 if (__ratelimit(&drbd_ratelimit_state)) {
1932 dev_err(DEV, "bdev->backing_bdev==NULL\n");
1937 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL)
1939 case DRBD_MD_INDEX_FLEX_EXT:
1945 drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w)
1947 unsigned long flags;
1948 spin_lock_irqsave(&q->q_lock, flags);
1949 list_add(&w->list, &q->q);
1950 up(&q->s); /* within the spinlock,
1951 see comment near end of drbd_worker() */
1952 spin_unlock_irqrestore(&q->q_lock, flags);
1956 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1958 unsigned long flags;
1959 spin_lock_irqsave(&q->q_lock, flags);
1960 list_add_tail(&w->list, &q->q);
1961 up(&q->s); /* within the spinlock,
1962 see comment near end of drbd_worker() */
1963 spin_unlock_irqrestore(&q->q_lock, flags);
1966 static inline void wake_asender(struct drbd_conf *mdev)
1968 if (test_bit(SIGNAL_ASENDER, &mdev->flags))
1969 force_sig(DRBD_SIG, mdev->asender.task);
1972 static inline void request_ping(struct drbd_conf *mdev)
1974 set_bit(SEND_PING, &mdev->flags);
1978 static inline int drbd_send_short_cmd(struct drbd_conf *mdev,
1979 enum drbd_packets cmd)
1981 struct p_header80 h;
1982 return drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, &h, sizeof(h));
1985 static inline int drbd_send_ping(struct drbd_conf *mdev)
1987 struct p_header80 h;
1988 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING, &h, sizeof(h));
1991 static inline int drbd_send_ping_ack(struct drbd_conf *mdev)
1993 struct p_header80 h;
1994 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING_ACK, &h, sizeof(h));
1997 static inline void drbd_thread_stop(struct drbd_thread *thi)
1999 _drbd_thread_stop(thi, FALSE, TRUE);
2002 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
2004 _drbd_thread_stop(thi, FALSE, FALSE);
2007 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
2009 _drbd_thread_stop(thi, TRUE, FALSE);
2012 /* counts how many answer packets packets we expect from our peer,
2013 * for either explicit application requests,
2014 * or implicit barrier packets as necessary.
2017 * _req_mod(req, queue_for_net_write or queue_for_net_read);
2018 * it is much easier and equally valid to count what we queue for the
2019 * worker, even before it actually was queued or send.
2020 * (drbd_make_request_common; recovery path on read io-error)
2022 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
2023 * _req_mod(req, data_received)
2024 * [from receive_DataReply]
2025 * _req_mod(req, write_acked_by_peer or recv_acked_by_peer or neg_acked)
2026 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
2027 * for some reason it is NOT decreased in got_NegAck,
2028 * but in the resulting cleanup code from report_params.
2029 * we should try to remember the reason for that...
2030 * _req_mod(req, send_failed or send_canceled)
2031 * _req_mod(req, connection_lost_while_pending)
2032 * [from tl_clear_barrier]
2034 static inline void inc_ap_pending(struct drbd_conf *mdev)
2036 atomic_inc(&mdev->ap_pending_cnt);
2039 #define ERR_IF_CNT_IS_NEGATIVE(which) \
2040 if (atomic_read(&mdev->which) < 0) \
2041 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
2042 __func__ , __LINE__ , \
2043 atomic_read(&mdev->which))
2045 #define dec_ap_pending(mdev) do { \
2046 typecheck(struct drbd_conf *, mdev); \
2047 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \
2048 wake_up(&mdev->misc_wait); \
2049 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
2051 /* counts how many resync-related answers we still expect from the peer
2053 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
2054 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK whith ID_SYNCER)
2055 * (or P_NEG_ACK with ID_SYNCER)
2057 static inline void inc_rs_pending(struct drbd_conf *mdev)
2059 atomic_inc(&mdev->rs_pending_cnt);
2062 #define dec_rs_pending(mdev) do { \
2063 typecheck(struct drbd_conf *, mdev); \
2064 atomic_dec(&mdev->rs_pending_cnt); \
2065 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
2067 /* counts how many answers we still need to send to the peer.
