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>
45 #include <linux/prefetch.h>
48 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
49 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
50 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
51 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
53 # define __protected_by(x)
54 # define __protected_read_by(x)
55 # define __protected_write_by(x)
56 # define __must_hold(x)
59 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
61 /* module parameter, defined in drbd_main.c */
62 extern unsigned int minor_count
;
63 extern int disable_sendpage
;
65 extern unsigned int cn_idx
;
67 #ifdef CONFIG_DRBD_FAULT_INJECTION
68 extern int enable_faults
;
69 extern int fault_rate
;
70 extern int fault_devs
;
73 extern char usermode_helper
[];
76 /* I don't remember why XCPU ...
77 * This is used to wake the asender,
78 * and to interrupt sending the sending task
81 #define DRBD_SIG SIGXCPU
83 /* This is used to stop/restart our threads.
84 * Cannot use SIGTERM nor SIGKILL, since these
85 * are sent out by init on runlevel changes
86 * I choose SIGHUP for now.
88 #define DRBD_SIGKILL SIGHUP
90 #define ID_IN_SYNC (4711ULL)
91 #define ID_OUT_OF_SYNC (4712ULL)
92 #define ID_SYNCER (-1ULL)
94 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
99 /* to shorten dev_warn(DEV, "msg"); and relatives statements */
100 #define DEV (disk_to_dev(mdev->vdisk))
102 #define D_ASSERT(exp) if (!(exp)) \
103 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
106 * expect - Make an assertion
108 * Unlike the assert macro, this macro returns a boolean result.
110 #define expect(exp) ({ \
111 bool _bool = (exp); \
113 dev_err(DEV, "ASSERTION %s FAILED in %s\n", \
118 /* Defines to control fault insertion */
120 DRBD_FAULT_MD_WR
= 0, /* meta data write */
121 DRBD_FAULT_MD_RD
= 1, /* read */
122 DRBD_FAULT_RS_WR
= 2, /* resync */
123 DRBD_FAULT_RS_RD
= 3,
124 DRBD_FAULT_DT_WR
= 4, /* data */
125 DRBD_FAULT_DT_RD
= 5,
126 DRBD_FAULT_DT_RA
= 6, /* data read ahead */
127 DRBD_FAULT_BM_ALLOC
= 7, /* bitmap allocation */
128 DRBD_FAULT_AL_EE
= 8, /* alloc ee */
129 DRBD_FAULT_RECEIVE
= 9, /* Changes some bytes upon receiving a [rs]data block */
135 _drbd_insert_fault(struct drbd_conf
*mdev
, unsigned int type
);
138 drbd_insert_fault(struct drbd_conf
*mdev
, unsigned int type
) {
139 #ifdef CONFIG_DRBD_FAULT_INJECTION
141 (enable_faults
& (1<<type
)) &&
142 _drbd_insert_fault(mdev
, type
);
148 /* integer division, round _UP_ to the next integer */
149 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
150 /* usual integer division */
151 #define div_floor(A, B) ((A)/(B))
153 /* drbd_meta-data.c (still in drbd_main.c) */
154 /* 4th incarnation of the disk layout. */
155 #define DRBD_MD_MAGIC (DRBD_MAGIC+4)
157 extern struct drbd_conf
**minor_table
;
158 extern struct ratelimit_state drbd_ratelimit_state
;
162 /* receiver (data socket) */
164 P_DATA_REPLY
= 0x01, /* Response to P_DATA_REQUEST */
165 P_RS_DATA_REPLY
= 0x02, /* Response to P_RS_DATA_REQUEST */
168 P_BECOME_SYNC_TARGET
= 0x05,
169 P_BECOME_SYNC_SOURCE
= 0x06,
170 P_UNPLUG_REMOTE
= 0x07, /* Used at various times to hint the peer */
171 P_DATA_REQUEST
= 0x08, /* Used to ask for a data block */
172 P_RS_DATA_REQUEST
= 0x09, /* Used to ask for a data block for resync */
179 P_AUTH_CHALLENGE
= 0x10,
180 P_AUTH_RESPONSE
= 0x11,
181 P_STATE_CHG_REQ
= 0x12,
183 /* asender (meta socket */
186 P_RECV_ACK
= 0x15, /* Used in protocol B */
187 P_WRITE_ACK
= 0x16, /* Used in protocol C */
188 P_RS_WRITE_ACK
= 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
189 P_DISCARD_ACK
= 0x18, /* Used in proto C, two-primaries conflict detection */
190 P_NEG_ACK
= 0x19, /* Sent if local disk is unusable */
191 P_NEG_DREPLY
= 0x1a, /* Local disk is broken... */
192 P_NEG_RS_DREPLY
= 0x1b, /* Local disk is broken... */
193 P_BARRIER_ACK
= 0x1c,
194 P_STATE_CHG_REPLY
= 0x1d,
196 /* "new" commands, no longer fitting into the ordering scheme above */
198 P_OV_REQUEST
= 0x1e, /* data socket */
200 P_OV_RESULT
= 0x20, /* meta socket */
201 P_CSUM_RS_REQUEST
= 0x21, /* data socket */
202 P_RS_IS_IN_SYNC
= 0x22, /* meta socket */
203 P_SYNC_PARAM89
= 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
204 P_COMPRESSED_BITMAP
= 0x24, /* compressed or otherwise encoded bitmap transfer */
205 /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */
206 /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */
207 P_DELAY_PROBE
= 0x27, /* is used on BOTH sockets */
208 P_OUT_OF_SYNC
= 0x28, /* Mark as out of sync (Outrunning), data socket */
209 P_RS_CANCEL
= 0x29, /* meta: Used to cancel RS_DATA_REQUEST packet by SyncSource */
212 P_MAY_IGNORE
= 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
213 P_MAX_OPT_CMD
= 0x101,
215 /* special command ids for handshake */
217 P_HAND_SHAKE_M
= 0xfff1, /* First Packet on the MetaSock */
218 P_HAND_SHAKE_S
= 0xfff2, /* First Packet on the Socket */
220 P_HAND_SHAKE
= 0xfffe /* FIXED for the next century! */
223 static inline const char *cmdname(enum drbd_packets cmd
)
225 /* THINK may need to become several global tables
226 * when we want to support more than
228 static const char *cmdnames
[] = {
230 [P_DATA_REPLY
] = "DataReply",
231 [P_RS_DATA_REPLY
] = "RSDataReply",
232 [P_BARRIER
] = "Barrier",
233 [P_BITMAP
] = "ReportBitMap",
234 [P_BECOME_SYNC_TARGET
] = "BecomeSyncTarget",
235 [P_BECOME_SYNC_SOURCE
] = "BecomeSyncSource",
236 [P_UNPLUG_REMOTE
] = "UnplugRemote",
237 [P_DATA_REQUEST
] = "DataRequest",
238 [P_RS_DATA_REQUEST
] = "RSDataRequest",
239 [P_SYNC_PARAM
] = "SyncParam",
240 [P_SYNC_PARAM89
] = "SyncParam89",
241 [P_PROTOCOL
] = "ReportProtocol",
242 [P_UUIDS
] = "ReportUUIDs",
243 [P_SIZES
] = "ReportSizes",
244 [P_STATE
] = "ReportState",
245 [P_SYNC_UUID
] = "ReportSyncUUID",
246 [P_AUTH_CHALLENGE
] = "AuthChallenge",
247 [P_AUTH_RESPONSE
] = "AuthResponse",
249 [P_PING_ACK
] = "PingAck",
250 [P_RECV_ACK
] = "RecvAck",
251 [P_WRITE_ACK
] = "WriteAck",
252 [P_RS_WRITE_ACK
] = "RSWriteAck",
253 [P_DISCARD_ACK
] = "DiscardAck",
254 [P_NEG_ACK
] = "NegAck",
255 [P_NEG_DREPLY
] = "NegDReply",
256 [P_NEG_RS_DREPLY
] = "NegRSDReply",
257 [P_BARRIER_ACK
] = "BarrierAck",
258 [P_STATE_CHG_REQ
] = "StateChgRequest",
259 [P_STATE_CHG_REPLY
] = "StateChgReply",
260 [P_OV_REQUEST
] = "OVRequest",
261 [P_OV_REPLY
] = "OVReply",
262 [P_OV_RESULT
] = "OVResult",
263 [P_CSUM_RS_REQUEST
] = "CsumRSRequest",
264 [P_RS_IS_IN_SYNC
] = "CsumRSIsInSync",
265 [P_COMPRESSED_BITMAP
] = "CBitmap",
266 [P_DELAY_PROBE
] = "DelayProbe",
267 [P_OUT_OF_SYNC
] = "OutOfSync",
271 if (cmd
== P_HAND_SHAKE_M
)
273 if (cmd
== P_HAND_SHAKE_S
)
275 if (cmd
== P_HAND_SHAKE
)
277 if (cmd
>= P_MAX_CMD
)
279 return cmdnames
[cmd
];
282 /* for sending/receiving the bitmap,
283 * possibly in some encoding scheme */
286 * stores total bits and long words
287 * of the bitmap, so we don't need to
288 * call the accessor functions over and again. */
289 unsigned long bm_bits
;
290 unsigned long bm_words
;
291 /* during xfer, current position within the bitmap */
292 unsigned long bit_offset
;
293 unsigned long word_offset
;
295 /* statistics; index: (h->command == P_BITMAP) */
300 extern void INFO_bm_xfer_stats(struct drbd_conf
*mdev
,
301 const char *direction
, struct bm_xfer_ctx
*c
);
303 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx
*c
)
305 /* word_offset counts "native long words" (32 or 64 bit),
307 * Encoded packet may end at an unaligned bit offset.
308 * In case a fallback clear text packet is transmitted in
309 * between, we adjust this offset back to the last 64bit
310 * aligned "native long word", which makes coding and decoding
311 * the plain text bitmap much more convenient. */
312 #if BITS_PER_LONG == 64
313 c
->word_offset
= c
->bit_offset
>> 6;
314 #elif BITS_PER_LONG == 32
315 c
->word_offset
= c
->bit_offset
>> 5;
316 c
->word_offset
&= ~(1UL);
318 # error "unsupported BITS_PER_LONG"
323 #define __packed __attribute__((packed))
326 /* This is the layout for a packet on the wire.
327 * The byteorder is the network byte order.
328 * (except block_id and barrier fields.
