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 adress 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, "%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 */
139 #ifdef CONFIG_DRBD_FAULT_INJECTION
141 _drbd_insert_fault(struct drbd_conf *mdev, unsigned int type);
143 drbd_insert_fault(struct drbd_conf *mdev, unsigned int type) {
145 (enable_faults & (1<<type)) &&
146 _drbd_insert_fault(mdev, type);
148 #define FAULT_ACTIVE(_m, _t) (drbd_insert_fault((_m), (_t)))
151 #define FAULT_ACTIVE(_m, _t) (0)
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 */
213 P_MAY_IGNORE = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
214 P_MAX_OPT_CMD = 0x101,
216 /* special command ids for handshake */
218 P_HAND_SHAKE_M = 0xfff1, /* First Packet on the MetaSock */
219 P_HAND_SHAKE_S = 0xfff2, /* First Packet on the Socket */
221 P_HAND_SHAKE = 0xfffe /* FIXED for the next century! */
224 static inline const char *cmdname(enum drbd_packets cmd)
226 /* THINK may need to become several global tables
227 * when we want to support more than
229 static const char *cmdnames[] = {
231 [P_DATA_REPLY] = "DataReply",
232 [P_RS_DATA_REPLY] = "RSDataReply",
233 [P_BARRIER] = "Barrier",
234 [P_BITMAP] = "ReportBitMap",
235 [P_BECOME_SYNC_TARGET] = "BecomeSyncTarget",
236 [P_BECOME_SYNC_SOURCE] = "BecomeSyncSource",
237 [P_UNPLUG_REMOTE] = "UnplugRemote",
238 [P_DATA_REQUEST] = "DataRequest",
239 [P_RS_DATA_REQUEST] = "RSDataRequest",
240 [P_SYNC_PARAM] = "SyncParam",
241 [P_SYNC_PARAM89] = "SyncParam89",
242 [P_PROTOCOL] = "ReportProtocol",
243 [P_UUIDS] = "ReportUUIDs",
244 [P_SIZES] = "ReportSizes",
245 [P_STATE] = "ReportState",
246 [P_SYNC_UUID] = "ReportSyncUUID",
247 [P_AUTH_CHALLENGE] = "AuthChallenge",
248 [P_AUTH_RESPONSE] = "AuthResponse",
250 [P_PING_ACK] = "PingAck",
251 [P_RECV_ACK] = "RecvAck",
252 [P_WRITE_ACK] = "WriteAck",
253 [P_RS_WRITE_ACK] = "RSWriteAck",
254 [P_DISCARD_ACK] = "DiscardAck",
255 [P_NEG_ACK] = "NegAck",
256 [P_NEG_DREPLY] = "NegDReply",
257 [P_NEG_RS_DREPLY] = "NegRSDReply",
258 [P_BARRIER_ACK] = "BarrierAck",
259 [P_STATE_CHG_REQ] = "StateChgRequest",
260 [P_STATE_CHG_REPLY] = "StateChgReply",
261 [P_OV_REQUEST] = "OVRequest",
262 [P_OV_REPLY] = "OVReply",
263 [P_OV_RESULT] = "OVResult",
267 if (cmd == P_HAND_SHAKE_M)
269 if (cmd == P_HAND_SHAKE_S)
271 if (cmd == P_HAND_SHAKE)
273 if (cmd >= P_MAX_CMD)
275 return cmdnames[cmd];
278 /* for sending/receiving the bitmap,
279 * possibly in some encoding scheme */
282 * stores total bits and long words
283 * of the bitmap, so we don't need to
284 * call the accessor functions over and again. */
285 unsigned long bm_bits;
286 unsigned long bm_words;
287 /* during xfer, current position within the bitmap */
288 unsigned long bit_offset;
289 unsigned long word_offset;
291 /* statistics; index: (h->command == P_BITMAP) */
296 extern void INFO_bm_xfer_stats(struct drbd_conf *mdev,
297 const char *direction, struct bm_xfer_ctx *c);
299 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
301 /* word_offset counts "native long words" (32 or 64 bit),
303 * Encoded packet may end at an unaligned bit offset.
304 * In case a fallback clear text packet is transmitted in
305 * between, we adjust this offset back to the last 64bit
306 * aligned "native long word", which makes coding and decoding
307 * the plain text bitmap much more convenient. */
308 #if BITS_PER_LONG == 64
309 c->word_offset = c->bit_offset >> 6;
310 #elif BITS_PER_LONG == 32
311 c->word_offset = c->bit_offset >> 5;
312 c->word_offset &= ~(1UL);
314 # error "unsupported BITS_PER_LONG"
319 #define __packed __attribute__((packed))
322 /* This is the layout for a packet on the wire.
323 * The byteorder is the network byte order.
324 * (except block_id and barrier fields.
325 * these are pointers to local structs
326 * and have no relevance for the partner,
327 * which just echoes them as received.)
329 * NOTE that the payload starts at a long aligned offset,
330 * regardless of 32 or 64 bit arch!
335 u16 length; /* bytes of data after this header */
338 /* 8 bytes. packet FIXED for the next century! */
341 * short commands, packets without payload, plain p_header:
344 * P_BECOME_SYNC_TARGET
345 * P_BECOME_SYNC_SOURCE
350 * commands with out-of-struct payload:
351 * P_BITMAP (no additional fields)
352 * P_DATA, P_DATA_REPLY (see p_data)
353 * P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
356 /* these defines must not be changed without changing the protocol version */
357 #define DP_HARDBARRIER 1
359 #define DP_MAY_SET_IN_SYNC 4
362 struct p_header head;
363 u64 sector; /* 64 bits sector number */
364 u64 block_id; /* to identify the request in protocol B&C */
370 * commands which share a struct:
372 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
373 * P_DISCARD_ACK (proto C, two-primaries conflict detection)
375 * P_DATA_REQUEST, P_RS_DATA_REQUEST
378 struct p_header head;
387 struct p_header head;
391 u32 pad; /* to multiple of 8 Byte */
395 * commands with their own struct for additional fields:
404 struct p_header head; /* 8 bytes */
409 /* should be more than enough for future enhancements
410 * for now, feature_flags and the reserverd array shall be zero.
416 /* 80 bytes, FIXED for the next century */
419 struct p_header head;
420 u32 barrier; /* barrier number _handle_ only */
421 u32 pad; /* to multiple of 8 Byte */
424 struct p_barrier_ack {
425 struct p_header head;
431 struct p_header head;
434 /* Since protocol version 88 and higher. */
438 struct p_rs_param_89 {
439 struct p_header head;
441 /* protocol version 89: */
442 char verify_alg[SHARED_SECRET_MAX];
443 char csums_alg[SHARED_SECRET_MAX];
447 struct p_header head;
455 /* Since protocol version 87 and higher. */
456 char integrity_alg[0];
461 struct p_header head;
462 u64 uuid[UI_EXTENDED_SIZE];
466 struct p_header head;
471 struct p_header head;
472 u64 d_size; /* size of disk */
473 u64 u_size; /* user requested size */
474 u64 c_size; /* current exported size */
475 u32 max_segment_size; /* Maximal size of a BIO */
476 u32 queue_order_type;
480 struct p_header head;
485 struct p_header head;
490 struct p_req_state_reply {
491 struct p_header head;
495 struct p_drbd06_param {
506 struct p_header head;
512 /* Valid values for the encoding field.
