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;
555 unsigned int seq_num;
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_wq_barrier {
745 struct completion done;
753 struct drbd_epoch_entry {
755 struct hlist_node colision;
756 struct drbd_epoch *epoch;
757 struct drbd_conf *mdev;
759 atomic_t pending_bios;
761 /* see comments on ee flag bits below */
768 * While corresponding bios are in flight, the only modification will be
769 * set_bit WAS_ERROR, which has to be atomic.
770 * If no bios are in flight yet, or all have been completed,
771 * non-atomic modification to ee->flags is ok.
774 __EE_CALL_AL_COMPLETE_IO,
775 __EE_MAY_SET_IN_SYNC,
777 /* This epoch entry closes an epoch using a barrier.
778 * On sucessful completion, the epoch is released,
779 * and the P_BARRIER_ACK send. */
782 /* In case a barrier failed,
783 * we need to resubmit without the barrier flag. */
786 /* we may have several bios per epoch entry.
787 * if any of those fail, we set this flag atomically
788 * from the endio callback */
791 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
792 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
793 #define EE_IS_BARRIER (1<<__EE_IS_BARRIER)
794 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
795 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
797 /* global flag bits */
799 CREATE_BARRIER, /* next P_DATA is preceeded by a P_BARRIER */
800 SIGNAL_ASENDER, /* whether asender wants to be interrupted */
801 SEND_PING, /* whether asender should send a ping asap */
803 STOP_SYNC_TIMER, /* tell timer to cancel itself */
804 UNPLUG_QUEUED, /* only relevant with kernel 2.4 */
805 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
806 MD_DIRTY, /* current uuids and flags not yet on disk */
807 DISCARD_CONCURRENT, /* Set on one node, cleared on the peer! */
808 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
809 CLUSTER_ST_CHANGE, /* Cluster wide state change going on... */
812 CRASHED_PRIMARY, /* This node was a crashed primary.
813 * Gets cleared when the state.conn
814 * goes into C_CONNECTED state. */
815 WRITE_BM_AFTER_RESYNC, /* A kmalloc() during resync failed */
816 NO_BARRIER_SUPP, /* underlying block device doesn't implement barriers */
819 MD_NO_BARRIER, /* meta data device does not support barriers,
821 SUSPEND_IO, /* suspend application io */
822 BITMAP_IO, /* suspend application io;
823 once no more io in flight, start bitmap io */
824 BITMAP_IO_QUEUED, /* Started bitmap IO */
825 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
826 NET_CONGESTED, /* The data socket is congested */
828 CONFIG_PENDING, /* serialization of (re)configuration requests.
829 * if set, also prevents the device from dying */
830 DEVICE_DYING, /* device became unconfigured,
831 * but worker thread is still handling the cleanup.
832 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
833 * while this is set. */
834 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
835 * the peer, if it changed there as well. */
836 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
837 GOT_PING_ACK, /* set when we receive a ping_ack packet, misc wait gets woken */
840 struct drbd_bitmap; /* opaque for drbd_conf */
842 /* TODO sort members for performance
843 * MAYBE group them further */
845 /* THINK maybe we actually want to use the default "event/%s" worker threads
846 * or similar in linux 2.6, which uses per cpu data and threads.
848 * To be general, this might need a spin_lock member.
849 * For now, please use the mdev->req_lock to protect list_head,
850 * see drbd_queue_work below.
852 struct drbd_work_queue {
854 struct semaphore s; /* producers up it, worker down()s it */
855 spinlock_t q_lock; /* to protect the list. */
859 struct drbd_work_queue work;
861 struct socket *socket;
862 /* this way we get our
863 * send/receive buffers off the stack */
864 union p_polymorph sbuf;
865 union p_polymorph rbuf;
869 u64 md_offset; /* sector offset to 'super' block */
871 u64 la_size_sect; /* last agreed size, unit sectors */
877 s32 al_offset; /* signed relative sector offset to al area */
878 s32 bm_offset; /* signed relative sector offset to bitmap */
880 /* u32 al_nr_extents; important for restoring the AL
881 * is stored into sync_conf.al_extents, which in turn
882 * gets applied to act_log->nr_elements
886 /* for sync_conf and other types... */
887 #define NL_PACKET(name, number, fields) struct name { fields };
888 #define NL_INTEGER(pn,pr,member) int member;
889 #define NL_INT64(pn,pr,member) __u64 member;
890 #define NL_BIT(pn,pr,member) unsigned member:1;
891 #define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
892 #include "linux/drbd_nl.h"
894 struct drbd_backing_dev {
895 struct block_device *backing_bdev;
896 struct block_device *md_bdev;
897 struct file *lo_file;
898 struct file *md_file;
900 struct disk_conf dc; /* The user provided config... */
901 sector_t known_size; /* last known size of that backing device */
905 struct drbd_conf *mdev;
906 struct completion event;
913 int (*io_fn)(struct drbd_conf *mdev);
914 void (*done)(struct drbd_conf *mdev, int rv);
917 enum write_ordering_e {
925 /* things that are stored as / read from meta data on disk */
928 /* configured by drbdsetup */
929 struct net_conf *net_conf; /* protected by get_net_conf() and put_net_conf() */
930 struct syncer_conf sync_conf;
931 struct drbd_backing_dev *ldev __protected_by(local);
933 sector_t p_size; /* partner's disk size */