329 * these are pointers to local structs
330 * and have no relevance for the partner,
331 * which just echoes them as received.)
333 * NOTE that the payload starts at a long aligned offset,
334 * regardless of 32 or 64 bit arch!
339 u16 length
; /* bytes of data after this header */
343 /* Header for big packets, Used for data packets exceeding 64kB */
345 u16 magic
; /* use DRBD_MAGIC_BIG here */
347 u32 length
; /* Use only 24 bits of that. Ignore the highest 8 bit. */
352 struct p_header80 h80
;
353 struct p_header95 h95
;
357 * short commands, packets without payload, plain p_header:
360 * P_BECOME_SYNC_TARGET
361 * P_BECOME_SYNC_SOURCE
366 * commands with out-of-struct payload:
367 * P_BITMAP (no additional fields)
368 * P_DATA, P_DATA_REPLY (see p_data)
369 * P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
372 /* these defines must not be changed without changing the protocol version */
373 #define DP_HARDBARRIER 1 /* depricated */
374 #define DP_RW_SYNC 2 /* equals REQ_SYNC */
375 #define DP_MAY_SET_IN_SYNC 4
376 #define DP_UNPLUG 8 /* not used anymore */
377 #define DP_FUA 16 /* equals REQ_FUA */
378 #define DP_FLUSH 32 /* equals REQ_FLUSH */
379 #define DP_DISCARD 64 /* equals REQ_DISCARD */
383 u64 sector
; /* 64 bits sector number */
384 u64 block_id
; /* to identify the request in protocol B&C */
390 * commands which share a struct:
392 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
393 * P_DISCARD_ACK (proto C, two-primaries conflict detection)
395 * P_DATA_REQUEST, P_RS_DATA_REQUEST
398 struct p_header80 head
;
407 struct p_header80 head
;
411 u32 pad
; /* to multiple of 8 Byte */
415 * commands with their own struct for additional fields:
424 struct p_header80 head
; /* 8 bytes */
429 /* should be more than enough for future enhancements
430 * for now, feature_flags and the reserverd array shall be zero.
436 /* 80 bytes, FIXED for the next century */
439 struct p_header80 head
;
440 u32 barrier
; /* barrier number _handle_ only */
441 u32 pad
; /* to multiple of 8 Byte */
444 struct p_barrier_ack
{
445 struct p_header80 head
;
451 struct p_header80 head
;
454 /* Since protocol version 88 and higher. */
458 struct p_rs_param_89
{
459 struct p_header80 head
;
461 /* protocol version 89: */
462 char verify_alg
[SHARED_SECRET_MAX
];
463 char csums_alg
[SHARED_SECRET_MAX
];
466 struct p_rs_param_95
{
467 struct p_header80 head
;
469 char verify_alg
[SHARED_SECRET_MAX
];
470 char csums_alg
[SHARED_SECRET_MAX
];
477 enum drbd_conn_flags
{
483 struct p_header80 head
;
491 /* Since protocol version 87 and higher. */
492 char integrity_alg
[0];
497 struct p_header80 head
;
498 u64 uuid
[UI_EXTENDED_SIZE
];
502 struct p_header80 head
;
507 struct p_header80 head
;
508 u64 d_size
; /* size of disk */
509 u64 u_size
; /* user requested size */
510 u64 c_size
; /* current exported size */
511 u32 max_bio_size
; /* Maximal size of a BIO */
512 u16 queue_order_type
; /* not yet implemented in DRBD*/
513 u16 dds_flags
; /* use enum dds_flags here. */
517 struct p_header80 head
;
522 struct p_header80 head
;
527 struct p_req_state_reply
{
528 struct p_header80 head
;
532 struct p_drbd06_param
{
543 struct p_header80 head
;
549 struct p_block_desc
{
550 struct p_header80 head
;
553 u32 pad
; /* to multiple of 8 Byte */
556 /* Valid values for the encoding field.
557 * Bump proto version when changing this. */
558 enum drbd_bitmap_code
{
559 /* RLE_VLI_Bytes = 0,
560 * and other bit variants had been defined during
561 * algorithm evaluation. */
565 struct p_compressed_bm
{
566 struct p_header80 head
;
567 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
568 * (encoding & 0x80): polarity (set/unset) of first runlength
569 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
570 * used to pad up to head.length bytes
577 struct p_delay_probe93
{
578 struct p_header80 head
;
579 u32 seq_num
; /* sequence number to match the two probe packets */
580 u32 offset
; /* usecs the probe got sent after the reference time point */
583 /* DCBP: Drbd Compressed Bitmap Packet ... */
584 static inline enum drbd_bitmap_code
585 DCBP_get_code(struct p_compressed_bm
*p
)
587 return (enum drbd_bitmap_code
)(p
->encoding
& 0x0f);
591 DCBP_set_code(struct p_compressed_bm
*p
, enum drbd_bitmap_code code
)
594 p
->encoding
= (p
->encoding
& ~0xf) | code
;
598 DCBP_get_start(struct p_compressed_bm
*p
)
600 return (p
->encoding
& 0x80) != 0;
604 DCBP_set_start(struct p_compressed_bm
*p
, int set
)
606 p
->encoding
= (p
->encoding
& ~0x80) | (set
? 0x80 : 0);
610 DCBP_get_pad_bits(struct p_compressed_bm
*p
)
612 return (p
->encoding
>> 4) & 0x7;
616 DCBP_set_pad_bits(struct p_compressed_bm
*p
, int n
)
619 p
->encoding
= (p
->encoding
& (~0x7 << 4)) | (n
<< 4);
622 /* one bitmap packet, including the p_header,
623 * should fit within one _architecture independend_ page.
624 * so we need to use the fixed size 4KiB page size
625 * most architectures have used for a long time.
627 #define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header80))
628 #define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
629 #define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm))
630 #if (PAGE_SIZE < 4096)
631 /* drbd_send_bitmap / receive_bitmap would break horribly */
632 #error "PAGE_SIZE too small"
636 union p_header header
;
637 struct p_handshake handshake
;
639 struct p_block_ack block_ack
;
640 struct p_barrier barrier
;
641 struct p_barrier_ack barrier_ack
;
642 struct p_rs_param_89 rs_param_89
;
643 struct p_rs_param_95 rs_param_95
;
644 struct p_protocol protocol
;
645 struct p_sizes sizes
;
646 struct p_uuids uuids
;
647 struct p_state state
;
648 struct p_req_state req_state
;
649 struct p_req_state_reply req_state_reply
;
650 struct p_block_req block_req
;
651 struct p_delay_probe93 delay_probe93
;
652 struct p_rs_uuid rs_uuid
;
653 struct p_block_desc block_desc
;
656 /**********************************************************************/
657 enum drbd_thread_state
{
666 struct task_struct
*task
;
667 struct completion stop
;
668 enum drbd_thread_state t_state
;
669 int (*function
) (struct drbd_thread
*);
670 struct drbd_conf
*mdev
;
674 static inline enum drbd_thread_state
get_t_state(struct drbd_thread
*thi
)
676 /* THINK testing the t_state seems to be uncritical in all cases
677 * (but thread_{start,stop}), so we can read it *without* the lock.
685 typedef int (*drbd_work_cb
)(struct drbd_conf
*, struct drbd_work
*, int cancel
);
687 struct list_head list
;
691 #include "drbd_interval.h"
693 struct drbd_request
{
695 struct drbd_conf
*mdev
;
697 /* if local IO is not allowed, will be NULL.
698 * if local IO _is_ allowed, holds the locally submitted bio clone,
699 * or, after local IO completion, the ERR_PTR(error).
700 * see drbd_endio_pri(). */
701 struct bio
*private_bio
;
703 struct drbd_interval i
;
704 unsigned int epoch
; /* barrier_nr */
706 /* barrier_nr: used to check on "completion" whether this req was in
707 * the current epoch, and we therefore have to close it,
708 * starting a new epoch...
711 struct list_head tl_requests
; /* ring list in the transfer log */
712 struct bio
*master_bio
; /* master bio pointer */
713 unsigned long rq_state
; /* see comments above _req_mod() */
715 unsigned long start_time
;
718 struct drbd_tl_epoch
{
720 struct list_head requests
; /* requests before */
721 struct drbd_tl_epoch
*next
; /* pointer to the next barrier */
722 unsigned int br_number
; /* the barriers identifier. */
723 int n_writes
; /* number of requests attached before this barrier */
728 /* These Tl_epoch_entries may be in one of 6 lists:
729 active_ee .. data packet being written
730 sync_ee .. syncer block being written
731 done_ee .. block written, need to send P_WRITE_ACK
732 read_ee .. [RS]P_DATA_REQUEST being read
736 struct list_head list
;
737 unsigned int barrier_nr
;
738 atomic_t epoch_size
; /* increased on every request added. */
739 atomic_t active
; /* increased on every req. added, and dec on every finished. */
743 /* drbd_epoch flag bits */
745 DE_HAVE_BARRIER_NUMBER
,
752 EV_CLEANUP
= 32, /* used as flag */
755 struct drbd_wq_barrier
{
757 struct completion done
;
765 struct drbd_epoch_entry
{
767 struct drbd_epoch
*epoch
; /* for writes */
768 struct drbd_conf
*mdev
;
770 atomic_t pending_bios
;
771 struct drbd_interval i
;
772 /* see comments on ee flag bits below */
776 struct digest_info
*digest
;
781 * While corresponding bios are in flight, the only modification will be
782 * set_bit WAS_ERROR, which has to be atomic.
783 * If no bios are in flight yet, or all have been completed,
784 * non-atomic modification to ee->flags is ok.
787 __EE_CALL_AL_COMPLETE_IO
,
788 __EE_MAY_SET_IN_SYNC
,
790 /* In case a barrier failed,
791 * we need to resubmit without the barrier flag. */
794 /* we may have several bios per epoch entry.
795 * if any of those fail, we set this flag atomically
796 * from the endio callback */
799 /* This ee has a pointer to a digest instead of a block id */
802 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
803 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
804 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
805 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
806 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
808 /* global flag bits */
810 CREATE_BARRIER
, /* next P_DATA is preceded by a P_BARRIER */
811 SIGNAL_ASENDER
, /* whether asender wants to be interrupted */
812 SEND_PING
, /* whether asender should send a ping asap */
814 UNPLUG_QUEUED
, /* only relevant with kernel 2.4 */
815 UNPLUG_REMOTE
, /* sending a "UnplugRemote" could help */
816 MD_DIRTY
, /* current uuids and flags not yet on disk */
817 DISCARD_CONCURRENT
, /* Set on one node, cleared on the peer! */
818 USE_DEGR_WFC_T
, /* degr-wfc-timeout instead of wfc-timeout. */
819 CLUSTER_ST_CHANGE
, /* Cluster wide state change going on... */
822 CRASHED_PRIMARY
, /* This node was a crashed primary.
