4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
6 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
7 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
8 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
10 drbd is free software; you can redistribute it and/or modify
11 it under the terms of the GNU General Public License as published by
12 the Free Software Foundation; either version 2, or (at your option)
15 drbd is distributed in the hope that it will be useful,
16 but WITHOUT ANY WARRANTY; without even the implied warranty of
17 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 GNU General Public License for more details.
20 You should have received a copy of the GNU General Public License
21 along with drbd; see the file COPYING. If not, write to
22 the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
29 #include <linux/compiler.h>
30 #include <linux/types.h>
31 #include <linux/version.h>
32 #include <linux/list.h>
33 #include <linux/sched.h>
34 #include <linux/bitops.h>
35 #include <linux/slab.h>
36 #include <linux/crypto.h>
37 #include <linux/ratelimit.h>
38 #include <linux/tcp.h>
39 #include <linux/mutex.h>
40 #include <linux/major.h>
41 #include <linux/blkdev.h>
42 #include <linux/genhd.h>
44 #include <linux/lru_cache.h>
47 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
48 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
49 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
50 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
52 # define __protected_by(x)
53 # define __protected_read_by(x)
54 # define __protected_write_by(x)
55 # define __must_hold(x)
58 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
60 /* module parameter, defined in drbd_main.c */
61 extern unsigned int minor_count;
62 extern int disable_sendpage;
64 extern unsigned int cn_idx;
66 #ifdef CONFIG_DRBD_FAULT_INJECTION
67 extern int enable_faults;
68 extern int fault_rate;
69 extern int fault_devs;
72 extern char usermode_helper[];
82 /* I don't remember why XCPU ...
83 * This is used to wake the asender,
84 * and to interrupt sending the sending task
87 #define DRBD_SIG SIGXCPU
89 /* This is used to stop/restart our threads.
90 * Cannot use SIGTERM nor SIGKILL, since these
91 * are sent out by init on runlevel changes
92 * I choose SIGHUP for now.
94 #define DRBD_SIGKILL SIGHUP
96 /* All EEs on the free list should have ID_VACANT (== 0)
97 * freshly allocated EEs get !ID_VACANT (== 1)
98 * so if it says "cannot dereference null pointer at address 0x00000001",
99 * it is most likely one of these :( */
101 #define ID_IN_SYNC (4711ULL)
102 #define ID_OUT_OF_SYNC (4712ULL)
104 #define ID_SYNCER (-1ULL)
106 #define is_syncer_block_id(id) ((id) == ID_SYNCER)
111 /* to shorten dev_warn(DEV, "msg"); and relatives statements */
112 #define DEV (disk_to_dev(mdev->vdisk))
114 #define D_ASSERT(exp) if (!(exp)) \
115 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
117 #define ERR_IF(exp) if (({ \
118 int _b = (exp) != 0; \
119 if (_b) dev_err(DEV, "%s: (%s) in %s:%d\n", \
120 __func__, #exp, __FILE__, __LINE__); \
124 /* Defines to control fault insertion */
126 DRBD_FAULT_MD_WR = 0, /* meta data write */
127 DRBD_FAULT_MD_RD = 1, /* read */
128 DRBD_FAULT_RS_WR = 2, /* resync */
129 DRBD_FAULT_RS_RD = 3,
130 DRBD_FAULT_DT_WR = 4, /* data */
131 DRBD_FAULT_DT_RD = 5,
132 DRBD_FAULT_DT_RA = 6, /* data read ahead */
133 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */
134 DRBD_FAULT_AL_EE = 8, /* alloc ee */
135 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
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 */
212 /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */
213 /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */
214 P_DELAY_PROBE = 0x27, /* is used on BOTH sockets */
217 P_MAY_IGNORE = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
218 P_MAX_OPT_CMD = 0x101,
220 /* special command ids for handshake */
222 P_HAND_SHAKE_M = 0xfff1, /* First Packet on the MetaSock */
223 P_HAND_SHAKE_S = 0xfff2, /* First Packet on the Socket */
225 P_HAND_SHAKE = 0xfffe /* FIXED for the next century! */
228 static inline const char *cmdname(enum drbd_packets cmd)
230 /* THINK may need to become several global tables
231 * when we want to support more than
233 static const char *cmdnames[] = {
235 [P_DATA_REPLY] = "DataReply",
236 [P_RS_DATA_REPLY] = "RSDataReply",
237 [P_BARRIER] = "Barrier",
238 [P_BITMAP] = "ReportBitMap",
239 [P_BECOME_SYNC_TARGET] = "BecomeSyncTarget",
240 [P_BECOME_SYNC_SOURCE] = "BecomeSyncSource",
241 [P_UNPLUG_REMOTE] = "UnplugRemote",
242 [P_DATA_REQUEST] = "DataRequest",
243 [P_RS_DATA_REQUEST] = "RSDataRequest",
244 [P_SYNC_PARAM] = "SyncParam",
245 [P_SYNC_PARAM89] = "SyncParam89",
246 [P_PROTOCOL] = "ReportProtocol",
247 [P_UUIDS] = "ReportUUIDs",
248 [P_SIZES] = "ReportSizes",
249 [P_STATE] = "ReportState",
250 [P_SYNC_UUID] = "ReportSyncUUID",
251 [P_AUTH_CHALLENGE] = "AuthChallenge",
252 [P_AUTH_RESPONSE] = "AuthResponse",
254 [P_PING_ACK] = "PingAck",
255 [P_RECV_ACK] = "RecvAck",
256 [P_WRITE_ACK] = "WriteAck",
257 [P_RS_WRITE_ACK] = "RSWriteAck",
258 [P_DISCARD_ACK] = "DiscardAck",
259 [P_NEG_ACK] = "NegAck",
260 [P_NEG_DREPLY] = "NegDReply",
261 [P_NEG_RS_DREPLY] = "NegRSDReply",
262 [P_BARRIER_ACK] = "BarrierAck",
263 [P_STATE_CHG_REQ] = "StateChgRequest",
264 [P_STATE_CHG_REPLY] = "StateChgReply",
265 [P_OV_REQUEST] = "OVRequest",
266 [P_OV_REPLY] = "OVReply",
267 [P_OV_RESULT] = "OVResult",
268 [P_CSUM_RS_REQUEST] = "CsumRSRequest",
269 [P_RS_IS_IN_SYNC] = "CsumRSIsInSync",
270 [P_COMPRESSED_BITMAP] = "CBitmap",
271 [P_DELAY_PROBE] = "DelayProbe",
275 if (cmd == P_HAND_SHAKE_M)
277 if (cmd == P_HAND_SHAKE_S)
279 if (cmd == P_HAND_SHAKE)
281 if (cmd >= P_MAX_CMD)
283 return cmdnames[cmd];
286 /* for sending/receiving the bitmap,
287 * possibly in some encoding scheme */
290 * stores total bits and long words
291 * of the bitmap, so we don't need to
292 * call the accessor functions over and again. */
293 unsigned long bm_bits;
294 unsigned long bm_words;
295 /* during xfer, current position within the bitmap */
296 unsigned long bit_offset;
297 unsigned long word_offset;
299 /* statistics; index: (h->command == P_BITMAP) */
304 extern void INFO_bm_xfer_stats(struct drbd_conf *mdev,
305 const char *direction, struct bm_xfer_ctx *c);
307 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
309 /* word_offset counts "native long words" (32 or 64 bit),
311 * Encoded packet may end at an unaligned bit offset.
312 * In case a fallback clear text packet is transmitted in
313 * between, we adjust this offset back to the last 64bit
314 * aligned "native long word", which makes coding and decoding
315 * the plain text bitmap much more convenient. */
316 #if BITS_PER_LONG == 64
317 c->word_offset = c->bit_offset >> 6;
318 #elif BITS_PER_LONG == 32
319 c->word_offset = c->bit_offset >> 5;
320 c->word_offset &= ~(1UL);
322 # error "unsupported BITS_PER_LONG"
327 #define __packed __attribute__((packed))
330 /* This is the layout for a packet on the wire.
331 * The byteorder is the network byte order.
332 * (except block_id and barrier fields.