934 struct request_queue *rq_queue;
935 struct block_device *this_bdev;
936 struct gendisk *vdisk;
938 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
939 struct drbd_socket meta; /* ping/ack (metadata) packets */
940 int agreed_pro_version; /* actually used protocol version */
941 unsigned long last_received; /* in jiffies, either socket */
942 unsigned int ko_count;
943 struct drbd_work resync_work,
948 struct timer_list resync_timer;
949 struct timer_list md_sync_timer;
950 struct timer_list delay_probe_timer;
952 /* Used after attach while negotiating new disk state. */
953 union drbd_state new_state_tmp;
955 union drbd_state state;
956 wait_queue_head_t misc_wait;
957 wait_queue_head_t state_wait; /* upon each state change. */
958 unsigned int send_cnt;
959 unsigned int recv_cnt;
960 unsigned int read_cnt;
961 unsigned int writ_cnt;
962 unsigned int al_writ_cnt;
963 unsigned int bm_writ_cnt;
964 atomic_t ap_bio_cnt; /* Requests we need to complete */
965 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
966 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
967 atomic_t unacked_cnt; /* Need to send replys for */
968 atomic_t local_cnt; /* Waiting for local completion */
969 atomic_t net_cnt; /* Users of net_conf */
971 struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */
972 struct drbd_tl_epoch *newest_tle;
973 struct drbd_tl_epoch *oldest_tle;
974 struct list_head out_of_sequence_requests;
975 struct hlist_head *tl_hash;
976 unsigned int tl_hash_s;
978 /* blocks to sync in this run [unit BM_BLOCK_SIZE] */
979 unsigned long rs_total;
980 /* number of sync IOs that failed in this run */
981 unsigned long rs_failed;
982 /* Syncer's start time [unit jiffies] */
983 unsigned long rs_start;
984 /* cumulated time in PausedSyncX state [unit jiffies] */
985 unsigned long rs_paused;
986 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
987 unsigned long rs_mark_left;
988 /* marks's time [unit jiffies] */
989 unsigned long rs_mark_time;
990 /* skipped because csum was equeal [unit BM_BLOCK_SIZE] */
991 unsigned long rs_same_csum;
993 /* where does the admin want us to start? (sector) */
994 sector_t ov_start_sector;
995 /* where are we now? (sector) */
996 sector_t ov_position;
997 /* Start sector of out of sync range (to merge printk reporting). */
998 sector_t ov_last_oos_start;
999 /* size of out-of-sync range in sectors. */
1000 sector_t ov_last_oos_size;
1001 unsigned long ov_left; /* in bits */
1002 struct crypto_hash *csums_tfm;
1003 struct crypto_hash *verify_tfm;
1005 struct drbd_thread receiver;
1006 struct drbd_thread worker;
1007 struct drbd_thread asender;
1008 struct drbd_bitmap *bitmap;
1009 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
1011 /* Used to track operations of resync... */
1012 struct lru_cache *resync;
1013 /* Number of locked elements in resync LRU */
1014 unsigned int resync_locked;
1015 /* resync extent number waiting for application requests */
1016 unsigned int resync_wenr;
1020 struct drbd_epoch *current_epoch;
1021 spinlock_t epoch_lock;
1022 unsigned int epochs;
1023 enum write_ordering_e write_ordering;
1024 struct list_head active_ee; /* IO in progress */
1025 struct list_head sync_ee; /* IO in progress */
1026 struct list_head done_ee; /* send ack */
1027 struct list_head read_ee; /* IO in progress */
1028 struct list_head net_ee; /* zero-copy network send in progress */
1029 struct hlist_head *ee_hash; /* is proteced by req_lock! */
1030 unsigned int ee_hash_s;
1032 /* this one is protected by ee_lock, single thread */
1033 struct drbd_epoch_entry *last_write_w_barrier;
1035 int next_barrier_nr;
1036 struct hlist_head *app_reads_hash; /* is proteced by req_lock */
1037 struct list_head resync_reads;
1039 wait_queue_head_t ee_wait;
1040 struct page *md_io_page; /* one page buffer for md_io */
1041 struct page *md_io_tmpp; /* for logical_block_size != 512 */
1042 struct mutex md_io_mutex; /* protects the md_io_buffer */
1044 wait_queue_head_t al_wait;
1045 struct lru_cache *act_log; /* activity log */
1046 unsigned int al_tr_number;
1048 int al_tr_pos; /* position of the next transaction in the journal */
1049 struct crypto_hash *cram_hmac_tfm;
1050 struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */
1051 struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */
1055 wait_queue_head_t seq_wait;
1056 atomic_t packet_seq;
1057 unsigned int peer_seq;
1058 spinlock_t peer_seq_lock;
1060 unsigned long comm_bm_set; /* communicated number of set bits. */
1061 cpumask_var_t cpu_mask;
1062 struct bm_io_work bm_io_work;
1063 u64 ed_uuid; /* UUID of the exposed data */
1064 struct mutex state_mutex;
1065 char congestion_reason; /* Why we where congested... */
1066 struct list_head delay_probes; /* protected by peer_seq_lock */
1067 int data_delay; /* Delay of packets on the data-sock behind meta-sock */
1068 unsigned int delay_seq; /* To generate sequence numbers of delay probes */
1069 struct timeval dps_time; /* delay-probes-start-time */
1070 unsigned int dp_volume_last; /* send_cnt of last delay probe */
1071 int c_sync_rate; /* current resync rate after delay_probe magic */
1072 atomic_t new_c_uuid;
1075 static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
1077 struct drbd_conf *mdev;
1079 mdev = minor < minor_count ? minor_table[minor] : NULL;
1084 static inline unsigned int mdev_to_minor(struct drbd_conf *mdev)
1089 /* returns 1 if it was successfull,
1090 * returns 0 if there was no data socket.
