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/tcp.h>
38 #include <linux/mutex.h>
39 #include <linux/major.h>
40 #include <linux/blkdev.h>
41 #include <linux/genhd.h>
43 #include <linux/lru_cache.h>
46 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
47 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
48 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
49 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
51 # define __protected_by(x)
52 # define __protected_read_by(x)
53 # define __protected_write_by(x)
54 # define __must_hold(x)
57 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
59 /* module parameter, defined in drbd_main.c */
60 extern unsigned int minor_count;
61 extern int disable_sendpage;
63 extern unsigned int cn_idx;
65 #ifdef CONFIG_DRBD_FAULT_INJECTION
66 extern int enable_faults;
67 extern int fault_rate;
68 extern int fault_devs;
71 extern char usermode_helper[];
81 /* I don't remember why XCPU ...
82 * This is used to wake the asender,
83 * and to interrupt sending the sending task
86 #define DRBD_SIG SIGXCPU
88 /* This is used to stop/restart our threads.
89 * Cannot use SIGTERM nor SIGKILL, since these
90 * are sent out by init on runlevel changes
91 * I choose SIGHUP for now.
93 #define DRBD_SIGKILL SIGHUP
95 /* All EEs on the free list should have ID_VACANT (== 0)
96 * freshly allocated EEs get !ID_VACANT (== 1)
97 * so if it says "cannot dereference null pointer at adress 0x00000001",
98 * it is most likely one of these :( */
100 #define ID_IN_SYNC (4711ULL)
101 #define ID_OUT_OF_SYNC (4712ULL)
103 #define ID_SYNCER (-1ULL)
105 #define is_syncer_block_id(id) ((id) == ID_SYNCER)
110 /* to shorten dev_warn(DEV, "msg"); and relatives statements */
111 #define DEV (disk_to_dev(mdev->vdisk))
113 #define D_ASSERT(exp) if (!(exp)) \
114 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
116 #define ERR_IF(exp) if (({ \
117 int _b = (exp) != 0; \
118 if (_b) dev_err(DEV, "%s: (%s) in %s:%d\n", \
119 __func__, #exp, __FILE__, __LINE__); \
123 /* Defines to control fault insertion */
125 DRBD_FAULT_MD_WR = 0, /* meta data write */
126 DRBD_FAULT_MD_RD = 1, /* read */
127 DRBD_FAULT_RS_WR = 2, /* resync */
128 DRBD_FAULT_RS_RD = 3,
129 DRBD_FAULT_DT_WR = 4, /* data */
130 DRBD_FAULT_DT_RD = 5,
131 DRBD_FAULT_DT_RA = 6, /* data read ahead */
132 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */
133 DRBD_FAULT_AL_EE = 8, /* alloc ee */
138 extern void trace_drbd_resync(struct drbd_conf *mdev, int level, const char *fmt, ...);
140 #ifdef CONFIG_DRBD_FAULT_INJECTION
142 _drbd_insert_fault(struct drbd_conf *mdev, unsigned int type);
144 drbd_insert_fault(struct drbd_conf *mdev, unsigned int type) {
146 (enable_faults & (1<<type)) &&
147 _drbd_insert_fault(mdev, type);
149 #define FAULT_ACTIVE(_m, _t) (drbd_insert_fault((_m), (_t)))
152 #define FAULT_ACTIVE(_m, _t) (0)
155 /* integer division, round _UP_ to the next integer */
156 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
157 /* usual integer division */
158 #define div_floor(A, B) ((A)/(B))
160 /* drbd_meta-data.c (still in drbd_main.c) */
161 /* 4th incarnation of the disk layout. */
162 #define DRBD_MD_MAGIC (DRBD_MAGIC+4)
164 extern struct drbd_conf **minor_table;
165 extern struct ratelimit_state drbd_ratelimit_state;
169 /* receiver (data socket) */
171 P_DATA_REPLY = 0x01, /* Response to P_DATA_REQUEST */
172 P_RS_DATA_REPLY = 0x02, /* Response to P_RS_DATA_REQUEST */
175 P_BECOME_SYNC_TARGET = 0x05,
176 P_BECOME_SYNC_SOURCE = 0x06,
177 P_UNPLUG_REMOTE = 0x07, /* Used at various times to hint the peer */
178 P_DATA_REQUEST = 0x08, /* Used to ask for a data block */
179 P_RS_DATA_REQUEST = 0x09, /* Used to ask for a data block for resync */
186 P_AUTH_CHALLENGE = 0x10,
187 P_AUTH_RESPONSE = 0x11,
188 P_STATE_CHG_REQ = 0x12,
190 /* asender (meta socket */
193 P_RECV_ACK = 0x15, /* Used in protocol B */
194 P_WRITE_ACK = 0x16, /* Used in protocol C */
195 P_RS_WRITE_ACK = 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
196 P_DISCARD_ACK = 0x18, /* Used in proto C, two-primaries conflict detection */
197 P_NEG_ACK = 0x19, /* Sent if local disk is unusable */
198 P_NEG_DREPLY = 0x1a, /* Local disk is broken... */
199 P_NEG_RS_DREPLY = 0x1b, /* Local disk is broken... */
200 P_BARRIER_ACK = 0x1c,
201 P_STATE_CHG_REPLY = 0x1d,
203 /* "new" commands, no longer fitting into the ordering scheme above */
205 P_OV_REQUEST = 0x1e, /* data socket */
207 P_OV_RESULT = 0x20, /* meta socket */
208 P_CSUM_RS_REQUEST = 0x21, /* data socket */
209 P_RS_IS_IN_SYNC = 0x22, /* meta socket */
210 P_SYNC_PARAM89 = 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
211 P_COMPRESSED_BITMAP = 0x24, /* compressed or otherwise encoded bitmap transfer */
214 P_MAY_IGNORE = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
215 P_MAX_OPT_CMD = 0x101,
217 /* special command ids for handshake */
219 P_HAND_SHAKE_M = 0xfff1, /* First Packet on the MetaSock */
220 P_HAND_SHAKE_S = 0xfff2, /* First Packet on the Socket */
222 P_HAND_SHAKE = 0xfffe /* FIXED for the next century! */
225 static inline const char *cmdname(enum drbd_packets cmd)
227 /* THINK may need to become several global tables
228 * when we want to support more than
230 static const char *cmdnames[] = {
232 [P_DATA_REPLY] = "DataReply",
233 [P_RS_DATA_REPLY] = "RSDataReply",
234 [P_BARRIER] = "Barrier",
235 [P_BITMAP] = "ReportBitMap",
236 [P_BECOME_SYNC_TARGET] = "BecomeSyncTarget",
237 [P_BECOME_SYNC_SOURCE] = "BecomeSyncSource",
238 [P_UNPLUG_REMOTE] = "UnplugRemote",
239 [P_DATA_REQUEST] = "DataRequest",
240 [P_RS_DATA_REQUEST] = "RSDataRequest",
241 [P_SYNC_PARAM] = "SyncParam",
242 [P_SYNC_PARAM89] = "SyncParam89",
243 [P_PROTOCOL] = "ReportProtocol",
244 [P_UUIDS] = "ReportUUIDs",
245 [P_SIZES] = "ReportSizes",
246 [P_STATE] = "ReportState",
247 [P_SYNC_UUID] = "ReportSyncUUID",
248 [P_AUTH_CHALLENGE] = "AuthChallenge",
249 [P_AUTH_RESPONSE] = "AuthResponse",
251 [P_PING_ACK] = "PingAck",
252 [P_RECV_ACK] = "RecvAck",
253 [P_WRITE_ACK] = "WriteAck",
254 [P_RS_WRITE_ACK] = "RSWriteAck",
255 [P_DISCARD_ACK] = "DiscardAck",
256 [P_NEG_ACK] = "NegAck",
257 [P_NEG_DREPLY] = "NegDReply",
258 [P_NEG_RS_DREPLY] = "NegRSDReply",
259 [P_BARRIER_ACK] = "BarrierAck",
260 [P_STATE_CHG_REQ] = "StateChgRequest",
261 [P_STATE_CHG_REPLY] = "StateChgReply",
262 [P_OV_REQUEST] = "OVRequest",
263 [P_OV_REPLY] = "OVReply",
264 [P_OV_RESULT] = "OVResult",
268 if (cmd == P_HAND_SHAKE_M)
270 if (cmd == P_HAND_SHAKE_S)
272 if (cmd == P_HAND_SHAKE)
274 if (cmd >= P_MAX_CMD)
276 return cmdnames[cmd];
279 /* for sending/receiving the bitmap,
280 * possibly in some encoding scheme */
283 * stores total bits and long words
284 * of the bitmap, so we don't need to
285 * call the accessor functions over and again. */
286 unsigned long bm_bits;
287 unsigned long bm_words;
288 /* during xfer, current position within the bitmap */
289 unsigned long bit_offset;
290 unsigned long word_offset;
292 /* statistics; index: (h->command == P_BITMAP) */
297 extern void INFO_bm_xfer_stats(struct drbd_conf *mdev,
298 const char *direction, struct bm_xfer_ctx *c);
300 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
302 /* word_offset counts "native long words" (32 or 64 bit),
304 * Encoded packet may end at an unaligned bit offset.