333 * these are pointers to local structs
334 * and have no relevance for the partner,
335 * which just echoes them as received.)
337 * NOTE that the payload starts at a long aligned offset,
338 * regardless of 32 or 64 bit arch!
343 u16 length; /* bytes of data after this header */
346 /* 8 bytes. packet FIXED for the next century! */
349 * short commands, packets without payload, plain p_header:
352 * P_BECOME_SYNC_TARGET
353 * P_BECOME_SYNC_SOURCE
358 * commands with out-of-struct payload:
359 * P_BITMAP (no additional fields)
360 * P_DATA, P_DATA_REPLY (see p_data)
361 * P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
364 /* these defines must not be changed without changing the protocol version */
365 #define DP_HARDBARRIER 1
367 #define DP_MAY_SET_IN_SYNC 4
370 struct p_header head;
371 u64 sector; /* 64 bits sector number */
372 u64 block_id; /* to identify the request in protocol B&C */
378 * commands which share a struct:
380 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
381 * P_DISCARD_ACK (proto C, two-primaries conflict detection)
383 * P_DATA_REQUEST, P_RS_DATA_REQUEST
386 struct p_header head;
395 struct p_header head;
399 u32 pad; /* to multiple of 8 Byte */
403 * commands with their own struct for additional fields:
412 struct p_header head; /* 8 bytes */
417 /* should be more than enough for future enhancements
418 * for now, feature_flags and the reserverd array shall be zero.
424 /* 80 bytes, FIXED for the next century */
427 struct p_header head;
428 u32 barrier; /* barrier number _handle_ only */
429 u32 pad; /* to multiple of 8 Byte */
432 struct p_barrier_ack {
433 struct p_header head;
439 struct p_header head;
442 /* Since protocol version 88 and higher. */
446 struct p_rs_param_89 {
447 struct p_header head;
449 /* protocol version 89: */
450 char verify_alg[SHARED_SECRET_MAX];
451 char csums_alg[SHARED_SECRET_MAX];
454 enum drbd_conn_flags {
460 struct p_header head;
468 /* Since protocol version 87 and higher. */
469 char integrity_alg[0];
474 struct p_header head;
475 u64 uuid[UI_EXTENDED_SIZE];
479 struct p_header head;
484 struct p_header head;
485 u64 d_size; /* size of disk */
486 u64 u_size; /* user requested size */
487 u64 c_size; /* current exported size */
488 u32 max_segment_size; /* Maximal size of a BIO */
489 u16 queue_order_type; /* not yet implemented in DRBD*/
490 u16 dds_flags; /* use enum dds_flags here. */
494 struct p_header head;
499 struct p_header head;
504 struct p_req_state_reply {
505 struct p_header head;
509 struct p_drbd06_param {
520 struct p_header head;
526 /* Valid values for the encoding field.
527 * Bump proto version when changing this. */
528 enum drbd_bitmap_code {
529 /* RLE_VLI_Bytes = 0,
530 * and other bit variants had been defined during
531 * algorithm evaluation. */
535 struct p_compressed_bm {
536 struct p_header head;
537 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
538 * (encoding & 0x80): polarity (set/unset) of first runlength
539 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
540 * used to pad up to head.length bytes
547 struct p_delay_probe {
548 struct p_header head;
549 u32 seq_num; /* sequence number to match the two probe packets */
550 u32 offset; /* usecs the probe got sent after the reference time point */
554 struct list_head list;
559 /* DCBP: Drbd Compressed Bitmap Packet ... */
560 static inline enum drbd_bitmap_code
561 DCBP_get_code(struct p_compressed_bm *p)
563 return (enum drbd_bitmap_code)(p->encoding & 0x0f);
567 DCBP_set_code(struct p_compressed_bm *p, enum drbd_bitmap_code code)
570 p->encoding = (p->encoding & ~0xf) | code;
574 DCBP_get_start(struct p_compressed_bm *p)
576 return (p->encoding & 0x80) != 0;
580 DCBP_set_start(struct p_compressed_bm *p, int set)
582 p->encoding = (p->encoding & ~0x80) | (set ? 0x80 : 0);
586 DCBP_get_pad_bits(struct p_compressed_bm *p)
588 return (p->encoding >> 4) & 0x7;
592 DCBP_set_pad_bits(struct p_compressed_bm *p, int n)
595 p->encoding = (p->encoding & (~0x7 << 4)) | (n << 4);
598 /* one bitmap packet, including the p_header,
599 * should fit within one _architecture independend_ page.
600 * so we need to use the fixed size 4KiB page size
601 * most architechtures have used for a long time.
603 #define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header))
604 #define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
605 #define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm))
606 #if (PAGE_SIZE < 4096)
607 /* drbd_send_bitmap / receive_bitmap would break horribly */
608 #error "PAGE_SIZE too small"
612 struct p_header header;
613 struct p_handshake handshake;
615 struct p_block_ack block_ack;
616 struct p_barrier barrier;
617 struct p_barrier_ack barrier_ack;
618 struct p_rs_param_89 rs_param_89;
619 struct p_protocol protocol;
620 struct p_sizes sizes;
621 struct p_uuids uuids;
622 struct p_state state;
623 struct p_req_state req_state;
624 struct p_req_state_reply req_state_reply;
625 struct p_block_req block_req;
628 /**********************************************************************/
629 enum drbd_thread_state {
638 struct task_struct *task;
639 struct completion stop;
640 enum drbd_thread_state t_state;
641 int (*function) (struct drbd_thread *);
642 struct drbd_conf *mdev;
646 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
648 /* THINK testing the t_state seems to be uncritical in all cases
649 * (but thread_{start,stop}), so we can read it *without* the lock.
658 * Having this as the first member of a struct provides sort of "inheritance".
659 * "derived" structs can be "drbd_queue_work()"ed.
660 * The callback should know and cast back to the descendant struct.
661 * drbd_request and drbd_epoch_entry are descendants of drbd_work.
664 typedef int (*drbd_work_cb)(struct drbd_conf *, struct drbd_work *, int cancel);
666 struct list_head list;
670 struct drbd_tl_epoch;
671 struct drbd_request {
673 struct drbd_conf *mdev;
675 /* if local IO is not allowed, will be NULL.
676 * if local IO _is_ allowed, holds the locally submitted bio clone,
677 * or, after local IO completion, the ERR_PTR(error).
678 * see drbd_endio_pri(). */
679 struct bio *private_bio;
681 struct hlist_node colision;
684 unsigned int epoch; /* barrier_nr */
686 /* barrier_nr: used to check on "completion" whether this req was in
687 * the current epoch, and we therefore have to close it,
688 * starting a new epoch...
691 /* up to here, the struct layout is identical to drbd_epoch_entry;
692 * we might be able to use that to our advantage... */
694 struct list_head tl_requests; /* ring list in the transfer log */
695 struct bio *master_bio; /* master bio pointer */
696 unsigned long rq_state; /* see comments above _req_mod() */
698 unsigned long start_time;
701 struct drbd_tl_epoch {
703 struct list_head requests; /* requests before */
704 struct drbd_tl_epoch *next; /* pointer to the next barrier */
705 unsigned int br_number; /* the barriers identifier. */
706 int n_req; /* number of requests attached before this barrier */
711 /* These Tl_epoch_entries may be in one of 6 lists:
712 active_ee .. data packet being written
713 sync_ee .. syncer block being written
714 done_ee .. block written, need to send P_WRITE_ACK
715 read_ee .. [RS]P_DATA_REQUEST being read
719 struct list_head list;
720 unsigned int barrier_nr;
721 atomic_t epoch_size; /* increased on every request added. */
722 atomic_t active; /* increased on every req. added, and dec on every finished. */
726 /* drbd_epoch flag bits */
728 DE_BARRIER_IN_NEXT_EPOCH_ISSUED,
729 DE_BARRIER_IN_NEXT_EPOCH_DONE,
730 DE_CONTAINS_A_BARRIER,
731 DE_HAVE_BARRIER_NUMBER,
740 EV_CLEANUP = 32, /* used as flag */
743 struct drbd_epoch_entry {
745 struct drbd_conf *mdev;
746 struct bio *private_bio;
747 struct hlist_node colision;
750 struct drbd_epoch *epoch;
752 /* up to here, the struct layout is identical to drbd_request;
753 * we might be able to use that to our advantage... */
759 struct drbd_wq_barrier {
761 struct completion done;
771 __EE_CALL_AL_COMPLETE_IO,
772 __EE_CONFLICT_PENDING,
773 __EE_MAY_SET_IN_SYNC,
776 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
777 #define EE_CONFLICT_PENDING (1<<__EE_CONFLICT_PENDING)
778 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
779 #define EE_IS_BARRIER (1<<__EE_IS_BARRIER)
