4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
6 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
7 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
8 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
10 drbd is free software; you can redistribute it and/or modify
11 it under the terms of the GNU General Public License as published by
12 the Free Software Foundation; either version 2, or (at your option)
15 drbd is distributed in the hope that it will be useful,
16 but WITHOUT ANY WARRANTY; without even the implied warranty of
17 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 GNU General Public License for more details.
20 You should have received a copy of the GNU General Public License
21 along with drbd; see the file COPYING. If not, write to
22 the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
29 #include <linux/compiler.h>
30 #include <linux/types.h>
31 #include <linux/version.h>
32 #include <linux/list.h>
33 #include <linux/sched.h>
34 #include <linux/bitops.h>
35 #include <linux/slab.h>
36 #include <linux/crypto.h>
37 #include <linux/ratelimit.h>
38 #include <linux/tcp.h>
39 #include <linux/mutex.h>
40 #include <linux/major.h>
41 #include <linux/blkdev.h>
42 #include <linux/genhd.h>
44 #include <linux/lru_cache.h>
47 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
48 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
49 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
50 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
52 # define __protected_by(x)
53 # define __protected_read_by(x)
54 # define __protected_write_by(x)
55 # define __must_hold(x)
58 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
60 /* module parameter, defined in drbd_main.c */
61 extern unsigned int minor_count
;
62 extern int disable_sendpage
;
64 extern unsigned int cn_idx
;
66 #ifdef CONFIG_DRBD_FAULT_INJECTION
67 extern int enable_faults
;
68 extern int fault_rate
;
69 extern int fault_devs
;
72 extern char usermode_helper
[];
82 /* I don't remember why XCPU ...
83 * This is used to wake the asender,
84 * and to interrupt sending the sending task
87 #define DRBD_SIG SIGXCPU
89 /* This is used to stop/restart our threads.
90 * Cannot use SIGTERM nor SIGKILL, since these
91 * are sent out by init on runlevel changes
92 * I choose SIGHUP for now.
94 #define DRBD_SIGKILL SIGHUP
96 /* All EEs on the free list should have ID_VACANT (== 0)
97 * freshly allocated EEs get !ID_VACANT (== 1)
98 * so if it says "cannot dereference null pointer at address 0x00000001",
99 * it is most likely one of these :( */
101 #define ID_IN_SYNC (4711ULL)
102 #define ID_OUT_OF_SYNC (4712ULL)
104 #define ID_SYNCER (-1ULL)
106 #define is_syncer_block_id(id) ((id) == ID_SYNCER)
111 /* to shorten dev_warn(DEV, "msg"); and relatives statements */
112 #define DEV (disk_to_dev(mdev->vdisk))
114 #define D_ASSERT(exp) if (!(exp)) \
115 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
117 #define ERR_IF(exp) if (({ \
118 int _b = (exp) != 0; \
119 if (_b) dev_err(DEV, "ASSERT FAILED: %s: (%s) in %s:%d\n", \
120 __func__, #exp, __FILE__, __LINE__); \
124 /* Defines to control fault insertion */
126 DRBD_FAULT_MD_WR
= 0, /* meta data write */
127 DRBD_FAULT_MD_RD
= 1, /* read */
128 DRBD_FAULT_RS_WR
= 2, /* resync */
129 DRBD_FAULT_RS_RD
= 3,
130 DRBD_FAULT_DT_WR
= 4, /* data */
131 DRBD_FAULT_DT_RD
= 5,
132 DRBD_FAULT_DT_RA
= 6, /* data read ahead */
133 DRBD_FAULT_BM_ALLOC
= 7, /* bitmap allocation */
134 DRBD_FAULT_AL_EE
= 8, /* alloc ee */
135 DRBD_FAULT_RECEIVE
= 9, /* Changes some bytes upon receiving a [rs]data block */
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 */
215 P_OUT_OF_SYNC
= 0x28, /* Mark as out of sync (Outrunning), data socket */
218 P_MAY_IGNORE
= 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
219 P_MAX_OPT_CMD
= 0x101,
221 /* special command ids for handshake */
223 P_HAND_SHAKE_M
= 0xfff1, /* First Packet on the MetaSock */
224 P_HAND_SHAKE_S
= 0xfff2, /* First Packet on the Socket */
226 P_HAND_SHAKE
= 0xfffe /* FIXED for the next century! */
229 static inline const char *cmdname(enum drbd_packets cmd
)
231 /* THINK may need to become several global tables
232 * when we want to support more than
234 static const char *cmdnames
[] = {
236 [P_DATA_REPLY
] = "DataReply",
237 [P_RS_DATA_REPLY
] = "RSDataReply",
238 [P_BARRIER
] = "Barrier",
239 [P_BITMAP
] = "ReportBitMap",
240 [P_BECOME_SYNC_TARGET
] = "BecomeSyncTarget",
241 [P_BECOME_SYNC_SOURCE
] = "BecomeSyncSource",
242 [P_UNPLUG_REMOTE
] = "UnplugRemote",
243 [P_DATA_REQUEST
] = "DataRequest",
244 [P_RS_DATA_REQUEST
] = "RSDataRequest",
245 [P_SYNC_PARAM
] = "SyncParam",
246 [P_SYNC_PARAM89
] = "SyncParam89",
247 [P_PROTOCOL
] = "ReportProtocol",
248 [P_UUIDS
] = "ReportUUIDs",
249 [P_SIZES
] = "ReportSizes",
250 [P_STATE
] = "ReportState",
251 [P_SYNC_UUID
] = "ReportSyncUUID",
252 [P_AUTH_CHALLENGE
] = "AuthChallenge",
253 [P_AUTH_RESPONSE
] = "AuthResponse",
255 [P_PING_ACK
] = "PingAck",
256 [P_RECV_ACK
] = "RecvAck",
257 [P_WRITE_ACK
] = "WriteAck",
258 [P_RS_WRITE_ACK
] = "RSWriteAck",
259 [P_DISCARD_ACK
] = "DiscardAck",
260 [P_NEG_ACK
] = "NegAck",
261 [P_NEG_DREPLY
] = "NegDReply",
262 [P_NEG_RS_DREPLY
] = "NegRSDReply",
263 [P_BARRIER_ACK
] = "BarrierAck",
264 [P_STATE_CHG_REQ
] = "StateChgRequest",
265 [P_STATE_CHG_REPLY
] = "StateChgReply",
266 [P_OV_REQUEST
] = "OVRequest",
267 [P_OV_REPLY
] = "OVReply",
268 [P_OV_RESULT
] = "OVResult",
269 [P_CSUM_RS_REQUEST
] = "CsumRSRequest",
270 [P_RS_IS_IN_SYNC
] = "CsumRSIsInSync",
271 [P_COMPRESSED_BITMAP
] = "CBitmap",
272 [P_DELAY_PROBE
] = "DelayProbe",
273 [P_OUT_OF_SYNC
] = "OutOfSync",
277 if (cmd
== P_HAND_SHAKE_M
)
279 if (cmd
== P_HAND_SHAKE_S
)
281 if (cmd
== P_HAND_SHAKE
)
283 if (cmd
>= P_MAX_CMD
)
285 return cmdnames
[cmd
];
288 /* for sending/receiving the bitmap,
289 * possibly in some encoding scheme */
292 * stores total bits and long words
293 * of the bitmap, so we don't need to
294 * call the accessor functions over and again. */
295 unsigned long bm_bits
;
296 unsigned long bm_words
;
297 /* during xfer, current position within the bitmap */
298 unsigned long bit_offset
;
299 unsigned long word_offset
;
301 /* statistics; index: (h->command == P_BITMAP) */
306 extern void INFO_bm_xfer_stats(struct drbd_conf
*mdev
,
307 const char *direction
, struct bm_xfer_ctx
*c
);
309 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx
*c
)
311 /* word_offset counts "native long words" (32 or 64 bit),
313 * Encoded packet may end at an unaligned bit offset.
314 * In case a fallback clear text packet is transmitted in
315 * between, we adjust this offset back to the last 64bit
316 * aligned "native long word", which makes coding and decoding
317 * the plain text bitmap much more convenient. */
318 #if BITS_PER_LONG == 64
319 c
->word_offset
= c
->bit_offset
>> 6;
320 #elif BITS_PER_LONG == 32
321 c
->word_offset
= c
->bit_offset
>> 5;
322 c
->word_offset
&= ~(1UL);
324 # error "unsupported BITS_PER_LONG"
329 #define __packed __attribute__((packed))
332 /* This is the layout for a packet on the wire.
333 * The byteorder is the network byte order.
334 * (except block_id and barrier fields.
335 * these are pointers to local structs
336 * and have no relevance for the partner,
337 * which just echoes them as received.)
339 * NOTE that the payload starts at a long aligned offset,
340 * regardless of 32 or 64 bit arch!
345 u16 length
; /* bytes of data after this header */
349 /* Header for big packets, Used for data packets exceeding 64kB */
351 u16 magic
; /* use DRBD_MAGIC_BIG here */
353 u32 length
; /* Use only 24 bits of that. Ignore the highest 8 bit. */
358 struct p_header80 h80
;
359 struct p_header95 h95
;
363 * short commands, packets without payload, plain p_header:
366 * P_BECOME_SYNC_TARGET
367 * P_BECOME_SYNC_SOURCE
372 * commands with out-of-struct payload:
373 * P_BITMAP (no additional fields)
374 * P_DATA, P_DATA_REPLY (see p_data)
375 * P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
378 /* these defines must not be changed without changing the protocol version */
379 #define DP_HARDBARRIER 1 /* depricated */
380 #define DP_RW_SYNC 2 /* equals REQ_SYNC */
381 #define DP_MAY_SET_IN_SYNC 4
382 #define DP_UNPLUG 8 /* not used anymore */
383 #define DP_FUA 16 /* equals REQ_FUA */
384 #define DP_FLUSH 32 /* equals REQ_FLUSH */
385 #define DP_DISCARD 64 /* equals REQ_DISCARD */
389 u64 sector
; /* 64 bits sector number */
390 u64 block_id
; /* to identify the request in protocol B&C */
396 * commands which share a struct:
398 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
399 * P_DISCARD_ACK (proto C, two-primaries conflict detection)
401 * P_DATA_REQUEST, P_RS_DATA_REQUEST
404 struct p_header80 head
;
413 struct p_header80 head
;
417 u32 pad
; /* to multiple of 8 Byte */
421 * commands with their own struct for additional fields:
430 struct p_header80 head
; /* 8 bytes */
435 /* should be more than enough for future enhancements
436 * for now, feature_flags and the reserverd array shall be zero.