1091 * so wherever you are going to use the data.socket, e.g. do
1092 * if (!drbd_get_data_sock(mdev))
1095 * drbd_put_data_sock(mdev);
1097 static inline int drbd_get_data_sock(struct drbd_conf *mdev)
1099 mutex_lock(&mdev->data.mutex);
1100 /* drbd_disconnect() could have called drbd_free_sock()
1101 * while we were waiting in down()... */
1102 if (unlikely(mdev->data.socket == NULL)) {
1103 mutex_unlock(&mdev->data.mutex);
1109 static inline void drbd_put_data_sock(struct drbd_conf *mdev)
1111 mutex_unlock(&mdev->data.mutex);
1115 * function declarations
1116 *************************/
1120 enum chg_state_flags {
1123 CS_WAIT_COMPLETE = 4,
1125 CS_ORDERED = CS_WAIT_COMPLETE + CS_SERIALIZE,
1130 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1133 extern void drbd_init_set_defaults(struct drbd_conf *mdev);
1134 extern int drbd_change_state(struct drbd_conf *mdev, enum chg_state_flags f,
1135 union drbd_state mask, union drbd_state val);
1136 extern void drbd_force_state(struct drbd_conf *, union drbd_state,
1138 extern int _drbd_request_state(struct drbd_conf *, union drbd_state,
1139 union drbd_state, enum chg_state_flags);
1140 extern int __drbd_set_state(struct drbd_conf *, union drbd_state,
1141 enum chg_state_flags, struct completion *done);
1142 extern void print_st_err(struct drbd_conf *, union drbd_state,
1143 union drbd_state, int);
1144 extern int drbd_thread_start(struct drbd_thread *thi);
1145 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1147 extern void drbd_thread_current_set_cpu(struct drbd_conf *mdev);
1148 extern void drbd_calc_cpu_mask(struct drbd_conf *mdev);
1150 #define drbd_thread_current_set_cpu(A) ({})
1151 #define drbd_calc_cpu_mask(A) ({})
1153 extern void drbd_free_resources(struct drbd_conf *mdev);
1154 extern void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr,
1155 unsigned int set_size);
1156 extern void tl_clear(struct drbd_conf *mdev);
1157 extern void _tl_add_barrier(struct drbd_conf *, struct drbd_tl_epoch *);
1158 extern void drbd_free_sock(struct drbd_conf *mdev);
1159 extern int drbd_send(struct drbd_conf *mdev, struct socket *sock,
1160 void *buf, size_t size, unsigned msg_flags);
1161 extern int drbd_send_protocol(struct drbd_conf *mdev);
1162 extern int drbd_send_uuids(struct drbd_conf *mdev);
1163 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
1164 extern int drbd_send_sync_uuid(struct drbd_conf *mdev, u64 val);
1165 extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags);
1166 extern int _drbd_send_state(struct drbd_conf *mdev);
1167 extern int drbd_send_state(struct drbd_conf *mdev);
1168 extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock,
1169 enum drbd_packets cmd, struct p_header *h,
1170 size_t size, unsigned msg_flags);
1171 #define USE_DATA_SOCKET 1
1172 #define USE_META_SOCKET 0
1173 extern int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket,
1174 enum drbd_packets cmd, struct p_header *h,
1176 extern int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd,
1177 char *data, size_t size);
1178 extern int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc);
1179 extern int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr,
1181 extern int drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd,
1182 struct drbd_epoch_entry *e);
1183 extern int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packets cmd,
1184 struct p_block_req *rp);
1185 extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd,
1187 extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd,
1188 sector_t sector, int blksize, u64 block_id);
1189 extern int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd,
1190 struct drbd_epoch_entry *e);
1191 extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
1192 extern int _drbd_send_barrier(struct drbd_conf *mdev,
1193 struct drbd_tl_epoch *barrier);
1194 extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1195 sector_t sector, int size, u64 block_id);
1196 extern int drbd_send_drequest_csum(struct drbd_conf *mdev,
1197 sector_t sector,int size,
1198 void *digest, int digest_size,
1199 enum drbd_packets cmd);
1200 extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size);
1202 extern int drbd_send_bitmap(struct drbd_conf *mdev);
1203 extern int _drbd_send_bitmap(struct drbd_conf *mdev);
1204 extern int drbd_send_sr_reply(struct drbd_conf *mdev, int retcode);
1205 extern void drbd_free_bc(struct drbd_backing_dev *ldev);
1206 extern void drbd_mdev_cleanup(struct drbd_conf *mdev);
1208 /* drbd_meta-data.c (still in drbd_main.c) */
1209 extern void drbd_md_sync(struct drbd_conf *mdev);
1210 extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
1211 /* maybe define them below as inline? */
1212 extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1213 extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1214 extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1215 extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1216 extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
1217 extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
1218 extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
1219 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1220 extern void drbd_md_mark_dirty(struct drbd_conf *mdev);
1221 extern void drbd_queue_bitmap_io(struct drbd_conf *mdev,
1222 int (*io_fn)(struct drbd_conf *),
1223 void (*done)(struct drbd_conf *, int),
1225 extern int drbd_bmio_set_n_write(struct drbd_conf *mdev);
1226 extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev);
1227 extern int drbd_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), char *why);
1231 We reserve a 128MB Block (4k aligned)
1232 * either at the end of the backing device
1233 * or on a separate meta data device. */
1235 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1236 /* The following numbers are sectors */
1237 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1238 #define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */
1239 /* Allows up to about 3.8TB */
1240 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1242 /* Since the smalles IO unit is usually 512 byte */
1243 #define MD_SECTOR_SHIFT 9
1244 #define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT)
1247 #define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1248 #define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */
1249 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1251 #if BITS_PER_LONG == 32
1253 #define cpu_to_lel(A) cpu_to_le32(A)
1254 #define lel_to_cpu(A) le32_to_cpu(A)
1255 #elif BITS_PER_LONG == 64
1257 #define cpu_to_lel(A) cpu_to_le64(A)
1258 #define lel_to_cpu(A) le64_to_cpu(A)
1260 #error "LN2 of BITS_PER_LONG unknown!"
1264 /* 16MB sized 'bitmap extent' to track syncer usage */
1266 int rs_left; /* number of bits set (out of sync) in this extent. */
1267 int rs_failed; /* number of failed resync requests in this extent. */
1268 unsigned long flags;
1269 struct lc_element lce;
1272 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1273 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1277 * We need to store one bit for a block.
1278 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1279 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1280 * Bit 1 ==> local node thinks this block needs to be synced.
1283 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1284 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1285 /* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1286 * per sector of on disk bitmap */
1287 #define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */
1288 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1290 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1291 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1294 /* thus many _storage_ sectors are described by one bit */
1295 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1296 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1297 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1299 /* bit to represented kilo byte conversion */
1300 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1302 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1303 * _storage_ sector is located in */
1304 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1306 /* how much _storage_ sectors we have per bitmap sector */
1307 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1308 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1310 /* in one sector of the bitmap, we have this many activity_log extents. */
1311 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1312 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1314 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1315 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1317 /* the extent in "PER_EXTENT" below is an activity log extent
1318 * we need that many (long words/bytes) to store the bitmap
1319 * of one AL_EXTENT_SIZE chunk of storage.
1320 * we can store the bitmap for that many AL_EXTENTS within
1321 * one sector of the _on_disk_ bitmap:
1322 * bit 0 bit 37 bit 38 bit (512*8)-1
1323 * ...|........|........|.. // ..|........|
1324 * sect. 0 `296 `304 ^(512*8*8)-1
1326 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1327 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1328 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1331 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1332 #define DRBD_MAX_SECTORS_BM \
1333 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1334 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1335 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1336 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1337 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1338 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1339 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1341 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1342 /* 16 TB in units of sectors */
1343 #if BITS_PER_LONG == 32
1344 /* adjust by one page worth of bitmap,
1345 * so we won't wrap around in drbd_bm_find_next_bit.