305 * In case a fallback clear text packet is transmitted in
306 * between, we adjust this offset back to the last 64bit
307 * aligned "native long word", which makes coding and decoding
308 * the plain text bitmap much more convenient. */
309 #if BITS_PER_LONG == 64
310 c->word_offset = c->bit_offset >> 6;
311 #elif BITS_PER_LONG == 32
312 c->word_offset = c->bit_offset >> 5;
313 c->word_offset &= ~(1UL);
315 # error "unsupported BITS_PER_LONG"
320 #define __packed __attribute__((packed))
323 /* This is the layout for a packet on the wire.
324 * The byteorder is the network byte order.
325 * (except block_id and barrier fields.
326 * these are pointers to local structs
327 * and have no relevance for the partner,
328 * which just echoes them as received.)
330 * NOTE that the payload starts at a long aligned offset,
331 * regardless of 32 or 64 bit arch!
336 u16 length; /* bytes of data after this header */
339 /* 8 bytes. packet FIXED for the next century! */
342 * short commands, packets without payload, plain p_header:
345 * P_BECOME_SYNC_TARGET
346 * P_BECOME_SYNC_SOURCE
351 * commands with out-of-struct payload:
352 * P_BITMAP (no additional fields)
353 * P_DATA, P_DATA_REPLY (see p_data)
354 * P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
357 /* these defines must not be changed without changing the protocol version */
358 #define DP_HARDBARRIER 1
360 #define DP_MAY_SET_IN_SYNC 4
363 struct p_header head;
364 u64 sector; /* 64 bits sector number */
365 u64 block_id; /* to identify the request in protocol B&C */
371 * commands which share a struct:
373 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
374 * P_DISCARD_ACK (proto C, two-primaries conflict detection)
376 * P_DATA_REQUEST, P_RS_DATA_REQUEST
379 struct p_header head;
388 struct p_header head;
392 u32 pad; /* to multiple of 8 Byte */
396 * commands with their own struct for additional fields:
405 struct p_header head; /* 8 bytes */
410 /* should be more than enough for future enhancements
411 * for now, feature_flags and the reserverd array shall be zero.
417 /* 80 bytes, FIXED for the next century */
420 struct p_header head;
421 u32 barrier; /* barrier number _handle_ only */
422 u32 pad; /* to multiple of 8 Byte */
425 struct p_barrier_ack {
426 struct p_header head;
432 struct p_header head;
435 /* Since protocol version 88 and higher. */
439 struct p_rs_param_89 {
440 struct p_header head;
442 /* protocol version 89: */
443 char verify_alg[SHARED_SECRET_MAX];
444 char csums_alg[SHARED_SECRET_MAX];
448 struct p_header head;
456 /* Since protocol version 87 and higher. */
457 char integrity_alg[0];
462 struct p_header head;
463 u64 uuid[UI_EXTENDED_SIZE];
467 struct p_header head;
472 struct p_header head;
473 u64 d_size; /* size of disk */
474 u64 u_size; /* user requested size */
475 u64 c_size; /* current exported size */
476 u32 max_segment_size; /* Maximal size of a BIO */
477 u32 queue_order_type;
481 struct p_header head;
486 struct p_header head;
491 struct p_req_state_reply {
492 struct p_header head;
496 struct p_drbd06_param {
507 struct p_header head;
513 /* Valid values for the encoding field.
514 * Bump proto version when changing this. */
515 enum drbd_bitmap_code {
516 /* RLE_VLI_Bytes = 0,
517 * and other bit variants had been defined during
518 * algorithm evaluation. */
522 struct p_compressed_bm {
523 struct p_header head;
524 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
525 * (encoding & 0x80): polarity (set/unset) of first runlength
526 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
527 * used to pad up to head.length bytes
534 /* DCBP: Drbd Compressed Bitmap Packet ... */
535 static inline enum drbd_bitmap_code
536 DCBP_get_code(struct p_compressed_bm *p)
538 return (enum drbd_bitmap_code)(p->encoding & 0x0f);
542 DCBP_set_code(struct p_compressed_bm *p, enum drbd_bitmap_code code)
545 p->encoding = (p->encoding & ~0xf) | code;
549 DCBP_get_start(struct p_compressed_bm *p)
551 return (p->encoding & 0x80) != 0;
555 DCBP_set_start(struct p_compressed_bm *p, int set)
557 p->encoding = (p->encoding & ~0x80) | (set ? 0x80 : 0);
561 DCBP_get_pad_bits(struct p_compressed_bm *p)
563 return (p->encoding >> 4) & 0x7;
567 DCBP_set_pad_bits(struct p_compressed_bm *p, int n)
570 p->encoding = (p->encoding & (~0x7 << 4)) | (n << 4);
573 /* one bitmap packet, including the p_header,
574 * should fit within one _architecture independend_ page.
575 * so we need to use the fixed size 4KiB page size
576 * most architechtures have used for a long time.
578 #define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header))
579 #define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
580 #define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm))
581 #if (PAGE_SIZE < 4096)
582 /* drbd_send_bitmap / receive_bitmap would break horribly */
583 #error "PAGE_SIZE too small"
587 struct p_header header;
588 struct p_handshake handshake;
590 struct p_block_ack block_ack;
591 struct p_barrier barrier;
592 struct p_barrier_ack barrier_ack;
593 struct p_rs_param_89 rs_param_89;
594 struct p_protocol protocol;
595 struct p_sizes sizes;
596 struct p_uuids uuids;
597 struct p_state state;
598 struct p_req_state req_state;
599 struct p_req_state_reply req_state_reply;
600 struct p_block_req block_req;
603 /**********************************************************************/
604 enum drbd_thread_state {
613 struct task_struct *task;
614 struct completion stop;
615 enum drbd_thread_state t_state;
616 int (*function) (struct drbd_thread *);
617 struct drbd_conf *mdev;
621 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
623 /* THINK testing the t_state seems to be uncritical in all cases
624 * (but thread_{start,stop}), so we can read it *without* the lock.
633 * Having this as the first member of a struct provides sort of "inheritance".
634 * "derived" structs can be "drbd_queue_work()"ed.
635 * The callback should know and cast back to the descendant struct.
636 * drbd_request and drbd_epoch_entry are descendants of drbd_work.
639 typedef int (*drbd_work_cb)(struct drbd_conf *, struct drbd_work *, int cancel);
641 struct list_head list;
645 struct drbd_tl_epoch;
646 struct drbd_request {
648 struct drbd_conf *mdev;
650 /* if local IO is not allowed, will be NULL.
651 * if local IO _is_ allowed, holds the locally submitted bio clone,
652 * or, after local IO completion, the ERR_PTR(error).
653 * see drbd_endio_pri(). */
654 struct bio *private_bio;
656 struct hlist_node colision;
659 unsigned int epoch; /* barrier_nr */
661 /* barrier_nr: used to check on "completion" whether this req was in
662 * the current epoch, and we therefore have to close it,
663 * starting a new epoch...