442 /* 80 bytes, FIXED for the next century */
445 struct p_header80 head
;
446 u32 barrier
; /* barrier number _handle_ only */
447 u32 pad
; /* to multiple of 8 Byte */
450 struct p_barrier_ack
{
451 struct p_header80 head
;
457 struct p_header80 head
;
460 /* Since protocol version 88 and higher. */
464 struct p_rs_param_89
{
465 struct p_header80 head
;
467 /* protocol version 89: */
468 char verify_alg
[SHARED_SECRET_MAX
];
469 char csums_alg
[SHARED_SECRET_MAX
];
472 struct p_rs_param_95
{
473 struct p_header80 head
;
475 char verify_alg
[SHARED_SECRET_MAX
];
476 char csums_alg
[SHARED_SECRET_MAX
];
483 enum drbd_conn_flags
{
489 struct p_header80 head
;
497 /* Since protocol version 87 and higher. */
498 char integrity_alg
[0];
503 struct p_header80 head
;
504 u64 uuid
[UI_EXTENDED_SIZE
];
508 struct p_header80 head
;
513 struct p_header80 head
;
514 u64 d_size
; /* size of disk */
515 u64 u_size
; /* user requested size */
516 u64 c_size
; /* current exported size */
517 u32 max_bio_size
; /* Maximal size of a BIO */
518 u16 queue_order_type
; /* not yet implemented in DRBD*/
519 u16 dds_flags
; /* use enum dds_flags here. */
523 struct p_header80 head
;
528 struct p_header80 head
;
533 struct p_req_state_reply
{
534 struct p_header80 head
;
538 struct p_drbd06_param
{
549 struct p_header80 head
;
555 struct p_block_desc
{
556 struct p_header80 head
;
559 u32 pad
; /* to multiple of 8 Byte */
562 /* Valid values for the encoding field.
563 * Bump proto version when changing this. */
564 enum drbd_bitmap_code
{
565 /* RLE_VLI_Bytes = 0,
566 * and other bit variants had been defined during
567 * algorithm evaluation. */
571 struct p_compressed_bm
{
572 struct p_header80 head
;
573 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
574 * (encoding & 0x80): polarity (set/unset) of first runlength
575 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
576 * used to pad up to head.length bytes
583 struct p_delay_probe93
{
584 struct p_header80 head
;
585 u32 seq_num
; /* sequence number to match the two probe packets */
586 u32 offset
; /* usecs the probe got sent after the reference time point */
589 /* DCBP: Drbd Compressed Bitmap Packet ... */
590 static inline enum drbd_bitmap_code
591 DCBP_get_code(struct p_compressed_bm
*p
)
593 return (enum drbd_bitmap_code
)(p
->encoding
& 0x0f);
597 DCBP_set_code(struct p_compressed_bm
*p
, enum drbd_bitmap_code code
)
600 p
->encoding
= (p
->encoding
& ~0xf) | code
;
604 DCBP_get_start(struct p_compressed_bm
*p
)
606 return (p
->encoding
& 0x80) != 0;
610 DCBP_set_start(struct p_compressed_bm
*p
, int set
)
612 p
->encoding
= (p
->encoding
& ~0x80) | (set
? 0x80 : 0);
616 DCBP_get_pad_bits(struct p_compressed_bm
*p
)
618 return (p
->encoding
>> 4) & 0x7;
622 DCBP_set_pad_bits(struct p_compressed_bm
*p
, int n
)
625 p
->encoding
= (p
->encoding
& (~0x7 << 4)) | (n
<< 4);
628 /* one bitmap packet, including the p_header,
629 * should fit within one _architecture independend_ page.
630 * so we need to use the fixed size 4KiB page size
631 * most architechtures have used for a long time.
633 #define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header80))
634 #define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
635 #define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm))
636 #if (PAGE_SIZE < 4096)
637 /* drbd_send_bitmap / receive_bitmap would break horribly */
638 #error "PAGE_SIZE too small"
642 union p_header header
;
643 struct p_handshake handshake
;
645 struct p_block_ack block_ack
;
646 struct p_barrier barrier
;
647 struct p_barrier_ack barrier_ack
;
648 struct p_rs_param_89 rs_param_89
;
649 struct p_rs_param_95 rs_param_95
;
650 struct p_protocol protocol
;
651 struct p_sizes sizes
;
652 struct p_uuids uuids
;
653 struct p_state state
;
654 struct p_req_state req_state
;
655 struct p_req_state_reply req_state_reply
;
656 struct p_block_req block_req
;
657 struct p_delay_probe93 delay_probe93
;
658 struct p_rs_uuid rs_uuid
;
659 struct p_block_desc block_desc
;
662 /**********************************************************************/
663 enum drbd_thread_state
{
672 struct task_struct
*task
;
673 struct completion stop
;
674 enum drbd_thread_state t_state
;
675 int (*function
) (struct drbd_thread
*);
676 struct drbd_conf
*mdev
;
680 static inline enum drbd_thread_state
get_t_state(struct drbd_thread
*thi
)
682 /* THINK testing the t_state seems to be uncritical in all cases
683 * (but thread_{start,stop}), so we can read it *without* the lock.
692 * Having this as the first member of a struct provides sort of "inheritance".
693 * "derived" structs can be "drbd_queue_work()"ed.
694 * The callback should know and cast back to the descendant struct.
695 * drbd_request and drbd_epoch_entry are descendants of drbd_work.
698 typedef int (*drbd_work_cb
)(struct drbd_conf
*, struct drbd_work
*, int cancel
);
700 struct list_head list
;
704 struct drbd_tl_epoch
;
705 struct drbd_request
{
707 struct drbd_conf
*mdev
;
709 /* if local IO is not allowed, will be NULL.
710 * if local IO _is_ allowed, holds the locally submitted bio clone,
711 * or, after local IO completion, the ERR_PTR(error).
712 * see drbd_endio_pri(). */
713 struct bio
*private_bio
;
715 struct hlist_node colision
;
718 unsigned int epoch
; /* barrier_nr */
720 /* barrier_nr: used to check on "completion" whether this req was in
721 * the current epoch, and we therefore have to close it,
722 * starting a new epoch...
725 /* up to here, the struct layout is identical to drbd_epoch_entry;
726 * we might be able to use that to our advantage... */
728 struct list_head tl_requests
; /* ring list in the transfer log */
729 struct bio
*master_bio
; /* master bio pointer */
730 unsigned long rq_state
; /* see comments above _req_mod() */
732 unsigned long start_time
;
735 struct drbd_tl_epoch
{
737 struct list_head requests
; /* requests before */
738 struct drbd_tl_epoch
*next
; /* pointer to the next barrier */
739 unsigned int br_number
; /* the barriers identifier. */
740 int n_writes
; /* number of requests attached before this barrier */
745 /* These Tl_epoch_entries may be in one of 6 lists:
746 active_ee .. data packet being written
747 sync_ee .. syncer block being written
748 done_ee .. block written, need to send P_WRITE_ACK
749 read_ee .. [RS]P_DATA_REQUEST being read
753 struct list_head list
;
754 unsigned int barrier_nr
;
755 atomic_t epoch_size
; /* increased on every request added. */
756 atomic_t active
; /* increased on every req. added, and dec on every finished. */
760 /* drbd_epoch flag bits */
762 DE_HAVE_BARRIER_NUMBER
,
769 EV_CLEANUP
= 32, /* used as flag */
772 struct drbd_wq_barrier
{
774 struct completion done
;
782 struct drbd_epoch_entry
{
784 struct hlist_node colision
;
785 struct drbd_epoch
*epoch
; /* for writes */
786 struct drbd_conf
*mdev
;
788 atomic_t pending_bios
;
790 /* see comments on ee flag bits below */
795 struct digest_info
*digest
;
800 * While corresponding bios are in flight, the only modification will be
801 * set_bit WAS_ERROR, which has to be atomic.
802 * If no bios are in flight yet, or all have been completed,
803 * non-atomic modification to ee->flags is ok.
806 __EE_CALL_AL_COMPLETE_IO
,
807 __EE_MAY_SET_IN_SYNC
,
809 /* In case a barrier failed,
810 * we need to resubmit without the barrier flag. */
813 /* we may have several bios per epoch entry.
814 * if any of those fail, we set this flag atomically
815 * from the endio callback */
818 /* This ee has a pointer to a digest instead of a block id */
821 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
822 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
823 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
824 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
825 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
827 /* global flag bits */
829 CREATE_BARRIER
, /* next P_DATA is preceeded by a P_BARRIER */
830 SIGNAL_ASENDER
, /* whether asender wants to be interrupted */
831 SEND_PING
, /* whether asender should send a ping asap */
833 UNPLUG_QUEUED
, /* only relevant with kernel 2.4 */
834 UNPLUG_REMOTE
, /* sending a "UnplugRemote" could help */
835 MD_DIRTY
, /* current uuids and flags not yet on disk */
836 DISCARD_CONCURRENT
, /* Set on one node, cleared on the peer! */
837 USE_DEGR_WFC_T
, /* degr-wfc-timeout instead of wfc-timeout. */
838 CLUSTER_ST_CHANGE
, /* Cluster wide state change going on... */
841 CRASHED_PRIMARY
, /* This node was a crashed primary.