1346 * you should use 64bit OS for that much storage, anyways. */
1347 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1349 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0x1LU << 32)
1353 /* Sector shift value for the "hash" functions of tl_hash and ee_hash tables.
1354 * With a value of 6 all IO in one 32K block make it to the same slot of the
1357 #define DRBD_MAX_SEGMENT_SIZE (1U<<(9+HT_SHIFT))
1359 /* Number of elements in the app_reads_hash */
1360 #define APP_R_HSIZE 15
1362 extern int drbd_bm_init(struct drbd_conf *mdev);
1363 extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits);
1364 extern void drbd_bm_cleanup(struct drbd_conf *mdev);
1365 extern void drbd_bm_set_all(struct drbd_conf *mdev);
1366 extern void drbd_bm_clear_all(struct drbd_conf *mdev);
1367 extern int drbd_bm_set_bits(
1368 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1369 extern int drbd_bm_clear_bits(
1370 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1371 /* bm_set_bits variant for use while holding drbd_bm_lock */
1372 extern void _drbd_bm_set_bits(struct drbd_conf *mdev,
1373 const unsigned long s, const unsigned long e);
1374 extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr);
1375 extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr);
1376 extern int drbd_bm_write_sect(struct drbd_conf *mdev, unsigned long enr) __must_hold(local);
1377 extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
1378 extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
1379 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev,
1380 unsigned long al_enr);
1381 extern size_t drbd_bm_words(struct drbd_conf *mdev);
1382 extern unsigned long drbd_bm_bits(struct drbd_conf *mdev);
1383 extern sector_t drbd_bm_capacity(struct drbd_conf *mdev);
1384 extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1385 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1386 extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1387 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo);
1388 extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev);
1389 extern int drbd_bm_rs_done(struct drbd_conf *mdev);
1390 /* for receive_bitmap */
1391 extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset,
1392 size_t number, unsigned long *buffer);
1393 /* for _drbd_send_bitmap and drbd_bm_write_sect */
1394 extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset,
1395 size_t number, unsigned long *buffer);
1397 extern void drbd_bm_lock(struct drbd_conf *mdev, char *why);
1398 extern void drbd_bm_unlock(struct drbd_conf *mdev);
1400 extern int drbd_bm_count_bits(struct drbd_conf *mdev, const unsigned long s, const unsigned long e);
1403 extern struct kmem_cache *drbd_request_cache;
1404 extern struct kmem_cache *drbd_ee_cache; /* epoch entries */
1405 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1406 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1407 extern mempool_t *drbd_request_mempool;
1408 extern mempool_t *drbd_ee_mempool;
1410 extern struct page *drbd_pp_pool; /* drbd's page pool */
1411 extern spinlock_t drbd_pp_lock;
1412 extern int drbd_pp_vacant;
1413 extern wait_queue_head_t drbd_pp_wait;
1415 extern rwlock_t global_state_lock;
1417 extern struct drbd_conf *drbd_new_device(unsigned int minor);
1418 extern void drbd_free_mdev(struct drbd_conf *mdev);
1420 extern int proc_details;
1423 extern int drbd_make_request_26(struct request_queue *q, struct bio *bio);
1424 extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req);
1425 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1426 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1430 extern void drbd_suspend_io(struct drbd_conf *mdev);
1431 extern void drbd_resume_io(struct drbd_conf *mdev);
1432 extern char *ppsize(char *buf, unsigned long long size);
1433 extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int);
1434 enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
1435 extern enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local);
1436 extern void resync_after_online_grow(struct drbd_conf *);
1437 extern void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int) __must_hold(local);
1438 extern int drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role,
1440 enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev);
1441 extern int drbd_khelper(struct drbd_conf *mdev, char *cmd);
1444 extern int drbd_worker(struct drbd_thread *thi);
1445 extern int drbd_alter_sa(struct drbd_conf *mdev, int na);
1446 extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side);
1447 extern void resume_next_sg(struct drbd_conf *mdev);
1448 extern void suspend_other_sg(struct drbd_conf *mdev);
1449 extern int drbd_resync_finished(struct drbd_conf *mdev);
1450 /* maybe rather drbd_main.c ? */
1451 extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
1452 struct drbd_backing_dev *bdev, sector_t sector, int rw);
1453 extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int);
1455 static inline void ov_oos_print(struct drbd_conf *mdev)
1457 if (mdev->ov_last_oos_size) {
1458 dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n",
1459 (unsigned long long)mdev->ov_last_oos_start,
1460 (unsigned long)mdev->ov_last_oos_size);
1462 mdev->ov_last_oos_size=0;
1466 extern void drbd_csum_bio(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
1467 extern void drbd_csum_ee(struct drbd_conf *, struct crypto_hash *, struct drbd_epoch_entry *, void *);
1468 /* worker callbacks */
1469 extern int w_req_cancel_conflict(struct drbd_conf *, struct drbd_work *, int);
1470 extern int w_read_retry_remote(struct drbd_conf *, struct drbd_work *, int);
1471 extern int w_e_end_data_req(struct drbd_conf *, struct drbd_work *, int);
1472 extern int w_e_end_rsdata_req(struct drbd_conf *, struct drbd_work *, int);
1473 extern int w_e_end_csum_rs_req(struct drbd_conf *, struct drbd_work *, int);
1474 extern int w_e_end_ov_reply(struct drbd_conf *, struct drbd_work *, int);
1475 extern int w_e_end_ov_req(struct drbd_conf *, struct drbd_work *, int);
1476 extern int w_ov_finished(struct drbd_conf *, struct drbd_work *, int);
1477 extern int w_resync_inactive(struct drbd_conf *, struct drbd_work *, int);
1478 extern int w_resume_next_sg(struct drbd_conf *, struct drbd_work *, int);
1479 extern int w_io_error(struct drbd_conf *, struct drbd_work *, int);
1480 extern int w_send_write_hint(struct drbd_conf *, struct drbd_work *, int);
1481 extern int w_make_resync_request(struct drbd_conf *, struct drbd_work *, int);
1482 extern int w_send_dblock(struct drbd_conf *, struct drbd_work *, int);
1483 extern int w_send_barrier(struct drbd_conf *, struct drbd_work *, int);
1484 extern int w_send_read_req(struct drbd_conf *, struct drbd_work *, int);
1485 extern int w_prev_work_done(struct drbd_conf *, struct drbd_work *, int);
1486 extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int);
1488 extern void resync_timer_fn(unsigned long data);
1490 /* drbd_receiver.c */
1491 extern int drbd_submit_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e,