513 * Bump proto version when changing this. */
514 enum drbd_bitmap_code {
515 /* RLE_VLI_Bytes = 0,
516 * and other bit variants had been defined during
517 * algorithm evaluation. */
521 struct p_compressed_bm {
522 struct p_header head;
523 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
524 * (encoding & 0x80): polarity (set/unset) of first runlength
525 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
526 * used to pad up to head.length bytes
533 /* DCBP: Drbd Compressed Bitmap Packet ... */
534 static inline enum drbd_bitmap_code
535 DCBP_get_code(struct p_compressed_bm *p)
537 return (enum drbd_bitmap_code)(p->encoding & 0x0f);
541 DCBP_set_code(struct p_compressed_bm *p, enum drbd_bitmap_code code)
544 p->encoding = (p->encoding & ~0xf) | code;
548 DCBP_get_start(struct p_compressed_bm *p)
550 return (p->encoding & 0x80) != 0;
554 DCBP_set_start(struct p_compressed_bm *p, int set)
556 p->encoding = (p->encoding & ~0x80) | (set ? 0x80 : 0);
560 DCBP_get_pad_bits(struct p_compressed_bm *p)
562 return (p->encoding >> 4) & 0x7;
566 DCBP_set_pad_bits(struct p_compressed_bm *p, int n)
569 p->encoding = (p->encoding & (~0x7 << 4)) | (n << 4);
572 /* one bitmap packet, including the p_header,
573 * should fit within one _architecture independend_ page.
574 * so we need to use the fixed size 4KiB page size
575 * most architechtures have used for a long time.
577 #define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header))
578 #define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
579 #define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm))
580 #if (PAGE_SIZE < 4096)
581 /* drbd_send_bitmap / receive_bitmap would break horribly */
582 #error "PAGE_SIZE too small"
586 struct p_header header;
587 struct p_handshake handshake;
589 struct p_block_ack block_ack;
590 struct p_barrier barrier;
591 struct p_barrier_ack barrier_ack;
592 struct p_rs_param_89 rs_param_89;
593 struct p_protocol protocol;
594 struct p_sizes sizes;
595 struct p_uuids uuids;
596 struct p_state state;
597 struct p_req_state req_state;
598 struct p_req_state_reply req_state_reply;
599 struct p_block_req block_req;
602 /**********************************************************************/
603 enum drbd_thread_state {
612 struct task_struct *task;
613 struct completion stop;
614 enum drbd_thread_state t_state;
615 int (*function) (struct drbd_thread *);
616 struct drbd_conf *mdev;
620 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
622 /* THINK testing the t_state seems to be uncritical in all cases
623 * (but thread_{start,stop}), so we can read it *without* the lock.
632 * Having this as the first member of a struct provides sort of "inheritance".
633 * "derived" structs can be "drbd_queue_work()"ed.
634 * The callback should know and cast back to the descendant struct.
635 * drbd_request and drbd_epoch_entry are descendants of drbd_work.
638 typedef int (*drbd_work_cb)(struct drbd_conf *, struct drbd_work *, int cancel);
640 struct list_head list;
644 struct drbd_tl_epoch;
645 struct drbd_request {
647 struct drbd_conf *mdev;
649 /* if local IO is not allowed, will be NULL.
650 * if local IO _is_ allowed, holds the locally submitted bio clone,
651 * or, after local IO completion, the ERR_PTR(error).
652 * see drbd_endio_pri(). */
653 struct bio *private_bio;
655 struct hlist_node colision;
658 unsigned int epoch; /* barrier_nr */
660 /* barrier_nr: used to check on "completion" whether this req was in
661 * the current epoch, and we therefore have to close it,
662 * starting a new epoch...
665 /* up to here, the struct layout is identical to drbd_epoch_entry;
666 * we might be able to use that to our advantage... */
668 struct list_head tl_requests; /* ring list in the transfer log */
669 struct bio *master_bio; /* master bio pointer */
670 unsigned long rq_state; /* see comments above _req_mod() */
672 unsigned long start_time;
675 struct drbd_tl_epoch {
677 struct list_head requests; /* requests before */
678 struct drbd_tl_epoch *next; /* pointer to the next barrier */
679 unsigned int br_number; /* the barriers identifier. */
680 int n_req; /* number of requests attached before this barrier */
685 /* These Tl_epoch_entries may be in one of 6 lists:
686 active_ee .. data packet being written
687 sync_ee .. syncer block being written
688 done_ee .. block written, need to send P_WRITE_ACK
689 read_ee .. [RS]P_DATA_REQUEST being read
693 struct list_head list;
694 unsigned int barrier_nr;
695 atomic_t epoch_size; /* increased on every request added. */
696 atomic_t active; /* increased on every req. added, and dec on every finished. */
700 /* drbd_epoch flag bits */
702 DE_BARRIER_IN_NEXT_EPOCH_ISSUED,
703 DE_BARRIER_IN_NEXT_EPOCH_DONE,
704 DE_CONTAINS_A_BARRIER,
705 DE_HAVE_BARRIER_NUMBER,
714 EV_CLEANUP = 32, /* used as flag */
717 struct drbd_epoch_entry {
719 struct drbd_conf *mdev;
720 struct bio *private_bio;
721 struct hlist_node colision;
724 struct drbd_epoch *epoch;
726 /* up to here, the struct layout is identical to drbd_request;
727 * we might be able to use that to our advantage... */
733 struct drbd_wq_barrier {
735 struct completion done;
745 __EE_CALL_AL_COMPLETE_IO,
746 __EE_CONFLICT_PENDING,
747 __EE_MAY_SET_IN_SYNC,
750 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
751 #define EE_CONFLICT_PENDING (1<<__EE_CONFLICT_PENDING)
752 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
753 #define EE_IS_BARRIER (1<<__EE_IS_BARRIER)
755 /* global flag bits */
757 CREATE_BARRIER, /* next P_DATA is preceeded by a P_BARRIER */
758 SIGNAL_ASENDER, /* whether asender wants to be interrupted */
759 SEND_PING, /* whether asender should send a ping asap */
761 STOP_SYNC_TIMER, /* tell timer to cancel itself */
762 UNPLUG_QUEUED, /* only relevant with kernel 2.4 */
763 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
764 MD_DIRTY, /* current uuids and flags not yet on disk */
765 DISCARD_CONCURRENT, /* Set on one node, cleared on the peer! */
766 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
767 CLUSTER_ST_CHANGE, /* Cluster wide state change going on... */
770 CRASHED_PRIMARY, /* This node was a crashed primary.
771 * Gets cleared when the state.conn
772 * goes into C_CONNECTED state. */
773 WRITE_BM_AFTER_RESYNC, /* A kmalloc() during resync failed */
774 NO_BARRIER_SUPP, /* underlying block device doesn't implement barriers */
777 MD_NO_BARRIER, /* meta data device does not support barriers,
779 SUSPEND_IO, /* suspend application io */
780 BITMAP_IO, /* suspend application io;
781 once no more io in flight, start bitmap io */
782 BITMAP_IO_QUEUED, /* Started bitmap IO */
783 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
784 NET_CONGESTED, /* The data socket is congested */
786 CONFIG_PENDING, /* serialization of (re)configuration requests.
787 * if set, also prevents the device from dying */
788 DEVICE_DYING, /* device became unconfigured,
789 * but worker thread is still handling the cleanup.
790 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
791 * while this is set. */
792 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
793 * the peer, if it changed there as well. */
796 struct drbd_bitmap; /* opaque for drbd_conf */
798 /* TODO sort members for performance
799 * MAYBE group them further */
801 /* THINK maybe we actually want to use the default "event/%s" worker threads
802 * or similar in linux 2.6, which uses per cpu data and threads.
804 * To be general, this might need a spin_lock member.