842 * Gets cleared when the state.conn
843 * goes into C_CONNECTED state. */
844 WRITE_BM_AFTER_RESYNC
, /* A kmalloc() during resync failed */
847 MD_NO_FUA
, /* Users wants us to not use FUA/FLUSH on meta data dev */
848 SUSPEND_IO
, /* suspend application io */
849 BITMAP_IO
, /* suspend application io;
850 once no more io in flight, start bitmap io */
851 BITMAP_IO_QUEUED
, /* Started bitmap IO */
852 GO_DISKLESS
, /* Disk is being detached, on io-error or admin request. */
853 WAS_IO_ERROR
, /* Local disk failed returned IO error */
854 RESYNC_AFTER_NEG
, /* Resync after online grow after the attach&negotiate finished. */
855 NET_CONGESTED
, /* The data socket is congested */
857 CONFIG_PENDING
, /* serialization of (re)configuration requests.
858 * if set, also prevents the device from dying */
859 DEVICE_DYING
, /* device became unconfigured,
860 * but worker thread is still handling the cleanup.
861 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
862 * while this is set. */
863 RESIZE_PENDING
, /* Size change detected locally, waiting for the response from
864 * the peer, if it changed there as well. */
865 CONN_DRY_RUN
, /* Expect disconnect after resync handshake. */
866 GOT_PING_ACK
, /* set when we receive a ping_ack packet, misc wait gets woken */
867 NEW_CUR_UUID
, /* Create new current UUID when thawing IO */
868 AL_SUSPENDED
, /* Activity logging is currently suspended. */
871 struct drbd_bitmap
; /* opaque for drbd_conf */
873 /* TODO sort members for performance
874 * MAYBE group them further */
876 /* THINK maybe we actually want to use the default "event/%s" worker threads
877 * or similar in linux 2.6, which uses per cpu data and threads.
879 struct drbd_work_queue
{
881 struct semaphore s
; /* producers up it, worker down()s it */
882 spinlock_t q_lock
; /* to protect the list. */
886 struct drbd_work_queue work
;
888 struct socket
*socket
;
889 /* this way we get our
890 * send/receive buffers off the stack */
891 union p_polymorph sbuf
;
892 union p_polymorph rbuf
;
896 u64 md_offset
; /* sector offset to 'super' block */
898 u64 la_size_sect
; /* last agreed size, unit sectors */
904 s32 al_offset
; /* signed relative sector offset to al area */
905 s32 bm_offset
; /* signed relative sector offset to bitmap */
907 /* u32 al_nr_extents; important for restoring the AL
908 * is stored into sync_conf.al_extents, which in turn
909 * gets applied to act_log->nr_elements
913 /* for sync_conf and other types... */
914 #define NL_PACKET(name, number, fields) struct name { fields };
915 #define NL_INTEGER(pn,pr,member) int member;
916 #define NL_INT64(pn,pr,member) __u64 member;
917 #define NL_BIT(pn,pr,member) unsigned member:1;
918 #define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
919 #include "linux/drbd_nl.h"
921 struct drbd_backing_dev
{
922 struct block_device
*backing_bdev
;
923 struct block_device
*md_bdev
;
925 struct disk_conf dc
; /* The user provided config... */
926 sector_t known_size
; /* last known size of that backing device */
930 struct drbd_conf
*mdev
;
931 struct completion event
;
938 int (*io_fn
)(struct drbd_conf
*mdev
);
939 void (*done
)(struct drbd_conf
*mdev
, int rv
);
942 enum write_ordering_e
{
950 unsigned int head_index
;
955 /* things that are stored as / read from meta data on disk */
958 /* configured by drbdsetup */
959 struct net_conf
*net_conf
; /* protected by get_net_conf() and put_net_conf() */
960 struct syncer_conf sync_conf
;
961 struct drbd_backing_dev
*ldev
__protected_by(local
);
963 sector_t p_size
; /* partner's disk size */
964 struct request_queue
*rq_queue
;
965 struct block_device
*this_bdev
;
966 struct gendisk
*vdisk
;
968 struct drbd_socket data
; /* data/barrier/cstate/parameter packets */
969 struct drbd_socket meta
; /* ping/ack (metadata) packets */
970 int agreed_pro_version
; /* actually used protocol version */
971 unsigned long last_received
; /* in jiffies, either socket */
972 unsigned int ko_count
;
973 struct drbd_work resync_work
,
978 struct timer_list resync_timer
;
979 struct timer_list md_sync_timer
;
980 #ifdef DRBD_DEBUG_MD_SYNC
984 } last_md_mark_dirty
;
987 /* Used after attach while negotiating new disk state. */
988 union drbd_state new_state_tmp
;
990 union drbd_state state
;
991 wait_queue_head_t misc_wait
;
992 wait_queue_head_t state_wait
; /* upon each state change. */
993 wait_queue_head_t net_cnt_wait
;
994 unsigned int send_cnt
;
995 unsigned int recv_cnt
;
996 unsigned int read_cnt
;
997 unsigned int writ_cnt
;
998 unsigned int al_writ_cnt
;
999 unsigned int bm_writ_cnt
;
1000 atomic_t ap_bio_cnt
; /* Requests we need to complete */
1001 atomic_t ap_pending_cnt
; /* AP data packets on the wire, ack expected */
1002 atomic_t rs_pending_cnt
; /* RS request/data packets on the wire */
1003 atomic_t unacked_cnt
; /* Need to send replys for */
1004 atomic_t local_cnt
; /* Waiting for local completion */
1005 atomic_t net_cnt
; /* Users of net_conf */
1006 spinlock_t req_lock
;
1007 struct drbd_tl_epoch
*unused_spare_tle
; /* for pre-allocation */
1008 struct drbd_tl_epoch
*newest_tle
;
1009 struct drbd_tl_epoch
*oldest_tle
;
1010 struct list_head out_of_sequence_requests
;
1011 struct hlist_head
*tl_hash
;
1012 unsigned int tl_hash_s
;
1014 /* blocks to sync in this run [unit BM_BLOCK_SIZE] */
1015 unsigned long rs_total
;
1016 /* number of sync IOs that failed in this run */
1017 unsigned long rs_failed
;
1018 /* Syncer's start time [unit jiffies] */
1019 unsigned long rs_start
;
1020 /* cumulated time in PausedSyncX state [unit jiffies] */
1021 unsigned long rs_paused
;
1022 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
1023 unsigned long rs_same_csum
;
1024 #define DRBD_SYNC_MARKS 8
1025 #define DRBD_SYNC_MARK_STEP (3*HZ)
1026 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
1027 unsigned long rs_mark_left
[DRBD_SYNC_MARKS
];
1028 /* marks's time [unit jiffies] */
1029 unsigned long rs_mark_time
[DRBD_SYNC_MARKS
];
1030 /* current index into rs_mark_{left,time} */
1033 /* where does the admin want us to start? (sector) */
1034 sector_t ov_start_sector
;
1035 /* where are we now? (sector) */
1036 sector_t ov_position
;
1037 /* Start sector of out of sync range (to merge printk reporting). */
1038 sector_t ov_last_oos_start
;
1039 /* size of out-of-sync range in sectors. */
1040 sector_t ov_last_oos_size
;
1041 unsigned long ov_left
; /* in bits */
1042 struct crypto_hash
*csums_tfm
;
1043 struct crypto_hash
*verify_tfm
;
1045 struct drbd_thread receiver
;
1046 struct drbd_thread worker
;
1047 struct drbd_thread asender
;
1048 struct drbd_bitmap
*bitmap
;
1049 unsigned long bm_resync_fo
; /* bit offset for drbd_bm_find_next */
1051 /* Used to track operations of resync... */
1052 struct lru_cache
*resync
;
1053 /* Number of locked elements in resync LRU */
1054 unsigned int resync_locked
;
1055 /* resync extent number waiting for application requests */
1056 unsigned int resync_wenr
;
1060 struct drbd_epoch
*current_epoch
;
1061 spinlock_t epoch_lock
;
1062 unsigned int epochs
;
1063 enum write_ordering_e write_ordering
;
1064 struct list_head active_ee
; /* IO in progress (P_DATA gets written to disk) */
1065 struct list_head sync_ee
; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
1066 struct list_head done_ee
; /* send ack */
1067 struct list_head read_ee
; /* IO in progress (any read) */
1068 struct list_head net_ee
; /* zero-copy network send in progress */
1069 struct hlist_head
*ee_hash
; /* is proteced by req_lock! */
1070 unsigned int ee_hash_s
;
1072 /* this one is protected by ee_lock, single thread */
1073 struct drbd_epoch_entry
*last_write_w_barrier
;
1075 int next_barrier_nr
;
1076 struct hlist_head
*app_reads_hash
; /* is proteced by req_lock */
1077 struct list_head resync_reads
;
1078 atomic_t pp_in_use
; /* allocated from page pool */
1079 atomic_t pp_in_use_by_net
; /* sendpage()d, still referenced by tcp */
1080 wait_queue_head_t ee_wait
;
1081 struct page
*md_io_page
; /* one page buffer for md_io */
1082 struct page
*md_io_tmpp
; /* for logical_block_size != 512 */
1083 struct mutex md_io_mutex
; /* protects the md_io_buffer */
1085 wait_queue_head_t al_wait
;
1086 struct lru_cache
*act_log
; /* activity log */
1087 unsigned int al_tr_number
;
1089 int al_tr_pos
; /* position of the next transaction in the journal */
1090 struct crypto_hash
*cram_hmac_tfm
;
1091 struct crypto_hash
*integrity_w_tfm
; /* to be used by the worker thread */
1092 struct crypto_hash
*integrity_r_tfm
; /* to be used by the receiver thread */
1096 wait_queue_head_t seq_wait
;
1097 atomic_t packet_seq
;
1098 unsigned int peer_seq
;
1099 spinlock_t peer_seq_lock
;
1101 unsigned long comm_bm_set
; /* communicated number of set bits. */
1102 cpumask_var_t cpu_mask
;
1103 struct bm_io_work bm_io_work
;
1104 u64 ed_uuid
; /* UUID of the exposed data */
1105 struct mutex state_mutex
;
1106 char congestion_reason
; /* Why we where congested... */
1107 atomic_t rs_sect_in
; /* for incoming resync data rate, SyncTarget */
1108 atomic_t rs_sect_ev
; /* for submitted resync data rate, both */
1109 int rs_last_sect_ev
; /* counter to compare with */
1110 int rs_last_events
; /* counter of read or write "events" (unit sectors)
1111 * on the lower level device when we last looked. */
1112 int c_sync_rate
; /* current resync rate after syncer throttle magic */
1113 struct fifo_buffer rs_plan_s
; /* correction values of resync planer */
1114 int rs_in_flight
; /* resync sectors in flight (to proxy, in proxy and from proxy) */
1115 int rs_planed
; /* resync sectors already planed */
1116 atomic_t ap_in_flight
; /* App sectors in flight (waiting for ack) */
1119 static inline struct drbd_conf
*minor_to_mdev(unsigned int minor
)
1121 struct drbd_conf
*mdev
;
1123 mdev
= minor
< minor_count
? minor_table
[minor
] : NULL
;
1128 static inline unsigned int mdev_to_minor(struct drbd_conf
*mdev
)
1133 /* returns 1 if it was successfull,
1134 * returns 0 if there was no data socket.