1492 const unsigned rw, const int fault_type);
1493 extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list);
1494 extern struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev,
1497 unsigned int data_size,
1498 gfp_t gfp_mask) __must_hold(local);
1499 extern void drbd_free_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e);
1500 extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1501 struct list_head *head);
1502 extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1503 struct list_head *head);
1504 extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled);
1505 extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed);
1506 extern void drbd_flush_workqueue(struct drbd_conf *mdev);
1508 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1509 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1510 static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1511 char __user *optval, int optlen)
1514 if (level == SOL_SOCKET)
1515 err = sock_setsockopt(sock, level, optname, optval, optlen);
1517 err = sock->ops->setsockopt(sock, level, optname, optval,
1522 static inline void drbd_tcp_cork(struct socket *sock)
1525 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1526 (char __user *)&val, sizeof(val));
1529 static inline void drbd_tcp_uncork(struct socket *sock)
1532 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1533 (char __user *)&val, sizeof(val));
1536 static inline void drbd_tcp_nodelay(struct socket *sock)
1539 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1540 (char __user *)&val, sizeof(val));
1543 static inline void drbd_tcp_quickack(struct socket *sock)
1546 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1547 (char __user *)&val, sizeof(val));
1550 void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo);
1553 extern struct proc_dir_entry *drbd_proc;
1554 extern const struct file_operations drbd_proc_fops;
1555 extern const char *drbd_conn_str(enum drbd_conns s);
1556 extern const char *drbd_role_str(enum drbd_role s);
1559 extern void drbd_al_begin_io(struct drbd_conf *mdev, sector_t sector);
1560 extern void drbd_al_complete_io(struct drbd_conf *mdev, sector_t sector);
1561 extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector);
1562 extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1563 extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1564 extern void drbd_rs_cancel_all(struct drbd_conf *mdev);
1565 extern int drbd_rs_del_all(struct drbd_conf *mdev);
1566 extern void drbd_rs_failed_io(struct drbd_conf *mdev,
1567 sector_t sector, int size);
1568 extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *);
1569 extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
1570 int size, const char *file, const unsigned int line);
1571 #define drbd_set_in_sync(mdev, sector, size) \
1572 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1573 extern void __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
1574 int size, const char *file, const unsigned int line);
1575 #define drbd_set_out_of_sync(mdev, sector, size) \
1576 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1577 extern void drbd_al_apply_to_bm(struct drbd_conf *mdev);
1578 extern void drbd_al_to_on_disk_bm(struct drbd_conf *mdev);
1579 extern void drbd_al_shrink(struct drbd_conf *mdev);
1584 void drbd_nl_cleanup(void);
1585 int __init drbd_nl_init(void);
1586 void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state);
1587 void drbd_bcast_sync_progress(struct drbd_conf *mdev);
1588 void drbd_bcast_ee(struct drbd_conf *mdev,
1589 const char *reason, const int dgs,
1590 const char* seen_hash, const char* calc_hash,
1591 const struct drbd_epoch_entry* e);
1595 * DOC: DRBD State macros
1597 * These macros are used to express state changes in easily readable form.
1599 * The NS macros expand to a mask and a value, that can be bit ored onto the
1600 * current state as soon as the spinlock (req_lock) was taken.
1602 * The _NS macros are used for state functions that get called with the
1603 * spinlock. These macros expand directly to the new state value.
1605 * Besides the basic forms NS() and _NS() additional _?NS[23] are defined
1606 * to express state changes that affect more than one aspect of the state.
1608 * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY)
1609 * Means that the network connection was established and that the peer
1610 * is in secondary role.
1612 #define role_MASK R_MASK
1613 #define peer_MASK R_MASK
1614 #define disk_MASK D_MASK
1615 #define pdsk_MASK D_MASK
1616 #define conn_MASK C_MASK
1618 #define user_isp_MASK 1
1619 #define aftr_isp_MASK 1
1622 ({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \
1623 ({ union drbd_state val; val.i = 0; val.T = (S); val; })
1624 #define NS2(T1, S1, T2, S2) \
1625 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1626 mask.T2 = T2##_MASK; mask; }), \
1627 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1628 val.T2 = (S2); val; })
1629 #define NS3(T1, S1, T2, S2, T3, S3) \
1630 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1631 mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \
1632 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1633 val.T2 = (S2); val.T3 = (S3); val; })
1635 #define _NS(D, T, S) \
1636 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; })
1637 #define _NS2(D, T1, S1, T2, S2) \
1638 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1639 __ns.T2 = (S2); __ns; })
1640 #define _NS3(D, T1, S1, T2, S2, T3, S3) \
1641 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1642 __ns.T2 = (S2); __ns.T3 = (S3); __ns; })
1645 * inline helper functions
1646 *************************/
1648 /* see also page_chain_add and friends in drbd_receiver.c */
1649 static inline struct page *page_chain_next(struct page *page)
1651 return (struct page *)page_private(page);
1653 #define page_chain_for_each(page) \
1654 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1655 page = page_chain_next(page))
1656 #define page_chain_for_each_safe(page, n) \
1657 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1659 static inline int drbd_bio_has_active_page(struct bio *bio)
1661 struct bio_vec *bvec;
1664 __bio_for_each_segment(bvec, bio, i, 0) {
1665 if (page_count(bvec->bv_page) > 1)
1672 static inline int drbd_ee_has_active_page(struct drbd_epoch_entry *e)
1674 struct page *page = e->pages;
1675 page_chain_for_each(page) {
1676 if (page_count(page) > 1)
1683 static inline void drbd_state_lock(struct drbd_conf *mdev)
1685 wait_event(mdev->misc_wait,
1686 !test_and_set_bit(CLUSTER_ST_CHANGE, &mdev->flags));
1689 static inline void drbd_state_unlock(struct drbd_conf *mdev)
1691 clear_bit(CLUSTER_ST_CHANGE, &mdev->flags);
1692 wake_up(&mdev->misc_wait);
1695 static inline int _drbd_set_state(struct drbd_conf *mdev,
1696 union drbd_state ns, enum chg_state_flags flags,
1697 struct completion *done)
1701 read_lock(&global_state_lock);
1702 rv = __drbd_set_state(mdev, ns, flags, done);
1703 read_unlock(&global_state_lock);
1709 * drbd_request_state() - Reqest a state change
1710 * @mdev: DRBD device.