666 /* up to here, the struct layout is identical to drbd_epoch_entry;
667 * we might be able to use that to our advantage... */
669 struct list_head tl_requests; /* ring list in the transfer log */
670 struct bio *master_bio; /* master bio pointer */
671 unsigned long rq_state; /* see comments above _req_mod() */
673 unsigned long start_time;
676 struct drbd_tl_epoch {
678 struct list_head requests; /* requests before */
679 struct drbd_tl_epoch *next; /* pointer to the next barrier */
680 unsigned int br_number; /* the barriers identifier. */
681 int n_req; /* number of requests attached before this barrier */
686 /* These Tl_epoch_entries may be in one of 6 lists:
687 active_ee .. data packet being written
688 sync_ee .. syncer block being written
689 done_ee .. block written, need to send P_WRITE_ACK
690 read_ee .. [RS]P_DATA_REQUEST being read
694 struct list_head list;
695 unsigned int barrier_nr;
696 atomic_t epoch_size; /* increased on every request added. */
697 atomic_t active; /* increased on every req. added, and dec on every finished. */
701 /* drbd_epoch flag bits */
703 DE_BARRIER_IN_NEXT_EPOCH_ISSUED,
704 DE_BARRIER_IN_NEXT_EPOCH_DONE,
705 DE_CONTAINS_A_BARRIER,
706 DE_HAVE_BARRIER_NUMBER,
715 EV_TRACE_FLUSH, /* TRACE_ are not real events, only used for tracing */
716 EV_TRACE_ADD_BARRIER, /* Doing the first write as a barrier write */
717 EV_TRACE_SETTING_BI, /* Barrier is expressed with the first write of the next epoch */
720 EV_CLEANUP = 32, /* used as flag */
723 struct drbd_epoch_entry {
725 struct drbd_conf *mdev;
726 struct bio *private_bio;
727 struct hlist_node colision;
730 struct drbd_epoch *epoch;
732 /* up to here, the struct layout is identical to drbd_request;
733 * we might be able to use that to our advantage... */
739 struct drbd_wq_barrier {
741 struct completion done;
751 __EE_CALL_AL_COMPLETE_IO,
752 __EE_CONFLICT_PENDING,
753 __EE_MAY_SET_IN_SYNC,
756 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
757 #define EE_CONFLICT_PENDING (1<<__EE_CONFLICT_PENDING)
758 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
759 #define EE_IS_BARRIER (1<<__EE_IS_BARRIER)
761 /* global flag bits */
763 CREATE_BARRIER, /* next P_DATA is preceeded by a P_BARRIER */
764 SIGNAL_ASENDER, /* whether asender wants to be interrupted */
765 SEND_PING, /* whether asender should send a ping asap */
767 STOP_SYNC_TIMER, /* tell timer to cancel itself */
768 UNPLUG_QUEUED, /* only relevant with kernel 2.4 */
769 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
770 MD_DIRTY, /* current uuids and flags not yet on disk */
771 DISCARD_CONCURRENT, /* Set on one node, cleared on the peer! */
772 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
773 CLUSTER_ST_CHANGE, /* Cluster wide state change going on... */
776 CRASHED_PRIMARY, /* This node was a crashed primary.
777 * Gets cleared when the state.conn
778 * goes into C_CONNECTED state. */
779 WRITE_BM_AFTER_RESYNC, /* A kmalloc() during resync failed */
780 NO_BARRIER_SUPP, /* underlying block device doesn't implement barriers */
783 MD_NO_BARRIER, /* meta data device does not support barriers,
785 SUSPEND_IO, /* suspend application io */
786 BITMAP_IO, /* suspend application io;
787 once no more io in flight, start bitmap io */
788 BITMAP_IO_QUEUED, /* Started bitmap IO */
789 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
790 NET_CONGESTED, /* The data socket is congested */
792 CONFIG_PENDING, /* serialization of (re)configuration requests.
793 * if set, also prevents the device from dying */
794 DEVICE_DYING, /* device became unconfigured,
795 * but worker thread is still handling the cleanup.
796 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
797 * while this is set. */
798 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
799 * the peer, if it changed there as well. */
802 struct drbd_bitmap; /* opaque for drbd_conf */
804 /* TODO sort members for performance
805 * MAYBE group them further */
807 /* THINK maybe we actually want to use the default "event/%s" worker threads
808 * or similar in linux 2.6, which uses per cpu data and threads.
810 * To be general, this might need a spin_lock member.
811 * For now, please use the mdev->req_lock to protect list_head,
812 * see drbd_queue_work below.
814 struct drbd_work_queue {
816 struct semaphore s; /* producers up it, worker down()s it */
817 spinlock_t q_lock; /* to protect the list. */
821 struct drbd_work_queue work;
823 struct socket *socket;
824 /* this way we get our
825 * send/receive buffers off the stack */
826 union p_polymorph sbuf;
827 union p_polymorph rbuf;
831 u64 md_offset; /* sector offset to 'super' block */
833 u64 la_size_sect; /* last agreed size, unit sectors */
839 s32 al_offset; /* signed relative sector offset to al area */
840 s32 bm_offset; /* signed relative sector offset to bitmap */
842 /* u32 al_nr_extents; important for restoring the AL
843 * is stored into sync_conf.al_extents, which in turn
844 * gets applied to act_log->nr_elements
848 /* for sync_conf and other types... */
849 #define NL_PACKET(name, number, fields) struct name { fields };
850 #define NL_INTEGER(pn,pr,member) int member;
851 #define NL_INT64(pn,pr,member) __u64 member;
852 #define NL_BIT(pn,pr,member) unsigned member:1;
853 #define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
854 #include "linux/drbd_nl.h"
856 struct drbd_backing_dev {
857 struct block_device *backing_bdev;
858 struct block_device *md_bdev;
859 struct file *lo_file;
860 struct file *md_file;
862 struct disk_conf dc; /* The user provided config... */
863 sector_t known_size; /* last known size of that backing device */
867 struct drbd_conf *mdev;
868 struct completion event;
875 int (*io_fn)(struct drbd_conf *mdev);
876 void (*done)(struct drbd_conf *mdev, int rv);
879 enum write_ordering_e {
887 /* things that are stored as / read from meta data on disk */
890 /* configured by drbdsetup */
891 struct net_conf *net_conf; /* protected by get_net_conf() and put_net_conf() */
892 struct syncer_conf sync_conf;
893 struct drbd_backing_dev *ldev __protected_by(local);
895 sector_t p_size; /* partner's disk size */
896 struct request_queue *rq_queue;
897 struct block_device *this_bdev;
898 struct gendisk *vdisk;
900 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
901 struct drbd_socket meta; /* ping/ack (metadata) packets */
902 int agreed_pro_version; /* actually used protocol version */
903 unsigned long last_received; /* in jiffies, either socket */
904 unsigned int ko_count;
905 struct drbd_work resync_work,
908 struct timer_list resync_timer;
909 struct timer_list md_sync_timer;
911 /* Used after attach while negotiating new disk state. */
912 union drbd_state new_state_tmp;
914 union drbd_state state;
915 wait_queue_head_t misc_wait;
916 wait_queue_head_t state_wait; /* upon each state change. */
917 unsigned int send_cnt;
918 unsigned int recv_cnt;
919 unsigned int read_cnt;
920 unsigned int writ_cnt;
921 unsigned int al_writ_cnt;
922 unsigned int bm_writ_cnt;
923 atomic_t ap_bio_cnt; /* Requests we need to complete */
924 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
925 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
926 atomic_t unacked_cnt; /* Need to send replys for */
927 atomic_t local_cnt; /* Waiting for local completion */
928 atomic_t net_cnt; /* Users of net_conf */
930 struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */
931 struct drbd_tl_epoch *newest_tle;
932 struct drbd_tl_epoch *oldest_tle;
933 struct list_head out_of_sequence_requests;
934 struct hlist_head *tl_hash;
935 unsigned int tl_hash_s;
937 /* blocks to sync in this run [unit BM_BLOCK_SIZE] */
938 unsigned long rs_total;
939 /* number of sync IOs that failed in this run */
940 unsigned long rs_failed;
941 /* Syncer's start time [unit jiffies] */
942 unsigned long rs_start;
943 /* cumulated time in PausedSyncX state [unit jiffies] */
944 unsigned long rs_paused;
945 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
946 unsigned long rs_mark_left;
947 /* marks's time [unit jiffies] */
948 unsigned long rs_mark_time;
949 /* skipped because csum was equeal [unit BM_BLOCK_SIZE] */
950 unsigned long rs_same_csum;
952 /* where does the admin want us to start? (sector) */
953 sector_t ov_start_sector;
954 /* where are we now? (sector) */
955 sector_t ov_position;
956 /* Start sector of out of sync range (to merge printk reporting). */
957 sector_t ov_last_oos_start;
958 /* size of out-of-sync range in sectors. */
959 sector_t ov_last_oos_size;
960 unsigned long ov_left; /* in bits */
961 struct crypto_hash *csums_tfm;
962 struct crypto_hash *verify_tfm;
964 struct drbd_thread receiver;
965 struct drbd_thread worker;
966 struct drbd_thread asender;
967 struct drbd_bitmap *bitmap;
968 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