1135 * so wherever you are going to use the data.socket, e.g. do
1136 * if (!drbd_get_data_sock(mdev))
1139 * drbd_put_data_sock(mdev);
1141 static inline int drbd_get_data_sock(struct drbd_conf
*mdev
)
1143 mutex_lock(&mdev
->data
.mutex
);
1144 /* drbd_disconnect() could have called drbd_free_sock()
1145 * while we were waiting in down()... */
1146 if (unlikely(mdev
->data
.socket
== NULL
)) {
1147 mutex_unlock(&mdev
->data
.mutex
);
1153 static inline void drbd_put_data_sock(struct drbd_conf
*mdev
)
1155 mutex_unlock(&mdev
->data
.mutex
);
1159 * function declarations
1160 *************************/
1164 enum chg_state_flags
{
1167 CS_WAIT_COMPLETE
= 4,
1169 CS_ORDERED
= CS_WAIT_COMPLETE
+ CS_SERIALIZE
,
1174 DDSF_NO_RESYNC
= 2, /* Do not run a resync for the new space */
1177 extern void drbd_init_set_defaults(struct drbd_conf
*mdev
);
1178 extern int drbd_change_state(struct drbd_conf
*mdev
, enum chg_state_flags f
,
1179 union drbd_state mask
, union drbd_state val
);
1180 extern void drbd_force_state(struct drbd_conf
*, union drbd_state
,
1182 extern int _drbd_request_state(struct drbd_conf
*, union drbd_state
,
1183 union drbd_state
, enum chg_state_flags
);
1184 extern int __drbd_set_state(struct drbd_conf
*, union drbd_state
,
1185 enum chg_state_flags
, struct completion
*done
);
1186 extern void print_st_err(struct drbd_conf
*, union drbd_state
,
1187 union drbd_state
, int);
1188 extern int drbd_thread_start(struct drbd_thread
*thi
);
1189 extern void _drbd_thread_stop(struct drbd_thread
*thi
, int restart
, int wait
);
1191 extern void drbd_thread_current_set_cpu(struct drbd_conf
*mdev
);
1192 extern void drbd_calc_cpu_mask(struct drbd_conf
*mdev
);
1194 #define drbd_thread_current_set_cpu(A) ({})
1195 #define drbd_calc_cpu_mask(A) ({})
1197 extern void drbd_free_resources(struct drbd_conf
*mdev
);
1198 extern void tl_release(struct drbd_conf
*mdev
, unsigned int barrier_nr
,
1199 unsigned int set_size
);
1200 extern void tl_clear(struct drbd_conf
*mdev
);
1201 enum drbd_req_event
;
1202 extern void tl_restart(struct drbd_conf
*mdev
, enum drbd_req_event what
);
1203 extern void _tl_add_barrier(struct drbd_conf
*, struct drbd_tl_epoch
*);
1204 extern void drbd_free_sock(struct drbd_conf
*mdev
);
1205 extern int drbd_send(struct drbd_conf
*mdev
, struct socket
*sock
,
1206 void *buf
, size_t size
, unsigned msg_flags
);
1207 extern int drbd_send_protocol(struct drbd_conf
*mdev
);
1208 extern int drbd_send_uuids(struct drbd_conf
*mdev
);
1209 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf
*mdev
);
1210 extern int drbd_send_sync_uuid(struct drbd_conf
*mdev
, u64 val
);
1211 extern int drbd_send_sizes(struct drbd_conf
*mdev
, int trigger_reply
, enum dds_flags flags
);
1212 extern int _drbd_send_state(struct drbd_conf
*mdev
);
1213 extern int drbd_send_state(struct drbd_conf
*mdev
);
1214 extern int _drbd_send_cmd(struct drbd_conf
*mdev
, struct socket
*sock
,
1215 enum drbd_packets cmd
, struct p_header80
*h
,
1216 size_t size
, unsigned msg_flags
);
1217 #define USE_DATA_SOCKET 1
1218 #define USE_META_SOCKET 0
1219 extern int drbd_send_cmd(struct drbd_conf
*mdev
, int use_data_socket
,
1220 enum drbd_packets cmd
, struct p_header80
*h
,
1222 extern int drbd_send_cmd2(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1223 char *data
, size_t size
);
1224 extern int drbd_send_sync_param(struct drbd_conf
*mdev
, struct syncer_conf
*sc
);
1225 extern int drbd_send_b_ack(struct drbd_conf
*mdev
, u32 barrier_nr
,
1227 extern int drbd_send_ack(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1228 struct drbd_epoch_entry
*e
);
1229 extern int drbd_send_ack_rp(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1230 struct p_block_req
*rp
);
1231 extern int drbd_send_ack_dp(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1232 struct p_data
*dp
, int data_size
);
1233 extern int drbd_send_ack_ex(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1234 sector_t sector
, int blksize
, u64 block_id
);
1235 extern int drbd_send_oos(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1236 extern int drbd_send_block(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1237 struct drbd_epoch_entry
*e
);
1238 extern int drbd_send_dblock(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1239 extern int drbd_send_drequest(struct drbd_conf
*mdev
, int cmd
,
1240 sector_t sector
, int size
, u64 block_id
);
1241 extern int drbd_send_drequest_csum(struct drbd_conf
*mdev
,
1242 sector_t sector
,int size
,
1243 void *digest
, int digest_size
,
1244 enum drbd_packets cmd
);
1245 extern int drbd_send_ov_request(struct drbd_conf
*mdev
,sector_t sector
,int size
);
1247 extern int drbd_send_bitmap(struct drbd_conf
*mdev
);
1248 extern int _drbd_send_bitmap(struct drbd_conf
*mdev
);
1249 extern int drbd_send_sr_reply(struct drbd_conf
*mdev
, int retcode
);
1250 extern void drbd_free_bc(struct drbd_backing_dev
*ldev
);
1251 extern void drbd_mdev_cleanup(struct drbd_conf
*mdev
);
1253 /* drbd_meta-data.c (still in drbd_main.c) */
1254 extern void drbd_md_sync(struct drbd_conf
*mdev
);
1255 extern int drbd_md_read(struct drbd_conf
*mdev
, struct drbd_backing_dev
*bdev
);
1256 /* maybe define them below as inline? */
1257 extern void drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1258 extern void _drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1259 extern void drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1260 extern void _drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1261 extern void drbd_uuid_set_bm(struct drbd_conf
*mdev
, u64 val
) __must_hold(local
);
1262 extern void drbd_md_set_flag(struct drbd_conf
*mdev
, int flags
) __must_hold(local
);
1263 extern void drbd_md_clear_flag(struct drbd_conf
*mdev
, int flags
)__must_hold(local
);
1264 extern int drbd_md_test_flag(struct drbd_backing_dev
*, int);
1265 #ifndef DRBD_DEBUG_MD_SYNC
1266 extern void drbd_md_mark_dirty(struct drbd_conf
*mdev
);
1268 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1269 extern void drbd_md_mark_dirty_(struct drbd_conf
*mdev
,
1270 unsigned int line
, const char *func
);
1272 extern void drbd_queue_bitmap_io(struct drbd_conf
*mdev
,
1273 int (*io_fn
)(struct drbd_conf
*),
1274 void (*done
)(struct drbd_conf
*, int),
1276 extern int drbd_bmio_set_n_write(struct drbd_conf
*mdev
);
1277 extern int drbd_bmio_clear_n_write(struct drbd_conf
*mdev
);
1278 extern int drbd_bitmap_io(struct drbd_conf
*mdev
, int (*io_fn
)(struct drbd_conf
*), char *why
);
1279 extern void drbd_go_diskless(struct drbd_conf
*mdev
);
1280 extern void drbd_ldev_destroy(struct drbd_conf
*mdev
);
1284 We reserve a 128MB Block (4k aligned)
1285 * either at the end of the backing device
1286 * or on a separate meta data device. */
1288 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1289 /* The following numbers are sectors */
1290 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1291 #define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */
1292 /* Allows up to about 3.8TB */
1293 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1295 /* Since the smalles IO unit is usually 512 byte */
1296 #define MD_SECTOR_SHIFT 9
1297 #define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT)
1300 #define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1301 #define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */
1302 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1304 #if BITS_PER_LONG == 32
1306 #define cpu_to_lel(A) cpu_to_le32(A)
1307 #define lel_to_cpu(A) le32_to_cpu(A)
1308 #elif BITS_PER_LONG == 64
1310 #define cpu_to_lel(A) cpu_to_le64(A)
1311 #define lel_to_cpu(A) le64_to_cpu(A)
1313 #error "LN2 of BITS_PER_LONG unknown!"
1317 /* 16MB sized 'bitmap extent' to track syncer usage */
1319 int rs_left
; /* number of bits set (out of sync) in this extent. */
1320 int rs_failed
; /* number of failed resync requests in this extent. */
1321 unsigned long flags
;
1322 struct lc_element lce
;
1325 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1326 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1327 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1331 * We need to store one bit for a block.