781 /* global flag bits */
783 CREATE_BARRIER, /* next P_DATA is preceeded by a P_BARRIER */
784 SIGNAL_ASENDER, /* whether asender wants to be interrupted */
785 SEND_PING, /* whether asender should send a ping asap */
787 STOP_SYNC_TIMER, /* tell timer to cancel itself */
788 UNPLUG_QUEUED, /* only relevant with kernel 2.4 */
789 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
790 MD_DIRTY, /* current uuids and flags not yet on disk */
791 DISCARD_CONCURRENT, /* Set on one node, cleared on the peer! */
792 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
793 CLUSTER_ST_CHANGE, /* Cluster wide state change going on... */
796 CRASHED_PRIMARY, /* This node was a crashed primary.
797 * Gets cleared when the state.conn
798 * goes into C_CONNECTED state. */
799 WRITE_BM_AFTER_RESYNC, /* A kmalloc() during resync failed */
800 NO_BARRIER_SUPP, /* underlying block device doesn't implement barriers */
803 MD_NO_BARRIER, /* meta data device does not support barriers,
805 SUSPEND_IO, /* suspend application io */
806 BITMAP_IO, /* suspend application io;
807 once no more io in flight, start bitmap io */
808 BITMAP_IO_QUEUED, /* Started bitmap IO */
809 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
810 NET_CONGESTED, /* The data socket is congested */
812 CONFIG_PENDING, /* serialization of (re)configuration requests.
813 * if set, also prevents the device from dying */
814 DEVICE_DYING, /* device became unconfigured,
815 * but worker thread is still handling the cleanup.
816 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
817 * while this is set. */
818 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
819 * the peer, if it changed there as well. */
820 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
821 GOT_PING_ACK, /* set when we receive a ping_ack packet, misc wait gets woken */
824 struct drbd_bitmap; /* opaque for drbd_conf */
826 /* TODO sort members for performance
827 * MAYBE group them further */
829 /* THINK maybe we actually want to use the default "event/%s" worker threads
830 * or similar in linux 2.6, which uses per cpu data and threads.
832 * To be general, this might need a spin_lock member.
833 * For now, please use the mdev->req_lock to protect list_head,
834 * see drbd_queue_work below.
836 struct drbd_work_queue {
838 struct semaphore s; /* producers up it, worker down()s it */
839 spinlock_t q_lock; /* to protect the list. */
843 struct drbd_work_queue work;
845 struct socket *socket;
846 /* this way we get our
847 * send/receive buffers off the stack */
848 union p_polymorph sbuf;
849 union p_polymorph rbuf;
853 u64 md_offset; /* sector offset to 'super' block */
855 u64 la_size_sect; /* last agreed size, unit sectors */
861 s32 al_offset; /* signed relative sector offset to al area */
862 s32 bm_offset; /* signed relative sector offset to bitmap */
864 /* u32 al_nr_extents; important for restoring the AL
865 * is stored into sync_conf.al_extents, which in turn
866 * gets applied to act_log->nr_elements
870 /* for sync_conf and other types... */
871 #define NL_PACKET(name, number, fields) struct name { fields };
872 #define NL_INTEGER(pn,pr,member) int member;
873 #define NL_INT64(pn,pr,member) __u64 member;
874 #define NL_BIT(pn,pr,member) unsigned member:1;
875 #define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
876 #include "linux/drbd_nl.h"
878 struct drbd_backing_dev {
879 struct block_device *backing_bdev;
880 struct block_device *md_bdev;
881 struct file *lo_file;
882 struct file *md_file;
884 struct disk_conf dc; /* The user provided config... */
885 sector_t known_size; /* last known size of that backing device */
889 struct drbd_conf *mdev;
890 struct completion event;
897 int (*io_fn)(struct drbd_conf *mdev);
898 void (*done)(struct drbd_conf *mdev, int rv);
901 enum write_ordering_e {
909 /* things that are stored as / read from meta data on disk */
912 /* configured by drbdsetup */
913 struct net_conf *net_conf; /* protected by get_net_conf() and put_net_conf() */
914 struct syncer_conf sync_conf;
915 struct drbd_backing_dev *ldev __protected_by(local);
917 sector_t p_size; /* partner's disk size */
918 struct request_queue *rq_queue;
919 struct block_device *this_bdev;
920 struct gendisk *vdisk;
922 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
923 struct drbd_socket meta; /* ping/ack (metadata) packets */
924 int agreed_pro_version; /* actually used protocol version */
925 unsigned long last_received; /* in jiffies, either socket */
926 unsigned int ko_count;
927 struct drbd_work resync_work,
931 struct timer_list resync_timer;
932 struct timer_list md_sync_timer;
933 struct timer_list delay_probe_timer;
935 /* Used after attach while negotiating new disk state. */
936 union drbd_state new_state_tmp;
938 union drbd_state state;
939 wait_queue_head_t misc_wait;
940 wait_queue_head_t state_wait; /* upon each state change. */
941 unsigned int send_cnt;
942 unsigned int recv_cnt;
943 unsigned int read_cnt;
944 unsigned int writ_cnt;
945 unsigned int al_writ_cnt;
946 unsigned int bm_writ_cnt;
947 atomic_t ap_bio_cnt; /* Requests we need to complete */
948 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
949 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
950 atomic_t unacked_cnt; /* Need to send replys for */
951 atomic_t local_cnt; /* Waiting for local completion */
952 atomic_t net_cnt; /* Users of net_conf */
954 struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */
955 struct drbd_tl_epoch *newest_tle;
956 struct drbd_tl_epoch *oldest_tle;
957 struct list_head out_of_sequence_requests;
958 struct hlist_head *tl_hash;
959 unsigned int tl_hash_s;
961 /* blocks to sync in this run [unit BM_BLOCK_SIZE] */
962 unsigned long rs_total;
963 /* number of sync IOs that failed in this run */
964 unsigned long rs_failed;
965 /* Syncer's start time [unit jiffies] */
966 unsigned long rs_start;
967 /* cumulated time in PausedSyncX state [unit jiffies] */
968 unsigned long rs_paused;
969 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
970 unsigned long rs_mark_left;
971 /* marks's time [unit jiffies] */
972 unsigned long rs_mark_time;
973 /* skipped because csum was equeal [unit BM_BLOCK_SIZE] */
974 unsigned long rs_same_csum;
976 /* where does the admin want us to start? (sector) */
977 sector_t ov_start_sector;
978 /* where are we now? (sector) */
979 sector_t ov_position;
980 /* Start sector of out of sync range (to merge printk reporting). */
981 sector_t ov_last_oos_start;
982 /* size of out-of-sync range in sectors. */
983 sector_t ov_last_oos_size;
984 unsigned long ov_left; /* in bits */
985 struct crypto_hash *csums_tfm;
986 struct crypto_hash *verify_tfm;
988 struct drbd_thread receiver;
989 struct drbd_thread worker;
990 struct drbd_thread asender;
991 struct drbd_bitmap *bitmap;
992 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
994 /* Used to track operations of resync... */
995 struct lru_cache *resync;
996 /* Number of locked elements in resync LRU */
997 unsigned int resync_locked;
998 /* resync extent number waiting for application requests */
999 unsigned int resync_wenr;
1003 struct drbd_epoch *current_epoch;
1004 spinlock_t epoch_lock;
1005 unsigned int epochs;
1006 enum write_ordering_e write_ordering;
1007 struct list_head active_ee; /* IO in progress */
1008 struct list_head sync_ee; /* IO in progress */
1009 struct list_head done_ee; /* send ack */
1010 struct list_head read_ee; /* IO in progress */
1011 struct list_head net_ee; /* zero-copy network send in progress */
1012 struct hlist_head *ee_hash; /* is proteced by req_lock! */
1013 unsigned int ee_hash_s;
1015 /* this one is protected by ee_lock, single thread */
1016 struct drbd_epoch_entry *last_write_w_barrier;
1018 int next_barrier_nr;
1019 struct hlist_head *app_reads_hash; /* is proteced by req_lock */
1020 struct list_head resync_reads;
1022 wait_queue_head_t ee_wait;
1023 struct page *md_io_page; /* one page buffer for md_io */
1024 struct page *md_io_tmpp; /* for logical_block_size != 512 */
1025 struct mutex md_io_mutex; /* protects the md_io_buffer */
1027 wait_queue_head_t al_wait;