970 /* Used to track operations of resync... */
971 struct lru_cache *resync;
972 /* Number of locked elements in resync LRU */
973 unsigned int resync_locked;
974 /* resync extent number waiting for application requests */
975 unsigned int resync_wenr;
979 struct drbd_epoch *current_epoch;
980 spinlock_t epoch_lock;
982 enum write_ordering_e write_ordering;
983 struct list_head active_ee; /* IO in progress */
984 struct list_head sync_ee; /* IO in progress */
985 struct list_head done_ee; /* send ack */
986 struct list_head read_ee; /* IO in progress */
987 struct list_head net_ee; /* zero-copy network send in progress */
988 struct hlist_head *ee_hash; /* is proteced by req_lock! */
989 unsigned int ee_hash_s;
991 /* this one is protected by ee_lock, single thread */
992 struct drbd_epoch_entry *last_write_w_barrier;
995 struct hlist_head *app_reads_hash; /* is proteced by req_lock */
996 struct list_head resync_reads;
998 wait_queue_head_t ee_wait;
999 struct page *md_io_page; /* one page buffer for md_io */
1000 struct page *md_io_tmpp; /* for logical_block_size != 512 */
1001 struct mutex md_io_mutex; /* protects the md_io_buffer */
1003 wait_queue_head_t al_wait;
1004 struct lru_cache *act_log; /* activity log */
1005 unsigned int al_tr_number;
1007 int al_tr_pos; /* position of the next transaction in the journal */
1008 struct crypto_hash *cram_hmac_tfm;
1009 struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */
1010 struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */
1014 wait_queue_head_t seq_wait;
1015 atomic_t packet_seq;
1016 unsigned int peer_seq;
1017 spinlock_t peer_seq_lock;
1019 unsigned long comm_bm_set; /* communicated number of set bits. */
1020 cpumask_var_t cpu_mask;
1021 struct bm_io_work bm_io_work;
1022 u64 ed_uuid; /* UUID of the exposed data */
1023 struct mutex state_mutex;
1024 char congestion_reason; /* Why we where congested... */
1027 static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
1029 struct drbd_conf *mdev;
1031 mdev = minor < minor_count ? minor_table[minor] : NULL;
1036 static inline unsigned int mdev_to_minor(struct drbd_conf *mdev)
1041 /* returns 1 if it was successfull,
1042 * returns 0 if there was no data socket.
1043 * so wherever you are going to use the data.socket, e.g. do
1044 * if (!drbd_get_data_sock(mdev))
1047 * drbd_put_data_sock(mdev);
1049 static inline int drbd_get_data_sock(struct drbd_conf *mdev)
1051 mutex_lock(&mdev->data.mutex);
1052 /* drbd_disconnect() could have called drbd_free_sock()
1053 * while we were waiting in down()... */
1054 if (unlikely(mdev->data.socket == NULL)) {
1055 mutex_unlock(&mdev->data.mutex);
1061 static inline void drbd_put_data_sock(struct drbd_conf *mdev)
1063 mutex_unlock(&mdev->data.mutex);
1067 * function declarations
1068 *************************/
1072 enum chg_state_flags {
1075 CS_WAIT_COMPLETE = 4,
1077 CS_ORDERED = CS_WAIT_COMPLETE + CS_SERIALIZE,
1080 extern void drbd_init_set_defaults(struct drbd_conf *mdev);
1081 extern int drbd_change_state(struct drbd_conf *mdev, enum chg_state_flags f,
1082 union drbd_state mask, union drbd_state val);
1083 extern void drbd_force_state(struct drbd_conf *, union drbd_state,
1085 extern int _drbd_request_state(struct drbd_conf *, union drbd_state,
1086 union drbd_state, enum chg_state_flags);
1087 extern int __drbd_set_state(struct drbd_conf *, union drbd_state,
1088 enum chg_state_flags, struct completion *done);
1089 extern void print_st_err(struct drbd_conf *, union drbd_state,
1090 union drbd_state, int);
1091 extern int drbd_thread_start(struct drbd_thread *thi);
1092 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1094 extern void drbd_thread_current_set_cpu(struct drbd_conf *mdev);
1095 extern void drbd_calc_cpu_mask(struct drbd_conf *mdev);
1097 #define drbd_thread_current_set_cpu(A) ({})
1098 #define drbd_calc_cpu_mask(A) ({})
1100 extern void drbd_free_resources(struct drbd_conf *mdev);
1101 extern void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr,
1102 unsigned int set_size);
1103 extern void tl_clear(struct drbd_conf *mdev);
1104 extern void _tl_add_barrier(struct drbd_conf *, struct drbd_tl_epoch *);
1105 extern void drbd_free_sock(struct drbd_conf *mdev);
1106 extern int drbd_send(struct drbd_conf *mdev, struct socket *sock,
1107 void *buf, size_t size, unsigned msg_flags);
1108 extern int drbd_send_protocol(struct drbd_conf *mdev);
1109 extern int drbd_send_uuids(struct drbd_conf *mdev);
1110 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
1111 extern int drbd_send_sync_uuid(struct drbd_conf *mdev, u64 val);
1112 extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply);
1113 extern int _drbd_send_state(struct drbd_conf *mdev);
1114 extern int drbd_send_state(struct drbd_conf *mdev);
1115 extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock,
1116 enum drbd_packets cmd, struct p_header *h,
1117 size_t size, unsigned msg_flags);
1118 #define USE_DATA_SOCKET 1
1119 #define USE_META_SOCKET 0
1120 extern int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket,
1121 enum drbd_packets cmd, struct p_header *h,
1123 extern int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd,
1124 char *data, size_t size);
1125 extern int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc);
1126 extern int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr,
1128 extern int drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd,
1129 struct drbd_epoch_entry *e);
1130 extern int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packets cmd,
1131 struct p_block_req *rp);
1132 extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd,
1134 extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd,
1135 sector_t sector, int blksize, u64 block_id);
1136 extern int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd,
1137 struct drbd_epoch_entry *e);
1138 extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
1139 extern int _drbd_send_barrier(struct drbd_conf *mdev,
1140 struct drbd_tl_epoch *barrier);
1141 extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1142 sector_t sector, int size, u64 block_id);
1143 extern int drbd_send_drequest_csum(struct drbd_conf *mdev,
1144 sector_t sector,int size,
1145 void *digest, int digest_size,
1146 enum drbd_packets cmd);
1147 extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size);
1149 extern int drbd_send_bitmap(struct drbd_conf *mdev);
1150 extern int _drbd_send_bitmap(struct drbd_conf *mdev);
1151 extern int drbd_send_sr_reply(struct drbd_conf *mdev, int retcode);
1152 extern void drbd_free_bc(struct drbd_backing_dev *ldev);
1153 extern void drbd_mdev_cleanup(struct drbd_conf *mdev);
1155 /* drbd_meta-data.c (still in drbd_main.c) */
1156 extern void drbd_md_sync(struct drbd_conf *mdev);
1157 extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
1158 /* maybe define them below as inline? */
1159 extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1160 extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1161 extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1162 extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1163 extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
1164 extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
1165 extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
1166 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1167 extern void drbd_md_mark_dirty(struct drbd_conf *mdev);
1168 extern void drbd_queue_bitmap_io(struct drbd_conf *mdev,
1169 int (*io_fn)(struct drbd_conf *),
1170 void (*done)(struct drbd_conf *, int),
1172 extern int drbd_bmio_set_n_write(struct drbd_conf *mdev);
1173 extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev);
1174 extern int drbd_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), char *why);
1178 We reserve a 128MB Block (4k aligned)
1179 * either at the end of the backing device
1180 * or on a seperate meta data device. */
1182 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1183 /* The following numbers are sectors */
1184 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1185 #define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */
1186 /* Allows up to about 3.8TB */
1187 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1189 /* Since the smalles IO unit is usually 512 byte */
1190 #define MD_SECTOR_SHIFT 9
1191 #define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT)
1194 #define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1195 #define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */
1196 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1198 #if BITS_PER_LONG == 32
1200 #define cpu_to_lel(A) cpu_to_le32(A)
1201 #define lel_to_cpu(A) le32_to_cpu(A)
1202 #elif BITS_PER_LONG == 64
1204 #define cpu_to_lel(A) cpu_to_le64(A)
1205 #define lel_to_cpu(A) le64_to_cpu(A)
1207 #error "LN2 of BITS_PER_LONG unknown!"