1332 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1333 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1334 * Bit 1 ==> local node thinks this block needs to be synced.
1337 #define SLEEP_TIME (HZ/10)
1339 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1340 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1341 /* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1342 * per sector of on disk bitmap */
1343 #define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */
1344 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1346 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1347 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1350 /* thus many _storage_ sectors are described by one bit */
1351 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1352 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1353 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1355 /* bit to represented kilo byte conversion */
1356 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1358 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1359 * _storage_ sector is located in */
1360 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1362 /* how much _storage_ sectors we have per bitmap sector */
1363 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1364 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1366 /* in one sector of the bitmap, we have this many activity_log extents. */
1367 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1368 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1370 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1371 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1373 /* the extent in "PER_EXTENT" below is an activity log extent
1374 * we need that many (long words/bytes) to store the bitmap
1375 * of one AL_EXTENT_SIZE chunk of storage.
1376 * we can store the bitmap for that many AL_EXTENTS within
1377 * one sector of the _on_disk_ bitmap:
1378 * bit 0 bit 37 bit 38 bit (512*8)-1
1379 * ...|........|........|.. // ..|........|
1380 * sect. 0 `296 `304 ^(512*8*8)-1
1382 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1383 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1384 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1387 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1388 #define DRBD_MAX_SECTORS_BM \
1389 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1390 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1391 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1392 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1393 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1394 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1395 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1397 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1398 /* 16 TB in units of sectors */
1399 #if BITS_PER_LONG == 32
1400 /* adjust by one page worth of bitmap,
1401 * so we won't wrap around in drbd_bm_find_next_bit.
1402 * you should use 64bit OS for that much storage, anyways. */
1403 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1405 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0x1LU << 32)
1409 /* Sector shift value for the "hash" functions of tl_hash and ee_hash tables.
1410 * With a value of 8 all IO in one 128K block make it to the same slot of the
1413 #define DRBD_MAX_BIO_SIZE (1U<<(9+HT_SHIFT))
1415 #define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */
1417 /* Number of elements in the app_reads_hash */
1418 #define APP_R_HSIZE 15
1420 extern int drbd_bm_init(struct drbd_conf
*mdev
);
1421 extern int drbd_bm_resize(struct drbd_conf
*mdev
, sector_t sectors
, int set_new_bits
);
1422 extern void drbd_bm_cleanup(struct drbd_conf
*mdev
);
1423 extern void drbd_bm_set_all(struct drbd_conf
*mdev
);
1424 extern void drbd_bm_clear_all(struct drbd_conf
*mdev
);
1425 extern int drbd_bm_set_bits(
1426 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1427 extern int drbd_bm_clear_bits(
1428 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1429 /* bm_set_bits variant for use while holding drbd_bm_lock */
1430 extern void _drbd_bm_set_bits(struct drbd_conf
*mdev
,
1431 const unsigned long s
, const unsigned long e
);
1432 extern int drbd_bm_test_bit(struct drbd_conf
*mdev
, unsigned long bitnr
);
1433 extern int drbd_bm_e_weight(struct drbd_conf
*mdev
, unsigned long enr
);
1434 extern int drbd_bm_write_sect(struct drbd_conf
*mdev
, unsigned long enr
) __must_hold(local
);
1435 extern int drbd_bm_read(struct drbd_conf
*mdev
) __must_hold(local
);
1436 extern int drbd_bm_write(struct drbd_conf
*mdev
) __must_hold(local
);
1437 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf
*mdev
,
1438 unsigned long al_enr
);
1439 extern size_t drbd_bm_words(struct drbd_conf
*mdev
);
1440 extern unsigned long drbd_bm_bits(struct drbd_conf
*mdev
);
1441 extern sector_t
drbd_bm_capacity(struct drbd_conf
*mdev
);
1442 extern unsigned long drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1443 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1444 extern unsigned long _drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1445 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1446 extern unsigned long _drbd_bm_total_weight(struct drbd_conf
*mdev
);
1447 extern unsigned long drbd_bm_total_weight(struct drbd_conf
*mdev
);
1448 extern int drbd_bm_rs_done(struct drbd_conf
*mdev
);
1449 /* for receive_bitmap */
1450 extern void drbd_bm_merge_lel(struct drbd_conf
*mdev
, size_t offset
,
1451 size_t number
, unsigned long *buffer
);
1452 /* for _drbd_send_bitmap and drbd_bm_write_sect */
1453 extern void drbd_bm_get_lel(struct drbd_conf
*mdev
, size_t offset
,
1454 size_t number
, unsigned long *buffer
);
1456 extern void drbd_bm_lock(struct drbd_conf
*mdev
, char *why
);
1457 extern void drbd_bm_unlock(struct drbd_conf
*mdev
);
1459 extern int drbd_bm_count_bits(struct drbd_conf
*mdev
, const unsigned long s
, const unsigned long e
);
1462 extern struct kmem_cache
*drbd_request_cache
;
1463 extern struct kmem_cache
*drbd_ee_cache
; /* epoch entries */
1464 extern struct kmem_cache
*drbd_bm_ext_cache
; /* bitmap extents */
1465 extern struct kmem_cache
*drbd_al_ext_cache
; /* activity log extents */
1466 extern mempool_t
*drbd_request_mempool
;
1467 extern mempool_t
*drbd_ee_mempool
;
1469 extern struct page
*drbd_pp_pool
; /* drbd's page pool */
1470 extern spinlock_t drbd_pp_lock
;
1471 extern int drbd_pp_vacant
;
1472 extern wait_queue_head_t drbd_pp_wait
;
1474 extern rwlock_t global_state_lock
;
1476 extern struct drbd_conf
*drbd_new_device(unsigned int minor
);
1477 extern void drbd_free_mdev(struct drbd_conf
*mdev
);
1479 extern int proc_details
;
1482 extern int drbd_make_request_26(struct request_queue
*q
, struct bio
*bio
);
1483 extern int drbd_read_remote(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1484 extern int drbd_merge_bvec(struct request_queue
*q
, struct bvec_merge_data
*bvm
, struct bio_vec
*bvec
);
1485 extern int is_valid_ar_handle(struct drbd_request
*, sector_t
);
1489 extern void drbd_suspend_io(struct drbd_conf
*mdev
);
1490 extern void drbd_resume_io(struct drbd_conf
*mdev
);
1491 extern char *ppsize(char *buf
, unsigned long long size
);
1492 extern sector_t
drbd_new_dev_size(struct drbd_conf
*, struct drbd_backing_dev
*, int);
1493 enum determine_dev_size
{ dev_size_error
= -1, unchanged
= 0, shrunk
= 1, grew
= 2 };
1494 extern enum determine_dev_size
drbd_determin_dev_size(struct drbd_conf
*, enum dds_flags
) __must_hold(local
);
1495 extern void resync_after_online_grow(struct drbd_conf
*);
1496 extern void drbd_setup_queue_param(struct drbd_conf
*mdev
, unsigned int) __must_hold(local
);
1497 extern int drbd_set_role(struct drbd_conf
*mdev
, enum drbd_role new_role
,
1499 extern enum drbd_disk_state
drbd_try_outdate_peer(struct drbd_conf
*mdev
);
1500 extern void drbd_try_outdate_peer_async(struct drbd_conf
*mdev
);
1501 extern int drbd_khelper(struct drbd_conf
*mdev
, char *cmd
);
1504 extern int drbd_worker(struct drbd_thread
*thi
);
1505 extern int drbd_alter_sa(struct drbd_conf
*mdev
, int na
);
1506 extern void drbd_start_resync(struct drbd_conf
*mdev
, enum drbd_conns side
);
1507 extern void resume_next_sg(struct drbd_conf
*mdev
);
1508 extern void suspend_other_sg(struct drbd_conf
*mdev
);
1509 extern int drbd_resync_finished(struct drbd_conf
*mdev
);
1510 /* maybe rather drbd_main.c ? */
1511 extern int drbd_md_sync_page_io(struct drbd_conf
*mdev
,
1512 struct drbd_backing_dev
*bdev
, sector_t sector
, int rw
);
1513 extern void drbd_ov_oos_found(struct drbd_conf
*, sector_t
, int);
1514 extern void drbd_rs_controller_reset(struct drbd_conf
*mdev
);
1516 static inline void ov_oos_print(struct drbd_conf
*mdev
)
1518 if (mdev
->ov_last_oos_size
) {
1519 dev_err(DEV
, "Out of sync: start=%llu, size=%lu (sectors)\n",
1520 (unsigned long long)mdev
->ov_last_oos_start
,
1521 (unsigned long)mdev
->ov_last_oos_size
);
1523 mdev
->ov_last_oos_size
=0;
1527 extern void drbd_csum_bio(struct drbd_conf
*, struct crypto_hash
*, struct bio
*, void *);
1528 extern void drbd_csum_ee(struct drbd_conf
*, struct crypto_hash
*, struct drbd_epoch_entry
*, void *);
1529 /* worker callbacks */
1530 extern int w_req_cancel_conflict(struct drbd_conf
*, struct drbd_work
*, int);
1531 extern int w_read_retry_remote(struct drbd_conf
*, struct drbd_work
*, int);
1532 extern int w_e_end_data_req(struct drbd_conf
*, struct drbd_work
*, int);