1711 * @mask: mask of state bits to change.
1712 * @val: value of new state bits.
1714 * This is the most graceful way of requesting a state change. It is verbose
1715 * quite verbose in case the state change is not possible, and all those
1716 * state changes are globally serialized.
1718 static inline int drbd_request_state(struct drbd_conf *mdev,
1719 union drbd_state mask,
1720 union drbd_state val)
1722 return _drbd_request_state(mdev, mask, val, CS_VERBOSE + CS_ORDERED);
1725 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1726 static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach, const char *where)
1728 switch (mdev->ldev->dc.on_io_error) {
1731 if (printk_ratelimit())
1732 dev_err(DEV, "Local IO failed in %s."
1733 "Passing error on...\n", where);
1736 /* NOTE fall through to detach case if forcedetach set */
1738 case EP_CALL_HELPER:
1739 if (mdev->state.disk > D_FAILED) {
1740 _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL);
1741 dev_err(DEV, "Local IO failed in %s."
1742 "Detaching...\n", where);
1749 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1750 * @mdev: DRBD device.
1751 * @error: Error code passed to the IO completion callback
1752 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1754 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1756 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1757 static inline void drbd_chk_io_error_(struct drbd_conf *mdev,
1758 int error, int forcedetach, const char *where)
1761 unsigned long flags;
1762 spin_lock_irqsave(&mdev->req_lock, flags);
1763 __drbd_chk_io_error_(mdev, forcedetach, where);
1764 spin_unlock_irqrestore(&mdev->req_lock, flags);
1770 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1771 * @bdev: Meta data block device.
1773 * BTW, for internal meta data, this happens to be the maximum capacity
1774 * we could agree upon with our peer node.
1776 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1778 switch (bdev->dc.meta_dev_idx) {
1779 case DRBD_MD_INDEX_INTERNAL:
1780 case DRBD_MD_INDEX_FLEX_INT:
1781 return bdev->md.md_offset + bdev->md.bm_offset;
1782 case DRBD_MD_INDEX_FLEX_EXT:
1784 return bdev->md.md_offset;
1789 * drbd_md_last_sector() - Return the last sector number of the meta data area
1790 * @bdev: Meta data block device.
1792 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1794 switch (bdev->dc.meta_dev_idx) {
1795 case DRBD_MD_INDEX_INTERNAL:
1796 case DRBD_MD_INDEX_FLEX_INT:
1797 return bdev->md.md_offset + MD_AL_OFFSET - 1;
1798 case DRBD_MD_INDEX_FLEX_EXT:
1800 return bdev->md.md_offset + bdev->md.md_size_sect;
1804 /* Returns the number of 512 byte sectors of the device */
1805 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1807 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1808 return bdev ? bdev->bd_inode->i_size >> 9 : 0;
1812 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1813 * @bdev: Meta data block device.
1815 * returns the capacity we announce to out peer. we clip ourselves at the
1816 * various MAX_SECTORS, because if we don't, current implementation will
1817 * oops sooner or later
1819 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1822 switch (bdev->dc.meta_dev_idx) {
1823 case DRBD_MD_INDEX_INTERNAL:
1824 case DRBD_MD_INDEX_FLEX_INT:
1825 s = drbd_get_capacity(bdev->backing_bdev)
1826 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1827 drbd_md_first_sector(bdev))
1830 case DRBD_MD_INDEX_FLEX_EXT:
1831 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1832 drbd_get_capacity(bdev->backing_bdev));
1833 /* clip at maximum size the meta device can support */
1834 s = min_t(sector_t, s,
1835 BM_EXT_TO_SECT(bdev->md.md_size_sect
1836 - bdev->md.bm_offset));
1839 s = min_t(sector_t, DRBD_MAX_SECTORS,
1840 drbd_get_capacity(bdev->backing_bdev));
1846 * drbd_md_ss__() - Return the sector number of our meta data super block
1847 * @mdev: DRBD device.
1848 * @bdev: Meta data block device.
1850 static inline sector_t drbd_md_ss__(struct drbd_conf *mdev,
1851 struct drbd_backing_dev *bdev)
1853 switch (bdev->dc.meta_dev_idx) {
1854 default: /* external, some index */
1855 return MD_RESERVED_SECT * bdev->dc.meta_dev_idx;
1856 case DRBD_MD_INDEX_INTERNAL:
1857 /* with drbd08, internal meta data is always "flexible" */
1858 case DRBD_MD_INDEX_FLEX_INT:
1859 /* sizeof(struct md_on_disk_07) == 4k
1860 * position: last 4k aligned block of 4k size */
1861 if (!bdev->backing_bdev) {
1862 if (__ratelimit(&drbd_ratelimit_state)) {
1863 dev_err(DEV, "bdev->backing_bdev==NULL\n");
1868 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL)
1870 case DRBD_MD_INDEX_FLEX_EXT:
1876 _drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1878 list_add_tail(&w->list, &q->q);
1883 drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w)
1885 unsigned long flags;
1886 spin_lock_irqsave(&q->q_lock, flags);
1887 list_add(&w->list, &q->q);
1888 up(&q->s); /* within the spinlock,
1889 see comment near end of drbd_worker() */
1890 spin_unlock_irqrestore(&q->q_lock, flags);
1894 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1896 unsigned long flags;
1897 spin_lock_irqsave(&q->q_lock, flags);
1898 list_add_tail(&w->list, &q->q);
1899 up(&q->s); /* within the spinlock,
1900 see comment near end of drbd_worker() */
1901 spin_unlock_irqrestore(&q->q_lock, flags);
1904 static inline void wake_asender(struct drbd_conf *mdev)
1906 if (test_bit(SIGNAL_ASENDER, &mdev->flags))
1907 force_sig(DRBD_SIG, mdev->asender.task);
1910 static inline void request_ping(struct drbd_conf *mdev)
1912 set_bit(SEND_PING, &mdev->flags);
1916 static inline int drbd_send_short_cmd(struct drbd_conf *mdev,
1917 enum drbd_packets cmd)
1920 return drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, &h, sizeof(h));
1923 static inline int drbd_send_ping(struct drbd_conf *mdev)
1926 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING, &h, sizeof(h));
1929 static inline int drbd_send_ping_ack(struct drbd_conf *mdev)
1932 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING_ACK, &h, sizeof(h));
1935 static inline void drbd_thread_stop(struct drbd_thread *thi)
1937 _drbd_thread_stop(thi, FALSE, TRUE);
1940 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
1942 _drbd_thread_stop(thi, FALSE, FALSE);
1945 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
1947 _drbd_thread_stop(thi, TRUE, FALSE);
1950 /* counts how many answer packets packets we expect from our peer,
1951 * for either explicit application requests,
1952 * or implicit barrier packets as necessary.