1028 struct lru_cache *act_log; /* activity log */
1029 unsigned int al_tr_number;
1031 int al_tr_pos; /* position of the next transaction in the journal */
1032 struct crypto_hash *cram_hmac_tfm;
1033 struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */
1034 struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */
1038 wait_queue_head_t seq_wait;
1039 atomic_t packet_seq;
1040 unsigned int peer_seq;
1041 spinlock_t peer_seq_lock;
1043 unsigned long comm_bm_set; /* communicated number of set bits. */
1044 cpumask_var_t cpu_mask;
1045 struct bm_io_work bm_io_work;
1046 u64 ed_uuid; /* UUID of the exposed data */
1047 struct mutex state_mutex;
1048 char congestion_reason; /* Why we where congested... */
1049 struct list_head delay_probes; /* protected by peer_seq_lock */
1050 int data_delay; /* Delay of packets on the data-sock behind meta-sock */
1051 atomic_t delay_seq; /* To generate sequence numbers of delay probes */
1052 struct timeval dps_time; /* delay-probes-start-time */
1053 int dp_volume_last; /* send_cnt of last delay probe */
1054 int c_sync_rate; /* current resync rate after delay_probe magic */
1057 static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
1059 struct drbd_conf *mdev;
1061 mdev = minor < minor_count ? minor_table[minor] : NULL;
1066 static inline unsigned int mdev_to_minor(struct drbd_conf *mdev)
1071 /* returns 1 if it was successfull,
1072 * returns 0 if there was no data socket.
1073 * so wherever you are going to use the data.socket, e.g. do
1074 * if (!drbd_get_data_sock(mdev))
1077 * drbd_put_data_sock(mdev);
1079 static inline int drbd_get_data_sock(struct drbd_conf *mdev)
1081 mutex_lock(&mdev->data.mutex);
1082 /* drbd_disconnect() could have called drbd_free_sock()
1083 * while we were waiting in down()... */
1084 if (unlikely(mdev->data.socket == NULL)) {
1085 mutex_unlock(&mdev->data.mutex);
1091 static inline void drbd_put_data_sock(struct drbd_conf *mdev)
1093 mutex_unlock(&mdev->data.mutex);
1097 * function declarations
1098 *************************/
1102 enum chg_state_flags {
1105 CS_WAIT_COMPLETE = 4,
1107 CS_ORDERED = CS_WAIT_COMPLETE + CS_SERIALIZE,
1112 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1115 extern void drbd_init_set_defaults(struct drbd_conf *mdev);
1116 extern int drbd_change_state(struct drbd_conf *mdev, enum chg_state_flags f,
1117 union drbd_state mask, union drbd_state val);
1118 extern void drbd_force_state(struct drbd_conf *, union drbd_state,
1120 extern int _drbd_request_state(struct drbd_conf *, union drbd_state,
1121 union drbd_state, enum chg_state_flags);
1122 extern int __drbd_set_state(struct drbd_conf *, union drbd_state,
1123 enum chg_state_flags, struct completion *done);
1124 extern void print_st_err(struct drbd_conf *, union drbd_state,
1125 union drbd_state, int);
1126 extern int drbd_thread_start(struct drbd_thread *thi);
1127 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1129 extern void drbd_thread_current_set_cpu(struct drbd_conf *mdev);
1130 extern void drbd_calc_cpu_mask(struct drbd_conf *mdev);
1132 #define drbd_thread_current_set_cpu(A) ({})
1133 #define drbd_calc_cpu_mask(A) ({})
1135 extern void drbd_free_resources(struct drbd_conf *mdev);
1136 extern void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr,
1137 unsigned int set_size);
1138 extern void tl_clear(struct drbd_conf *mdev);
1139 extern void _tl_add_barrier(struct drbd_conf *, struct drbd_tl_epoch *);
1140 extern void drbd_free_sock(struct drbd_conf *mdev);
1141 extern int drbd_send(struct drbd_conf *mdev, struct socket *sock,
1142 void *buf, size_t size, unsigned msg_flags);
1143 extern int drbd_send_protocol(struct drbd_conf *mdev);
1144 extern int drbd_send_uuids(struct drbd_conf *mdev);
1145 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
1146 extern int drbd_send_sync_uuid(struct drbd_conf *mdev, u64 val);
1147 extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags);
1148 extern int _drbd_send_state(struct drbd_conf *mdev);
1149 extern int drbd_send_state(struct drbd_conf *mdev);
1150 extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock,
1151 enum drbd_packets cmd, struct p_header *h,
1152 size_t size, unsigned msg_flags);
1153 #define USE_DATA_SOCKET 1
1154 #define USE_META_SOCKET 0
1155 extern int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket,
1156 enum drbd_packets cmd, struct p_header *h,
1158 extern int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd,
1159 char *data, size_t size);
1160 extern int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc);
1161 extern int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr,
1163 extern int drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd,
1164 struct drbd_epoch_entry *e);
1165 extern int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packets cmd,
1166 struct p_block_req *rp);
1167 extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd,
1169 extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd,
1170 sector_t sector, int blksize, u64 block_id);
1171 extern int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd,
1172 struct drbd_epoch_entry *e);
1173 extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
1174 extern int _drbd_send_barrier(struct drbd_conf *mdev,
1175 struct drbd_tl_epoch *barrier);
1176 extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1177 sector_t sector, int size, u64 block_id);
1178 extern int drbd_send_drequest_csum(struct drbd_conf *mdev,
1179 sector_t sector,int size,
1180 void *digest, int digest_size,
1181 enum drbd_packets cmd);
1182 extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size);
1184 extern int drbd_send_bitmap(struct drbd_conf *mdev);
1185 extern int _drbd_send_bitmap(struct drbd_conf *mdev);
1186 extern int drbd_send_sr_reply(struct drbd_conf *mdev, int retcode);
1187 extern void drbd_free_bc(struct drbd_backing_dev *ldev);
1188 extern void drbd_mdev_cleanup(struct drbd_conf *mdev);
1190 /* drbd_meta-data.c (still in drbd_main.c) */
1191 extern void drbd_md_sync(struct drbd_conf *mdev);
1192 extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
1193 /* maybe define them below as inline? */
1194 extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1195 extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1196 extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1197 extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1198 extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
1199 extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
1200 extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
1201 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1202 extern void drbd_md_mark_dirty(struct drbd_conf *mdev);
1203 extern void drbd_queue_bitmap_io(struct drbd_conf *mdev,
1204 int (*io_fn)(struct drbd_conf *),
1205 void (*done)(struct drbd_conf *, int),
1207 extern int drbd_bmio_set_n_write(struct drbd_conf *mdev);
1208 extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev);
1209 extern int drbd_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), char *why);
1213 We reserve a 128MB Block (4k aligned)
1214 * either at the end of the backing device
1215 * or on a separate meta data device. */
1217 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1218 /* The following numbers are sectors */
1219 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1220 #define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */
1221 /* Allows up to about 3.8TB */
1222 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1224 /* Since the smalles IO unit is usually 512 byte */
1225 #define MD_SECTOR_SHIFT 9
1226 #define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT)
1229 #define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1230 #define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */
1231 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1233 #if BITS_PER_LONG == 32
1235 #define cpu_to_lel(A) cpu_to_le32(A)
1236 #define lel_to_cpu(A) le32_to_cpu(A)
1237 #elif BITS_PER_LONG == 64
1239 #define cpu_to_lel(A) cpu_to_le64(A)
1240 #define lel_to_cpu(A) le64_to_cpu(A)
1242 #error "LN2 of BITS_PER_LONG unknown!"