805 * For now, please use the mdev->req_lock to protect list_head,
806 * see drbd_queue_work below.
808 struct drbd_work_queue {
810 struct semaphore s; /* producers up it, worker down()s it */
811 spinlock_t q_lock; /* to protect the list. */
815 struct drbd_work_queue work;
817 struct socket *socket;
818 /* this way we get our
819 * send/receive buffers off the stack */
820 union p_polymorph sbuf;
821 union p_polymorph rbuf;
825 u64 md_offset; /* sector offset to 'super' block */
827 u64 la_size_sect; /* last agreed size, unit sectors */
833 s32 al_offset; /* signed relative sector offset to al area */
834 s32 bm_offset; /* signed relative sector offset to bitmap */
836 /* u32 al_nr_extents; important for restoring the AL
837 * is stored into sync_conf.al_extents, which in turn
838 * gets applied to act_log->nr_elements
842 /* for sync_conf and other types... */
843 #define NL_PACKET(name, number, fields) struct name { fields };
844 #define NL_INTEGER(pn,pr,member) int member;
845 #define NL_INT64(pn,pr,member) __u64 member;
846 #define NL_BIT(pn,pr,member) unsigned member:1;
847 #define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
848 #include "linux/drbd_nl.h"
850 struct drbd_backing_dev {
851 struct block_device *backing_bdev;
852 struct block_device *md_bdev;
853 struct file *lo_file;
854 struct file *md_file;
856 struct disk_conf dc; /* The user provided config... */
857 sector_t known_size; /* last known size of that backing device */
861 struct drbd_conf *mdev;
862 struct completion event;
869 int (*io_fn)(struct drbd_conf *mdev);
870 void (*done)(struct drbd_conf *mdev, int rv);
873 enum write_ordering_e {
881 /* things that are stored as / read from meta data on disk */
884 /* configured by drbdsetup */
885 struct net_conf *net_conf; /* protected by get_net_conf() and put_net_conf() */
886 struct syncer_conf sync_conf;
887 struct drbd_backing_dev *ldev __protected_by(local);
889 sector_t p_size; /* partner's disk size */
890 struct request_queue *rq_queue;
891 struct block_device *this_bdev;
892 struct gendisk *vdisk;
894 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
895 struct drbd_socket meta; /* ping/ack (metadata) packets */
896 int agreed_pro_version; /* actually used protocol version */
897 unsigned long last_received; /* in jiffies, either socket */
898 unsigned int ko_count;
899 struct drbd_work resync_work,
902 struct timer_list resync_timer;
903 struct timer_list md_sync_timer;
905 /* Used after attach while negotiating new disk state. */
906 union drbd_state new_state_tmp;
908 union drbd_state state;
909 wait_queue_head_t misc_wait;
910 wait_queue_head_t state_wait; /* upon each state change. */
911 unsigned int send_cnt;
912 unsigned int recv_cnt;
913 unsigned int read_cnt;
914 unsigned int writ_cnt;
915 unsigned int al_writ_cnt;
916 unsigned int bm_writ_cnt;
917 atomic_t ap_bio_cnt; /* Requests we need to complete */
918 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
919 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
920 atomic_t unacked_cnt; /* Need to send replys for */
921 atomic_t local_cnt; /* Waiting for local completion */
922 atomic_t net_cnt; /* Users of net_conf */
924 struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */
925 struct drbd_tl_epoch *newest_tle;
926 struct drbd_tl_epoch *oldest_tle;
927 struct list_head out_of_sequence_requests;
928 struct hlist_head *tl_hash;
929 unsigned int tl_hash_s;
931 /* blocks to sync in this run [unit BM_BLOCK_SIZE] */
932 unsigned long rs_total;
933 /* number of sync IOs that failed in this run */
934 unsigned long rs_failed;
935 /* Syncer's start time [unit jiffies] */
936 unsigned long rs_start;
937 /* cumulated time in PausedSyncX state [unit jiffies] */
938 unsigned long rs_paused;
939 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
940 unsigned long rs_mark_left;
941 /* marks's time [unit jiffies] */
942 unsigned long rs_mark_time;
943 /* skipped because csum was equeal [unit BM_BLOCK_SIZE] */
944 unsigned long rs_same_csum;
946 /* where does the admin want us to start? (sector) */
947 sector_t ov_start_sector;
948 /* where are we now? (sector) */
949 sector_t ov_position;
950 /* Start sector of out of sync range (to merge printk reporting). */
951 sector_t ov_last_oos_start;
952 /* size of out-of-sync range in sectors. */
953 sector_t ov_last_oos_size;
954 unsigned long ov_left; /* in bits */
955 struct crypto_hash *csums_tfm;
956 struct crypto_hash *verify_tfm;
958 struct drbd_thread receiver;
959 struct drbd_thread worker;
960 struct drbd_thread asender;
961 struct drbd_bitmap *bitmap;
962 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
964 /* Used to track operations of resync... */
965 struct lru_cache *resync;
966 /* Number of locked elements in resync LRU */
967 unsigned int resync_locked;
968 /* resync extent number waiting for application requests */
969 unsigned int resync_wenr;
973 struct drbd_epoch *current_epoch;
974 spinlock_t epoch_lock;
976 enum write_ordering_e write_ordering;
977 struct list_head active_ee; /* IO in progress */
978 struct list_head sync_ee; /* IO in progress */
979 struct list_head done_ee; /* send ack */
980 struct list_head read_ee; /* IO in progress */
981 struct list_head net_ee; /* zero-copy network send in progress */
982 struct hlist_head *ee_hash; /* is proteced by req_lock! */
983 unsigned int ee_hash_s;
985 /* this one is protected by ee_lock, single thread */
986 struct drbd_epoch_entry *last_write_w_barrier;
989 struct hlist_head *app_reads_hash; /* is proteced by req_lock */
990 struct list_head resync_reads;
992 wait_queue_head_t ee_wait;
993 struct page *md_io_page; /* one page buffer for md_io */
994 struct page *md_io_tmpp; /* for logical_block_size != 512 */
995 struct mutex md_io_mutex; /* protects the md_io_buffer */
997 wait_queue_head_t al_wait;
998 struct lru_cache *act_log; /* activity log */
999 unsigned int al_tr_number;
1001 int al_tr_pos; /* position of the next transaction in the journal */
1002 struct crypto_hash *cram_hmac_tfm;
1003 struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */
1004 struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */
1008 wait_queue_head_t seq_wait;
1009 atomic_t packet_seq;
1010 unsigned int peer_seq;
1011 spinlock_t peer_seq_lock;
1013 unsigned long comm_bm_set; /* communicated number of set bits. */
1014 cpumask_var_t cpu_mask;
1015 struct bm_io_work bm_io_work;
1016 u64 ed_uuid; /* UUID of the exposed data */
1017 struct mutex state_mutex;
1018 char congestion_reason; /* Why we where congested... */
1021 static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
1023 struct drbd_conf *mdev;
1025 mdev = minor < minor_count ? minor_table[minor] : NULL;
1030 static inline unsigned int mdev_to_minor(struct drbd_conf *mdev)
1035 /* returns 1 if it was successfull,
1036 * returns 0 if there was no data socket.