1211 /* 16MB sized 'bitmap extent' to track syncer usage */
1213 int rs_left; /* number of bits set (out of sync) in this extent. */
1214 int rs_failed; /* number of failed resync requests in this extent. */
1215 unsigned long flags;
1216 struct lc_element lce;
1219 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1220 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1224 * We need to store one bit for a block.
1225 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1226 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1227 * Bit 1 ==> local node thinks this block needs to be synced.
1230 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1231 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1232 /* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1233 * per sector of on disk bitmap */
1234 #define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */
1235 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1237 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1238 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1241 /* thus many _storage_ sectors are described by one bit */
1242 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1243 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1244 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1246 /* bit to represented kilo byte conversion */
1247 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1249 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1250 * _storage_ sector is located in */
1251 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1253 /* how much _storage_ sectors we have per bitmap sector */
1254 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1255 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1257 /* in one sector of the bitmap, we have this many activity_log extents. */
1258 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1259 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1261 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1262 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1264 /* the extent in "PER_EXTENT" below is an activity log extent
1265 * we need that many (long words/bytes) to store the bitmap
1266 * of one AL_EXTENT_SIZE chunk of storage.
1267 * we can store the bitmap for that many AL_EXTENTS within
1268 * one sector of the _on_disk_ bitmap:
1269 * bit 0 bit 37 bit 38 bit (512*8)-1
1270 * ...|........|........|.. // ..|........|
1271 * sect. 0 `296 `304 ^(512*8*8)-1
1273 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1274 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1275 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1278 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1279 #define DRBD_MAX_SECTORS_BM \
1280 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1281 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1282 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1283 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1284 #elif !defined(CONFIG_LBD) && BITS_PER_LONG == 32
1285 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1286 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1288 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1289 /* 16 TB in units of sectors */
1290 #if BITS_PER_LONG == 32
1291 /* adjust by one page worth of bitmap,
1292 * so we won't wrap around in drbd_bm_find_next_bit.
1293 * you should use 64bit OS for that much storage, anyways. */
1294 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1296 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0x1LU << 32)
1300 /* Sector shift value for the "hash" functions of tl_hash and ee_hash tables.
1301 * With a value of 6 all IO in one 32K block make it to the same slot of the
1304 #define DRBD_MAX_SEGMENT_SIZE (1U<<(9+HT_SHIFT))
1306 /* Number of elements in the app_reads_hash */
1307 #define APP_R_HSIZE 15
1309 extern int drbd_bm_init(struct drbd_conf *mdev);
1310 extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors);
1311 extern void drbd_bm_cleanup(struct drbd_conf *mdev);
1312 extern void drbd_bm_set_all(struct drbd_conf *mdev);
1313 extern void drbd_bm_clear_all(struct drbd_conf *mdev);
1314 extern int drbd_bm_set_bits(
1315 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1316 extern int drbd_bm_clear_bits(
1317 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1318 /* bm_set_bits variant for use while holding drbd_bm_lock */
1319 extern void _drbd_bm_set_bits(struct drbd_conf *mdev,
1320 const unsigned long s, const unsigned long e);
1321 extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr);
1322 extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr);
1323 extern int drbd_bm_write_sect(struct drbd_conf *mdev, unsigned long enr) __must_hold(local);
1324 extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
1325 extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
1326 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev,
1327 unsigned long al_enr);
1328 extern size_t drbd_bm_words(struct drbd_conf *mdev);
1329 extern unsigned long drbd_bm_bits(struct drbd_conf *mdev);
1330 extern sector_t drbd_bm_capacity(struct drbd_conf *mdev);
1331 extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1332 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1333 extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1334 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo);
1335 extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev);
1336 extern int drbd_bm_rs_done(struct drbd_conf *mdev);
1337 /* for receive_bitmap */
1338 extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset,
1339 size_t number, unsigned long *buffer);
1340 /* for _drbd_send_bitmap and drbd_bm_write_sect */
1341 extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset,
1342 size_t number, unsigned long *buffer);
1344 extern void drbd_bm_lock(struct drbd_conf *mdev, char *why);
1345 extern void drbd_bm_unlock(struct drbd_conf *mdev);
1347 extern int drbd_bm_count_bits(struct drbd_conf *mdev, const unsigned long s, const unsigned long e);
1350 extern struct kmem_cache *drbd_request_cache;
1351 extern struct kmem_cache *drbd_ee_cache; /* epoch entries */
1352 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1353 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1354 extern mempool_t *drbd_request_mempool;
1355 extern mempool_t *drbd_ee_mempool;
1357 extern struct page *drbd_pp_pool; /* drbd's page pool */
1358 extern spinlock_t drbd_pp_lock;
1359 extern int drbd_pp_vacant;
1360 extern wait_queue_head_t drbd_pp_wait;
1362 extern rwlock_t global_state_lock;
1364 extern struct drbd_conf *drbd_new_device(unsigned int minor);
1365 extern void drbd_free_mdev(struct drbd_conf *mdev);
1367 extern int proc_details;
1370 extern int drbd_make_request_26(struct request_queue *q, struct bio *bio);
1371 extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req);
1372 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1373 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1377 extern void drbd_suspend_io(struct drbd_conf *mdev);
1378 extern void drbd_resume_io(struct drbd_conf *mdev);
1379 extern char *ppsize(char *buf, unsigned long long size);
1380 extern sector_t drbd_new_dev_size(struct drbd_conf *,
1381 struct drbd_backing_dev *);
1382 enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
1383 extern enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *) __must_hold(local);
1384 extern void resync_after_online_grow(struct drbd_conf *);
1385 extern void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int) __must_hold(local);
1386 extern int drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role,
1388 enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev);
1389 extern int drbd_khelper(struct drbd_conf *mdev, char *cmd);
1392 extern int drbd_worker(struct drbd_thread *thi);
1393 extern int drbd_alter_sa(struct drbd_conf *mdev, int na);
1394 extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side);
1395 extern void resume_next_sg(struct drbd_conf *mdev);
1396 extern void suspend_other_sg(struct drbd_conf *mdev);
1397 extern int drbd_resync_finished(struct drbd_conf *mdev);
1398 /* maybe rather drbd_main.c ? */
1399 extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
1400 struct drbd_backing_dev *bdev, sector_t sector, int rw);
1401 extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int);
1403 static inline void ov_oos_print(struct drbd_conf *mdev)
1405 if (mdev->ov_last_oos_size) {
1406 dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n",
1407 (unsigned long long)mdev->ov_last_oos_start,
1408 (unsigned long)mdev->ov_last_oos_size);
1410 mdev->ov_last_oos_size=0;
1414 extern void drbd_csum(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
1415 /* worker callbacks */
1416 extern int w_req_cancel_conflict(struct drbd_conf *, struct drbd_work *, int);
1417 extern int w_read_retry_remote(struct drbd_conf *, struct drbd_work *, int);
1418 extern int w_e_end_data_req(struct drbd_conf *, struct drbd_work *, int);
1419 extern int w_e_end_rsdata_req(struct drbd_conf *, struct drbd_work *, int);
1420 extern int w_e_end_csum_rs_req(struct drbd_conf *, struct drbd_work *, int);