1533 extern int w_e_end_rsdata_req(struct drbd_conf
*, struct drbd_work
*, int);
1534 extern int w_e_end_csum_rs_req(struct drbd_conf
*, struct drbd_work
*, int);
1535 extern int w_e_end_ov_reply(struct drbd_conf
*, struct drbd_work
*, int);
1536 extern int w_e_end_ov_req(struct drbd_conf
*, struct drbd_work
*, int);
1537 extern int w_ov_finished(struct drbd_conf
*, struct drbd_work
*, int);
1538 extern int w_resync_inactive(struct drbd_conf
*, struct drbd_work
*, int);
1539 extern int w_resume_next_sg(struct drbd_conf
*, struct drbd_work
*, int);
1540 extern int w_send_write_hint(struct drbd_conf
*, struct drbd_work
*, int);
1541 extern int w_send_dblock(struct drbd_conf
*, struct drbd_work
*, int);
1542 extern int w_send_barrier(struct drbd_conf
*, struct drbd_work
*, int);
1543 extern int w_send_read_req(struct drbd_conf
*, struct drbd_work
*, int);
1544 extern int w_prev_work_done(struct drbd_conf
*, struct drbd_work
*, int);
1545 extern int w_e_reissue(struct drbd_conf
*, struct drbd_work
*, int);
1546 extern int w_restart_disk_io(struct drbd_conf
*, struct drbd_work
*, int);
1547 extern int w_send_oos(struct drbd_conf
*, struct drbd_work
*, int);
1548 extern int w_start_resync(struct drbd_conf
*, struct drbd_work
*, int);
1550 extern void resync_timer_fn(unsigned long data
);
1552 /* drbd_receiver.c */
1553 extern int drbd_rs_should_slow_down(struct drbd_conf
*mdev
, sector_t sector
);
1554 extern int drbd_submit_ee(struct drbd_conf
*mdev
, struct drbd_epoch_entry
*e
,
1555 const unsigned rw
, const int fault_type
);
1556 extern int drbd_release_ee(struct drbd_conf
*mdev
, struct list_head
*list
);
1557 extern struct drbd_epoch_entry
*drbd_alloc_ee(struct drbd_conf
*mdev
,
1560 unsigned int data_size
,
1561 gfp_t gfp_mask
) __must_hold(local
);
1562 extern void drbd_free_some_ee(struct drbd_conf
*mdev
, struct drbd_epoch_entry
*e
,
1564 #define drbd_free_ee(m,e) drbd_free_some_ee(m, e, 0)
1565 #define drbd_free_net_ee(m,e) drbd_free_some_ee(m, e, 1)
1566 extern void drbd_wait_ee_list_empty(struct drbd_conf
*mdev
,
1567 struct list_head
*head
);
1568 extern void _drbd_wait_ee_list_empty(struct drbd_conf
*mdev
,
1569 struct list_head
*head
);
1570 extern void drbd_set_recv_tcq(struct drbd_conf
*mdev
, int tcq_enabled
);
1571 extern void _drbd_clear_done_ee(struct drbd_conf
*mdev
, struct list_head
*to_be_freed
);
1572 extern void drbd_flush_workqueue(struct drbd_conf
*mdev
);
1573 extern void drbd_free_tl_hash(struct drbd_conf
*mdev
);
1575 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1576 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1577 static inline int drbd_setsockopt(struct socket
*sock
, int level
, int optname
,
1578 char __user
*optval
, int optlen
)
1581 if (level
== SOL_SOCKET
)
1582 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
1584 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1589 static inline void drbd_tcp_cork(struct socket
*sock
)
1592 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1593 (char __user
*)&val
, sizeof(val
));
1596 static inline void drbd_tcp_uncork(struct socket
*sock
)
1599 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1600 (char __user
*)&val
, sizeof(val
));
1603 static inline void drbd_tcp_nodelay(struct socket
*sock
)
1606 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_NODELAY
,
1607 (char __user
*)&val
, sizeof(val
));
1610 static inline void drbd_tcp_quickack(struct socket
*sock
)
1613 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_QUICKACK
,
1614 (char __user
*)&val
, sizeof(val
));
1617 void drbd_bump_write_ordering(struct drbd_conf
*mdev
, enum write_ordering_e wo
);
1620 extern struct proc_dir_entry
*drbd_proc
;
1621 extern const struct file_operations drbd_proc_fops
;
1622 extern const char *drbd_conn_str(enum drbd_conns s
);
1623 extern const char *drbd_role_str(enum drbd_role s
);
1626 extern void drbd_al_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1627 extern void drbd_al_complete_io(struct drbd_conf
*mdev
, sector_t sector
);
1628 extern void drbd_rs_complete_io(struct drbd_conf
*mdev
, sector_t sector
);
1629 extern int drbd_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1630 extern int drbd_try_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1631 extern void drbd_rs_cancel_all(struct drbd_conf
*mdev
);
1632 extern int drbd_rs_del_all(struct drbd_conf
*mdev
);
1633 extern void drbd_rs_failed_io(struct drbd_conf
*mdev
,
1634 sector_t sector
, int size
);
1635 extern int drbd_al_read_log(struct drbd_conf
*mdev
, struct drbd_backing_dev
*);
1636 extern void drbd_advance_rs_marks(struct drbd_conf
*mdev
, unsigned long still_to_go
);
1637 extern void __drbd_set_in_sync(struct drbd_conf
*mdev
, sector_t sector
,
1638 int size
, const char *file
, const unsigned int line
);
1639 #define drbd_set_in_sync(mdev, sector, size) \
1640 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1641 extern int __drbd_set_out_of_sync(struct drbd_conf
*mdev
, sector_t sector
,
1642 int size
, const char *file
, const unsigned int line
);
1643 #define drbd_set_out_of_sync(mdev, sector, size) \
1644 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1645 extern void drbd_al_apply_to_bm(struct drbd_conf
*mdev
);
1646 extern void drbd_al_to_on_disk_bm(struct drbd_conf
*mdev
);
1647 extern void drbd_al_shrink(struct drbd_conf
*mdev
);
1652 void drbd_nl_cleanup(void);
1653 int __init
drbd_nl_init(void);
1654 void drbd_bcast_state(struct drbd_conf
*mdev
, union drbd_state
);
1655 void drbd_bcast_sync_progress(struct drbd_conf
*mdev
);
1656 void drbd_bcast_ee(struct drbd_conf
*mdev
,
1657 const char *reason
, const int dgs
,
1658 const char* seen_hash
, const char* calc_hash
,
1659 const struct drbd_epoch_entry
* e
);
1663 * DOC: DRBD State macros
1665 * These macros are used to express state changes in easily readable form.
1667 * The NS macros expand to a mask and a value, that can be bit ored onto the
1668 * current state as soon as the spinlock (req_lock) was taken.
1670 * The _NS macros are used for state functions that get called with the
1671 * spinlock. These macros expand directly to the new state value.
1673 * Besides the basic forms NS() and _NS() additional _?NS[23] are defined
1674 * to express state changes that affect more than one aspect of the state.
1676 * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY)
1677 * Means that the network connection was established and that the peer
1678 * is in secondary role.
1680 #define role_MASK R_MASK
1681 #define peer_MASK R_MASK
1682 #define disk_MASK D_MASK
1683 #define pdsk_MASK D_MASK
1684 #define conn_MASK C_MASK
1686 #define user_isp_MASK 1
1687 #define aftr_isp_MASK 1
1688 #define susp_nod_MASK 1
1689 #define susp_fen_MASK 1
1692 ({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \
1693 ({ union drbd_state val; val.i = 0; val.T = (S); val; })
1694 #define NS2(T1, S1, T2, S2) \
1695 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1696 mask.T2 = T2##_MASK; mask; }), \
1697 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1698 val.T2 = (S2); val; })
1699 #define NS3(T1, S1, T2, S2, T3, S3) \
1700 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1701 mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \
1702 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1703 val.T2 = (S2); val.T3 = (S3); val; })
1705 #define _NS(D, T, S) \
1706 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; })
1707 #define _NS2(D, T1, S1, T2, S2) \
1708 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1709 __ns.T2 = (S2); __ns; })
1710 #define _NS3(D, T1, S1, T2, S2, T3, S3) \
1711 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1712 __ns.T2 = (S2); __ns.T3 = (S3); __ns; })
1715 * inline helper functions
1716 *************************/
1718 /* see also page_chain_add and friends in drbd_receiver.c */
1719 static inline struct page
*page_chain_next(struct page
*page
)
1721 return (struct page
*)page_private(page
);
1723 #define page_chain_for_each(page) \
1724 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1725 page = page_chain_next(page))
1726 #define page_chain_for_each_safe(page, n) \
1727 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1729 static inline int drbd_bio_has_active_page(struct bio
*bio
)
1731 struct bio_vec
*bvec
;
1734 __bio_for_each_segment(bvec
, bio
, i
, 0) {
1735 if (page_count(bvec
->bv_page
) > 1)
1742 static inline int drbd_ee_has_active_page(struct drbd_epoch_entry
*e
)
1744 struct page
*page
= e
->pages
;
1745 page_chain_for_each(page
) {
1746 if (page_count(page
) > 1)
1753 static inline void drbd_state_lock(struct drbd_conf
*mdev
)
1755 wait_event(mdev
->misc_wait
,
1756 !test_and_set_bit(CLUSTER_ST_CHANGE
, &mdev
->flags
));
1759 static inline void drbd_state_unlock(struct drbd_conf
*mdev
)
1761 clear_bit(CLUSTER_ST_CHANGE
, &mdev
->flags
);
1762 wake_up(&mdev
->misc_wait
);
1765 static inline int _drbd_set_state(struct drbd_conf
*mdev
,
1766 union drbd_state ns
, enum chg_state_flags flags
,
1767 struct completion
*done
)
1771 read_lock(&global_state_lock
);
1772 rv
= __drbd_set_state(mdev
, ns
, flags
, done
);
1773 read_unlock(&global_state_lock
);
1779 * drbd_request_state() - Reqest a state change
1780 * @mdev: DRBD device.
1781 * @mask: mask of state bits to change.
1782 * @val: value of new state bits.
1784 * This is the most graceful way of requesting a state change. It is verbose
1785 * quite verbose in case the state change is not possible, and all those
1786 * state changes are globally serialized.