1955 * _req_mod(req, queue_for_net_write or queue_for_net_read);
1956 * it is much easier and equally valid to count what we queue for the
1957 * worker, even before it actually was queued or send.
1958 * (drbd_make_request_common; recovery path on read io-error)
1960 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1961 * _req_mod(req, data_received)
1962 * [from receive_DataReply]
1963 * _req_mod(req, write_acked_by_peer or recv_acked_by_peer or neg_acked)
1964 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1965 * for some reason it is NOT decreased in got_NegAck,
1966 * but in the resulting cleanup code from report_params.
1967 * we should try to remember the reason for that...
1968 * _req_mod(req, send_failed or send_canceled)
1969 * _req_mod(req, connection_lost_while_pending)
1970 * [from tl_clear_barrier]
1972 static inline void inc_ap_pending(struct drbd_conf *mdev)
1974 atomic_inc(&mdev->ap_pending_cnt);
1977 #define ERR_IF_CNT_IS_NEGATIVE(which) \
1978 if (atomic_read(&mdev->which) < 0) \
1979 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
1980 __func__ , __LINE__ , \
1981 atomic_read(&mdev->which))
1983 #define dec_ap_pending(mdev) do { \
1984 typecheck(struct drbd_conf *, mdev); \
1985 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \
1986 wake_up(&mdev->misc_wait); \
1987 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
1989 /* counts how many resync-related answers we still expect from the peer
1991 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1992 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK whith ID_SYNCER)
1993 * (or P_NEG_ACK with ID_SYNCER)
1995 static inline void inc_rs_pending(struct drbd_conf *mdev)
1997 atomic_inc(&mdev->rs_pending_cnt);
2000 #define dec_rs_pending(mdev) do { \
2001 typecheck(struct drbd_conf *, mdev); \
2002 atomic_dec(&mdev->rs_pending_cnt); \
2003 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
2005 /* counts how many answers we still need to send to the peer.
2007 * receive_Data unless protocol A;
2008 * we need to send a P_RECV_ACK (proto B)
2009 * or P_WRITE_ACK (proto C)
2010 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2011 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2012 * receive_Barrier_* we need to send a P_BARRIER_ACK
2014 static inline void inc_unacked(struct drbd_conf *mdev)
2016 atomic_inc(&mdev->unacked_cnt);
2019 #define dec_unacked(mdev) do { \
2020 typecheck(struct drbd_conf *, mdev); \
2021 atomic_dec(&mdev->unacked_cnt); \
2022 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2024 #define sub_unacked(mdev, n) do { \
2025 typecheck(struct drbd_conf *, mdev); \
2026 atomic_sub(n, &mdev->unacked_cnt); \
2027 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2030 static inline void put_net_conf(struct drbd_conf *mdev)
2032 if (atomic_dec_and_test(&mdev->net_cnt))
2033 wake_up(&mdev->misc_wait);
2037 * get_net_conf() - Increase ref count on mdev->net_conf; Returns 0 if nothing there
2038 * @mdev: DRBD device.
2040 * You have to call put_net_conf() when finished working with mdev->net_conf.
2042 static inline int get_net_conf(struct drbd_conf *mdev)
2046 atomic_inc(&mdev->net_cnt);
2047 have_net_conf = mdev->state.conn >= C_UNCONNECTED;
2050 return have_net_conf;
2054 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2057 * You have to call put_ldev() when finished working with mdev->ldev.
2059 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2060 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2062 static inline void put_ldev(struct drbd_conf *mdev)
2065 if (atomic_dec_and_test(&mdev->local_cnt))
2066 wake_up(&mdev->misc_wait);
2067 D_ASSERT(atomic_read(&mdev->local_cnt) >= 0);
2071 static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
2075 atomic_inc(&mdev->local_cnt);
2076 io_allowed = (mdev->state.disk >= mins);
2082 extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins);
2085 /* you must have an "get_ldev" reference */
2086 static inline void drbd_get_syncer_progress(struct drbd_conf *mdev,
2087 unsigned long *bits_left, unsigned int *per_mil_done)
2090 * this is to break it at compile time when we change that
2091 * (we may feel 4TB maximum storage per drbd is not enough)
2093 typecheck(unsigned long, mdev->rs_total);
2095 /* note: both rs_total and rs_left are in bits, i.e. in
2096 * units of BM_BLOCK_SIZE.
2097 * for the percentage, we don't care. */
2099 *bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
2100 /* >> 10 to prevent overflow,
2101 * +1 to prevent division by zero */
2102 if (*bits_left > mdev->rs_total) {
2103 /* doh. maybe a logic bug somewhere.
2104 * may also be just a race condition
2105 * between this and a disconnect during sync.
2106 * for now, just prevent in-kernel buffer overflow.