823 * Gets cleared when the state.conn
824 * goes into C_CONNECTED state. */
825 NO_BARRIER_SUPP
, /* underlying block device doesn't implement barriers */
828 MD_NO_FUA
, /* Users wants us to not use FUA/FLUSH on meta data dev */
829 SUSPEND_IO
, /* suspend application io */
830 BITMAP_IO
, /* suspend application io;
831 once no more io in flight, start bitmap io */
832 BITMAP_IO_QUEUED
, /* Started bitmap IO */
833 GO_DISKLESS
, /* Disk is being detached, on io-error or admin request. */
834 WAS_IO_ERROR
, /* Local disk failed returned IO error */
835 RESYNC_AFTER_NEG
, /* Resync after online grow after the attach&negotiate finished. */
836 NET_CONGESTED
, /* The data socket is congested */
838 CONFIG_PENDING
, /* serialization of (re)configuration requests.
839 * if set, also prevents the device from dying */
840 DEVICE_DYING
, /* device became unconfigured,
841 * but worker thread is still handling the cleanup.
842 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
843 * while this is set. */
844 RESIZE_PENDING
, /* Size change detected locally, waiting for the response from
845 * the peer, if it changed there as well. */
846 CONN_DRY_RUN
, /* Expect disconnect after resync handshake. */
847 GOT_PING_ACK
, /* set when we receive a ping_ack packet, misc wait gets woken */
848 NEW_CUR_UUID
, /* Create new current UUID when thawing IO */
849 AL_SUSPENDED
, /* Activity logging is currently suspended. */
850 AHEAD_TO_SYNC_SOURCE
, /* Ahead -> SyncSource queued */
853 struct drbd_bitmap
; /* opaque for drbd_conf */
855 /* definition of bits in bm_flags to be used in drbd_bm_lock
856 * and drbd_bitmap_io and friends. */
858 /* do we need to kfree, or vfree bm_pages? */
859 BM_P_VMALLOCED
= 0x10000, /* internal use only, will be masked out */
861 /* currently locked for bulk operation */
862 BM_LOCKED_MASK
= 0x7,
864 /* in detail, that is: */
869 /* (test bit, count bit) allowed (common case) */
870 BM_LOCKED_TEST_ALLOWED
= 0x3,
872 /* testing bits, as well as setting new bits allowed, but clearing bits
873 * would be unexpected. Used during bitmap receive. Setting new bits
874 * requires sending of "out-of-sync" information, though. */
875 BM_LOCKED_SET_ALLOWED
= 0x1,
877 /* clear is not expected while bitmap is locked for bulk operation */
881 /* TODO sort members for performance
882 * MAYBE group them further */
884 /* THINK maybe we actually want to use the default "event/%s" worker threads
885 * or similar in linux 2.6, which uses per cpu data and threads.
887 struct drbd_work_queue
{
889 struct semaphore s
; /* producers up it, worker down()s it */
890 spinlock_t q_lock
; /* to protect the list. */
894 struct drbd_work_queue work
;
896 struct socket
*socket
;
897 /* this way we get our
898 * send/receive buffers off the stack */
899 union p_polymorph sbuf
;
900 union p_polymorph rbuf
;
904 u64 md_offset
; /* sector offset to 'super' block */
906 u64 la_size_sect
; /* last agreed size, unit sectors */
912 s32 al_offset
; /* signed relative sector offset to al area */
913 s32 bm_offset
; /* signed relative sector offset to bitmap */
915 /* u32 al_nr_extents; important for restoring the AL
916 * is stored into sync_conf.al_extents, which in turn
917 * gets applied to act_log->nr_elements
921 /* for sync_conf and other types... */
922 #define NL_PACKET(name, number, fields) struct name { fields };
923 #define NL_INTEGER(pn,pr,member) int member;
924 #define NL_INT64(pn,pr,member) __u64 member;
925 #define NL_BIT(pn,pr,member) unsigned member:1;
926 #define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
927 #include "linux/drbd_nl.h"
929 struct drbd_backing_dev
{
930 struct block_device
*backing_bdev
;
931 struct block_device
*md_bdev
;
933 struct disk_conf dc
; /* The user provided config... */
934 sector_t known_size
; /* last known size of that backing device */
938 struct drbd_conf
*mdev
;
939 struct completion event
;
947 int (*io_fn
)(struct drbd_conf
*mdev
);
948 void (*done
)(struct drbd_conf
*mdev
, int rv
);
951 enum write_ordering_e
{
959 unsigned int head_index
;
964 /* things that are stored as / read from meta data on disk */
967 /* configured by drbdsetup */
968 struct net_conf
*net_conf
; /* protected by get_net_conf() and put_net_conf() */
969 struct syncer_conf sync_conf
;
970 struct drbd_backing_dev
*ldev
__protected_by(local
);
972 sector_t p_size
; /* partner's disk size */
973 struct request_queue
*rq_queue
;
974 struct block_device
*this_bdev
;
975 struct gendisk
*vdisk
;
977 struct drbd_socket data
; /* data/barrier/cstate/parameter packets */
978 struct drbd_socket meta
; /* ping/ack (metadata) packets */
979 int agreed_pro_version
; /* actually used protocol version */
980 unsigned long last_received
; /* in jiffies, either socket */
981 unsigned int ko_count
;
982 struct drbd_work resync_work
,
987 struct timer_list resync_timer
;
988 struct timer_list md_sync_timer
;
989 struct timer_list start_resync_timer
;
990 struct timer_list request_timer
;
991 #ifdef DRBD_DEBUG_MD_SYNC
995 } last_md_mark_dirty
;
998 /* Used after attach while negotiating new disk state. */
999 union drbd_state new_state_tmp
;
1001 union drbd_state state
;
1002 wait_queue_head_t misc_wait
;
1003 wait_queue_head_t state_wait
; /* upon each state change. */
1004 wait_queue_head_t net_cnt_wait
;
1005 unsigned int send_cnt
;
1006 unsigned int recv_cnt
;
1007 unsigned int read_cnt
;
1008 unsigned int writ_cnt
;
1009 unsigned int al_writ_cnt
;
1010 unsigned int bm_writ_cnt
;
1011 atomic_t ap_bio_cnt
; /* Requests we need to complete */
1012 atomic_t ap_pending_cnt
; /* AP data packets on the wire, ack expected */
1013 atomic_t rs_pending_cnt
; /* RS request/data packets on the wire */
1014 atomic_t unacked_cnt
; /* Need to send replys for */
1015 atomic_t local_cnt
; /* Waiting for local completion */
1016 atomic_t net_cnt
; /* Users of net_conf */
1017 spinlock_t req_lock
;
1018 struct drbd_tl_epoch
*unused_spare_tle
; /* for pre-allocation */
1019 struct drbd_tl_epoch
*newest_tle
;
1020 struct drbd_tl_epoch
*oldest_tle
;
1021 struct list_head out_of_sequence_requests
;
1023 /* Interval tree of pending local requests */
1024 struct rb_root read_requests
;
1025 struct rb_root write_requests
;
1027 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
1028 unsigned long rs_total
;
1029 /* number of resync blocks that failed in this run */
1030 unsigned long rs_failed
;
1031 /* Syncer's start time [unit jiffies] */
1032 unsigned long rs_start
;
1033 /* cumulated time in PausedSyncX state [unit jiffies] */
1034 unsigned long rs_paused
;
1035 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
1036 unsigned long rs_same_csum
;
1037 #define DRBD_SYNC_MARKS 8
1038 #define DRBD_SYNC_MARK_STEP (3*HZ)
1039 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
1040 unsigned long rs_mark_left
[DRBD_SYNC_MARKS
];
1041 /* marks's time [unit jiffies] */
1042 unsigned long rs_mark_time
[DRBD_SYNC_MARKS
];
1043 /* current index into rs_mark_{left,time} */
1046 /* where does the admin want us to start? (sector) */
1047 sector_t ov_start_sector
;
1048 /* where are we now? (sector) */
1049 sector_t ov_position
;
1050 /* Start sector of out of sync range (to merge printk reporting). */
1051 sector_t ov_last_oos_start
;
1052 /* size of out-of-sync range in sectors. */
1053 sector_t ov_last_oos_size
;
1054 unsigned long ov_left
; /* in bits */
1055 struct crypto_hash
*csums_tfm
;
1056 struct crypto_hash
*verify_tfm
;
1058 struct drbd_thread receiver
;
1059 struct drbd_thread worker
;
1060 struct drbd_thread asender
;
1061 struct drbd_bitmap
*bitmap
;
1062 unsigned long bm_resync_fo
; /* bit offset for drbd_bm_find_next */
1064 /* Used to track operations of resync... */
1065 struct lru_cache
*resync
;
1066 /* Number of locked elements in resync LRU */
1067 unsigned int resync_locked
;
1068 /* resync extent number waiting for application requests */
1069 unsigned int resync_wenr
;
1073 struct drbd_epoch
*current_epoch
;
1074 spinlock_t epoch_lock
;
1075 unsigned int epochs
;
1076 enum write_ordering_e write_ordering
;
1077 struct list_head active_ee
; /* IO in progress (P_DATA gets written to disk) */
1078 struct list_head sync_ee
; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
1079 struct list_head done_ee
; /* send ack */
1080 struct list_head read_ee
; /* IO in progress (any read) */
1081 struct list_head net_ee
; /* zero-copy network send in progress */
1083 /* Interval tree of pending remote write requests (struct drbd_epoch_entry) */
1084 struct rb_root epoch_entries
;
1086 /* this one is protected by ee_lock, single thread */
1087 struct drbd_epoch_entry
*last_write_w_barrier
;
1089 int next_barrier_nr
;
1090 struct list_head resync_reads
;
1091 atomic_t pp_in_use
; /* allocated from page pool */
1092 atomic_t pp_in_use_by_net
; /* sendpage()d, still referenced by tcp */
1093 wait_queue_head_t ee_wait
;
1094 struct page
*md_io_page
; /* one page buffer for md_io */
1095 struct page
*md_io_tmpp
; /* for logical_block_size != 512 */
1096 struct mutex md_io_mutex
; /* protects the md_io_buffer */
1098 wait_queue_head_t al_wait
;
1099 struct lru_cache
*act_log
; /* activity log */
1100 unsigned int al_tr_number
;
1102 int al_tr_pos
; /* position of the next transaction in the journal */
1103 struct crypto_hash
*cram_hmac_tfm
;
1104 struct crypto_hash
*integrity_w_tfm
; /* to be used by the worker thread */
1105 struct crypto_hash
*integrity_r_tfm
; /* to be used by the receiver thread */
1109 wait_queue_head_t seq_wait
;
1110 atomic_t packet_seq
;
1111 unsigned int peer_seq
;
1112 spinlock_t peer_seq_lock
;
1114 unsigned long comm_bm_set
; /* communicated number of set bits. */
1115 cpumask_var_t cpu_mask
;
1116 struct bm_io_work bm_io_work
;
1117 u64 ed_uuid
; /* UUID of the exposed data */
1118 struct mutex state_mutex
;
1119 char congestion_reason
; /* Why we where congested... */
1120 atomic_t rs_sect_in
; /* for incoming resync data rate, SyncTarget */
1121 atomic_t rs_sect_ev
; /* for submitted resync data rate, both */
1122 int rs_last_sect_ev
; /* counter to compare with */
1123 int rs_last_events
; /* counter of read or write "events" (unit sectors)
1124 * on the lower level device when we last looked. */
1125 int c_sync_rate
; /* current resync rate after syncer throttle magic */
1126 struct fifo_buffer rs_plan_s
; /* correction values of resync planer */
1127 int rs_in_flight
; /* resync sectors in flight (to proxy, in proxy and from proxy) */
1128 int rs_planed
; /* resync sectors already planned */
1129 atomic_t ap_in_flight
; /* App sectors in flight (waiting for ack) */
1130 int peer_max_bio_size
;
1131 int local_max_bio_size
;
1134 static inline struct drbd_conf
*minor_to_mdev(unsigned int minor
)
1136 struct drbd_conf
*mdev
;
1138 mdev
= minor
< minor_count
? minor_table
[minor
] : NULL
;
1143 static inline unsigned int mdev_to_minor(struct drbd_conf
*mdev
)
1148 /* returns 1 if it was successful,
1149 * returns 0 if there was no data socket.