1037 * so wherever you are going to use the data.socket, e.g. do
1038 * if (!drbd_get_data_sock(mdev))
1041 * drbd_put_data_sock(mdev);
1043 static inline int drbd_get_data_sock(struct drbd_conf *mdev)
1045 mutex_lock(&mdev->data.mutex);
1046 /* drbd_disconnect() could have called drbd_free_sock()
1047 * while we were waiting in down()... */
1048 if (unlikely(mdev->data.socket == NULL)) {
1049 mutex_unlock(&mdev->data.mutex);
1055 static inline void drbd_put_data_sock(struct drbd_conf *mdev)
1057 mutex_unlock(&mdev->data.mutex);
1061 * function declarations
1062 *************************/
1066 enum chg_state_flags {
1069 CS_WAIT_COMPLETE = 4,
1071 CS_ORDERED = CS_WAIT_COMPLETE + CS_SERIALIZE,
1074 extern void drbd_init_set_defaults(struct drbd_conf *mdev);
1075 extern int drbd_change_state(struct drbd_conf *mdev, enum chg_state_flags f,
1076 union drbd_state mask, union drbd_state val);
1077 extern void drbd_force_state(struct drbd_conf *, union drbd_state,
1079 extern int _drbd_request_state(struct drbd_conf *, union drbd_state,
1080 union drbd_state, enum chg_state_flags);
1081 extern int __drbd_set_state(struct drbd_conf *, union drbd_state,
1082 enum chg_state_flags, struct completion *done);
1083 extern void print_st_err(struct drbd_conf *, union drbd_state,
1084 union drbd_state, int);
1085 extern int drbd_thread_start(struct drbd_thread *thi);
1086 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1088 extern void drbd_thread_current_set_cpu(struct drbd_conf *mdev);
1089 extern void drbd_calc_cpu_mask(struct drbd_conf *mdev);
1091 #define drbd_thread_current_set_cpu(A) ({})
1092 #define drbd_calc_cpu_mask(A) ({})
1094 extern void drbd_free_resources(struct drbd_conf *mdev);
1095 extern void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr,
1096 unsigned int set_size);
1097 extern void tl_clear(struct drbd_conf *mdev);
1098 extern void _tl_add_barrier(struct drbd_conf *, struct drbd_tl_epoch *);
1099 extern void drbd_free_sock(struct drbd_conf *mdev);
1100 extern int drbd_send(struct drbd_conf *mdev, struct socket *sock,
1101 void *buf, size_t size, unsigned msg_flags);
1102 extern int drbd_send_protocol(struct drbd_conf *mdev);
1103 extern int drbd_send_uuids(struct drbd_conf *mdev);
1104 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
1105 extern int drbd_send_sync_uuid(struct drbd_conf *mdev, u64 val);
1106 extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply);
1107 extern int _drbd_send_state(struct drbd_conf *mdev);
1108 extern int drbd_send_state(struct drbd_conf *mdev);
1109 extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock,
1110 enum drbd_packets cmd, struct p_header *h,
1111 size_t size, unsigned msg_flags);
1112 #define USE_DATA_SOCKET 1
1113 #define USE_META_SOCKET 0
1114 extern int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket,
1115 enum drbd_packets cmd, struct p_header *h,
1117 extern int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd,
1118 char *data, size_t size);
1119 extern int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc);
1120 extern int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr,
1122 extern int drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd,
1123 struct drbd_epoch_entry *e);
1124 extern int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packets cmd,
1125 struct p_block_req *rp);
1126 extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd,
1128 extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd,
1129 sector_t sector, int blksize, u64 block_id);
1130 extern int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd,
1131 struct drbd_epoch_entry *e);
1132 extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
1133 extern int _drbd_send_barrier(struct drbd_conf *mdev,
1134 struct drbd_tl_epoch *barrier);
1135 extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1136 sector_t sector, int size, u64 block_id);
1137 extern int drbd_send_drequest_csum(struct drbd_conf *mdev,
1138 sector_t sector,int size,
1139 void *digest, int digest_size,
1140 enum drbd_packets cmd);
1141 extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size);
1143 extern int drbd_send_bitmap(struct drbd_conf *mdev);
1144 extern int _drbd_send_bitmap(struct drbd_conf *mdev);
1145 extern int drbd_send_sr_reply(struct drbd_conf *mdev, int retcode);
1146 extern void drbd_free_bc(struct drbd_backing_dev *ldev);
1147 extern void drbd_mdev_cleanup(struct drbd_conf *mdev);
1149 /* drbd_meta-data.c (still in drbd_main.c) */
1150 extern void drbd_md_sync(struct drbd_conf *mdev);
1151 extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
1152 /* maybe define them below as inline? */
1153 extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1154 extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1155 extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1156 extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1157 extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
1158 extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
1159 extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
1160 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1161 extern void drbd_md_mark_dirty(struct drbd_conf *mdev);
1162 extern void drbd_queue_bitmap_io(struct drbd_conf *mdev,
1163 int (*io_fn)(struct drbd_conf *),
1164 void (*done)(struct drbd_conf *, int),
1166 extern int drbd_bmio_set_n_write(struct drbd_conf *mdev);
1167 extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev);
1168 extern int drbd_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), char *why);
1172 We reserve a 128MB Block (4k aligned)
1173 * either at the end of the backing device
1174 * or on a seperate meta data device. */
1176 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1177 /* The following numbers are sectors */
1178 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1179 #define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */
1180 /* Allows up to about 3.8TB */
1181 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1183 /* Since the smalles IO unit is usually 512 byte */
1184 #define MD_SECTOR_SHIFT 9
1185 #define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT)
1188 #define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1189 #define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */
1190 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1192 #if BITS_PER_LONG == 32
1194 #define cpu_to_lel(A) cpu_to_le32(A)
1195 #define lel_to_cpu(A) le32_to_cpu(A)
1196 #elif BITS_PER_LONG == 64
1198 #define cpu_to_lel(A) cpu_to_le64(A)
1199 #define lel_to_cpu(A) le64_to_cpu(A)
1201 #error "LN2 of BITS_PER_LONG unknown!"
1205 /* 16MB sized 'bitmap extent' to track syncer usage */
1207 int rs_left; /* number of bits set (out of sync) in this extent. */
1208 int rs_failed; /* number of failed resync requests in this extent. */
1209 unsigned long flags;
1210 struct lc_element lce;
1213 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1214 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1218 * We need to store one bit for a block.
1219 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1220 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1221 * Bit 1 ==> local node thinks this block needs to be synced.
1224 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1225 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1226 /* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1227 * per sector of on disk bitmap */
1228 #define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */
1229 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1231 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1232 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1235 /* thus many _storage_ sectors are described by one bit */
1236 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1237 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1238 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1240 /* bit to represented kilo byte conversion */
1241 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1243 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1244 * _storage_ sector is located in */
1245 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1247 /* how much _storage_ sectors we have per bitmap sector */
1248 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1249 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1251 /* in one sector of the bitmap, we have this many activity_log extents. */
1252 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1253 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1255 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1256 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1258 /* the extent in "PER_EXTENT" below is an activity log extent
1259 * we need that many (long words/bytes) to store the bitmap
1260 * of one AL_EXTENT_SIZE chunk of storage.
1261 * we can store the bitmap for that many AL_EXTENTS within
1262 * one sector of the _on_disk_ bitmap:
1263 * bit 0 bit 37 bit 38 bit (512*8)-1
1264 * ...|........|........|.. // ..|........|
1265 * sect. 0 `296 `304 ^(512*8*8)-1
1267 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1268 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1269 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1272 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1273 #define DRBD_MAX_SECTORS_BM \
1274 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1275 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1276 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1277 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1278 #elif !defined(CONFIG_LBD) && BITS_PER_LONG == 32
1279 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1280 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1282 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1283 /* 16 TB in units of sectors */
1284 #if BITS_PER_LONG == 32
1285 /* adjust by one page worth of bitmap,
1286 * so we won't wrap around in drbd_bm_find_next_bit.