2109 dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2110 drbd_conn_str(mdev->state.conn),
2111 *bits_left, mdev->rs_total, mdev->rs_failed);
2114 /* make sure the calculation happens in long context */
2115 unsigned long tmp = 1000UL -
2116 (*bits_left >> 10)*1000UL
2117 / ((mdev->rs_total >> 10) + 1UL);
2118 *per_mil_done = tmp;
2123 /* this throttles on-the-fly application requests
2124 * according to max_buffers settings;
2125 * maybe re-implement using semaphores? */
2126 static inline int drbd_get_max_buffers(struct drbd_conf *mdev)
2128 int mxb = 1000000; /* arbitrary limit on open requests */
2129 if (get_net_conf(mdev)) {
2130 mxb = mdev->net_conf->max_buffers;
2136 static inline int drbd_state_is_stable(union drbd_state s)
2139 /* DO NOT add a default clause, we want the compiler to warn us
2140 * for any newly introduced state we may have forgotten to add here */
2142 switch ((enum drbd_conns)s.conn) {
2143 /* new io only accepted when there is no connection, ... */
2145 case C_WF_CONNECTION:
2146 /* ... or there is a well established connection. */
2152 case C_PAUSED_SYNC_S:
2153 case C_PAUSED_SYNC_T:
2154 /* maybe stable, look at the disk state */
2157 /* no new io accepted during tansitional states
2158 * like handshake or teardown */
2159 case C_DISCONNECTING:
2163 case C_NETWORK_FAILURE:
2164 case C_PROTOCOL_ERROR:
2166 case C_WF_REPORT_PARAMS:
2167 case C_STARTING_SYNC_S:
2168 case C_STARTING_SYNC_T:
2171 case C_WF_SYNC_UUID:
2177 switch ((enum drbd_disk_state)s.disk) {
2179 case D_INCONSISTENT:
2183 /* disk state is stable as well. */
2186 /* no new io accepted during tansitional states */
2199 static inline int __inc_ap_bio_cond(struct drbd_conf *mdev)
2201 int mxb = drbd_get_max_buffers(mdev);
2203 if (mdev->state.susp)
2205 if (test_bit(SUSPEND_IO, &mdev->flags))
2208 /* to avoid potential deadlock or bitmap corruption,
2209 * in various places, we only allow new application io
2210 * to start during "stable" states. */
2212 /* no new io accepted when attaching or detaching the disk */
2213 if (!drbd_state_is_stable(mdev->state))
2216 /* since some older kernels don't have atomic_add_unless,
2217 * and we are within the spinlock anyways, we have this workaround. */
2218 if (atomic_read(&mdev->ap_bio_cnt) > mxb)
2220 if (test_bit(BITMAP_IO, &mdev->flags))
2222 if (atomic_read(&mdev->new_c_uuid))
2227 /* I'd like to use wait_event_lock_irq,
2228 * but I'm not sure when it got introduced,
2229 * and not sure when it has 3 or 4 arguments */
2230 static inline void inc_ap_bio(struct drbd_conf *mdev, int count)
2232 /* compare with after_state_ch,
2233 * os.conn != C_WF_BITMAP_S && ns.conn == C_WF_BITMAP_S */
2237 * as long as the device is suspended
2238 * until the bitmap is no longer on the fly during connection
2239 * handshake as long as we would exeed the max_buffer limit.
2241 * to avoid races with the reconnect code,
2242 * we need to atomic_inc within the spinlock. */
2244 if (atomic_read(&mdev->new_c_uuid) && atomic_add_unless(&mdev->new_c_uuid, -1, 1))
2245 drbd_queue_work_front(&mdev->data.work, &mdev->uuid_work);
2247 spin_lock_irq(&mdev->req_lock);
2248 while (!__inc_ap_bio_cond(mdev)) {
2249 prepare_to_wait(&mdev->misc_wait, &wait, TASK_UNINTERRUPTIBLE);
2250 spin_unlock_irq(&mdev->req_lock);
2252 finish_wait(&mdev->misc_wait, &wait);
2253 spin_lock_irq(&mdev->req_lock);
2255 atomic_add(count, &mdev->ap_bio_cnt);
2256 spin_unlock_irq(&mdev->req_lock);
2259 static inline void dec_ap_bio(struct drbd_conf *mdev)
2261 int mxb = drbd_get_max_buffers(mdev);
2262 int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt);
2264 D_ASSERT(ap_bio >= 0);
2265 /* this currently does wake_up for every dec_ap_bio!
2266 * maybe rather introduce some type of hysteresis?
2267 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2269 wake_up(&mdev->misc_wait);
2270 if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) {
2271 if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags))
2272 drbd_queue_work(&mdev->data.work, &mdev->bm_io_work.w);
2276 static inline void drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val)
2278 mdev->ed_uuid = val;
2281 static inline int seq_cmp(u32 a, u32 b)
2283 /* we assume wrap around at 32bit.
2284 * for wrap around at 24bit (old atomic_t),
2288 return (s32)(a) - (s32)(b);
2290 #define seq_lt(a, b) (seq_cmp((a), (b)) < 0)
2291 #define seq_gt(a, b) (seq_cmp((a), (b)) > 0)
2292 #define seq_ge(a, b) (seq_cmp((a), (b)) >= 0)
2293 #define seq_le(a, b) (seq_cmp((a), (b)) <= 0)
2294 /* CAUTION: please no side effects in arguments! */
2295 #define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b)))
2297 static inline void update_peer_seq(struct drbd_conf *mdev, unsigned int new_seq)
2300 spin_lock(&mdev->peer_seq_lock);
2301 m = seq_max(mdev->peer_seq, new_seq);
2303 spin_unlock(&mdev->peer_seq_lock);
2305 wake_up(&mdev->seq_wait);
2308 static inline void drbd_update_congested(struct drbd_conf *mdev)
2310 struct sock *sk = mdev->data.socket->sk;
2311 if (sk->sk_wmem_queued > sk->sk_sndbuf * 4 / 5)
2312 set_bit(NET_CONGESTED, &mdev->flags);
2315 static inline int drbd_queue_order_type(struct drbd_conf *mdev)
2317 /* sorry, we currently have no working implementation
2318 * of distributed TCQ stuff */
2319 #ifndef QUEUE_ORDERED_NONE
2320 #define QUEUE_ORDERED_NONE 0
2322 return QUEUE_ORDERED_NONE;
2325 static inline void drbd_blk_run_queue(struct request_queue *q)
2327 if (q && q->unplug_fn)
2331 static inline void drbd_kick_lo(struct drbd_conf *mdev)
2333 if (get_ldev(mdev)) {
2334 drbd_blk_run_queue(bdev_get_queue(mdev->ldev->backing_bdev));
2339 static inline void drbd_md_flush(struct drbd_conf *mdev)
2343 if (test_bit(MD_NO_BARRIER, &mdev->flags))
2346 r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL,
2349 set_bit(MD_NO_BARRIER, &mdev->flags);
2350 dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r);