1150 * so wherever you are going to use the data.socket, e.g. do
1151 * if (!drbd_get_data_sock(mdev))
1154 * drbd_put_data_sock(mdev);
1156 static inline int drbd_get_data_sock(struct drbd_conf
*mdev
)
1158 mutex_lock(&mdev
->data
.mutex
);
1159 /* drbd_disconnect() could have called drbd_free_sock()
1160 * while we were waiting in down()... */
1161 if (unlikely(mdev
->data
.socket
== NULL
)) {
1162 mutex_unlock(&mdev
->data
.mutex
);
1168 static inline void drbd_put_data_sock(struct drbd_conf
*mdev
)
1170 mutex_unlock(&mdev
->data
.mutex
);
1174 * function declarations
1175 *************************/
1179 enum chg_state_flags
{
1182 CS_WAIT_COMPLETE
= 4,
1184 CS_ORDERED
= CS_WAIT_COMPLETE
+ CS_SERIALIZE
,
1189 DDSF_NO_RESYNC
= 2, /* Do not run a resync for the new space */
1192 extern void drbd_init_set_defaults(struct drbd_conf
*mdev
);
1193 extern enum drbd_state_rv
drbd_change_state(struct drbd_conf
*mdev
,
1194 enum chg_state_flags f
,
1195 union drbd_state mask
,
1196 union drbd_state val
);
1197 extern void drbd_force_state(struct drbd_conf
*, union drbd_state
,
1199 extern enum drbd_state_rv
_drbd_request_state(struct drbd_conf
*,
1202 enum chg_state_flags
);
1203 extern enum drbd_state_rv
__drbd_set_state(struct drbd_conf
*, union drbd_state
,
1204 enum chg_state_flags
,
1205 struct completion
*done
);
1206 extern void print_st_err(struct drbd_conf
*, union drbd_state
,
1207 union drbd_state
, int);
1208 extern int drbd_thread_start(struct drbd_thread
*thi
);
1209 extern void _drbd_thread_stop(struct drbd_thread
*thi
, int restart
, int wait
);
1211 extern void drbd_thread_current_set_cpu(struct drbd_conf
*mdev
);
1212 extern void drbd_calc_cpu_mask(struct drbd_conf
*mdev
);
1214 #define drbd_thread_current_set_cpu(A) ({})
1215 #define drbd_calc_cpu_mask(A) ({})
1217 extern void drbd_free_resources(struct drbd_conf
*mdev
);
1218 extern void tl_release(struct drbd_conf
*mdev
, unsigned int barrier_nr
,
1219 unsigned int set_size
);
1220 extern void tl_clear(struct drbd_conf
*mdev
);
1221 extern void _tl_add_barrier(struct drbd_conf
*, struct drbd_tl_epoch
*);
1222 extern void drbd_free_sock(struct drbd_conf
*mdev
);
1223 extern int drbd_send(struct drbd_conf
*mdev
, struct socket
*sock
,
1224 void *buf
, size_t size
, unsigned msg_flags
);
1225 extern int drbd_send_protocol(struct drbd_conf
*mdev
);
1226 extern int drbd_send_uuids(struct drbd_conf
*mdev
);
1227 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf
*mdev
);
1228 extern int drbd_gen_and_send_sync_uuid(struct drbd_conf
*mdev
);
1229 extern int drbd_send_sizes(struct drbd_conf
*mdev
, int trigger_reply
, enum dds_flags flags
);
1230 extern int _drbd_send_state(struct drbd_conf
*mdev
);
1231 extern int drbd_send_state(struct drbd_conf
*mdev
);
1232 extern int _drbd_send_cmd(struct drbd_conf
*mdev
, struct socket
*sock
,
1233 enum drbd_packets cmd
, struct p_header80
*h
,
1234 size_t size
, unsigned msg_flags
);
1235 #define USE_DATA_SOCKET 1
1236 #define USE_META_SOCKET 0
1237 extern int drbd_send_cmd(struct drbd_conf
*mdev
, int use_data_socket
,
1238 enum drbd_packets cmd
, struct p_header80
*h
,
1240 extern int drbd_send_cmd2(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1241 char *data
, size_t size
);
1242 extern int drbd_send_sync_param(struct drbd_conf
*mdev
, struct syncer_conf
*sc
);
1243 extern int drbd_send_b_ack(struct drbd_conf
*mdev
, u32 barrier_nr
,
1245 extern int drbd_send_ack(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1246 struct drbd_epoch_entry
*e
);
1247 extern int drbd_send_ack_rp(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1248 struct p_block_req
*rp
);
1249 extern int drbd_send_ack_dp(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1250 struct p_data
*dp
, int data_size
);
1251 extern int drbd_send_ack_ex(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1252 sector_t sector
, int blksize
, u64 block_id
);
1253 extern int drbd_send_oos(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1254 extern int drbd_send_block(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1255 struct drbd_epoch_entry
*e
);
1256 extern int drbd_send_dblock(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1257 extern int drbd_send_drequest(struct drbd_conf
*mdev
, int cmd
,
1258 sector_t sector
, int size
, u64 block_id
);
1259 extern int drbd_send_drequest_csum(struct drbd_conf
*mdev
,
1260 sector_t sector
,int size
,
1261 void *digest
, int digest_size
,
1262 enum drbd_packets cmd
);
1263 extern int drbd_send_ov_request(struct drbd_conf
*mdev
,sector_t sector
,int size
);
1265 extern int drbd_send_bitmap(struct drbd_conf
*mdev
);
1266 extern int _drbd_send_bitmap(struct drbd_conf
*mdev
);
1267 extern int drbd_send_sr_reply(struct drbd_conf
*mdev
, enum drbd_state_rv retcode
);
1268 extern void drbd_free_bc(struct drbd_backing_dev
*ldev
);
1269 extern void drbd_mdev_cleanup(struct drbd_conf
*mdev
);
1270 void drbd_print_uuids(struct drbd_conf
*mdev
, const char *text
);
1272 extern void drbd_md_sync(struct drbd_conf
*mdev
);
1273 extern int drbd_md_read(struct drbd_conf
*mdev
, struct drbd_backing_dev
*bdev
);
1274 extern void drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1275 extern void _drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1276 extern void drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1277 extern void _drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1278 extern void drbd_uuid_set_bm(struct drbd_conf
*mdev
, u64 val
) __must_hold(local
);
1279 extern void drbd_md_set_flag(struct drbd_conf
*mdev
, int flags
) __must_hold(local
);
1280 extern void drbd_md_clear_flag(struct drbd_conf
*mdev
, int flags
)__must_hold(local
);
1281 extern int drbd_md_test_flag(struct drbd_backing_dev
*, int);
1282 #ifndef DRBD_DEBUG_MD_SYNC
1283 extern void drbd_md_mark_dirty(struct drbd_conf
*mdev
);
1285 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1286 extern void drbd_md_mark_dirty_(struct drbd_conf
*mdev
,
1287 unsigned int line
, const char *func
);
1289 extern void drbd_queue_bitmap_io(struct drbd_conf
*mdev
,
1290 int (*io_fn
)(struct drbd_conf
*),
1291 void (*done
)(struct drbd_conf
*, int),
1292 char *why
, enum bm_flag flags
);
1293 extern int drbd_bitmap_io(struct drbd_conf
*mdev
,
1294 int (*io_fn
)(struct drbd_conf
*),
1295 char *why
, enum bm_flag flags
);
1296 extern int drbd_bmio_set_n_write(struct drbd_conf
*mdev
);
1297 extern int drbd_bmio_clear_n_write(struct drbd_conf
*mdev
);
1298 extern void drbd_go_diskless(struct drbd_conf
*mdev
);
1299 extern void drbd_ldev_destroy(struct drbd_conf
*mdev
);
1303 We reserve a 128MB Block (4k aligned)
1304 * either at the end of the backing device
1305 * or on a separate meta data device. */
1307 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1308 /* The following numbers are sectors */
1309 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1310 #define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */
1311 /* Allows up to about 3.8TB */
1312 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1314 /* Since the smalles IO unit is usually 512 byte */
1315 #define MD_SECTOR_SHIFT 9
1316 #define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT)
1319 #define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1320 #define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */
1321 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1323 #if BITS_PER_LONG == 32
1325 #define cpu_to_lel(A) cpu_to_le32(A)
1326 #define lel_to_cpu(A) le32_to_cpu(A)
1327 #elif BITS_PER_LONG == 64
1329 #define cpu_to_lel(A) cpu_to_le64(A)
1330 #define lel_to_cpu(A) le64_to_cpu(A)
1332 #error "LN2 of BITS_PER_LONG unknown!"