1246 /* 16MB sized 'bitmap extent' to track syncer usage */
1248 int rs_left; /* number of bits set (out of sync) in this extent. */
1249 int rs_failed; /* number of failed resync requests in this extent. */
1250 unsigned long flags;
1251 struct lc_element lce;
1254 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1255 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1259 * We need to store one bit for a block.
1260 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1261 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1262 * Bit 1 ==> local node thinks this block needs to be synced.
1265 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1266 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1267 /* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1268 * per sector of on disk bitmap */
1269 #define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */
1270 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1272 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1273 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1276 /* thus many _storage_ sectors are described by one bit */
1277 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1278 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1279 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1281 /* bit to represented kilo byte conversion */
1282 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1284 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1285 * _storage_ sector is located in */
1286 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1288 /* how much _storage_ sectors we have per bitmap sector */
1289 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1290 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1292 /* in one sector of the bitmap, we have this many activity_log extents. */
1293 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1294 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1296 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1297 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1299 /* the extent in "PER_EXTENT" below is an activity log extent
1300 * we need that many (long words/bytes) to store the bitmap
1301 * of one AL_EXTENT_SIZE chunk of storage.
1302 * we can store the bitmap for that many AL_EXTENTS within
1303 * one sector of the _on_disk_ bitmap:
1304 * bit 0 bit 37 bit 38 bit (512*8)-1
1305 * ...|........|........|.. // ..|........|
1306 * sect. 0 `296 `304 ^(512*8*8)-1
1308 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1309 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1310 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1313 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1314 #define DRBD_MAX_SECTORS_BM \
1315 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1316 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1317 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1318 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1319 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1320 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1321 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1323 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1324 /* 16 TB in units of sectors */
1325 #if BITS_PER_LONG == 32
1326 /* adjust by one page worth of bitmap,
1327 * so we won't wrap around in drbd_bm_find_next_bit.
1328 * you should use 64bit OS for that much storage, anyways. */
1329 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1331 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0x1LU << 32)
1335 /* Sector shift value for the "hash" functions of tl_hash and ee_hash tables.
1336 * With a value of 6 all IO in one 32K block make it to the same slot of the
1339 #define DRBD_MAX_SEGMENT_SIZE (1U<<(9+HT_SHIFT))
1341 /* Number of elements in the app_reads_hash */
1342 #define APP_R_HSIZE 15
1344 extern int drbd_bm_init(struct drbd_conf *mdev);
1345 extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits);
1346 extern void drbd_bm_cleanup(struct drbd_conf *mdev);
1347 extern void drbd_bm_set_all(struct drbd_conf *mdev);
1348 extern void drbd_bm_clear_all(struct drbd_conf *mdev);
1349 extern int drbd_bm_set_bits(
1350 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1351 extern int drbd_bm_clear_bits(
1352 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1353 /* bm_set_bits variant for use while holding drbd_bm_lock */
1354 extern void _drbd_bm_set_bits(struct drbd_conf *mdev,
1355 const unsigned long s, const unsigned long e);
1356 extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr);
1357 extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr);
1358 extern int drbd_bm_write_sect(struct drbd_conf *mdev, unsigned long enr) __must_hold(local);
1359 extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
1360 extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
1361 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev,
1362 unsigned long al_enr);
1363 extern size_t drbd_bm_words(struct drbd_conf *mdev);
1364 extern unsigned long drbd_bm_bits(struct drbd_conf *mdev);
1365 extern sector_t drbd_bm_capacity(struct drbd_conf *mdev);
1366 extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1367 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1368 extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1369 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo);
1370 extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev);
1371 extern int drbd_bm_rs_done(struct drbd_conf *mdev);
1372 /* for receive_bitmap */
1373 extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset,
1374 size_t number, unsigned long *buffer);
1375 /* for _drbd_send_bitmap and drbd_bm_write_sect */
1376 extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset,
1377 size_t number, unsigned long *buffer);
1379 extern void drbd_bm_lock(struct drbd_conf *mdev, char *why);
1380 extern void drbd_bm_unlock(struct drbd_conf *mdev);
1382 extern int drbd_bm_count_bits(struct drbd_conf *mdev, const unsigned long s, const unsigned long e);
1385 extern struct kmem_cache *drbd_request_cache;
1386 extern struct kmem_cache *drbd_ee_cache; /* epoch entries */
1387 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1388 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1389 extern mempool_t *drbd_request_mempool;
1390 extern mempool_t *drbd_ee_mempool;
1392 extern struct page *drbd_pp_pool; /* drbd's page pool */
1393 extern spinlock_t drbd_pp_lock;
1394 extern int drbd_pp_vacant;
1395 extern wait_queue_head_t drbd_pp_wait;
1397 extern rwlock_t global_state_lock;
1399 extern struct drbd_conf *drbd_new_device(unsigned int minor);
1400 extern void drbd_free_mdev(struct drbd_conf *mdev);
1402 extern int proc_details;
1405 extern int drbd_make_request_26(struct request_queue *q, struct bio *bio);
1406 extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req);
1407 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1408 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1412 extern void drbd_suspend_io(struct drbd_conf *mdev);
1413 extern void drbd_resume_io(struct drbd_conf *mdev);
1414 extern char *ppsize(char *buf, unsigned long long size);
1415 extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int);
1416 enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
1417 extern enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local);
1418 extern void resync_after_online_grow(struct drbd_conf *);
1419 extern void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int) __must_hold(local);
1420 extern int drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role,
1422 enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev);
1423 extern int drbd_khelper(struct drbd_conf *mdev, char *cmd);
1426 extern int drbd_worker(struct drbd_thread *thi);
1427 extern int drbd_alter_sa(struct drbd_conf *mdev, int na);
1428 extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side);
1429 extern void resume_next_sg(struct drbd_conf *mdev);
1430 extern void suspend_other_sg(struct drbd_conf *mdev);
1431 extern int drbd_resync_finished(struct drbd_conf *mdev);
1432 /* maybe rather drbd_main.c ? */
1433 extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
1434 struct drbd_backing_dev *bdev, sector_t sector, int rw);
1435 extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int);
1437 static inline void ov_oos_print(struct drbd_conf *mdev)
1439 if (mdev->ov_last_oos_size) {
1440 dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n",
1441 (unsigned long long)mdev->ov_last_oos_start,
1442 (unsigned long)mdev->ov_last_oos_size);
1444 mdev->ov_last_oos_size=0;
1448 extern void drbd_csum(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
1449 /* worker callbacks */
1450 extern int w_req_cancel_conflict(struct drbd_conf *, struct drbd_work *, int);
1451 extern int w_read_retry_remote(struct drbd_conf *, struct drbd_work *, int);
1452 extern int w_e_end_data_req(struct drbd_conf *, struct drbd_work *, int);
1453 extern int w_e_end_rsdata_req(struct drbd_conf *, struct drbd_work *, int);
1454 extern int w_e_end_csum_rs_req(struct drbd_conf *, struct drbd_work *, int);
1455 extern int w_e_end_ov_reply(struct drbd_conf *, struct drbd_work *, int);
1456 extern int w_e_end_ov_req(struct drbd_conf *, struct drbd_work *, int);
1457 extern int w_ov_finished(struct drbd_conf *, struct drbd_work *, int);