1421 extern int w_e_end_ov_reply(struct drbd_conf *, struct drbd_work *, int);
1422 extern int w_e_end_ov_req(struct drbd_conf *, struct drbd_work *, int);
1423 extern int w_ov_finished(struct drbd_conf *, struct drbd_work *, int);
1424 extern int w_resync_inactive(struct drbd_conf *, struct drbd_work *, int);
1425 extern int w_resume_next_sg(struct drbd_conf *, struct drbd_work *, int);
1426 extern int w_io_error(struct drbd_conf *, struct drbd_work *, int);
1427 extern int w_send_write_hint(struct drbd_conf *, struct drbd_work *, int);
1428 extern int w_make_resync_request(struct drbd_conf *, struct drbd_work *, int);
1429 extern int w_send_dblock(struct drbd_conf *, struct drbd_work *, int);
1430 extern int w_send_barrier(struct drbd_conf *, struct drbd_work *, int);
1431 extern int w_send_read_req(struct drbd_conf *, struct drbd_work *, int);
1432 extern int w_prev_work_done(struct drbd_conf *, struct drbd_work *, int);
1433 extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int);
1435 extern void resync_timer_fn(unsigned long data);
1437 /* drbd_receiver.c */
1438 extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list);
1439 extern struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev,
1442 unsigned int data_size,
1443 gfp_t gfp_mask) __must_hold(local);
1444 extern void drbd_free_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e);
1445 extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1446 struct list_head *head);
1447 extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1448 struct list_head *head);
1449 extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled);
1450 extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed);
1451 extern void drbd_flush_workqueue(struct drbd_conf *mdev);
1453 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1454 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1455 static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1456 char __user *optval, int optlen)
1459 if (level == SOL_SOCKET)
1460 err = sock_setsockopt(sock, level, optname, optval, optlen);
1462 err = sock->ops->setsockopt(sock, level, optname, optval,
1467 static inline void drbd_tcp_cork(struct socket *sock)
1470 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1471 (char __user *)&val, sizeof(val));
1474 static inline void drbd_tcp_uncork(struct socket *sock)
1477 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1478 (char __user *)&val, sizeof(val));
1481 static inline void drbd_tcp_nodelay(struct socket *sock)
1484 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1485 (char __user *)&val, sizeof(val));
1488 static inline void drbd_tcp_quickack(struct socket *sock)
1491 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1492 (char __user *)&val, sizeof(val));
1495 void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo);
1498 extern struct proc_dir_entry *drbd_proc;
1499 extern struct file_operations drbd_proc_fops;
1500 extern const char *drbd_conn_str(enum drbd_conns s);
1501 extern const char *drbd_role_str(enum drbd_role s);
1504 extern void drbd_al_begin_io(struct drbd_conf *mdev, sector_t sector);
1505 extern void drbd_al_complete_io(struct drbd_conf *mdev, sector_t sector);
1506 extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector);
1507 extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1508 extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1509 extern void drbd_rs_cancel_all(struct drbd_conf *mdev);
1510 extern int drbd_rs_del_all(struct drbd_conf *mdev);
1511 extern void drbd_rs_failed_io(struct drbd_conf *mdev,
1512 sector_t sector, int size);
1513 extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *);
1514 extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
1515 int size, const char *file, const unsigned int line);
1516 #define drbd_set_in_sync(mdev, sector, size) \
1517 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1518 extern void __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
1519 int size, const char *file, const unsigned int line);
1520 #define drbd_set_out_of_sync(mdev, sector, size) \
1521 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1522 extern void drbd_al_apply_to_bm(struct drbd_conf *mdev);
1523 extern void drbd_al_to_on_disk_bm(struct drbd_conf *mdev);
1524 extern void drbd_al_shrink(struct drbd_conf *mdev);
1529 void drbd_nl_cleanup(void);
1530 int __init drbd_nl_init(void);
1531 void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state);
1532 void drbd_bcast_sync_progress(struct drbd_conf *mdev);
1533 void drbd_bcast_ee(struct drbd_conf *mdev,
1534 const char *reason, const int dgs,
1535 const char* seen_hash, const char* calc_hash,
1536 const struct drbd_epoch_entry* e);
1540 * DOC: DRBD State macros
1542 * These macros are used to express state changes in easily readable form.
1544 * The NS macros expand to a mask and a value, that can be bit ored onto the
1545 * current state as soon as the spinlock (req_lock) was taken.
1547 * The _NS macros are used for state functions that get called with the
1548 * spinlock. These macros expand directly to the new state value.
1550 * Besides the basic forms NS() and _NS() additional _?NS[23] are defined
1551 * to express state changes that affect more than one aspect of the state.
1553 * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY)
1554 * Means that the network connection was established and that the peer
1555 * is in secondary role.
1557 #define role_MASK R_MASK
1558 #define peer_MASK R_MASK
1559 #define disk_MASK D_MASK
1560 #define pdsk_MASK D_MASK
1561 #define conn_MASK C_MASK
1563 #define user_isp_MASK 1
1564 #define aftr_isp_MASK 1
1567 ({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \
1568 ({ union drbd_state val; val.i = 0; val.T = (S); val; })
1569 #define NS2(T1, S1, T2, S2) \
1570 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1571 mask.T2 = T2##_MASK; mask; }), \
1572 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1573 val.T2 = (S2); val; })
1574 #define NS3(T1, S1, T2, S2, T3, S3) \
1575 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1576 mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \
1577 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1578 val.T2 = (S2); val.T3 = (S3); val; })
1580 #define _NS(D, T, S) \
1581 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; })
1582 #define _NS2(D, T1, S1, T2, S2) \
1583 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1584 __ns.T2 = (S2); __ns; })
1585 #define _NS3(D, T1, S1, T2, S2, T3, S3) \
1586 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1587 __ns.T2 = (S2); __ns.T3 = (S3); __ns; })
1590 * inline helper functions
1591 *************************/
1593 static inline void drbd_state_lock(struct drbd_conf *mdev)
1595 wait_event(mdev->misc_wait,
1596 !test_and_set_bit(CLUSTER_ST_CHANGE, &mdev->flags));
1599 static inline void drbd_state_unlock(struct drbd_conf *mdev)
1601 clear_bit(CLUSTER_ST_CHANGE, &mdev->flags);
1602 wake_up(&mdev->misc_wait);
1605 static inline int _drbd_set_state(struct drbd_conf *mdev,
1606 union drbd_state ns, enum chg_state_flags flags,
1607 struct completion *done)
1611 read_lock(&global_state_lock);
1612 rv = __drbd_set_state(mdev, ns, flags, done);
1613 read_unlock(&global_state_lock);
1619 * drbd_request_state() - Reqest a state change
1620 * @mdev: DRBD device.
1621 * @mask: mask of state bits to change.
1622 * @val: value of new state bits.
1624 * This is the most graceful way of requesting a state change. It is verbose
1625 * quite verbose in case the state change is not possible, and all those
1626 * state changes are globally serialized.
1628 static inline int drbd_request_state(struct drbd_conf *mdev,
1629 union drbd_state mask,
1630 union drbd_state val)
1632 return _drbd_request_state(mdev, mask, val, CS_VERBOSE + CS_ORDERED);
1635 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1636 static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach, const char *where)
1638 switch (mdev->ldev->dc.on_io_error) {
1641 if (printk_ratelimit())
1642 dev_err(DEV, "Local IO failed in %s."
1643 "Passing error on...\n", where);
1646 /* NOTE fall through to detach case if forcedetach set */
1648 case EP_CALL_HELPER:
1649 if (mdev->state.disk > D_FAILED) {
1650 _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL);
1651 dev_err(DEV, "Local IO failed in %s."
1652 "Detaching...\n", where);
1659 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1660 * @mdev: DRBD device.
1661 * @error: Error code passed to the IO completion callback
1662 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1664 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1666 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1667 static inline void drbd_chk_io_error_(struct drbd_conf *mdev,
1668 int error, int forcedetach, const char *where)
1671 unsigned long flags;
1672 spin_lock_irqsave(&mdev->req_lock, flags);
1673 __drbd_chk_io_error_(mdev, forcedetach, where);
1674 spin_unlock_irqrestore(&mdev->req_lock, flags);
1680 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1681 * @bdev: Meta data block device.
1683 * BTW, for internal meta data, this happens to be the maximum capacity
1684 * we could agree upon with our peer node.