1336 /* 16MB sized 'bitmap extent' to track syncer usage */
1338 int rs_left
; /* number of bits set (out of sync) in this extent. */
1339 int rs_failed
; /* number of failed resync requests in this extent. */
1340 unsigned long flags
;
1341 struct lc_element lce
;
1344 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1345 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1346 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1350 * We need to store one bit for a block.
1351 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1352 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1353 * Bit 1 ==> local node thinks this block needs to be synced.
1356 #define SLEEP_TIME (HZ/10)
1358 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1359 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1360 /* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1361 * per sector of on disk bitmap */
1362 #define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */
1363 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1365 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1366 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1369 /* thus many _storage_ sectors are described by one bit */
1370 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1371 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1372 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1374 /* bit to represented kilo byte conversion */
1375 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1377 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1378 * _storage_ sector is located in */
1379 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1381 /* how much _storage_ sectors we have per bitmap sector */
1382 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1383 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1385 /* in one sector of the bitmap, we have this many activity_log extents. */
1386 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1387 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1389 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1390 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1392 /* the extent in "PER_EXTENT" below is an activity log extent
1393 * we need that many (long words/bytes) to store the bitmap
1394 * of one AL_EXTENT_SIZE chunk of storage.
1395 * we can store the bitmap for that many AL_EXTENTS within
1396 * one sector of the _on_disk_ bitmap:
1397 * bit 0 bit 37 bit 38 bit (512*8)-1
1398 * ...|........|........|.. // ..|........|
1399 * sect. 0 `296 `304 ^(512*8*8)-1
1401 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1402 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1403 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1406 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1407 #define DRBD_MAX_SECTORS_BM \
1408 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1409 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1410 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1411 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1412 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1413 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1414 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1416 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1417 /* 16 TB in units of sectors */
1418 #if BITS_PER_LONG == 32
1419 /* adjust by one page worth of bitmap,
1420 * so we won't wrap around in drbd_bm_find_next_bit.
1421 * you should use 64bit OS for that much storage, anyways. */
1422 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1424 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1425 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1426 /* corresponds to (1UL << 38) bits right now. */
1431 #define DRBD_MAX_BIO_SIZE (1U<<(9+HT_SHIFT))
1432 #define DRBD_MAX_BIO_SIZE_SAFE (1 << 12) /* Works always = 4k */
1434 #define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */
1436 extern int drbd_bm_init(struct drbd_conf
*mdev
);
1437 extern int drbd_bm_resize(struct drbd_conf
*mdev
, sector_t sectors
, int set_new_bits
);
1438 extern void drbd_bm_cleanup(struct drbd_conf
*mdev
);
1439 extern void drbd_bm_set_all(struct drbd_conf
*mdev
);
1440 extern void drbd_bm_clear_all(struct drbd_conf
*mdev
);
1441 /* set/clear/test only a few bits at a time */
1442 extern int drbd_bm_set_bits(
1443 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1444 extern int drbd_bm_clear_bits(
1445 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1446 extern int drbd_bm_count_bits(
1447 struct drbd_conf
*mdev
, const unsigned long s
, const unsigned long e
);
1448 /* bm_set_bits variant for use while holding drbd_bm_lock,
1449 * may process the whole bitmap in one go */
1450 extern void _drbd_bm_set_bits(struct drbd_conf
*mdev
,
1451 const unsigned long s
, const unsigned long e
);
1452 extern int drbd_bm_test_bit(struct drbd_conf
*mdev
, unsigned long bitnr
);
1453 extern int drbd_bm_e_weight(struct drbd_conf
*mdev
, unsigned long enr
);
1454 extern int drbd_bm_write_page(struct drbd_conf
*mdev
, unsigned int idx
) __must_hold(local
);
1455 extern int drbd_bm_read(struct drbd_conf
*mdev
) __must_hold(local
);
1456 extern int drbd_bm_write(struct drbd_conf
*mdev
) __must_hold(local
);
1457 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf
*mdev
,
1458 unsigned long al_enr
);
1459 extern size_t drbd_bm_words(struct drbd_conf
*mdev
);
1460 extern unsigned long drbd_bm_bits(struct drbd_conf
*mdev
);
1461 extern sector_t
drbd_bm_capacity(struct drbd_conf
*mdev
);
1463 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1464 extern unsigned long drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1465 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1466 extern unsigned long _drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1467 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1468 extern unsigned long _drbd_bm_total_weight(struct drbd_conf
*mdev
);
1469 extern unsigned long drbd_bm_total_weight(struct drbd_conf
*mdev
);
1470 extern int drbd_bm_rs_done(struct drbd_conf
*mdev
);
1471 /* for receive_bitmap */
1472 extern void drbd_bm_merge_lel(struct drbd_conf
*mdev
, size_t offset
,
1473 size_t number
, unsigned long *buffer
);
1474 /* for _drbd_send_bitmap */
1475 extern void drbd_bm_get_lel(struct drbd_conf
*mdev
, size_t offset
,
1476 size_t number
, unsigned long *buffer
);
1478 extern void drbd_bm_lock(struct drbd_conf
*mdev
, char *why
, enum bm_flag flags
);
1479 extern void drbd_bm_unlock(struct drbd_conf
*mdev
);
1482 extern struct kmem_cache
*drbd_request_cache
;
1483 extern struct kmem_cache
*drbd_ee_cache
; /* epoch entries */
1484 extern struct kmem_cache
*drbd_bm_ext_cache
; /* bitmap extents */
1485 extern struct kmem_cache
*drbd_al_ext_cache
; /* activity log extents */
1486 extern mempool_t
*drbd_request_mempool
;
1487 extern mempool_t
*drbd_ee_mempool
;
1489 extern struct page
*drbd_pp_pool
; /* drbd's page pool */
1490 extern spinlock_t drbd_pp_lock
;
1491 extern int drbd_pp_vacant
;
1492 extern wait_queue_head_t drbd_pp_wait
;
1494 extern rwlock_t global_state_lock
;
1496 extern struct drbd_conf
*drbd_new_device(unsigned int minor
);
1497 extern void drbd_free_mdev(struct drbd_conf
*mdev
);
1499 extern int proc_details
;
1502 extern int drbd_make_request(struct request_queue
*q
, struct bio
*bio
);
1503 extern int drbd_read_remote(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1504 extern int drbd_merge_bvec(struct request_queue
*q
, struct bvec_merge_data
*bvm
, struct bio_vec
*bvec
);
1505 extern int is_valid_ar_handle(struct drbd_request
*, sector_t
);
1509 extern void drbd_suspend_io(struct drbd_conf
*mdev
);
1510 extern void drbd_resume_io(struct drbd_conf
*mdev
);
1511 extern char *ppsize(char *buf
, unsigned long long size
);
1512 extern sector_t
drbd_new_dev_size(struct drbd_conf
*, struct drbd_backing_dev
*, int);
1513 enum determine_dev_size
{ dev_size_error
= -1, unchanged
= 0, shrunk
= 1, grew
= 2 };
1514 extern enum determine_dev_size
drbd_determine_dev_size(struct drbd_conf
*, enum dds_flags
) __must_hold(local
);
1515 extern void resync_after_online_grow(struct drbd_conf
*);
1516 extern void drbd_reconsider_max_bio_size(struct drbd_conf
*mdev
);
1517 extern enum drbd_state_rv
drbd_set_role(struct drbd_conf
*mdev
,
1518 enum drbd_role new_role
,
1520 extern enum drbd_disk_state
drbd_try_outdate_peer(struct drbd_conf
*mdev
);
1521 extern void drbd_try_outdate_peer_async(struct drbd_conf
*mdev
);
1522 extern int drbd_khelper(struct drbd_conf
*mdev
, char *cmd
);
1525 extern int drbd_worker(struct drbd_thread
*thi
);
1526 extern int drbd_alter_sa(struct drbd_conf
*mdev
, int na
);
1527 extern void drbd_start_resync(struct drbd_conf
*mdev
, enum drbd_conns side
);
1528 extern void resume_next_sg(struct drbd_conf
*mdev
);
1529 extern void suspend_other_sg(struct drbd_conf
*mdev
);
1530 extern int drbd_resync_finished(struct drbd_conf
*mdev
);
1531 /* maybe rather drbd_main.c ? */
1532 extern int drbd_md_sync_page_io(struct drbd_conf
*mdev
,
1533 struct drbd_backing_dev
*bdev
, sector_t sector
, int rw
);
1534 extern void drbd_ov_oos_found(struct drbd_conf
*, sector_t
, int);
1535 extern void drbd_rs_controller_reset(struct drbd_conf
*mdev
);
1537 static inline void ov_oos_print(struct drbd_conf
*mdev
)
1539 if (mdev
->ov_last_oos_size
) {
1540 dev_err(DEV
, "Out of sync: start=%llu, size=%lu (sectors)\n",
1541 (unsigned long long)mdev
->ov_last_oos_start
,
1542 (unsigned long)mdev
->ov_last_oos_size
);
1544 mdev
->ov_last_oos_size
=0;
1548 extern void drbd_csum_bio(struct drbd_conf
*, struct crypto_hash
*, struct bio
*, void *);
1549 extern void drbd_csum_ee(struct drbd_conf
*, struct crypto_hash
*, struct drbd_epoch_entry
*, void *);