1287 * you should use 64bit OS for that much storage, anyways. */
1288 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1290 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0x1LU << 32)
1294 /* Sector shift value for the "hash" functions of tl_hash and ee_hash tables.
1295 * With a value of 6 all IO in one 32K block make it to the same slot of the
1298 #define DRBD_MAX_SEGMENT_SIZE (1U<<(9+HT_SHIFT))
1300 /* Number of elements in the app_reads_hash */
1301 #define APP_R_HSIZE 15
1303 extern int drbd_bm_init(struct drbd_conf *mdev);
1304 extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors);
1305 extern void drbd_bm_cleanup(struct drbd_conf *mdev);
1306 extern void drbd_bm_set_all(struct drbd_conf *mdev);
1307 extern void drbd_bm_clear_all(struct drbd_conf *mdev);
1308 extern int drbd_bm_set_bits(
1309 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1310 extern int drbd_bm_clear_bits(
1311 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1312 /* bm_set_bits variant for use while holding drbd_bm_lock */
1313 extern void _drbd_bm_set_bits(struct drbd_conf *mdev,
1314 const unsigned long s, const unsigned long e);
1315 extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr);
1316 extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr);
1317 extern int drbd_bm_write_sect(struct drbd_conf *mdev, unsigned long enr) __must_hold(local);
1318 extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
1319 extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
1320 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev,
1321 unsigned long al_enr);
1322 extern size_t drbd_bm_words(struct drbd_conf *mdev);
1323 extern unsigned long drbd_bm_bits(struct drbd_conf *mdev);
1324 extern sector_t drbd_bm_capacity(struct drbd_conf *mdev);
1325 extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1326 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1327 extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1328 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo);
1329 extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev);
1330 extern int drbd_bm_rs_done(struct drbd_conf *mdev);
1331 /* for receive_bitmap */
1332 extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset,
1333 size_t number, unsigned long *buffer);
1334 /* for _drbd_send_bitmap and drbd_bm_write_sect */
1335 extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset,
1336 size_t number, unsigned long *buffer);
1338 extern void drbd_bm_lock(struct drbd_conf *mdev, char *why);
1339 extern void drbd_bm_unlock(struct drbd_conf *mdev);
1341 extern int drbd_bm_count_bits(struct drbd_conf *mdev, const unsigned long s, const unsigned long e);
1344 extern struct kmem_cache *drbd_request_cache;
1345 extern struct kmem_cache *drbd_ee_cache; /* epoch entries */
1346 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1347 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1348 extern mempool_t *drbd_request_mempool;
1349 extern mempool_t *drbd_ee_mempool;
1351 extern struct page *drbd_pp_pool; /* drbd's page pool */
1352 extern spinlock_t drbd_pp_lock;
1353 extern int drbd_pp_vacant;
1354 extern wait_queue_head_t drbd_pp_wait;
1356 extern rwlock_t global_state_lock;
1358 extern struct drbd_conf *drbd_new_device(unsigned int minor);
1359 extern void drbd_free_mdev(struct drbd_conf *mdev);
1361 extern int proc_details;
1364 extern int drbd_make_request_26(struct request_queue *q, struct bio *bio);
1365 extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req);
1366 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1367 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1371 extern void drbd_suspend_io(struct drbd_conf *mdev);
1372 extern void drbd_resume_io(struct drbd_conf *mdev);
1373 extern char *ppsize(char *buf, unsigned long long size);
1374 extern sector_t drbd_new_dev_size(struct drbd_conf *,
1375 struct drbd_backing_dev *);
1376 enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
1377 extern enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *) __must_hold(local);
1378 extern void resync_after_online_grow(struct drbd_conf *);
1379 extern void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int) __must_hold(local);
1380 extern int drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role,
1382 enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev);
1383 extern int drbd_khelper(struct drbd_conf *mdev, char *cmd);
1386 extern int drbd_worker(struct drbd_thread *thi);
1387 extern int drbd_alter_sa(struct drbd_conf *mdev, int na);
1388 extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side);
1389 extern void resume_next_sg(struct drbd_conf *mdev);
1390 extern void suspend_other_sg(struct drbd_conf *mdev);
1391 extern int drbd_resync_finished(struct drbd_conf *mdev);
1392 /* maybe rather drbd_main.c ? */
1393 extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
1394 struct drbd_backing_dev *bdev, sector_t sector, int rw);
1395 extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int);
1397 static inline void ov_oos_print(struct drbd_conf *mdev)
1399 if (mdev->ov_last_oos_size) {
1400 dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n",
1401 (unsigned long long)mdev->ov_last_oos_start,
1402 (unsigned long)mdev->ov_last_oos_size);
1404 mdev->ov_last_oos_size=0;
1408 extern void drbd_csum(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
1409 /* worker callbacks */
1410 extern int w_req_cancel_conflict(struct drbd_conf *, struct drbd_work *, int);
1411 extern int w_read_retry_remote(struct drbd_conf *, struct drbd_work *, int);
1412 extern int w_e_end_data_req(struct drbd_conf *, struct drbd_work *, int);
1413 extern int w_e_end_rsdata_req(struct drbd_conf *, struct drbd_work *, int);
1414 extern int w_e_end_csum_rs_req(struct drbd_conf *, struct drbd_work *, int);
1415 extern int w_e_end_ov_reply(struct drbd_conf *, struct drbd_work *, int);
1416 extern int w_e_end_ov_req(struct drbd_conf *, struct drbd_work *, int);
1417 extern int w_ov_finished(struct drbd_conf *, struct drbd_work *, int);
1418 extern int w_resync_inactive(struct drbd_conf *, struct drbd_work *, int);
1419 extern int w_resume_next_sg(struct drbd_conf *, struct drbd_work *, int);
1420 extern int w_io_error(struct drbd_conf *, struct drbd_work *, int);
1421 extern int w_send_write_hint(struct drbd_conf *, struct drbd_work *, int);
1422 extern int w_make_resync_request(struct drbd_conf *, struct drbd_work *, int);
1423 extern int w_send_dblock(struct drbd_conf *, struct drbd_work *, int);
1424 extern int w_send_barrier(struct drbd_conf *, struct drbd_work *, int);
1425 extern int w_send_read_req(struct drbd_conf *, struct drbd_work *, int);
1426 extern int w_prev_work_done(struct drbd_conf *, struct drbd_work *, int);
1427 extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int);
1429 extern void resync_timer_fn(unsigned long data);
1431 /* drbd_receiver.c */
1432 extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list);
1433 extern struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev,
1436 unsigned int data_size,
1437 gfp_t gfp_mask) __must_hold(local);
1438 extern void drbd_free_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e);
1439 extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1440 struct list_head *head);
1441 extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1442 struct list_head *head);
1443 extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled);
1444 extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed);
1445 extern void drbd_flush_workqueue(struct drbd_conf *mdev);
1447 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1448 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1449 static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1450 char __user *optval, int optlen)
1453 if (level == SOL_SOCKET)
1454 err = sock_setsockopt(sock, level, optname, optval, optlen);
1456 err = sock->ops->setsockopt(sock, level, optname, optval,
1461 static inline void drbd_tcp_cork(struct socket *sock)
1464 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1465 (char __user *)&val, sizeof(val));
1468 static inline void drbd_tcp_uncork(struct socket *sock)
1471 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1472 (char __user *)&val, sizeof(val));
1475 static inline void drbd_tcp_nodelay(struct socket *sock)
1478 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1479 (char __user *)&val, sizeof(val));
1482 static inline void drbd_tcp_quickack(struct socket *sock)
1485 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1486 (char __user *)&val, sizeof(val));
1489 void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo);
1492 extern struct proc_dir_entry *drbd_proc;
1493 extern const struct file_operations drbd_proc_fops;
1494 extern const char *drbd_conn_str(enum drbd_conns s);
1495 extern const char *drbd_role_str(enum drbd_role s);
1498 extern void drbd_al_begin_io(struct drbd_conf *mdev, sector_t sector);
1499 extern void drbd_al_complete_io(struct drbd_conf *mdev, sector_t sector);
1500 extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector);
1501 extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1502 extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1503 extern void drbd_rs_cancel_all(struct drbd_conf *mdev);
1504 extern int drbd_rs_del_all(struct drbd_conf *mdev);
1505 extern void drbd_rs_failed_io(struct drbd_conf *mdev,
1506 sector_t sector, int size);
1507 extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *);
1508 extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
1509 int size, const char *file, const unsigned int line);
1510 #define drbd_set_in_sync(mdev, sector, size) \
1511 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1512 extern void __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
1513 int size, const char *file, const unsigned int line);
1514 #define drbd_set_out_of_sync(mdev, sector, size) \
1515 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1516 extern void drbd_al_apply_to_bm(struct drbd_conf *mdev);
1517 extern void drbd_al_to_on_disk_bm(struct drbd_conf *mdev);
1518 extern void drbd_al_shrink(struct drbd_conf *mdev);
1523 void drbd_nl_cleanup(void);
1524 int __init drbd_nl_init(void);
1525 void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state);
1526 void drbd_bcast_sync_progress(struct drbd_conf *mdev);
1527 void drbd_bcast_ee(struct drbd_conf *mdev,
1528 const char *reason, const int dgs,
1529 const char* seen_hash, const char* calc_hash,
1530 const struct drbd_epoch_entry* e);
1534 * DOC: DRBD State macros
1536 * These macros are used to express state changes in easily readable form.