1686 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1688 switch (bdev->dc.meta_dev_idx) {
1689 case DRBD_MD_INDEX_INTERNAL:
1690 case DRBD_MD_INDEX_FLEX_INT:
1691 return bdev->md.md_offset + bdev->md.bm_offset;
1692 case DRBD_MD_INDEX_FLEX_EXT:
1694 return bdev->md.md_offset;
1699 * drbd_md_last_sector() - Return the last sector number of the meta data area
1700 * @bdev: Meta data block device.
1702 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1704 switch (bdev->dc.meta_dev_idx) {
1705 case DRBD_MD_INDEX_INTERNAL:
1706 case DRBD_MD_INDEX_FLEX_INT:
1707 return bdev->md.md_offset + MD_AL_OFFSET - 1;
1708 case DRBD_MD_INDEX_FLEX_EXT:
1710 return bdev->md.md_offset + bdev->md.md_size_sect;
1714 /* Returns the number of 512 byte sectors of the device */
1715 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1717 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1718 return bdev ? bdev->bd_inode->i_size >> 9 : 0;
1722 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1723 * @bdev: Meta data block device.
1725 * returns the capacity we announce to out peer. we clip ourselves at the
1726 * various MAX_SECTORS, because if we don't, current implementation will
1727 * oops sooner or later
1729 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1732 switch (bdev->dc.meta_dev_idx) {
1733 case DRBD_MD_INDEX_INTERNAL:
1734 case DRBD_MD_INDEX_FLEX_INT:
1735 s = drbd_get_capacity(bdev->backing_bdev)
1736 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1737 drbd_md_first_sector(bdev))
1740 case DRBD_MD_INDEX_FLEX_EXT:
1741 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1742 drbd_get_capacity(bdev->backing_bdev));
1743 /* clip at maximum size the meta device can support */
1744 s = min_t(sector_t, s,
1745 BM_EXT_TO_SECT(bdev->md.md_size_sect
1746 - bdev->md.bm_offset));
1749 s = min_t(sector_t, DRBD_MAX_SECTORS,
1750 drbd_get_capacity(bdev->backing_bdev));
1756 * drbd_md_ss__() - Return the sector number of our meta data super block
1757 * @mdev: DRBD device.
1758 * @bdev: Meta data block device.
1760 static inline sector_t drbd_md_ss__(struct drbd_conf *mdev,
1761 struct drbd_backing_dev *bdev)
1763 switch (bdev->dc.meta_dev_idx) {
1764 default: /* external, some index */
1765 return MD_RESERVED_SECT * bdev->dc.meta_dev_idx;
1766 case DRBD_MD_INDEX_INTERNAL:
1767 /* with drbd08, internal meta data is always "flexible" */
1768 case DRBD_MD_INDEX_FLEX_INT:
1769 /* sizeof(struct md_on_disk_07) == 4k
1770 * position: last 4k aligned block of 4k size */
1771 if (!bdev->backing_bdev) {
1772 if (__ratelimit(&drbd_ratelimit_state)) {
1773 dev_err(DEV, "bdev->backing_bdev==NULL\n");
1778 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL)
1780 case DRBD_MD_INDEX_FLEX_EXT:
1786 _drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1788 list_add_tail(&w->list, &q->q);
1793 drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w)
1795 unsigned long flags;
1796 spin_lock_irqsave(&q->q_lock, flags);
1797 list_add(&w->list, &q->q);
1798 up(&q->s); /* within the spinlock,
1799 see comment near end of drbd_worker() */
1800 spin_unlock_irqrestore(&q->q_lock, flags);
1804 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1806 unsigned long flags;
1807 spin_lock_irqsave(&q->q_lock, flags);
1808 list_add_tail(&w->list, &q->q);
1809 up(&q->s); /* within the spinlock,
1810 see comment near end of drbd_worker() */
1811 spin_unlock_irqrestore(&q->q_lock, flags);
1814 static inline void wake_asender(struct drbd_conf *mdev)
1816 if (test_bit(SIGNAL_ASENDER, &mdev->flags))
1817 force_sig(DRBD_SIG, mdev->asender.task);
1820 static inline void request_ping(struct drbd_conf *mdev)
1822 set_bit(SEND_PING, &mdev->flags);
1826 static inline int drbd_send_short_cmd(struct drbd_conf *mdev,
1827 enum drbd_packets cmd)
1830 return drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, &h, sizeof(h));
1833 static inline int drbd_send_ping(struct drbd_conf *mdev)
1836 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING, &h, sizeof(h));
1839 static inline int drbd_send_ping_ack(struct drbd_conf *mdev)
1842 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING_ACK, &h, sizeof(h));
1845 static inline void drbd_thread_stop(struct drbd_thread *thi)
1847 _drbd_thread_stop(thi, FALSE, TRUE);
1850 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
1852 _drbd_thread_stop(thi, FALSE, FALSE);
1855 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
1857 _drbd_thread_stop(thi, TRUE, FALSE);
1860 /* counts how many answer packets packets we expect from our peer,
1861 * for either explicit application requests,
1862 * or implicit barrier packets as necessary.
1865 * _req_mod(req, queue_for_net_write or queue_for_net_read);
1866 * it is much easier and equally valid to count what we queue for the
1867 * worker, even before it actually was queued or send.
1868 * (drbd_make_request_common; recovery path on read io-error)
1870 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1871 * _req_mod(req, data_received)
1872 * [from receive_DataReply]
1873 * _req_mod(req, write_acked_by_peer or recv_acked_by_peer or neg_acked)
1874 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1875 * for some reason it is NOT decreased in got_NegAck,
1876 * but in the resulting cleanup code from report_params.
1877 * we should try to remember the reason for that...
1878 * _req_mod(req, send_failed or send_canceled)
1879 * _req_mod(req, connection_lost_while_pending)
1880 * [from tl_clear_barrier]
1882 static inline void inc_ap_pending(struct drbd_conf *mdev)
1884 atomic_inc(&mdev->ap_pending_cnt);
1887 #define ERR_IF_CNT_IS_NEGATIVE(which) \
1888 if (atomic_read(&mdev->which) < 0) \
1889 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
1890 __func__ , __LINE__ , \
1891 atomic_read(&mdev->which))
1893 #define dec_ap_pending(mdev) do { \
1894 typecheck(struct drbd_conf *, mdev); \
1895 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \
1896 wake_up(&mdev->misc_wait); \
1897 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
1899 /* counts how many resync-related answers we still expect from the peer
1901 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1902 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK whith ID_SYNCER)
1903 * (or P_NEG_ACK with ID_SYNCER)
1905 static inline void inc_rs_pending(struct drbd_conf *mdev)
1907 atomic_inc(&mdev->rs_pending_cnt);
1910 #define dec_rs_pending(mdev) do { \
1911 typecheck(struct drbd_conf *, mdev); \
1912 atomic_dec(&mdev->rs_pending_cnt); \
1913 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
1915 /* counts how many answers we still need to send to the peer.
1917 * receive_Data unless protocol A;
1918 * we need to send a P_RECV_ACK (proto B)
1919 * or P_WRITE_ACK (proto C)
1920 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1921 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1922 * receive_Barrier_* we need to send a P_BARRIER_ACK
1924 static inline void inc_unacked(struct drbd_conf *mdev)
1926 atomic_inc(&mdev->unacked_cnt);
1929 #define dec_unacked(mdev) do { \
1930 typecheck(struct drbd_conf *, mdev); \
1931 atomic_dec(&mdev->unacked_cnt); \
1932 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
1934 #define sub_unacked(mdev, n) do { \
1935 typecheck(struct drbd_conf *, mdev); \
1936 atomic_sub(n, &mdev->unacked_cnt); \
1937 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
1940 static inline void put_net_conf(struct drbd_conf *mdev)
1942 if (atomic_dec_and_test(&mdev->net_cnt))
1943 wake_up(&mdev->misc_wait);
1947 * get_net_conf() - Increase ref count on mdev->net_conf; Returns 0 if nothing there
1948 * @mdev: DRBD device.
1950 * You have to call put_net_conf() when finished working with mdev->net_conf.
1952 static inline int get_net_conf(struct drbd_conf *mdev)
1956 atomic_inc(&mdev->net_cnt);
1957 have_net_conf = mdev->state.conn >= C_UNCONNECTED;
1960 return have_net_conf;
1964 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
1967 * You have to call put_ldev() when finished working with mdev->ldev.