1788 static inline int drbd_request_state(struct drbd_conf
*mdev
,
1789 union drbd_state mask
,
1790 union drbd_state val
)
1792 return _drbd_request_state(mdev
, mask
, val
, CS_VERBOSE
+ CS_ORDERED
);
1795 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1796 static inline void __drbd_chk_io_error_(struct drbd_conf
*mdev
, int forcedetach
, const char *where
)
1798 switch (mdev
->ldev
->dc
.on_io_error
) {
1801 if (__ratelimit(&drbd_ratelimit_state
))
1802 dev_err(DEV
, "Local IO failed in %s.\n", where
);
1805 /* NOTE fall through to detach case if forcedetach set */
1807 case EP_CALL_HELPER
:
1808 set_bit(WAS_IO_ERROR
, &mdev
->flags
);
1809 if (mdev
->state
.disk
> D_FAILED
) {
1810 _drbd_set_state(_NS(mdev
, disk
, D_FAILED
), CS_HARD
, NULL
);
1812 "Local IO failed in %s. Detaching...\n", where
);
1819 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1820 * @mdev: DRBD device.
1821 * @error: Error code passed to the IO completion callback
1822 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1824 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1826 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1827 static inline void drbd_chk_io_error_(struct drbd_conf
*mdev
,
1828 int error
, int forcedetach
, const char *where
)
1831 unsigned long flags
;
1832 spin_lock_irqsave(&mdev
->req_lock
, flags
);
1833 __drbd_chk_io_error_(mdev
, forcedetach
, where
);
1834 spin_unlock_irqrestore(&mdev
->req_lock
, flags
);
1840 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1841 * @bdev: Meta data block device.
1843 * BTW, for internal meta data, this happens to be the maximum capacity
1844 * we could agree upon with our peer node.
1846 static inline sector_t
drbd_md_first_sector(struct drbd_backing_dev
*bdev
)
1848 switch (bdev
->dc
.meta_dev_idx
) {
1849 case DRBD_MD_INDEX_INTERNAL
:
1850 case DRBD_MD_INDEX_FLEX_INT
:
1851 return bdev
->md
.md_offset
+ bdev
->md
.bm_offset
;
1852 case DRBD_MD_INDEX_FLEX_EXT
:
1854 return bdev
->md
.md_offset
;
1859 * drbd_md_last_sector() - Return the last sector number of the meta data area
1860 * @bdev: Meta data block device.
1862 static inline sector_t
drbd_md_last_sector(struct drbd_backing_dev
*bdev
)
1864 switch (bdev
->dc
.meta_dev_idx
) {
1865 case DRBD_MD_INDEX_INTERNAL
:
1866 case DRBD_MD_INDEX_FLEX_INT
:
1867 return bdev
->md
.md_offset
+ MD_AL_OFFSET
- 1;
1868 case DRBD_MD_INDEX_FLEX_EXT
:
1870 return bdev
->md
.md_offset
+ bdev
->md
.md_size_sect
;
1874 /* Returns the number of 512 byte sectors of the device */
1875 static inline sector_t
drbd_get_capacity(struct block_device
*bdev
)
1877 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1878 return bdev
? i_size_read(bdev
->bd_inode
) >> 9 : 0;
1882 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1883 * @bdev: Meta data block device.
1885 * returns the capacity we announce to out peer. we clip ourselves at the
1886 * various MAX_SECTORS, because if we don't, current implementation will
1887 * oops sooner or later
1889 static inline sector_t
drbd_get_max_capacity(struct drbd_backing_dev
*bdev
)
1892 switch (bdev
->dc
.meta_dev_idx
) {
1893 case DRBD_MD_INDEX_INTERNAL
:
1894 case DRBD_MD_INDEX_FLEX_INT
:
1895 s
= drbd_get_capacity(bdev
->backing_bdev
)
1896 ? min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1897 drbd_md_first_sector(bdev
))
1900 case DRBD_MD_INDEX_FLEX_EXT
:
1901 s
= min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1902 drbd_get_capacity(bdev
->backing_bdev
));
1903 /* clip at maximum size the meta device can support */
1904 s
= min_t(sector_t
, s
,
1905 BM_EXT_TO_SECT(bdev
->md
.md_size_sect
1906 - bdev
->md
.bm_offset
));
1909 s
= min_t(sector_t
, DRBD_MAX_SECTORS
,
1910 drbd_get_capacity(bdev
->backing_bdev
));
1916 * drbd_md_ss__() - Return the sector number of our meta data super block
1917 * @mdev: DRBD device.
1918 * @bdev: Meta data block device.
1920 static inline sector_t
drbd_md_ss__(struct drbd_conf
*mdev
,
1921 struct drbd_backing_dev
*bdev
)
1923 switch (bdev
->dc
.meta_dev_idx
) {
1924 default: /* external, some index */
1925 return MD_RESERVED_SECT
* bdev
->dc
.meta_dev_idx
;
1926 case DRBD_MD_INDEX_INTERNAL
:
1927 /* with drbd08, internal meta data is always "flexible" */
1928 case DRBD_MD_INDEX_FLEX_INT
:
1929 /* sizeof(struct md_on_disk_07) == 4k
1930 * position: last 4k aligned block of 4k size */
1931 if (!bdev
->backing_bdev
) {
1932 if (__ratelimit(&drbd_ratelimit_state
)) {
1933 dev_err(DEV
, "bdev->backing_bdev==NULL\n");
1938 return (drbd_get_capacity(bdev
->backing_bdev
) & ~7ULL)
1940 case DRBD_MD_INDEX_FLEX_EXT
:
1946 drbd_queue_work_front(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1948 unsigned long flags
;
1949 spin_lock_irqsave(&q
->q_lock
, flags
);
1950 list_add(&w
->list
, &q
->q
);
1951 up(&q
->s
); /* within the spinlock,
1952 see comment near end of drbd_worker() */
1953 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1957 drbd_queue_work(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1959 unsigned long flags
;
1960 spin_lock_irqsave(&q
->q_lock
, flags
);
1961 list_add_tail(&w
->list
, &q
->q
);
1962 up(&q
->s
); /* within the spinlock,
1963 see comment near end of drbd_worker() */
1964 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1967 static inline void wake_asender(struct drbd_conf
*mdev
)
1969 if (test_bit(SIGNAL_ASENDER
, &mdev
->flags
))
1970 force_sig(DRBD_SIG
, mdev
->asender
.task
);
1973 static inline void request_ping(struct drbd_conf
*mdev
)
1975 set_bit(SEND_PING
, &mdev
->flags
);
1979 static inline int drbd_send_short_cmd(struct drbd_conf
*mdev
,
1980 enum drbd_packets cmd
)
1982 struct p_header80 h
;
1983 return drbd_send_cmd(mdev
, USE_DATA_SOCKET
, cmd
, &h
, sizeof(h
));
1986 static inline int drbd_send_ping(struct drbd_conf
*mdev
)
1988 struct p_header80 h
;
1989 return drbd_send_cmd(mdev
, USE_META_SOCKET
, P_PING
, &h
, sizeof(h
));
1992 static inline int drbd_send_ping_ack(struct drbd_conf
*mdev
)
1994 struct p_header80 h
;
1995 return drbd_send_cmd(mdev
, USE_META_SOCKET
, P_PING_ACK
, &h
, sizeof(h
));
1998 static inline void drbd_thread_stop(struct drbd_thread
*thi
)
2000 _drbd_thread_stop(thi
, FALSE
, TRUE
);
2003 static inline void drbd_thread_stop_nowait(struct drbd_thread
*thi
)
2005 _drbd_thread_stop(thi
, FALSE
, FALSE
);
2008 static inline void drbd_thread_restart_nowait(struct drbd_thread
*thi
)
2010 _drbd_thread_stop(thi
, TRUE
, FALSE
);
2013 /* counts how many answer packets packets we expect from our peer,
2014 * for either explicit application requests,
2015 * or implicit barrier packets as necessary.
2018 * _req_mod(req, queue_for_net_write or queue_for_net_read);
2019 * it is much easier and equally valid to count what we queue for the
2020 * worker, even before it actually was queued or send.
2021 * (drbd_make_request_common; recovery path on read io-error)
2023 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
2024 * _req_mod(req, data_received)
2025 * [from receive_DataReply]
2026 * _req_mod(req, write_acked_by_peer or recv_acked_by_peer or neg_acked)
2027 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
2028 * for some reason it is NOT decreased in got_NegAck,
2029 * but in the resulting cleanup code from report_params.
2030 * we should try to remember the reason for that...
2031 * _req_mod(req, send_failed or send_canceled)
2032 * _req_mod(req, connection_lost_while_pending)
2033 * [from tl_clear_barrier]
2035 static inline void inc_ap_pending(struct drbd_conf
*mdev
)
2037 atomic_inc(&mdev
->ap_pending_cnt
);
2040 #define ERR_IF_CNT_IS_NEGATIVE(which) \
2041 if (atomic_read(&mdev->which) < 0) \
2042 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
2043 __func__ , __LINE__ , \
2044 atomic_read(&mdev->which))
2046 #define dec_ap_pending(mdev) do { \
2047 typecheck(struct drbd_conf *, mdev); \
2048 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \
2049 wake_up(&mdev->misc_wait); \
2050 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
2052 /* counts how many resync-related answers we still expect from the peer
2054 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
2055 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK whith ID_SYNCER)
2056 * (or P_NEG_ACK with ID_SYNCER)
2058 static inline void inc_rs_pending(struct drbd_conf
*mdev
)
2060 atomic_inc(&mdev
->rs_pending_cnt
);
2063 #define dec_rs_pending(mdev) do { \
2064 typecheck(struct drbd_conf *, mdev); \
2065 atomic_dec(&mdev->rs_pending_cnt); \
2066 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
2068 /* counts how many answers we still need to send to the peer.
2070 * receive_Data unless protocol A;
2071 * we need to send a P_RECV_ACK (proto B)
2072 * or P_WRITE_ACK (proto C)
2073 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2074 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2075 * receive_Barrier_* we need to send a P_BARRIER_ACK
2077 static inline void inc_unacked(struct drbd_conf
*mdev
)
2079 atomic_inc(&mdev
->unacked_cnt
);
2082 #define dec_unacked(mdev) do { \
2083 typecheck(struct drbd_conf *, mdev); \
2084 atomic_dec(&mdev->unacked_cnt); \
2085 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2087 #define sub_unacked(mdev, n) do { \
2088 typecheck(struct drbd_conf *, mdev); \
2089 atomic_sub(n, &mdev->unacked_cnt); \
2090 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2093 static inline void put_net_conf(struct drbd_conf
*mdev
)
2095 if (atomic_dec_and_test(&mdev
->net_cnt
))
2096 wake_up(&mdev
->net_cnt_wait
);
2100 * get_net_conf() - Increase ref count on mdev->net_conf; Returns 0 if nothing there
2101 * @mdev: DRBD device.