1538 * The NS macros expand to a mask and a value, that can be bit ored onto the
1539 * current state as soon as the spinlock (req_lock) was taken.
1541 * The _NS macros are used for state functions that get called with the
1542 * spinlock. These macros expand directly to the new state value.
1544 * Besides the basic forms NS() and _NS() additional _?NS[23] are defined
1545 * to express state changes that affect more than one aspect of the state.
1547 * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY)
1548 * Means that the network connection was established and that the peer
1549 * is in secondary role.
1551 #define role_MASK R_MASK
1552 #define peer_MASK R_MASK
1553 #define disk_MASK D_MASK
1554 #define pdsk_MASK D_MASK
1555 #define conn_MASK C_MASK
1557 #define user_isp_MASK 1
1558 #define aftr_isp_MASK 1
1561 ({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \
1562 ({ union drbd_state val; val.i = 0; val.T = (S); val; })
1563 #define NS2(T1, S1, T2, S2) \
1564 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1565 mask.T2 = T2##_MASK; mask; }), \
1566 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1567 val.T2 = (S2); val; })
1568 #define NS3(T1, S1, T2, S2, T3, S3) \
1569 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1570 mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \
1571 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1572 val.T2 = (S2); val.T3 = (S3); val; })
1574 #define _NS(D, T, S) \
1575 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; })
1576 #define _NS2(D, T1, S1, T2, S2) \
1577 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1578 __ns.T2 = (S2); __ns; })
1579 #define _NS3(D, T1, S1, T2, S2, T3, S3) \
1580 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1581 __ns.T2 = (S2); __ns.T3 = (S3); __ns; })
1584 * inline helper functions
1585 *************************/
1587 static inline void drbd_state_lock(struct drbd_conf *mdev)
1589 wait_event(mdev->misc_wait,
1590 !test_and_set_bit(CLUSTER_ST_CHANGE, &mdev->flags));
1593 static inline void drbd_state_unlock(struct drbd_conf *mdev)
1595 clear_bit(CLUSTER_ST_CHANGE, &mdev->flags);
1596 wake_up(&mdev->misc_wait);
1599 static inline int _drbd_set_state(struct drbd_conf *mdev,
1600 union drbd_state ns, enum chg_state_flags flags,
1601 struct completion *done)
1605 read_lock(&global_state_lock);
1606 rv = __drbd_set_state(mdev, ns, flags, done);
1607 read_unlock(&global_state_lock);
1613 * drbd_request_state() - Reqest a state change
1614 * @mdev: DRBD device.
1615 * @mask: mask of state bits to change.
1616 * @val: value of new state bits.
1618 * This is the most graceful way of requesting a state change. It is verbose
1619 * quite verbose in case the state change is not possible, and all those
1620 * state changes are globally serialized.
1622 static inline int drbd_request_state(struct drbd_conf *mdev,
1623 union drbd_state mask,
1624 union drbd_state val)
1626 return _drbd_request_state(mdev, mask, val, CS_VERBOSE + CS_ORDERED);
1629 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1630 static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach, const char *where)
1632 switch (mdev->ldev->dc.on_io_error) {
1635 if (printk_ratelimit())
1636 dev_err(DEV, "Local IO failed in %s."
1637 "Passing error on...\n", where);
1640 /* NOTE fall through to detach case if forcedetach set */
1642 case EP_CALL_HELPER:
1643 if (mdev->state.disk > D_FAILED) {
1644 _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL);
1645 dev_err(DEV, "Local IO failed in %s."
1646 "Detaching...\n", where);
1653 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1654 * @mdev: DRBD device.
1655 * @error: Error code passed to the IO completion callback
1656 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1658 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1660 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1661 static inline void drbd_chk_io_error_(struct drbd_conf *mdev,
1662 int error, int forcedetach, const char *where)
1665 unsigned long flags;
1666 spin_lock_irqsave(&mdev->req_lock, flags);
1667 __drbd_chk_io_error_(mdev, forcedetach, where);
1668 spin_unlock_irqrestore(&mdev->req_lock, flags);
1674 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1675 * @bdev: Meta data block device.
1677 * BTW, for internal meta data, this happens to be the maximum capacity
1678 * we could agree upon with our peer node.
1680 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1682 switch (bdev->dc.meta_dev_idx) {
1683 case DRBD_MD_INDEX_INTERNAL:
1684 case DRBD_MD_INDEX_FLEX_INT:
1685 return bdev->md.md_offset + bdev->md.bm_offset;
1686 case DRBD_MD_INDEX_FLEX_EXT:
1688 return bdev->md.md_offset;
1693 * drbd_md_last_sector() - Return the last sector number of the meta data area
1694 * @bdev: Meta data block device.
1696 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1698 switch (bdev->dc.meta_dev_idx) {
1699 case DRBD_MD_INDEX_INTERNAL:
1700 case DRBD_MD_INDEX_FLEX_INT:
1701 return bdev->md.md_offset + MD_AL_OFFSET - 1;
1702 case DRBD_MD_INDEX_FLEX_EXT:
1704 return bdev->md.md_offset + bdev->md.md_size_sect;
1708 /* Returns the number of 512 byte sectors of the device */
1709 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1711 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1712 return bdev ? bdev->bd_inode->i_size >> 9 : 0;
1716 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1717 * @bdev: Meta data block device.