1969 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
1970 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
1972 static inline void put_ldev(struct drbd_conf *mdev)
1975 if (atomic_dec_and_test(&mdev->local_cnt))
1976 wake_up(&mdev->misc_wait);
1977 D_ASSERT(atomic_read(&mdev->local_cnt) >= 0);
1981 static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
1985 atomic_inc(&mdev->local_cnt);
1986 io_allowed = (mdev->state.disk >= mins);
1992 extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins);
1995 /* you must have an "get_ldev" reference */
1996 static inline void drbd_get_syncer_progress(struct drbd_conf *mdev,
1997 unsigned long *bits_left, unsigned int *per_mil_done)
2000 * this is to break it at compile time when we change that
2001 * (we may feel 4TB maximum storage per drbd is not enough)
2003 typecheck(unsigned long, mdev->rs_total);
2005 /* note: both rs_total and rs_left are in bits, i.e. in
2006 * units of BM_BLOCK_SIZE.
2007 * for the percentage, we don't care. */
2009 *bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
2010 /* >> 10 to prevent overflow,
2011 * +1 to prevent division by zero */
2012 if (*bits_left > mdev->rs_total) {
2013 /* doh. maybe a logic bug somewhere.
2014 * may also be just a race condition
2015 * between this and a disconnect during sync.
2016 * for now, just prevent in-kernel buffer overflow.
2019 dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2020 drbd_conn_str(mdev->state.conn),
2021 *bits_left, mdev->rs_total, mdev->rs_failed);
2024 /* make sure the calculation happens in long context */
2025 unsigned long tmp = 1000UL -
2026 (*bits_left >> 10)*1000UL
2027 / ((mdev->rs_total >> 10) + 1UL);
2028 *per_mil_done = tmp;
2033 /* this throttles on-the-fly application requests
2034 * according to max_buffers settings;
2035 * maybe re-implement using semaphores? */
2036 static inline int drbd_get_max_buffers(struct drbd_conf *mdev)
2038 int mxb = 1000000; /* arbitrary limit on open requests */
2039 if (get_net_conf(mdev)) {
2040 mxb = mdev->net_conf->max_buffers;
2046 static inline int drbd_state_is_stable(union drbd_state s)
2049 /* DO NOT add a default clause, we want the compiler to warn us
2050 * for any newly introduced state we may have forgotten to add here */
2052 switch ((enum drbd_conns)s.conn) {
2053 /* new io only accepted when there is no connection, ... */
2055 case C_WF_CONNECTION:
2056 /* ... or there is a well established connection. */
2062 case C_PAUSED_SYNC_S:
2063 case C_PAUSED_SYNC_T:
2064 /* maybe stable, look at the disk state */
2067 /* no new io accepted during tansitional states
2068 * like handshake or teardown */
2069 case C_DISCONNECTING:
2073 case C_NETWORK_FAILURE:
2074 case C_PROTOCOL_ERROR:
2076 case C_WF_REPORT_PARAMS:
2077 case C_STARTING_SYNC_S:
2078 case C_STARTING_SYNC_T:
2081 case C_WF_SYNC_UUID:
2087 switch ((enum drbd_disk_state)s.disk) {
2089 case D_INCONSISTENT:
2093 /* disk state is stable as well. */
2096 /* no new io accepted during tansitional states */
2109 static inline int __inc_ap_bio_cond(struct drbd_conf *mdev)
2111 int mxb = drbd_get_max_buffers(mdev);
2113 if (mdev->state.susp)
2115 if (test_bit(SUSPEND_IO, &mdev->flags))
2118 /* to avoid potential deadlock or bitmap corruption,
2119 * in various places, we only allow new application io
2120 * to start during "stable" states. */
2122 /* no new io accepted when attaching or detaching the disk */
2123 if (!drbd_state_is_stable(mdev->state))
2126 /* since some older kernels don't have atomic_add_unless,
2127 * and we are within the spinlock anyways, we have this workaround. */
2128 if (atomic_read(&mdev->ap_bio_cnt) > mxb)
2130 if (test_bit(BITMAP_IO, &mdev->flags))
2135 /* I'd like to use wait_event_lock_irq,
2136 * but I'm not sure when it got introduced,
2137 * and not sure when it has 3 or 4 arguments */
2138 static inline void inc_ap_bio(struct drbd_conf *mdev, int one_or_two)
2140 /* compare with after_state_ch,
2141 * os.conn != C_WF_BITMAP_S && ns.conn == C_WF_BITMAP_S */
2145 * as long as the device is suspended
2146 * until the bitmap is no longer on the fly during connection
2147 * handshake as long as we would exeed the max_buffer limit.
2149 * to avoid races with the reconnect code,
2150 * we need to atomic_inc within the spinlock. */
2152 spin_lock_irq(&mdev->req_lock);
2153 while (!__inc_ap_bio_cond(mdev)) {
2154 prepare_to_wait(&mdev->misc_wait, &wait, TASK_UNINTERRUPTIBLE);
2155 spin_unlock_irq(&mdev->req_lock);
2157 finish_wait(&mdev->misc_wait, &wait);
2158 spin_lock_irq(&mdev->req_lock);
2160 atomic_add(one_or_two, &mdev->ap_bio_cnt);
2161 spin_unlock_irq(&mdev->req_lock);
2164 static inline void dec_ap_bio(struct drbd_conf *mdev)
2166 int mxb = drbd_get_max_buffers(mdev);
2167 int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt);
2169 D_ASSERT(ap_bio >= 0);
2170 /* this currently does wake_up for every dec_ap_bio!
2171 * maybe rather introduce some type of hysteresis?
2172 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2174 wake_up(&mdev->misc_wait);
2175 if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) {
2176 if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags))
2177 drbd_queue_work(&mdev->data.work, &mdev->bm_io_work.w);
2181 static inline void drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val)
2183 mdev->ed_uuid = val;
2186 static inline int seq_cmp(u32 a, u32 b)
2188 /* we assume wrap around at 32bit.
2189 * for wrap around at 24bit (old atomic_t),
2193 return (s32)(a) - (s32)(b);
2195 #define seq_lt(a, b) (seq_cmp((a), (b)) < 0)
2196 #define seq_gt(a, b) (seq_cmp((a), (b)) > 0)
2197 #define seq_ge(a, b) (seq_cmp((a), (b)) >= 0)
2198 #define seq_le(a, b) (seq_cmp((a), (b)) <= 0)
2199 /* CAUTION: please no side effects in arguments! */
2200 #define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b)))
2202 static inline void update_peer_seq(struct drbd_conf *mdev, unsigned int new_seq)
2205 spin_lock(&mdev->peer_seq_lock);
2206 m = seq_max(mdev->peer_seq, new_seq);
2208 spin_unlock(&mdev->peer_seq_lock);
2210 wake_up(&mdev->seq_wait);
2213 static inline void drbd_update_congested(struct drbd_conf *mdev)
2215 struct sock *sk = mdev->data.socket->sk;
2216 if (sk->sk_wmem_queued > sk->sk_sndbuf * 4 / 5)
2217 set_bit(NET_CONGESTED, &mdev->flags);
2220 static inline int drbd_queue_order_type(struct drbd_conf *mdev)
2222 /* sorry, we currently have no working implementation
2223 * of distributed TCQ stuff */
2224 #ifndef QUEUE_ORDERED_NONE
2225 #define QUEUE_ORDERED_NONE 0
2227 return QUEUE_ORDERED_NONE;
2230 static inline void drbd_blk_run_queue(struct request_queue *q)
2232 if (q && q->unplug_fn)
2236 static inline void drbd_kick_lo(struct drbd_conf *mdev)
2238 if (get_ldev(mdev)) {
2239 drbd_blk_run_queue(bdev_get_queue(mdev->ldev->backing_bdev));
2244 static inline void drbd_md_flush(struct drbd_conf *mdev)
2248 if (test_bit(MD_NO_BARRIER, &mdev->flags))
2251 r = blkdev_issue_flush(mdev->ldev->md_bdev, NULL);
2253 set_bit(MD_NO_BARRIER, &mdev->flags);
2254 dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r);