2069 * receive_Data unless protocol A;
2070 * we need to send a P_RECV_ACK (proto B)
2071 * or P_WRITE_ACK (proto C)
2072 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2073 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2074 * receive_Barrier_* we need to send a P_BARRIER_ACK
2076 static inline void inc_unacked(struct drbd_conf *mdev)
2078 atomic_inc(&mdev->unacked_cnt);
2081 #define dec_unacked(mdev) do { \
2082 typecheck(struct drbd_conf *, mdev); \
2083 atomic_dec(&mdev->unacked_cnt); \
2084 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2086 #define sub_unacked(mdev, n) do { \
2087 typecheck(struct drbd_conf *, mdev); \
2088 atomic_sub(n, &mdev->unacked_cnt); \
2089 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2092 static inline void put_net_conf(struct drbd_conf *mdev)
2094 if (atomic_dec_and_test(&mdev->net_cnt))
2095 wake_up(&mdev->net_cnt_wait);
2099 * get_net_conf() - Increase ref count on mdev->net_conf; Returns 0 if nothing there
2100 * @mdev: DRBD device.
2102 * You have to call put_net_conf() when finished working with mdev->net_conf.
2104 static inline int get_net_conf(struct drbd_conf *mdev)
2108 atomic_inc(&mdev->net_cnt);
2109 have_net_conf = mdev->state.conn >= C_UNCONNECTED;
2112 return have_net_conf;
2116 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2119 * You have to call put_ldev() when finished working with mdev->ldev.
2121 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2122 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2124 static inline void put_ldev(struct drbd_conf *mdev)
2126 int i = atomic_dec_return(&mdev->local_cnt);
2130 if (mdev->state.disk == D_DISKLESS)
2131 /* even internal references gone, safe to destroy */
2132 drbd_ldev_destroy(mdev);
2133 if (mdev->state.disk == D_FAILED)
2134 /* all application IO references gone. */
2135 drbd_go_diskless(mdev);
2136 wake_up(&mdev->misc_wait);
2141 static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
2145 /* never get a reference while D_DISKLESS */
2146 if (mdev->state.disk == D_DISKLESS)
2149 atomic_inc(&mdev->local_cnt);
2150 io_allowed = (mdev->state.disk >= mins);
2156 extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins);
2159 /* you must have an "get_ldev" reference */
2160 static inline void drbd_get_syncer_progress(struct drbd_conf *mdev,
2161 unsigned long *bits_left, unsigned int *per_mil_done)
2164 * this is to break it at compile time when we change that
2165 * (we may feel 4TB maximum storage per drbd is not enough)
2167 typecheck(unsigned long, mdev->rs_total);
2169 /* note: both rs_total and rs_left are in bits, i.e. in
2170 * units of BM_BLOCK_SIZE.
2171 * for the percentage, we don't care. */
2173 if (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T)
2174 *bits_left = mdev->ov_left;
2176 *bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
2177 /* >> 10 to prevent overflow,
2178 * +1 to prevent division by zero */
2179 if (*bits_left > mdev->rs_total) {
2180 /* doh. maybe a logic bug somewhere.
2181 * may also be just a race condition
2182 * between this and a disconnect during sync.
2183 * for now, just prevent in-kernel buffer overflow.
2186 dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2187 drbd_conn_str(mdev->state.conn),
2188 *bits_left, mdev->rs_total, mdev->rs_failed);
2191 /* make sure the calculation happens in long context */
2192 unsigned long tmp = 1000UL -
2193 (*bits_left >> 10)*1000UL
2194 / ((mdev->rs_total >> 10) + 1UL);
2195 *per_mil_done = tmp;
2200 /* this throttles on-the-fly application requests
2201 * according to max_buffers settings;
2202 * maybe re-implement using semaphores? */
2203 static inline int drbd_get_max_buffers(struct drbd_conf *mdev)
2205 int mxb = 1000000; /* arbitrary limit on open requests */
2206 if (get_net_conf(mdev)) {
2207 mxb = mdev->net_conf->max_buffers;
2213 static inline int drbd_state_is_stable(struct drbd_conf *mdev)
2215 union drbd_state s = mdev->state;
2217 /* DO NOT add a default clause, we want the compiler to warn us
2218 * for any newly introduced state we may have forgotten to add here */
2220 switch ((enum drbd_conns)s.conn) {
2221 /* new io only accepted when there is no connection, ... */
2223 case C_WF_CONNECTION:
2224 /* ... or there is a well established connection. */
2230 case C_PAUSED_SYNC_S:
2231 case C_PAUSED_SYNC_T:
2234 /* transitional states, IO allowed */
2235 case C_DISCONNECTING:
2239 case C_NETWORK_FAILURE:
2240 case C_PROTOCOL_ERROR:
2242 case C_WF_REPORT_PARAMS:
2243 case C_STARTING_SYNC_S:
2244 case C_STARTING_SYNC_T:
2247 /* Allow IO in BM exchange states with new protocols */
2249 if (mdev->agreed_pro_version < 96)
2253 /* no new io accepted in these states */
2255 case C_WF_SYNC_UUID:
2261 switch ((enum drbd_disk_state)s.disk) {
2263 case D_INCONSISTENT:
2267 /* disk state is stable as well. */
2270 /* no new io accepted during tansitional states */
2283 static inline int is_susp(union drbd_state s)
2285 return s.susp || s.susp_nod || s.susp_fen;
2288 static inline int __inc_ap_bio_cond(struct drbd_conf *mdev)
2290 int mxb = drbd_get_max_buffers(mdev);
2292 if (is_susp(mdev->state))
2294 if (test_bit(SUSPEND_IO, &mdev->flags))
2297 /* to avoid potential deadlock or bitmap corruption,
2298 * in various places, we only allow new application io
2299 * to start during "stable" states. */
2301 /* no new io accepted when attaching or detaching the disk */
2302 if (!drbd_state_is_stable(mdev))
2305 /* since some older kernels don't have atomic_add_unless,
2306 * and we are within the spinlock anyways, we have this workaround. */
2307 if (atomic_read(&mdev->ap_bio_cnt) > mxb)
2309 if (test_bit(BITMAP_IO, &mdev->flags))
2314 static inline int _inc_ap_bio_cond(struct drbd_conf *mdev, int count)
2318 spin_lock_irq(&mdev->req_lock);
2319 rv = __inc_ap_bio_cond(mdev);
2321 atomic_add(count, &mdev->ap_bio_cnt);
2322 spin_unlock_irq(&mdev->req_lock);
2327 static inline void inc_ap_bio(struct drbd_conf *mdev, int count)
2330 * as long as the device is suspended
2331 * until the bitmap is no longer on the fly during connection
2332 * handshake as long as we would exeed the max_buffer limit.
2334 * to avoid races with the reconnect code,
2335 * we need to atomic_inc within the spinlock. */
2337 wait_event(mdev->misc_wait, _inc_ap_bio_cond(mdev, count));
2340 static inline void dec_ap_bio(struct drbd_conf *mdev)
2342 int mxb = drbd_get_max_buffers(mdev);
2343 int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt);
2345 D_ASSERT(ap_bio >= 0);
2346 /* this currently does wake_up for every dec_ap_bio!
2347 * maybe rather introduce some type of hysteresis?
2348 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2350 wake_up(&mdev->misc_wait);
2351 if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) {
2352 if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags))
2353 drbd_queue_work(&mdev->data.work, &mdev->bm_io_work.w);
2357 static inline void drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val)
2359 mdev->ed_uuid = val;
2362 static inline int seq_cmp(u32 a, u32 b)
2364 /* we assume wrap around at 32bit.
2365 * for wrap around at 24bit (old atomic_t),
2369 return (s32)(a) - (s32)(b);
2371 #define seq_lt(a, b) (seq_cmp((a), (b)) < 0)
2372 #define seq_gt(a, b) (seq_cmp((a), (b)) > 0)
2373 #define seq_ge(a, b) (seq_cmp((a), (b)) >= 0)
2374 #define seq_le(a, b) (seq_cmp((a), (b)) <= 0)
2375 /* CAUTION: please no side effects in arguments! */
2376 #define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b)))
2378 static inline void update_peer_seq(struct drbd_conf *mdev, unsigned int new_seq)
2381 spin_lock(&mdev->peer_seq_lock);
2382 m = seq_max(mdev->peer_seq, new_seq);
2384 spin_unlock(&mdev->peer_seq_lock);
2386 wake_up(&mdev->seq_wait);
2389 static inline void drbd_update_congested(struct drbd_conf *mdev)
2391 struct sock *sk = mdev->data.socket->sk;
2392 if (sk->sk_wmem_queued > sk->sk_sndbuf * 4 / 5)
2393 set_bit(NET_CONGESTED, &mdev->flags);
2396 static inline int drbd_queue_order_type(struct drbd_conf *mdev)
2398 /* sorry, we currently have no working implementation
2399 * of distributed TCQ stuff */
2400 #ifndef QUEUE_ORDERED_NONE
2401 #define QUEUE_ORDERED_NONE 0
2403 return QUEUE_ORDERED_NONE;
2406 static inline void drbd_md_flush(struct drbd_conf *mdev)
2410 if (test_bit(MD_NO_FUA, &mdev->flags))
2413 r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL);
2415 set_bit(MD_NO_FUA, &mdev->flags);
2416 dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r);