1550 /* worker callbacks */
1551 extern int w_req_cancel_conflict(struct drbd_conf
*, struct drbd_work
*, int);
1552 extern int w_read_retry_remote(struct drbd_conf
*, struct drbd_work
*, int);
1553 extern int w_e_end_data_req(struct drbd_conf
*, struct drbd_work
*, int);
1554 extern int w_e_end_rsdata_req(struct drbd_conf
*, struct drbd_work
*, int);
1555 extern int w_e_end_csum_rs_req(struct drbd_conf
*, struct drbd_work
*, int);
1556 extern int w_e_end_ov_reply(struct drbd_conf
*, struct drbd_work
*, int);
1557 extern int w_e_end_ov_req(struct drbd_conf
*, struct drbd_work
*, int);
1558 extern int w_ov_finished(struct drbd_conf
*, struct drbd_work
*, int);
1559 extern int w_resync_timer(struct drbd_conf
*, struct drbd_work
*, int);
1560 extern int w_resume_next_sg(struct drbd_conf
*, struct drbd_work
*, int);
1561 extern int w_send_write_hint(struct drbd_conf
*, struct drbd_work
*, int);
1562 extern int w_send_dblock(struct drbd_conf
*, struct drbd_work
*, int);
1563 extern int w_send_barrier(struct drbd_conf
*, struct drbd_work
*, int);
1564 extern int w_send_read_req(struct drbd_conf
*, struct drbd_work
*, int);
1565 extern int w_prev_work_done(struct drbd_conf
*, struct drbd_work
*, int);
1566 extern int w_e_reissue(struct drbd_conf
*, struct drbd_work
*, int);
1567 extern int w_restart_disk_io(struct drbd_conf
*, struct drbd_work
*, int);
1568 extern int w_send_oos(struct drbd_conf
*, struct drbd_work
*, int);
1569 extern int w_start_resync(struct drbd_conf
*, struct drbd_work
*, int);
1571 extern void resync_timer_fn(unsigned long data
);
1572 extern void start_resync_timer_fn(unsigned long data
);
1574 /* drbd_receiver.c */
1575 extern int drbd_rs_should_slow_down(struct drbd_conf
*mdev
, sector_t sector
);
1576 extern int drbd_submit_ee(struct drbd_conf
*mdev
, struct drbd_epoch_entry
*e
,
1577 const unsigned rw
, const int fault_type
);
1578 extern int drbd_release_ee(struct drbd_conf
*mdev
, struct list_head
*list
);
1579 extern struct drbd_epoch_entry
*drbd_alloc_ee(struct drbd_conf
*mdev
,
1582 unsigned int data_size
,
1583 gfp_t gfp_mask
) __must_hold(local
);
1584 extern void drbd_free_some_ee(struct drbd_conf
*mdev
, struct drbd_epoch_entry
*e
,
1586 #define drbd_free_ee(m,e) drbd_free_some_ee(m, e, 0)
1587 #define drbd_free_net_ee(m,e) drbd_free_some_ee(m, e, 1)
1588 extern void drbd_wait_ee_list_empty(struct drbd_conf
*mdev
,
1589 struct list_head
*head
);
1590 extern void _drbd_wait_ee_list_empty(struct drbd_conf
*mdev
,
1591 struct list_head
*head
);
1592 extern void drbd_set_recv_tcq(struct drbd_conf
*mdev
, int tcq_enabled
);
1593 extern void _drbd_clear_done_ee(struct drbd_conf
*mdev
, struct list_head
*to_be_freed
);
1594 extern void drbd_flush_workqueue(struct drbd_conf
*mdev
);
1596 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1597 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1598 static inline int drbd_setsockopt(struct socket
*sock
, int level
, int optname
,
1599 char __user
*optval
, int optlen
)
1602 if (level
== SOL_SOCKET
)
1603 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
1605 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1610 static inline void drbd_tcp_cork(struct socket
*sock
)
1613 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1614 (char __user
*)&val
, sizeof(val
));
1617 static inline void drbd_tcp_uncork(struct socket
*sock
)
1620 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1621 (char __user
*)&val
, sizeof(val
));
1624 static inline void drbd_tcp_nodelay(struct socket
*sock
)
1627 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_NODELAY
,
1628 (char __user
*)&val
, sizeof(val
));
1631 static inline void drbd_tcp_quickack(struct socket
*sock
)
1634 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_QUICKACK
,
1635 (char __user
*)&val
, sizeof(val
));
1638 void drbd_bump_write_ordering(struct drbd_conf
*mdev
, enum write_ordering_e wo
);
1641 extern struct proc_dir_entry
*drbd_proc
;
1642 extern const struct file_operations drbd_proc_fops
;
1643 extern const char *drbd_conn_str(enum drbd_conns s
);
1644 extern const char *drbd_role_str(enum drbd_role s
);
1647 extern void drbd_al_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1648 extern void drbd_al_complete_io(struct drbd_conf
*mdev
, sector_t sector
);
1649 extern void drbd_rs_complete_io(struct drbd_conf
*mdev
, sector_t sector
);
1650 extern int drbd_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1651 extern int drbd_try_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1652 extern void drbd_rs_cancel_all(struct drbd_conf
*mdev
);
1653 extern int drbd_rs_del_all(struct drbd_conf
*mdev
);
1654 extern void drbd_rs_failed_io(struct drbd_conf
*mdev
,
1655 sector_t sector
, int size
);
1656 extern int drbd_al_read_log(struct drbd_conf
*mdev
, struct drbd_backing_dev
*);
1657 extern void drbd_advance_rs_marks(struct drbd_conf
*mdev
, unsigned long still_to_go
);
1658 extern void __drbd_set_in_sync(struct drbd_conf
*mdev
, sector_t sector
,
1659 int size
, const char *file
, const unsigned int line
);
1660 #define drbd_set_in_sync(mdev, sector, size) \
1661 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1662 extern int __drbd_set_out_of_sync(struct drbd_conf
*mdev
, sector_t sector
,
1663 int size
, const char *file
, const unsigned int line
);
1664 #define drbd_set_out_of_sync(mdev, sector, size) \
1665 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1666 extern void drbd_al_apply_to_bm(struct drbd_conf
*mdev
);
1667 extern void drbd_al_shrink(struct drbd_conf
*mdev
);
1672 void drbd_nl_cleanup(void);
1673 int __init
drbd_nl_init(void);
1674 void drbd_bcast_state(struct drbd_conf
*mdev
, union drbd_state
);
1675 void drbd_bcast_sync_progress(struct drbd_conf
*mdev
);
1676 void drbd_bcast_ee(struct drbd_conf
*mdev
,
1677 const char *reason
, const int dgs
,
1678 const char* seen_hash
, const char* calc_hash
,
1679 const struct drbd_epoch_entry
* e
);
1683 * DOC: DRBD State macros
1685 * These macros are used to express state changes in easily readable form.
1687 * The NS macros expand to a mask and a value, that can be bit ored onto the
1688 * current state as soon as the spinlock (req_lock) was taken.
1690 * The _NS macros are used for state functions that get called with the
1691 * spinlock. These macros expand directly to the new state value.
1693 * Besides the basic forms NS() and _NS() additional _?NS[23] are defined
1694 * to express state changes that affect more than one aspect of the state.
1696 * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY)
1697 * Means that the network connection was established and that the peer
1698 * is in secondary role.
1700 #define role_MASK R_MASK
1701 #define peer_MASK R_MASK
1702 #define disk_MASK D_MASK
1703 #define pdsk_MASK D_MASK
1704 #define conn_MASK C_MASK
1706 #define user_isp_MASK 1
1707 #define aftr_isp_MASK 1
1708 #define susp_nod_MASK 1
1709 #define susp_fen_MASK 1
1712 ({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \
1713 ({ union drbd_state val; val.i = 0; val.T = (S); val; })
1714 #define NS2(T1, S1, T2, S2) \
1715 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1716 mask.T2 = T2##_MASK; mask; }), \
1717 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1718 val.T2 = (S2); val; })
1719 #define NS3(T1, S1, T2, S2, T3, S3) \
1720 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1721 mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \
1722 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1723 val.T2 = (S2); val.T3 = (S3); val; })
1725 #define _NS(D, T, S) \
1726 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; })
1727 #define _NS2(D, T1, S1, T2, S2) \
1728 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1729 __ns.T2 = (S2); __ns; })
1730 #define _NS3(D, T1, S1, T2, S2, T3, S3) \
1731 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1732 __ns.T2 = (S2); __ns.T3 = (S3); __ns; })
1735 * inline helper functions
1736 *************************/
1738 /* see also page_chain_add and friends in drbd_receiver.c */
1739 static inline struct page
*page_chain_next(struct page
*page
)
1741 return (struct page
*)page_private(page
);
1743 #define page_chain_for_each(page) \
1744 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1745 page = page_chain_next(page))
1746 #define page_chain_for_each_safe(page, n) \
1747 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1749 static inline int drbd_bio_has_active_page(struct bio
*bio
)
1751 struct bio_vec
*bvec
;
1754 __bio_for_each_segment(bvec
, bio
, i
, 0) {
1755 if (page_count(bvec
->bv_page
) > 1)
1762 static inline int drbd_ee_has_active_page(struct drbd_epoch_entry
*e
)
1764 struct page
*page
= e
->pages
;
1765 page_chain_for_each(page
) {
1766 if (page_count(page
) > 1)
1773 static inline void drbd_state_lock(struct drbd_conf
*mdev
)
1775 wait_event(mdev
->misc_wait
,
1776 !test_and_set_bit(CLUSTER_ST_CHANGE
, &mdev
->flags
));
1779 static inline void drbd_state_unlock(struct drbd_conf
*mdev
)
1781 clear_bit(CLUSTER_ST_CHANGE
, &mdev
->flags
);
1782 wake_up(&mdev
->misc_wait
);
1785 static inline enum drbd_state_rv
1786 _drbd_set_state(struct drbd_conf
*mdev
, union drbd_state ns
,
1787 enum chg_state_flags flags
, struct completion
*done
)
1789 enum drbd_state_rv rv
;
1791 read_lock(&global_state_lock
);
1792 rv
= __drbd_set_state(mdev
, ns
, flags
, done
);
1793 read_unlock(&global_state_lock
);
1799 * drbd_request_state() - Reqest a state change
1800 * @mdev: DRBD device.
1801 * @mask: mask of state bits to change.
1802 * @val: value of new state bits.
1804 * This is the most graceful way of requesting a state change. It is verbose
1805 * quite verbose in case the state change is not possible, and all those
1806 * state changes are globally serialized.