1458 extern int w_resync_inactive(struct drbd_conf *, struct drbd_work *, int);
1459 extern int w_resume_next_sg(struct drbd_conf *, struct drbd_work *, int);
1460 extern int w_io_error(struct drbd_conf *, struct drbd_work *, int);
1461 extern int w_send_write_hint(struct drbd_conf *, struct drbd_work *, int);
1462 extern int w_make_resync_request(struct drbd_conf *, struct drbd_work *, int);
1463 extern int w_send_dblock(struct drbd_conf *, struct drbd_work *, int);
1464 extern int w_send_barrier(struct drbd_conf *, struct drbd_work *, int);
1465 extern int w_send_read_req(struct drbd_conf *, struct drbd_work *, int);
1466 extern int w_prev_work_done(struct drbd_conf *, struct drbd_work *, int);
1467 extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int);
1469 extern void resync_timer_fn(unsigned long data);
1471 /* drbd_receiver.c */
1472 extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list);
1473 extern struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev,
1476 unsigned int data_size,
1477 gfp_t gfp_mask) __must_hold(local);
1478 extern void drbd_free_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e);
1479 extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1480 struct list_head *head);
1481 extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1482 struct list_head *head);
1483 extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled);
1484 extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed);
1485 extern void drbd_flush_workqueue(struct drbd_conf *mdev);
1487 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1488 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1489 static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1490 char __user *optval, int optlen)
1493 if (level == SOL_SOCKET)
1494 err = sock_setsockopt(sock, level, optname, optval, optlen);
1496 err = sock->ops->setsockopt(sock, level, optname, optval,
1501 static inline void drbd_tcp_cork(struct socket *sock)
1504 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1505 (char __user *)&val, sizeof(val));
1508 static inline void drbd_tcp_uncork(struct socket *sock)
1511 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1512 (char __user *)&val, sizeof(val));
1515 static inline void drbd_tcp_nodelay(struct socket *sock)
1518 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1519 (char __user *)&val, sizeof(val));
1522 static inline void drbd_tcp_quickack(struct socket *sock)
1525 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1526 (char __user *)&val, sizeof(val));
1529 void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo);
1532 extern struct proc_dir_entry *drbd_proc;
1533 extern const struct file_operations drbd_proc_fops;
1534 extern const char *drbd_conn_str(enum drbd_conns s);
1535 extern const char *drbd_role_str(enum drbd_role s);
1538 extern void drbd_al_begin_io(struct drbd_conf *mdev, sector_t sector);
1539 extern void drbd_al_complete_io(struct drbd_conf *mdev, sector_t sector);
1540 extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector);
1541 extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1542 extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1543 extern void drbd_rs_cancel_all(struct drbd_conf *mdev);
1544 extern int drbd_rs_del_all(struct drbd_conf *mdev);
1545 extern void drbd_rs_failed_io(struct drbd_conf *mdev,
1546 sector_t sector, int size);
1547 extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *);
1548 extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
1549 int size, const char *file, const unsigned int line);
1550 #define drbd_set_in_sync(mdev, sector, size) \
1551 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1552 extern void __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
1553 int size, const char *file, const unsigned int line);
1554 #define drbd_set_out_of_sync(mdev, sector, size) \
1555 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1556 extern void drbd_al_apply_to_bm(struct drbd_conf *mdev);
1557 extern void drbd_al_to_on_disk_bm(struct drbd_conf *mdev);
1558 extern void drbd_al_shrink(struct drbd_conf *mdev);
1563 void drbd_nl_cleanup(void);
1564 int __init drbd_nl_init(void);
1565 void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state);
1566 void drbd_bcast_sync_progress(struct drbd_conf *mdev);
1567 void drbd_bcast_ee(struct drbd_conf *mdev,
1568 const char *reason, const int dgs,
1569 const char* seen_hash, const char* calc_hash,
1570 const struct drbd_epoch_entry* e);
1574 * DOC: DRBD State macros
1576 * These macros are used to express state changes in easily readable form.
1578 * The NS macros expand to a mask and a value, that can be bit ored onto the
1579 * current state as soon as the spinlock (req_lock) was taken.
1581 * The _NS macros are used for state functions that get called with the
1582 * spinlock. These macros expand directly to the new state value.
1584 * Besides the basic forms NS() and _NS() additional _?NS[23] are defined
1585 * to express state changes that affect more than one aspect of the state.
1587 * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY)
1588 * Means that the network connection was established and that the peer
1589 * is in secondary role.
1591 #define role_MASK R_MASK
1592 #define peer_MASK R_MASK
1593 #define disk_MASK D_MASK
1594 #define pdsk_MASK D_MASK
1595 #define conn_MASK C_MASK
1597 #define user_isp_MASK 1
1598 #define aftr_isp_MASK 1
1601 ({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \
1602 ({ union drbd_state val; val.i = 0; val.T = (S); val; })
1603 #define NS2(T1, S1, T2, S2) \
1604 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1605 mask.T2 = T2##_MASK; mask; }), \
1606 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1607 val.T2 = (S2); val; })
1608 #define NS3(T1, S1, T2, S2, T3, S3) \
1609 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1610 mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \
1611 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1612 val.T2 = (S2); val.T3 = (S3); val; })
1614 #define _NS(D, T, S) \
1615 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; })
1616 #define _NS2(D, T1, S1, T2, S2) \
1617 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1618 __ns.T2 = (S2); __ns; })
1619 #define _NS3(D, T1, S1, T2, S2, T3, S3) \
1620 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1621 __ns.T2 = (S2); __ns.T3 = (S3); __ns; })
1624 * inline helper functions
1625 *************************/
1627 static inline void drbd_state_lock(struct drbd_conf *mdev)
1629 wait_event(mdev->misc_wait,
1630 !test_and_set_bit(CLUSTER_ST_CHANGE, &mdev->flags));
1633 static inline void drbd_state_unlock(struct drbd_conf *mdev)
1635 clear_bit(CLUSTER_ST_CHANGE, &mdev->flags);
1636 wake_up(&mdev->misc_wait);
1639 static inline int _drbd_set_state(struct drbd_conf *mdev,
1640 union drbd_state ns, enum chg_state_flags flags,
1641 struct completion *done)
1645 read_lock(&global_state_lock);
1646 rv = __drbd_set_state(mdev, ns, flags, done);
1647 read_unlock(&global_state_lock);
1653 * drbd_request_state() - Reqest a state change
1654 * @mdev: DRBD device.
1655 * @mask: mask of state bits to change.
1656 * @val: value of new state bits.
1658 * This is the most graceful way of requesting a state change. It is verbose
1659 * quite verbose in case the state change is not possible, and all those
1660 * state changes are globally serialized.
1662 static inline int drbd_request_state(struct drbd_conf *mdev,
1663 union drbd_state mask,
1664 union drbd_state val)
1666 return _drbd_request_state(mdev, mask, val, CS_VERBOSE + CS_ORDERED);
1669 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1670 static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach, const char *where)
1672 switch (mdev->ldev->dc.on_io_error) {
1675 if (printk_ratelimit())
1676 dev_err(DEV, "Local IO failed in %s."
1677 "Passing error on...\n", where);
1680 /* NOTE fall through to detach case if forcedetach set */
1682 case EP_CALL_HELPER:
1683 if (mdev->state.disk > D_FAILED) {
1684 _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL);
1685 dev_err(DEV, "Local IO failed in %s."
1686 "Detaching...\n", where);
1693 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1694 * @mdev: DRBD device.
1695 * @error: Error code passed to the IO completion callback
1696 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1698 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1700 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1701 static inline void drbd_chk_io_error_(struct drbd_conf *mdev,
1702 int error, int forcedetach, const char *where)
1705 unsigned long flags;
1706 spin_lock_irqsave(&mdev->req_lock, flags);
1707 __drbd_chk_io_error_(mdev, forcedetach, where);
1708 spin_unlock_irqrestore(&mdev->req_lock, flags);
1714 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1715 * @bdev: Meta data block device.
1717 * BTW, for internal meta data, this happens to be the maximum capacity
1718 * we could agree upon with our peer node.
1720 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1722 switch (bdev->dc.meta_dev_idx) {
1723 case DRBD_MD_INDEX_INTERNAL:
1724 case DRBD_MD_INDEX_FLEX_INT:
1725 return bdev->md.md_offset + bdev->md.bm_offset;
1726 case DRBD_MD_INDEX_FLEX_EXT:
1728 return bdev->md.md_offset;
1733 * drbd_md_last_sector() - Return the last sector number of the meta data area
1734 * @bdev: Meta data block device.