1719 * returns the capacity we announce to out peer. we clip ourselves at the
1720 * various MAX_SECTORS, because if we don't, current implementation will
1721 * oops sooner or later
1723 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1726 switch (bdev->dc.meta_dev_idx) {
1727 case DRBD_MD_INDEX_INTERNAL:
1728 case DRBD_MD_INDEX_FLEX_INT:
1729 s = drbd_get_capacity(bdev->backing_bdev)
1730 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1731 drbd_md_first_sector(bdev))
1734 case DRBD_MD_INDEX_FLEX_EXT:
1735 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1736 drbd_get_capacity(bdev->backing_bdev));
1737 /* clip at maximum size the meta device can support */
1738 s = min_t(sector_t, s,
1739 BM_EXT_TO_SECT(bdev->md.md_size_sect
1740 - bdev->md.bm_offset));
1743 s = min_t(sector_t, DRBD_MAX_SECTORS,
1744 drbd_get_capacity(bdev->backing_bdev));
1750 * drbd_md_ss__() - Return the sector number of our meta data super block
1751 * @mdev: DRBD device.
1752 * @bdev: Meta data block device.
1754 static inline sector_t drbd_md_ss__(struct drbd_conf *mdev,
1755 struct drbd_backing_dev *bdev)
1757 switch (bdev->dc.meta_dev_idx) {
1758 default: /* external, some index */
1759 return MD_RESERVED_SECT * bdev->dc.meta_dev_idx;
1760 case DRBD_MD_INDEX_INTERNAL:
1761 /* with drbd08, internal meta data is always "flexible" */
1762 case DRBD_MD_INDEX_FLEX_INT:
1763 /* sizeof(struct md_on_disk_07) == 4k
1764 * position: last 4k aligned block of 4k size */
1765 if (!bdev->backing_bdev) {
1766 if (__ratelimit(&drbd_ratelimit_state)) {
1767 dev_err(DEV, "bdev->backing_bdev==NULL\n");
1772 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL)
1774 case DRBD_MD_INDEX_FLEX_EXT:
1780 _drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1782 list_add_tail(&w->list, &q->q);
1787 drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w)
1789 unsigned long flags;
1790 spin_lock_irqsave(&q->q_lock, flags);
1791 list_add(&w->list, &q->q);
1792 up(&q->s); /* within the spinlock,
1793 see comment near end of drbd_worker() */
1794 spin_unlock_irqrestore(&q->q_lock, flags);
1798 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1800 unsigned long flags;
1801 spin_lock_irqsave(&q->q_lock, flags);
1802 list_add_tail(&w->list, &q->q);
1803 up(&q->s); /* within the spinlock,
1804 see comment near end of drbd_worker() */
1805 spin_unlock_irqrestore(&q->q_lock, flags);
1808 static inline void wake_asender(struct drbd_conf *mdev)
1810 if (test_bit(SIGNAL_ASENDER, &mdev->flags))
1811 force_sig(DRBD_SIG, mdev->asender.task);
1814 static inline void request_ping(struct drbd_conf *mdev)
1816 set_bit(SEND_PING, &mdev->flags);
1820 static inline int drbd_send_short_cmd(struct drbd_conf *mdev,
1821 enum drbd_packets cmd)
1824 return drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, &h, sizeof(h));
1827 static inline int drbd_send_ping(struct drbd_conf *mdev)
1830 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING, &h, sizeof(h));
1833 static inline int drbd_send_ping_ack(struct drbd_conf *mdev)
1836 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING_ACK, &h, sizeof(h));
1839 static inline void drbd_thread_stop(struct drbd_thread *thi)
1841 _drbd_thread_stop(thi, FALSE, TRUE);
1844 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
1846 _drbd_thread_stop(thi, FALSE, FALSE);
1849 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
1851 _drbd_thread_stop(thi, TRUE, FALSE);
1854 /* counts how many answer packets packets we expect from our peer,
1855 * for either explicit application requests,
1856 * or implicit barrier packets as necessary.
1859 * _req_mod(req, queue_for_net_write or queue_for_net_read);
1860 * it is much easier and equally valid to count what we queue for the
1861 * worker, even before it actually was queued or send.
1862 * (drbd_make_request_common; recovery path on read io-error)
1864 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1865 * _req_mod(req, data_received)
1866 * [from receive_DataReply]
1867 * _req_mod(req, write_acked_by_peer or recv_acked_by_peer or neg_acked)
1868 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1869 * for some reason it is NOT decreased in got_NegAck,
1870 * but in the resulting cleanup code from report_params.
1871 * we should try to remember the reason for that...
1872 * _req_mod(req, send_failed or send_canceled)
1873 * _req_mod(req, connection_lost_while_pending)
1874 * [from tl_clear_barrier]
1876 static inline void inc_ap_pending(struct drbd_conf *mdev)
1878 atomic_inc(&mdev->ap_pending_cnt);
1881 #define ERR_IF_CNT_IS_NEGATIVE(which) \
1882 if (atomic_read(&mdev->which) < 0) \
1883 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
1884 __func__ , __LINE__ , \
1885 atomic_read(&mdev->which))
1887 #define dec_ap_pending(mdev) do { \
1888 typecheck(struct drbd_conf *, mdev); \
1889 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \
1890 wake_up(&mdev->misc_wait); \
1891 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
1893 /* counts how many resync-related answers we still expect from the peer
1895 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1896 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK whith ID_SYNCER)
1897 * (or P_NEG_ACK with ID_SYNCER)
1899 static inline void inc_rs_pending(struct drbd_conf *mdev)
1901 atomic_inc(&mdev->rs_pending_cnt);
1904 #define dec_rs_pending(mdev) do { \
1905 typecheck(struct drbd_conf *, mdev); \
1906 atomic_dec(&mdev->rs_pending_cnt); \
1907 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
1909 /* counts how many answers we still need to send to the peer.
1911 * receive_Data unless protocol A;
1912 * we need to send a P_RECV_ACK (proto B)
1913 * or P_WRITE_ACK (proto C)
1914 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1915 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1916 * receive_Barrier_* we need to send a P_BARRIER_ACK
1918 static inline void inc_unacked(struct drbd_conf *mdev)
1920 atomic_inc(&mdev->unacked_cnt);
1923 #define dec_unacked(mdev) do { \
1924 typecheck(struct drbd_conf *, mdev); \
1925 atomic_dec(&mdev->unacked_cnt); \
1926 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
1928 #define sub_unacked(mdev, n) do { \
1929 typecheck(struct drbd_conf *, mdev); \
1930 atomic_sub(n, &mdev->unacked_cnt); \
1931 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
1934 static inline void put_net_conf(struct drbd_conf *mdev)
1936 if (atomic_dec_and_test(&mdev->net_cnt))
1937 wake_up(&mdev->misc_wait);
1941 * get_net_conf() - Increase ref count on mdev->net_conf; Returns 0 if nothing there
1942 * @mdev: DRBD device.
1944 * You have to call put_net_conf() when finished working with mdev->net_conf.
1946 static inline int get_net_conf(struct drbd_conf *mdev)
1950 atomic_inc(&mdev->net_cnt);
1951 have_net_conf = mdev->state.conn >= C_UNCONNECTED;
1954 return have_net_conf;
1958 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
1961 * You have to call put_ldev() when finished working with mdev->ldev.
1963 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
1964 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
1966 static inline void put_ldev(struct drbd_conf *mdev)
1969 if (atomic_dec_and_test(&mdev->local_cnt))
1970 wake_up(&mdev->misc_wait);
1971 D_ASSERT(atomic_read(&mdev->local_cnt) >= 0);
1975 static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
1979 atomic_inc(&mdev->local_cnt);
1980 io_allowed = (mdev->state.disk >= mins);
1986 extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins);
1989 /* you must have an "get_ldev" reference */
1990 static inline void drbd_get_syncer_progress(struct drbd_conf *mdev,
1991 unsigned long *bits_left, unsigned int *per_mil_done)
1994 * this is to break it at compile time when we change that
1995 * (we may feel 4TB maximum storage per drbd is not enough)
1997 typecheck(unsigned long, mdev->rs_total);
1999 /* note: both rs_total and rs_left are in bits, i.e. in
2000 * units of BM_BLOCK_SIZE.