1808 static inline int drbd_request_state(struct drbd_conf
*mdev
,
1809 union drbd_state mask
,
1810 union drbd_state val
)
1812 return _drbd_request_state(mdev
, mask
, val
, CS_VERBOSE
+ CS_ORDERED
);
1815 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1816 static inline void __drbd_chk_io_error_(struct drbd_conf
*mdev
, int forcedetach
, const char *where
)
1818 switch (mdev
->ldev
->dc
.on_io_error
) {
1821 if (__ratelimit(&drbd_ratelimit_state
))
1822 dev_err(DEV
, "Local IO failed in %s.\n", where
);
1823 if (mdev
->state
.disk
> D_INCONSISTENT
)
1824 _drbd_set_state(_NS(mdev
, disk
, D_INCONSISTENT
), CS_HARD
, NULL
);
1827 /* NOTE fall through to detach case if forcedetach set */
1829 case EP_CALL_HELPER
:
1830 set_bit(WAS_IO_ERROR
, &mdev
->flags
);
1831 if (mdev
->state
.disk
> D_FAILED
) {
1832 _drbd_set_state(_NS(mdev
, disk
, D_FAILED
), CS_HARD
, NULL
);
1834 "Local IO failed in %s. Detaching...\n", where
);
1841 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1842 * @mdev: DRBD device.
1843 * @error: Error code passed to the IO completion callback
1844 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1846 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1848 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1849 static inline void drbd_chk_io_error_(struct drbd_conf
*mdev
,
1850 int error
, int forcedetach
, const char *where
)
1853 unsigned long flags
;
1854 spin_lock_irqsave(&mdev
->req_lock
, flags
);
1855 __drbd_chk_io_error_(mdev
, forcedetach
, where
);
1856 spin_unlock_irqrestore(&mdev
->req_lock
, flags
);
1862 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1863 * @bdev: Meta data block device.
1865 * BTW, for internal meta data, this happens to be the maximum capacity
1866 * we could agree upon with our peer node.
1868 static inline sector_t
drbd_md_first_sector(struct drbd_backing_dev
*bdev
)
1870 switch (bdev
->dc
.meta_dev_idx
) {
1871 case DRBD_MD_INDEX_INTERNAL
:
1872 case DRBD_MD_INDEX_FLEX_INT
:
1873 return bdev
->md
.md_offset
+ bdev
->md
.bm_offset
;
1874 case DRBD_MD_INDEX_FLEX_EXT
:
1876 return bdev
->md
.md_offset
;
1881 * drbd_md_last_sector() - Return the last sector number of the meta data area
1882 * @bdev: Meta data block device.
1884 static inline sector_t
drbd_md_last_sector(struct drbd_backing_dev
*bdev
)
1886 switch (bdev
->dc
.meta_dev_idx
) {
1887 case DRBD_MD_INDEX_INTERNAL
:
1888 case DRBD_MD_INDEX_FLEX_INT
:
1889 return bdev
->md
.md_offset
+ MD_AL_OFFSET
- 1;
1890 case DRBD_MD_INDEX_FLEX_EXT
:
1892 return bdev
->md
.md_offset
+ bdev
->md
.md_size_sect
;
1896 /* Returns the number of 512 byte sectors of the device */
1897 static inline sector_t
drbd_get_capacity(struct block_device
*bdev
)
1899 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1900 return bdev
? i_size_read(bdev
->bd_inode
) >> 9 : 0;
1904 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1905 * @bdev: Meta data block device.
1907 * returns the capacity we announce to out peer. we clip ourselves at the
1908 * various MAX_SECTORS, because if we don't, current implementation will
1909 * oops sooner or later
1911 static inline sector_t
drbd_get_max_capacity(struct drbd_backing_dev
*bdev
)
1914 switch (bdev
->dc
.meta_dev_idx
) {
1915 case DRBD_MD_INDEX_INTERNAL
:
1916 case DRBD_MD_INDEX_FLEX_INT
:
1917 s
= drbd_get_capacity(bdev
->backing_bdev
)
1918 ? min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1919 drbd_md_first_sector(bdev
))
1922 case DRBD_MD_INDEX_FLEX_EXT
:
1923 s
= min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1924 drbd_get_capacity(bdev
->backing_bdev
));
1925 /* clip at maximum size the meta device can support */
1926 s
= min_t(sector_t
, s
,
1927 BM_EXT_TO_SECT(bdev
->md
.md_size_sect
1928 - bdev
->md
.bm_offset
));
1931 s
= min_t(sector_t
, DRBD_MAX_SECTORS
,
1932 drbd_get_capacity(bdev
->backing_bdev
));
1938 * drbd_md_ss__() - Return the sector number of our meta data super block
1939 * @mdev: DRBD device.
1940 * @bdev: Meta data block device.
1942 static inline sector_t
drbd_md_ss__(struct drbd_conf
*mdev
,
1943 struct drbd_backing_dev
*bdev
)
1945 switch (bdev
->dc
.meta_dev_idx
) {
1946 default: /* external, some index */
1947 return MD_RESERVED_SECT
* bdev
->dc
.meta_dev_idx
;
1948 case DRBD_MD_INDEX_INTERNAL
:
1949 /* with drbd08, internal meta data is always "flexible" */
1950 case DRBD_MD_INDEX_FLEX_INT
:
1951 /* sizeof(struct md_on_disk_07) == 4k
1952 * position: last 4k aligned block of 4k size */
1953 if (!bdev
->backing_bdev
) {
1954 if (__ratelimit(&drbd_ratelimit_state
)) {
1955 dev_err(DEV
, "bdev->backing_bdev==NULL\n");
1960 return (drbd_get_capacity(bdev
->backing_bdev
) & ~7ULL)
1962 case DRBD_MD_INDEX_FLEX_EXT
:
1968 drbd_queue_work_front(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1970 unsigned long flags
;
1971 spin_lock_irqsave(&q
->q_lock
, flags
);
1972 list_add(&w
->list
, &q
->q
);
1973 up(&q
->s
); /* within the spinlock,
1974 see comment near end of drbd_worker() */
1975 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1979 drbd_queue_work(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1981 unsigned long flags
;
1982 spin_lock_irqsave(&q
->q_lock
, flags
);
1983 list_add_tail(&w
->list
, &q
->q
);
1984 up(&q
->s
); /* within the spinlock,
1985 see comment near end of drbd_worker() */
1986 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1989 static inline void wake_asender(struct drbd_conf
*mdev
)
1991 if (test_bit(SIGNAL_ASENDER
, &mdev
->flags
))
1992 force_sig(DRBD_SIG
, mdev
->asender
.task
);
1995 static inline void request_ping(struct drbd_conf
*mdev
)
1997 set_bit(SEND_PING
, &mdev
->flags
);
2001 static inline int drbd_send_short_cmd(struct drbd_conf
*mdev
,
2002 enum drbd_packets cmd
)
2004 struct p_header80 h
;
2005 return drbd_send_cmd(mdev
, USE_DATA_SOCKET
, cmd
, &h
, sizeof(h
));
2008 static inline int drbd_send_ping(struct drbd_conf
*mdev
)
2010 struct p_header80 h
;
2011 return drbd_send_cmd(mdev
, USE_META_SOCKET
, P_PING
, &h
, sizeof(h
));
2014 static inline int drbd_send_ping_ack(struct drbd_conf
*mdev
)
2016 struct p_header80 h
;
2017 return drbd_send_cmd(mdev
, USE_META_SOCKET
, P_PING_ACK
, &h
, sizeof(h
));
2020 static inline void drbd_thread_stop(struct drbd_thread
*thi
)
2022 _drbd_thread_stop(thi
, false, true);
2025 static inline void drbd_thread_stop_nowait(struct drbd_thread
*thi
)
2027 _drbd_thread_stop(thi
, false, false);
2030 static inline void drbd_thread_restart_nowait(struct drbd_thread
*thi
)
2032 _drbd_thread_stop(thi
, true, false);
2035 /* counts how many answer packets packets we expect from our peer,
2036 * for either explicit application requests,
2037 * or implicit barrier packets as necessary.
2040 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
2041 * it is much easier and equally valid to count what we queue for the
2042 * worker, even before it actually was queued or send.
2043 * (drbd_make_request_common; recovery path on read io-error)
2045 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
2046 * _req_mod(req, DATA_RECEIVED)
2047 * [from receive_DataReply]
2048 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
2049 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
2050 * for some reason it is NOT decreased in got_NegAck,
2051 * but in the resulting cleanup code from report_params.
2052 * we should try to remember the reason for that...
2053 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
2054 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
2055 * [from tl_clear_barrier]
2057 static inline void inc_ap_pending(struct drbd_conf
*mdev
)
2059 atomic_inc(&mdev
->ap_pending_cnt
);
2062 #define ERR_IF_CNT_IS_NEGATIVE(which) \
2063 if (atomic_read(&mdev->which) < 0) \
2064 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
2065 __func__ , __LINE__ , \
2066 atomic_read(&mdev->which))
2068 #define dec_ap_pending(mdev) do { \
2069 typecheck(struct drbd_conf *, mdev); \
2070 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \
2071 wake_up(&mdev->misc_wait); \
2072 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
2074 /* counts how many resync-related answers we still expect from the peer
2076 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
2077 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
2078 * (or P_NEG_ACK with ID_SYNCER)
2080 static inline void inc_rs_pending(struct drbd_conf
*mdev
)
2082 atomic_inc(&mdev
->rs_pending_cnt
);
2085 #define dec_rs_pending(mdev) do { \
2086 typecheck(struct drbd_conf *, mdev); \
2087 atomic_dec(&mdev->rs_pending_cnt); \
2088 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
2090 /* counts how many answers we still need to send to the peer.
2092 * receive_Data unless protocol A;
2093 * we need to send a P_RECV_ACK (proto B)
2094 * or P_WRITE_ACK (proto C)
2095 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2096 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2097 * receive_Barrier_* we need to send a P_BARRIER_ACK
2099 static inline void inc_unacked(struct drbd_conf
*mdev
)
2101 atomic_inc(&mdev
->unacked_cnt
);
2104 #define dec_unacked(mdev) do { \
2105 typecheck(struct drbd_conf *, mdev); \
2106 atomic_dec(&mdev->unacked_cnt); \
2107 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2109 #define sub_unacked(mdev, n) do { \
2110 typecheck(struct drbd_conf *, mdev); \
2111 atomic_sub(n, &mdev->unacked_cnt); \
2112 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2115 static inline void put_net_conf(struct drbd_conf
*mdev
)
2117 if (atomic_dec_and_test(&mdev
->net_cnt
))
2118 wake_up(&mdev
->net_cnt_wait
);
2122 * get_net_conf() - Increase ref count on mdev->net_conf; Returns 0 if nothing there
2123 * @mdev: DRBD device.