1736 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1738 switch (bdev->dc.meta_dev_idx) {
1739 case DRBD_MD_INDEX_INTERNAL:
1740 case DRBD_MD_INDEX_FLEX_INT:
1741 return bdev->md.md_offset + MD_AL_OFFSET - 1;
1742 case DRBD_MD_INDEX_FLEX_EXT:
1744 return bdev->md.md_offset + bdev->md.md_size_sect;
1748 /* Returns the number of 512 byte sectors of the device */
1749 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1751 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1752 return bdev ? bdev->bd_inode->i_size >> 9 : 0;
1756 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1757 * @bdev: Meta data block device.
1759 * returns the capacity we announce to out peer. we clip ourselves at the
1760 * various MAX_SECTORS, because if we don't, current implementation will
1761 * oops sooner or later
1763 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1766 switch (bdev->dc.meta_dev_idx) {
1767 case DRBD_MD_INDEX_INTERNAL:
1768 case DRBD_MD_INDEX_FLEX_INT:
1769 s = drbd_get_capacity(bdev->backing_bdev)
1770 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1771 drbd_md_first_sector(bdev))
1774 case DRBD_MD_INDEX_FLEX_EXT:
1775 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1776 drbd_get_capacity(bdev->backing_bdev));
1777 /* clip at maximum size the meta device can support */
1778 s = min_t(sector_t, s,
1779 BM_EXT_TO_SECT(bdev->md.md_size_sect
1780 - bdev->md.bm_offset));
1783 s = min_t(sector_t, DRBD_MAX_SECTORS,
1784 drbd_get_capacity(bdev->backing_bdev));
1790 * drbd_md_ss__() - Return the sector number of our meta data super block
1791 * @mdev: DRBD device.
1792 * @bdev: Meta data block device.
1794 static inline sector_t drbd_md_ss__(struct drbd_conf *mdev,
1795 struct drbd_backing_dev *bdev)
1797 switch (bdev->dc.meta_dev_idx) {
1798 default: /* external, some index */
1799 return MD_RESERVED_SECT * bdev->dc.meta_dev_idx;
1800 case DRBD_MD_INDEX_INTERNAL:
1801 /* with drbd08, internal meta data is always "flexible" */
1802 case DRBD_MD_INDEX_FLEX_INT:
1803 /* sizeof(struct md_on_disk_07) == 4k
1804 * position: last 4k aligned block of 4k size */
1805 if (!bdev->backing_bdev) {
1806 if (__ratelimit(&drbd_ratelimit_state)) {
1807 dev_err(DEV, "bdev->backing_bdev==NULL\n");
1812 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL)
1814 case DRBD_MD_INDEX_FLEX_EXT:
1820 _drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1822 list_add_tail(&w->list, &q->q);
1827 drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w)
1829 unsigned long flags;
1830 spin_lock_irqsave(&q->q_lock, flags);
1831 list_add(&w->list, &q->q);
1832 up(&q->s); /* within the spinlock,
1833 see comment near end of drbd_worker() */
1834 spin_unlock_irqrestore(&q->q_lock, flags);
1838 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1840 unsigned long flags;
1841 spin_lock_irqsave(&q->q_lock, flags);
1842 list_add_tail(&w->list, &q->q);
1843 up(&q->s); /* within the spinlock,
1844 see comment near end of drbd_worker() */
1845 spin_unlock_irqrestore(&q->q_lock, flags);
1848 static inline void wake_asender(struct drbd_conf *mdev)
1850 if (test_bit(SIGNAL_ASENDER, &mdev->flags))
1851 force_sig(DRBD_SIG, mdev->asender.task);
1854 static inline void request_ping(struct drbd_conf *mdev)
1856 set_bit(SEND_PING, &mdev->flags);
1860 static inline int drbd_send_short_cmd(struct drbd_conf *mdev,
1861 enum drbd_packets cmd)
1864 return drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, &h, sizeof(h));
1867 static inline int drbd_send_ping(struct drbd_conf *mdev)
1870 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING, &h, sizeof(h));
1873 static inline int drbd_send_ping_ack(struct drbd_conf *mdev)
1876 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING_ACK, &h, sizeof(h));
1879 static inline void drbd_thread_stop(struct drbd_thread *thi)
1881 _drbd_thread_stop(thi, FALSE, TRUE);
1884 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
1886 _drbd_thread_stop(thi, FALSE, FALSE);
1889 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
1891 _drbd_thread_stop(thi, TRUE, FALSE);
1894 /* counts how many answer packets packets we expect from our peer,
1895 * for either explicit application requests,
1896 * or implicit barrier packets as necessary.
1899 * _req_mod(req, queue_for_net_write or queue_for_net_read);
1900 * it is much easier and equally valid to count what we queue for the
1901 * worker, even before it actually was queued or send.
1902 * (drbd_make_request_common; recovery path on read io-error)
1904 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1905 * _req_mod(req, data_received)
1906 * [from receive_DataReply]
1907 * _req_mod(req, write_acked_by_peer or recv_acked_by_peer or neg_acked)
1908 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1909 * for some reason it is NOT decreased in got_NegAck,
1910 * but in the resulting cleanup code from report_params.
1911 * we should try to remember the reason for that...
1912 * _req_mod(req, send_failed or send_canceled)
1913 * _req_mod(req, connection_lost_while_pending)
1914 * [from tl_clear_barrier]
1916 static inline void inc_ap_pending(struct drbd_conf *mdev)
1918 atomic_inc(&mdev->ap_pending_cnt);
1921 #define ERR_IF_CNT_IS_NEGATIVE(which) \
1922 if (atomic_read(&mdev->which) < 0) \
1923 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
1924 __func__ , __LINE__ , \
1925 atomic_read(&mdev->which))
1927 #define dec_ap_pending(mdev) do { \
1928 typecheck(struct drbd_conf *, mdev); \
1929 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \
1930 wake_up(&mdev->misc_wait); \
1931 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
1933 /* counts how many resync-related answers we still expect from the peer
1935 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1936 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK whith ID_SYNCER)
1937 * (or P_NEG_ACK with ID_SYNCER)
1939 static inline void inc_rs_pending(struct drbd_conf *mdev)
1941 atomic_inc(&mdev->rs_pending_cnt);
1944 #define dec_rs_pending(mdev) do { \
1945 typecheck(struct drbd_conf *, mdev); \
1946 atomic_dec(&mdev->rs_pending_cnt); \
1947 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
1949 /* counts how many answers we still need to send to the peer.
1951 * receive_Data unless protocol A;
1952 * we need to send a P_RECV_ACK (proto B)
1953 * or P_WRITE_ACK (proto C)
1954 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1955 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1956 * receive_Barrier_* we need to send a P_BARRIER_ACK
1958 static inline void inc_unacked(struct drbd_conf *mdev)
1960 atomic_inc(&mdev->unacked_cnt);
1963 #define dec_unacked(mdev) do { \
1964 typecheck(struct drbd_conf *, mdev); \
1965 atomic_dec(&mdev->unacked_cnt); \
1966 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
1968 #define sub_unacked(mdev, n) do { \
1969 typecheck(struct drbd_conf *, mdev); \
1970 atomic_sub(n, &mdev->unacked_cnt); \
1971 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
1974 static inline void put_net_conf(struct drbd_conf *mdev)
1976 if (atomic_dec_and_test(&mdev->net_cnt))
1977 wake_up(&mdev->misc_wait);
1981 * get_net_conf() - Increase ref count on mdev->net_conf; Returns 0 if nothing there
1982 * @mdev: DRBD device.
1984 * You have to call put_net_conf() when finished working with mdev->net_conf.
1986 static inline int get_net_conf(struct drbd_conf *mdev)
1990 atomic_inc(&mdev->net_cnt);
1991 have_net_conf = mdev->state.conn >= C_UNCONNECTED;
1994 return have_net_conf;
1998 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2001 * You have to call put_ldev() when finished working with mdev->ldev.