2001 * for the percentage, we don't care. */
2003 *bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
2004 /* >> 10 to prevent overflow,
2005 * +1 to prevent division by zero */
2006 if (*bits_left > mdev->rs_total) {
2007 /* doh. maybe a logic bug somewhere.
2008 * may also be just a race condition
2009 * between this and a disconnect during sync.
2010 * for now, just prevent in-kernel buffer overflow.
2013 dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2014 drbd_conn_str(mdev->state.conn),
2015 *bits_left, mdev->rs_total, mdev->rs_failed);
2018 /* make sure the calculation happens in long context */
2019 unsigned long tmp = 1000UL -
2020 (*bits_left >> 10)*1000UL
2021 / ((mdev->rs_total >> 10) + 1UL);
2022 *per_mil_done = tmp;
2027 /* this throttles on-the-fly application requests
2028 * according to max_buffers settings;
2029 * maybe re-implement using semaphores? */
2030 static inline int drbd_get_max_buffers(struct drbd_conf *mdev)
2032 int mxb = 1000000; /* arbitrary limit on open requests */
2033 if (get_net_conf(mdev)) {
2034 mxb = mdev->net_conf->max_buffers;
2040 static inline int drbd_state_is_stable(union drbd_state s)
2043 /* DO NOT add a default clause, we want the compiler to warn us
2044 * for any newly introduced state we may have forgotten to add here */
2046 switch ((enum drbd_conns)s.conn) {
2047 /* new io only accepted when there is no connection, ... */
2049 case C_WF_CONNECTION:
2050 /* ... or there is a well established connection. */
2056 case C_PAUSED_SYNC_S:
2057 case C_PAUSED_SYNC_T:
2058 /* maybe stable, look at the disk state */
2061 /* no new io accepted during tansitional states
2062 * like handshake or teardown */
2063 case C_DISCONNECTING:
2067 case C_NETWORK_FAILURE:
2068 case C_PROTOCOL_ERROR:
2070 case C_WF_REPORT_PARAMS:
2071 case C_STARTING_SYNC_S:
2072 case C_STARTING_SYNC_T:
2075 case C_WF_SYNC_UUID:
2081 switch ((enum drbd_disk_state)s.disk) {
2083 case D_INCONSISTENT:
2087 /* disk state is stable as well. */
2090 /* no new io accepted during tansitional states */
2103 static inline int __inc_ap_bio_cond(struct drbd_conf *mdev)
2105 int mxb = drbd_get_max_buffers(mdev);
2107 if (mdev->state.susp)
2109 if (test_bit(SUSPEND_IO, &mdev->flags))
2112 /* to avoid potential deadlock or bitmap corruption,
2113 * in various places, we only allow new application io
2114 * to start during "stable" states. */
2116 /* no new io accepted when attaching or detaching the disk */
2117 if (!drbd_state_is_stable(mdev->state))
2120 /* since some older kernels don't have atomic_add_unless,
2121 * and we are within the spinlock anyways, we have this workaround. */
2122 if (atomic_read(&mdev->ap_bio_cnt) > mxb)
2124 if (test_bit(BITMAP_IO, &mdev->flags))
2129 /* I'd like to use wait_event_lock_irq,
2130 * but I'm not sure when it got introduced,
2131 * and not sure when it has 3 or 4 arguments */
2132 static inline void inc_ap_bio(struct drbd_conf *mdev, int one_or_two)
2134 /* compare with after_state_ch,
2135 * os.conn != C_WF_BITMAP_S && ns.conn == C_WF_BITMAP_S */
2139 * as long as the device is suspended
2140 * until the bitmap is no longer on the fly during connection
2141 * handshake as long as we would exeed the max_buffer limit.
2143 * to avoid races with the reconnect code,
2144 * we need to atomic_inc within the spinlock. */
2146 spin_lock_irq(&mdev->req_lock);
2147 while (!__inc_ap_bio_cond(mdev)) {
2148 prepare_to_wait(&mdev->misc_wait, &wait, TASK_UNINTERRUPTIBLE);
2149 spin_unlock_irq(&mdev->req_lock);
2151 finish_wait(&mdev->misc_wait, &wait);
2152 spin_lock_irq(&mdev->req_lock);
2154 atomic_add(one_or_two, &mdev->ap_bio_cnt);
2155 spin_unlock_irq(&mdev->req_lock);
2158 static inline void dec_ap_bio(struct drbd_conf *mdev)
2160 int mxb = drbd_get_max_buffers(mdev);
2161 int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt);
2163 D_ASSERT(ap_bio >= 0);
2164 /* this currently does wake_up for every dec_ap_bio!
2165 * maybe rather introduce some type of hysteresis?
2166 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2168 wake_up(&mdev->misc_wait);
2169 if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) {
2170 if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags))
2171 drbd_queue_work(&mdev->data.work, &mdev->bm_io_work.w);
2175 static inline void drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val)
2177 mdev->ed_uuid = val;
2180 static inline int seq_cmp(u32 a, u32 b)
2182 /* we assume wrap around at 32bit.
2183 * for wrap around at 24bit (old atomic_t),
2187 return (s32)(a) - (s32)(b);
2189 #define seq_lt(a, b) (seq_cmp((a), (b)) < 0)
2190 #define seq_gt(a, b) (seq_cmp((a), (b)) > 0)
2191 #define seq_ge(a, b) (seq_cmp((a), (b)) >= 0)
2192 #define seq_le(a, b) (seq_cmp((a), (b)) <= 0)
2193 /* CAUTION: please no side effects in arguments! */
2194 #define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b)))
2196 static inline void update_peer_seq(struct drbd_conf *mdev, unsigned int new_seq)
2199 spin_lock(&mdev->peer_seq_lock);
2200 m = seq_max(mdev->peer_seq, new_seq);
2202 spin_unlock(&mdev->peer_seq_lock);
2204 wake_up(&mdev->seq_wait);
2207 static inline void drbd_update_congested(struct drbd_conf *mdev)
2209 struct sock *sk = mdev->data.socket->sk;
2210 if (sk->sk_wmem_queued > sk->sk_sndbuf * 4 / 5)
2211 set_bit(NET_CONGESTED, &mdev->flags);
2214 static inline int drbd_queue_order_type(struct drbd_conf *mdev)
2216 /* sorry, we currently have no working implementation
2217 * of distributed TCQ stuff */
2218 #ifndef QUEUE_ORDERED_NONE
2219 #define QUEUE_ORDERED_NONE 0
2221 return QUEUE_ORDERED_NONE;
2224 static inline void drbd_blk_run_queue(struct request_queue *q)
2226 if (q && q->unplug_fn)
2230 static inline void drbd_kick_lo(struct drbd_conf *mdev)
2232 if (get_ldev(mdev)) {
2233 drbd_blk_run_queue(bdev_get_queue(mdev->ldev->backing_bdev));
2238 static inline void drbd_md_flush(struct drbd_conf *mdev)
2242 if (test_bit(MD_NO_BARRIER, &mdev->flags))
2245 r = blkdev_issue_flush(mdev->ldev->md_bdev, NULL);
2247 set_bit(MD_NO_BARRIER, &mdev->flags);
2248 dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r);