2125 * You have to call put_net_conf() when finished working with mdev->net_conf.
2127 static inline int get_net_conf(struct drbd_conf
*mdev
)
2131 atomic_inc(&mdev
->net_cnt
);
2132 have_net_conf
= mdev
->state
.conn
>= C_UNCONNECTED
;
2135 return have_net_conf
;
2139 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2142 * You have to call put_ldev() when finished working with mdev->ldev.
2144 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2145 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2147 static inline void put_ldev(struct drbd_conf
*mdev
)
2149 int i
= atomic_dec_return(&mdev
->local_cnt
);
2151 /* This may be called from some endio handler,
2152 * so we must not sleep here. */
2157 if (mdev
->state
.disk
== D_DISKLESS
)
2158 /* even internal references gone, safe to destroy */
2159 drbd_ldev_destroy(mdev
);
2160 if (mdev
->state
.disk
== D_FAILED
)
2161 /* all application IO references gone. */
2162 drbd_go_diskless(mdev
);
2163 wake_up(&mdev
->misc_wait
);
2168 static inline int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
)
2172 /* never get a reference while D_DISKLESS */
2173 if (mdev
->state
.disk
== D_DISKLESS
)
2176 atomic_inc(&mdev
->local_cnt
);
2177 io_allowed
= (mdev
->state
.disk
>= mins
);
2183 extern int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
);
2186 /* you must have an "get_ldev" reference */
2187 static inline void drbd_get_syncer_progress(struct drbd_conf
*mdev
,
2188 unsigned long *bits_left
, unsigned int *per_mil_done
)
2190 /* this is to break it at compile time when we change that, in case we
2191 * want to support more than (1<<32) bits on a 32bit arch. */
2192 typecheck(unsigned long, mdev
->rs_total
);
2194 /* note: both rs_total and rs_left are in bits, i.e. in
2195 * units of BM_BLOCK_SIZE.
2196 * for the percentage, we don't care. */
2198 if (mdev
->state
.conn
== C_VERIFY_S
|| mdev
->state
.conn
== C_VERIFY_T
)
2199 *bits_left
= mdev
->ov_left
;
2201 *bits_left
= drbd_bm_total_weight(mdev
) - mdev
->rs_failed
;
2202 /* >> 10 to prevent overflow,
2203 * +1 to prevent division by zero */
2204 if (*bits_left
> mdev
->rs_total
) {
2205 /* doh. maybe a logic bug somewhere.
2206 * may also be just a race condition
2207 * between this and a disconnect during sync.
2208 * for now, just prevent in-kernel buffer overflow.
2211 dev_warn(DEV
, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2212 drbd_conn_str(mdev
->state
.conn
),
2213 *bits_left
, mdev
->rs_total
, mdev
->rs_failed
);
2216 /* Make sure the division happens in long context.
2217 * We allow up to one petabyte storage right now,
2218 * at a granularity of 4k per bit that is 2**38 bits.
2219 * After shift right and multiplication by 1000,
2220 * this should still fit easily into a 32bit long,
2221 * so we don't need a 64bit division on 32bit arch.
2222 * Note: currently we don't support such large bitmaps on 32bit
2223 * arch anyways, but no harm done to be prepared for it here.
2225 unsigned int shift
= mdev
->rs_total
>= (1ULL << 32) ? 16 : 10;
2226 unsigned long left
= *bits_left
>> shift
;
2227 unsigned long total
= 1UL + (mdev
->rs_total
>> shift
);
2228 unsigned long tmp
= 1000UL - left
* 1000UL/total
;
2229 *per_mil_done
= tmp
;
2234 /* this throttles on-the-fly application requests
2235 * according to max_buffers settings;
2236 * maybe re-implement using semaphores? */
2237 static inline int drbd_get_max_buffers(struct drbd_conf
*mdev
)
2239 int mxb
= 1000000; /* arbitrary limit on open requests */
2240 if (get_net_conf(mdev
)) {
2241 mxb
= mdev
->net_conf
->max_buffers
;
2247 static inline int drbd_state_is_stable(struct drbd_conf
*mdev
)
2249 union drbd_state s
= mdev
->state
;
2251 /* DO NOT add a default clause, we want the compiler to warn us
2252 * for any newly introduced state we may have forgotten to add here */
2254 switch ((enum drbd_conns
)s
.conn
) {
2255 /* new io only accepted when there is no connection, ... */
2257 case C_WF_CONNECTION
:
2258 /* ... or there is a well established connection. */
2264 case C_PAUSED_SYNC_S
:
2265 case C_PAUSED_SYNC_T
:
2268 /* transitional states, IO allowed */
2269 case C_DISCONNECTING
:
2273 case C_NETWORK_FAILURE
:
2274 case C_PROTOCOL_ERROR
:
2276 case C_WF_REPORT_PARAMS
:
2277 case C_STARTING_SYNC_S
:
2278 case C_STARTING_SYNC_T
:
2281 /* Allow IO in BM exchange states with new protocols */
2283 if (mdev
->agreed_pro_version
< 96)
2287 /* no new io accepted in these states */
2289 case C_WF_SYNC_UUID
:
2295 switch ((enum drbd_disk_state
)s
.disk
) {
2297 case D_INCONSISTENT
:
2301 /* disk state is stable as well. */
2304 /* no new io accepted during tansitional states */
2317 static inline int is_susp(union drbd_state s
)
2319 return s
.susp
|| s
.susp_nod
|| s
.susp_fen
;
2322 static inline bool may_inc_ap_bio(struct drbd_conf
*mdev
)
2324 int mxb
= drbd_get_max_buffers(mdev
);
2326 if (is_susp(mdev
->state
))
2328 if (test_bit(SUSPEND_IO
, &mdev
->flags
))
2331 /* to avoid potential deadlock or bitmap corruption,
2332 * in various places, we only allow new application io
2333 * to start during "stable" states. */
2335 /* no new io accepted when attaching or detaching the disk */
2336 if (!drbd_state_is_stable(mdev
))
2339 /* since some older kernels don't have atomic_add_unless,
2340 * and we are within the spinlock anyways, we have this workaround. */
2341 if (atomic_read(&mdev
->ap_bio_cnt
) > mxb
)
2343 if (test_bit(BITMAP_IO
, &mdev
->flags
))
2348 static inline bool inc_ap_bio_cond(struct drbd_conf
*mdev
, int count
)
2352 spin_lock_irq(&mdev
->req_lock
);
2353 rv
= may_inc_ap_bio(mdev
);
2355 atomic_add(count
, &mdev
->ap_bio_cnt
);
2356 spin_unlock_irq(&mdev
->req_lock
);
2361 static inline void inc_ap_bio(struct drbd_conf
*mdev
, int count
)
2364 * as long as the device is suspended
2365 * until the bitmap is no longer on the fly during connection
2366 * handshake as long as we would exeed the max_buffer limit.
2368 * to avoid races with the reconnect code,
2369 * we need to atomic_inc within the spinlock. */
2371 wait_event(mdev
->misc_wait
, inc_ap_bio_cond(mdev
, count
));
2374 static inline void dec_ap_bio(struct drbd_conf
*mdev
)
2376 int mxb
= drbd_get_max_buffers(mdev
);
2377 int ap_bio
= atomic_dec_return(&mdev
->ap_bio_cnt
);
2379 D_ASSERT(ap_bio
>= 0);
2380 /* this currently does wake_up for every dec_ap_bio!
2381 * maybe rather introduce some type of hysteresis?
2382 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2384 wake_up(&mdev
->misc_wait
);
2385 if (ap_bio
== 0 && test_bit(BITMAP_IO
, &mdev
->flags
)) {
2386 if (!test_and_set_bit(BITMAP_IO_QUEUED
, &mdev
->flags
))
2387 drbd_queue_work(&mdev
->data
.work
, &mdev
->bm_io_work
.w
);
2391 static inline int drbd_set_ed_uuid(struct drbd_conf
*mdev
, u64 val
)
2393 int changed
= mdev
->ed_uuid
!= val
;
2394 mdev
->ed_uuid
= val
;
2398 static inline int seq_cmp(u32 a
, u32 b
)
2400 /* we assume wrap around at 32bit.
2401 * for wrap around at 24bit (old atomic_t),
2405 return (s32
)(a
) - (s32
)(b
);
2407 #define seq_lt(a, b) (seq_cmp((a), (b)) < 0)
2408 #define seq_gt(a, b) (seq_cmp((a), (b)) > 0)
2409 #define seq_ge(a, b) (seq_cmp((a), (b)) >= 0)
2410 #define seq_le(a, b) (seq_cmp((a), (b)) <= 0)
2411 /* CAUTION: please no side effects in arguments! */
2412 #define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b)))
2414 static inline void update_peer_seq(struct drbd_conf
*mdev
, unsigned int new_seq
)
2417 spin_lock(&mdev
->peer_seq_lock
);
2418 m
= seq_max(mdev
->peer_seq
, new_seq
);
2420 spin_unlock(&mdev
->peer_seq_lock
);
2422 wake_up(&mdev
->seq_wait
);
2425 static inline void drbd_update_congested(struct drbd_conf
*mdev
)
2427 struct sock
*sk
= mdev
->data
.socket
->sk
;
2428 if (sk
->sk_wmem_queued
> sk
->sk_sndbuf
* 4 / 5)
2429 set_bit(NET_CONGESTED
, &mdev
->flags
);
2432 static inline int drbd_queue_order_type(struct drbd_conf
*mdev
)
2434 /* sorry, we currently have no working implementation
2435 * of distributed TCQ stuff */
2436 #ifndef QUEUE_ORDERED_NONE
2437 #define QUEUE_ORDERED_NONE 0
2439 return QUEUE_ORDERED_NONE
;
2442 static inline void drbd_md_flush(struct drbd_conf
*mdev
)
2446 if (test_bit(MD_NO_FUA
, &mdev
->flags
))
2449 r
= blkdev_issue_flush(mdev
->ldev
->md_bdev
, GFP_KERNEL
, NULL
);
2451 set_bit(MD_NO_FUA
, &mdev
->flags
);
2452 dev_err(DEV
, "meta data flush failed with status %d, disabling md-flushes\n", r
);