2003 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2004 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2006 static inline void put_ldev(struct drbd_conf *mdev)
2009 if (atomic_dec_and_test(&mdev->local_cnt))
2010 wake_up(&mdev->misc_wait);
2011 D_ASSERT(atomic_read(&mdev->local_cnt) >= 0);
2015 static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
2019 atomic_inc(&mdev->local_cnt);
2020 io_allowed = (mdev->state.disk >= mins);
2026 extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins);
2029 /* you must have an "get_ldev" reference */
2030 static inline void drbd_get_syncer_progress(struct drbd_conf *mdev,
2031 unsigned long *bits_left, unsigned int *per_mil_done)
2034 * this is to break it at compile time when we change that
2035 * (we may feel 4TB maximum storage per drbd is not enough)
2037 typecheck(unsigned long, mdev->rs_total);
2039 /* note: both rs_total and rs_left are in bits, i.e. in
2040 * units of BM_BLOCK_SIZE.
2041 * for the percentage, we don't care. */
2043 *bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
2044 /* >> 10 to prevent overflow,
2045 * +1 to prevent division by zero */
2046 if (*bits_left > mdev->rs_total) {
2047 /* doh. maybe a logic bug somewhere.
2048 * may also be just a race condition
2049 * between this and a disconnect during sync.
2050 * for now, just prevent in-kernel buffer overflow.
2053 dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2054 drbd_conn_str(mdev->state.conn),
2055 *bits_left, mdev->rs_total, mdev->rs_failed);
2058 /* make sure the calculation happens in long context */
2059 unsigned long tmp = 1000UL -
2060 (*bits_left >> 10)*1000UL
2061 / ((mdev->rs_total >> 10) + 1UL);
2062 *per_mil_done = tmp;
2067 /* this throttles on-the-fly application requests
2068 * according to max_buffers settings;
2069 * maybe re-implement using semaphores? */
2070 static inline int drbd_get_max_buffers(struct drbd_conf *mdev)
2072 int mxb = 1000000; /* arbitrary limit on open requests */
2073 if (get_net_conf(mdev)) {
2074 mxb = mdev->net_conf->max_buffers;
2080 static inline int drbd_state_is_stable(union drbd_state s)
2083 /* DO NOT add a default clause, we want the compiler to warn us
2084 * for any newly introduced state we may have forgotten to add here */
2086 switch ((enum drbd_conns)s.conn) {
2087 /* new io only accepted when there is no connection, ... */
2089 case C_WF_CONNECTION:
2090 /* ... or there is a well established connection. */
2096 case C_PAUSED_SYNC_S:
2097 case C_PAUSED_SYNC_T:
2098 /* maybe stable, look at the disk state */
2101 /* no new io accepted during tansitional states
2102 * like handshake or teardown */
2103 case C_DISCONNECTING:
2107 case C_NETWORK_FAILURE:
2108 case C_PROTOCOL_ERROR:
2110 case C_WF_REPORT_PARAMS:
2111 case C_STARTING_SYNC_S:
2112 case C_STARTING_SYNC_T:
2115 case C_WF_SYNC_UUID:
2121 switch ((enum drbd_disk_state)s.disk) {
2123 case D_INCONSISTENT:
2127 /* disk state is stable as well. */
2130 /* no new io accepted during tansitional states */
2143 static inline int __inc_ap_bio_cond(struct drbd_conf *mdev)
2145 int mxb = drbd_get_max_buffers(mdev);
2147 if (mdev->state.susp)
2149 if (test_bit(SUSPEND_IO, &mdev->flags))
2152 /* to avoid potential deadlock or bitmap corruption,
2153 * in various places, we only allow new application io
2154 * to start during "stable" states. */
2156 /* no new io accepted when attaching or detaching the disk */
2157 if (!drbd_state_is_stable(mdev->state))
2160 /* since some older kernels don't have atomic_add_unless,
2161 * and we are within the spinlock anyways, we have this workaround. */
2162 if (atomic_read(&mdev->ap_bio_cnt) > mxb)
2164 if (test_bit(BITMAP_IO, &mdev->flags))
2169 /* I'd like to use wait_event_lock_irq,
2170 * but I'm not sure when it got introduced,
2171 * and not sure when it has 3 or 4 arguments */
2172 static inline void inc_ap_bio(struct drbd_conf *mdev, int one_or_two)
2174 /* compare with after_state_ch,
2175 * os.conn != C_WF_BITMAP_S && ns.conn == C_WF_BITMAP_S */
2179 * as long as the device is suspended
2180 * until the bitmap is no longer on the fly during connection
2181 * handshake as long as we would exeed the max_buffer limit.
2183 * to avoid races with the reconnect code,
2184 * we need to atomic_inc within the spinlock. */
2186 spin_lock_irq(&mdev->req_lock);
2187 while (!__inc_ap_bio_cond(mdev)) {
2188 prepare_to_wait(&mdev->misc_wait, &wait, TASK_UNINTERRUPTIBLE);
2189 spin_unlock_irq(&mdev->req_lock);
2191 finish_wait(&mdev->misc_wait, &wait);
2192 spin_lock_irq(&mdev->req_lock);
2194 atomic_add(one_or_two, &mdev->ap_bio_cnt);
2195 spin_unlock_irq(&mdev->req_lock);
2198 static inline void dec_ap_bio(struct drbd_conf *mdev)
2200 int mxb = drbd_get_max_buffers(mdev);
2201 int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt);
2203 D_ASSERT(ap_bio >= 0);
2204 /* this currently does wake_up for every dec_ap_bio!
2205 * maybe rather introduce some type of hysteresis?
2206 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2208 wake_up(&mdev->misc_wait);
2209 if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) {
2210 if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags))
2211 drbd_queue_work(&mdev->data.work, &mdev->bm_io_work.w);
2215 static inline void drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val)
2217 mdev->ed_uuid = val;
2220 static inline int seq_cmp(u32 a, u32 b)
2222 /* we assume wrap around at 32bit.
2223 * for wrap around at 24bit (old atomic_t),
2227 return (s32)(a) - (s32)(b);
2229 #define seq_lt(a, b) (seq_cmp((a), (b)) < 0)
2230 #define seq_gt(a, b) (seq_cmp((a), (b)) > 0)
2231 #define seq_ge(a, b) (seq_cmp((a), (b)) >= 0)
2232 #define seq_le(a, b) (seq_cmp((a), (b)) <= 0)
2233 /* CAUTION: please no side effects in arguments! */
2234 #define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b)))
2236 static inline void update_peer_seq(struct drbd_conf *mdev, unsigned int new_seq)
2239 spin_lock(&mdev->peer_seq_lock);
2240 m = seq_max(mdev->peer_seq, new_seq);
2242 spin_unlock(&mdev->peer_seq_lock);
2244 wake_up(&mdev->seq_wait);
2247 static inline void drbd_update_congested(struct drbd_conf *mdev)
2249 struct sock *sk = mdev->data.socket->sk;
2250 if (sk->sk_wmem_queued > sk->sk_sndbuf * 4 / 5)
2251 set_bit(NET_CONGESTED, &mdev->flags);
2254 static inline int drbd_queue_order_type(struct drbd_conf *mdev)
2256 /* sorry, we currently have no working implementation
2257 * of distributed TCQ stuff */
2258 #ifndef QUEUE_ORDERED_NONE
2259 #define QUEUE_ORDERED_NONE 0
2261 return QUEUE_ORDERED_NONE;
2264 static inline void drbd_blk_run_queue(struct request_queue *q)
2266 if (q && q->unplug_fn)
2270 static inline void drbd_kick_lo(struct drbd_conf *mdev)
2272 if (get_ldev(mdev)) {
2273 drbd_blk_run_queue(bdev_get_queue(mdev->ldev->backing_bdev));
2278 static inline void drbd_md_flush(struct drbd_conf *mdev)
2282 if (test_bit(MD_NO_BARRIER, &mdev->flags))
2285 r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL,
2288 set_bit(MD_NO_BARRIER, &mdev->flags);
2289 dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r);