2103 * You have to call put_net_conf() when finished working with mdev->net_conf.
2105 static inline int get_net_conf(struct drbd_conf
*mdev
)
2109 atomic_inc(&mdev
->net_cnt
);
2110 have_net_conf
= mdev
->state
.conn
>= C_UNCONNECTED
;
2113 return have_net_conf
;
2117 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2120 * You have to call put_ldev() when finished working with mdev->ldev.
2122 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2123 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2125 static inline void put_ldev(struct drbd_conf
*mdev
)
2127 int i
= atomic_dec_return(&mdev
->local_cnt
);
2131 if (mdev
->state
.disk
== D_DISKLESS
)
2132 /* even internal references gone, safe to destroy */
2133 drbd_ldev_destroy(mdev
);
2134 if (mdev
->state
.disk
== D_FAILED
)
2135 /* all application IO references gone. */
2136 drbd_go_diskless(mdev
);
2137 wake_up(&mdev
->misc_wait
);
2142 static inline int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
)
2146 /* never get a reference while D_DISKLESS */
2147 if (mdev
->state
.disk
== D_DISKLESS
)
2150 atomic_inc(&mdev
->local_cnt
);
2151 io_allowed
= (mdev
->state
.disk
>= mins
);
2157 extern int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
);
2160 /* you must have an "get_ldev" reference */
2161 static inline void drbd_get_syncer_progress(struct drbd_conf
*mdev
,
2162 unsigned long *bits_left
, unsigned int *per_mil_done
)
2165 * this is to break it at compile time when we change that
2166 * (we may feel 4TB maximum storage per drbd is not enough)
2168 typecheck(unsigned long, mdev
->rs_total
);
2170 /* note: both rs_total and rs_left are in bits, i.e. in
2171 * units of BM_BLOCK_SIZE.
2172 * for the percentage, we don't care. */
2174 if (mdev
->state
.conn
== C_VERIFY_S
|| mdev
->state
.conn
== C_VERIFY_T
)
2175 *bits_left
= mdev
->ov_left
;
2177 *bits_left
= drbd_bm_total_weight(mdev
) - mdev
->rs_failed
;
2178 /* >> 10 to prevent overflow,
2179 * +1 to prevent division by zero */
2180 if (*bits_left
> mdev
->rs_total
) {
2181 /* doh. maybe a logic bug somewhere.
2182 * may also be just a race condition
2183 * between this and a disconnect during sync.
2184 * for now, just prevent in-kernel buffer overflow.
2187 dev_warn(DEV
, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2188 drbd_conn_str(mdev
->state
.conn
),
2189 *bits_left
, mdev
->rs_total
, mdev
->rs_failed
);
2192 /* make sure the calculation happens in long context */
2193 unsigned long tmp
= 1000UL -
2194 (*bits_left
>> 10)*1000UL
2195 / ((mdev
->rs_total
>> 10) + 1UL);
2196 *per_mil_done
= tmp
;
2201 /* this throttles on-the-fly application requests
2202 * according to max_buffers settings;
2203 * maybe re-implement using semaphores? */
2204 static inline int drbd_get_max_buffers(struct drbd_conf
*mdev
)
2206 int mxb
= 1000000; /* arbitrary limit on open requests */
2207 if (get_net_conf(mdev
)) {
2208 mxb
= mdev
->net_conf
->max_buffers
;
2214 static inline int drbd_state_is_stable(struct drbd_conf
*mdev
)
2216 union drbd_state s
= mdev
->state
;
2218 /* DO NOT add a default clause, we want the compiler to warn us
2219 * for any newly introduced state we may have forgotten to add here */
2221 switch ((enum drbd_conns
)s
.conn
) {
2222 /* new io only accepted when there is no connection, ... */
2224 case C_WF_CONNECTION
:
2225 /* ... or there is a well established connection. */
2231 case C_PAUSED_SYNC_S
:
2232 case C_PAUSED_SYNC_T
:
2235 /* transitional states, IO allowed */
2236 case C_DISCONNECTING
:
2240 case C_NETWORK_FAILURE
:
2241 case C_PROTOCOL_ERROR
:
2243 case C_WF_REPORT_PARAMS
:
2244 case C_STARTING_SYNC_S
:
2245 case C_STARTING_SYNC_T
:
2248 /* Allow IO in BM exchange states with new protocols */
2250 if (mdev
->agreed_pro_version
< 96)
2254 /* no new io accepted in these states */
2256 case C_WF_SYNC_UUID
:
2262 switch ((enum drbd_disk_state
)s
.disk
) {
2264 case D_INCONSISTENT
:
2268 /* disk state is stable as well. */
2271 /* no new io accepted during tansitional states */
2284 static inline int is_susp(union drbd_state s
)
2286 return s
.susp
|| s
.susp_nod
|| s
.susp_fen
;
2289 static inline int __inc_ap_bio_cond(struct drbd_conf
*mdev
)
2291 int mxb
= drbd_get_max_buffers(mdev
);
2293 if (is_susp(mdev
->state
))
2295 if (test_bit(SUSPEND_IO
, &mdev
->flags
))
2298 /* to avoid potential deadlock or bitmap corruption,
2299 * in various places, we only allow new application io
2300 * to start during "stable" states. */
2302 /* no new io accepted when attaching or detaching the disk */
2303 if (!drbd_state_is_stable(mdev
))
2306 /* since some older kernels don't have atomic_add_unless,
2307 * and we are within the spinlock anyways, we have this workaround. */
2308 if (atomic_read(&mdev
->ap_bio_cnt
) > mxb
)
2310 if (test_bit(BITMAP_IO
, &mdev
->flags
))
2315 static inline int _inc_ap_bio_cond(struct drbd_conf
*mdev
, int count
)
2319 spin_lock_irq(&mdev
->req_lock
);
2320 rv
= __inc_ap_bio_cond(mdev
);
2322 atomic_add(count
, &mdev
->ap_bio_cnt
);
2323 spin_unlock_irq(&mdev
->req_lock
);
2328 static inline void inc_ap_bio(struct drbd_conf
*mdev
, int count
)
2331 * as long as the device is suspended
2332 * until the bitmap is no longer on the fly during connection
2333 * handshake as long as we would exeed the max_buffer limit.
2335 * to avoid races with the reconnect code,
2336 * we need to atomic_inc within the spinlock. */
2338 wait_event(mdev
->misc_wait
, _inc_ap_bio_cond(mdev
, count
));
2341 static inline void dec_ap_bio(struct drbd_conf
*mdev
)
2343 int mxb
= drbd_get_max_buffers(mdev
);
2344 int ap_bio
= atomic_dec_return(&mdev
->ap_bio_cnt
);
2346 D_ASSERT(ap_bio
>= 0);
2347 /* this currently does wake_up for every dec_ap_bio!
2348 * maybe rather introduce some type of hysteresis?
2349 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2351 wake_up(&mdev
->misc_wait
);
2352 if (ap_bio
== 0 && test_bit(BITMAP_IO
, &mdev
->flags
)) {
2353 if (!test_and_set_bit(BITMAP_IO_QUEUED
, &mdev
->flags
))
2354 drbd_queue_work(&mdev
->data
.work
, &mdev
->bm_io_work
.w
);
2358 static inline void drbd_set_ed_uuid(struct drbd_conf
*mdev
, u64 val
)
2360 mdev
->ed_uuid
= val
;
2363 static inline int seq_cmp(u32 a
, u32 b
)
2365 /* we assume wrap around at 32bit.
2366 * for wrap around at 24bit (old atomic_t),
2370 return (s32
)(a
) - (s32
)(b
);
2372 #define seq_lt(a, b) (seq_cmp((a), (b)) < 0)
2373 #define seq_gt(a, b) (seq_cmp((a), (b)) > 0)
2374 #define seq_ge(a, b) (seq_cmp((a), (b)) >= 0)
2375 #define seq_le(a, b) (seq_cmp((a), (b)) <= 0)
2376 /* CAUTION: please no side effects in arguments! */
2377 #define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b)))
2379 static inline void update_peer_seq(struct drbd_conf
*mdev
, unsigned int new_seq
)
2382 spin_lock(&mdev
->peer_seq_lock
);
2383 m
= seq_max(mdev
->peer_seq
, new_seq
);
2385 spin_unlock(&mdev
->peer_seq_lock
);
2387 wake_up(&mdev
->seq_wait
);
2390 static inline void drbd_update_congested(struct drbd_conf
*mdev
)
2392 struct sock
*sk
= mdev
->data
.socket
->sk
;
2393 if (sk
->sk_wmem_queued
> sk
->sk_sndbuf
* 4 / 5)
2394 set_bit(NET_CONGESTED
, &mdev
->flags
);
2397 static inline int drbd_queue_order_type(struct drbd_conf
*mdev
)
2399 /* sorry, we currently have no working implementation
2400 * of distributed TCQ stuff */
2401 #ifndef QUEUE_ORDERED_NONE
2402 #define QUEUE_ORDERED_NONE 0
2404 return QUEUE_ORDERED_NONE
;
2407 static inline void drbd_md_flush(struct drbd_conf
*mdev
)
2411 if (test_bit(MD_NO_FUA
, &mdev
->flags
))
2414 r
= blkdev_issue_flush(mdev
->ldev
->md_bdev
, GFP_KERNEL
, NULL
);
2416 set_bit(MD_NO_FUA
, &mdev
->flags
);
2417 dev_err(DEV
, "meta data flush failed with status %d, disabling md-flushes\n", r
);