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/tcp.h>
38
#include <linux/mutex.h>
39
#include <linux/major.h>
40
#include <linux/blkdev.h>
41
#include <linux/bio.h>
43
#include "lru_cache.h"
46
# define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
47
# define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
48
# define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
49
# define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
51
# define __protected_by(x)
52
# define __protected_read_by(x)
53
# define __protected_write_by(x)
54
# define __must_hold(x)
57
#define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
59
/* module parameter, defined in drbd_main.c */
60
extern unsigned int minor_count;
62
extern unsigned int cn_idx;
64
#ifdef DRBD_ENABLE_FAULTS
65
extern int enable_faults;
66
extern int fault_rate;
67
extern int fault_devs;
70
extern char usermode_helper[];
80
/* I don't remember why XCPU ...
81
* This is used to wake the asender,
82
* and to interrupt sending the sending task
85
#define DRBD_SIG SIGXCPU
87
/* This is used to stop/restart our threads.
88
* Cannot use SIGTERM nor SIGKILL, since these
89
* are sent out by init on runlevel changes
90
* I choose SIGHUP for now.
92
#define DRBD_SIGKILL SIGHUP
94
/* All EEs on the free list should have ID_VACANT (== 0)
95
* freshly allocated EEs get !ID_VACANT (== 1)
96
* so if it says "cannot dereference null pointer at adress 0x00000001",
97
* it is most likely one of these :( */
99
#define ID_IN_SYNC (4711ULL)
100
#define ID_OUT_OF_SYNC (4712ULL)
102
#define ID_SYNCER (-1ULL)
104
#define is_syncer_block_id(id) ((id) == ID_SYNCER)
108
#ifdef DBG_ALL_SYMBOLS
111
# define STATIC static
115
* Some Message Macros
116
*************************/
118
#define DUMPP(A) ERR(#A " = %p in %s:%d\n", (A), __FILE__, __LINE__);
119
#define DUMPLU(A) ERR(#A " = %lu in %s:%d\n", (unsigned long)(A), __FILE__, __LINE__);
120
#define DUMPLLU(A) ERR(#A " = %llu in %s:%d\n", (unsigned long long)(A), __FILE__, __LINE__);
121
#define DUMPLX(A) ERR(#A " = %lx in %s:%d\n", (A), __FILE__, __LINE__);
122
#define DUMPI(A) ERR(#A " = %d in %s:%d\n", (int)(A), __FILE__, __LINE__);
125
#define PRINTK(level, fmt, args...) \
126
printk(level "drbd%d: " fmt, \
127
mdev->minor , ##args)
129
#define ALERT(fmt, args...) PRINTK(KERN_ALERT, fmt , ##args)
130
#define ERR(fmt, args...) PRINTK(KERN_ERR, fmt , ##args)
131
/* nowadays, WARN() is defined as BUG() without crash in bug.h */
132
#define drbd_WARN(fmt, args...) PRINTK(KERN_WARNING, fmt , ##args)
133
#define INFO(fmt, args...) PRINTK(KERN_INFO, fmt , ##args)
134
#define DBG(fmt, args...) PRINTK(KERN_DEBUG, fmt , ##args)
136
#define D_ASSERT(exp) if (!(exp)) \
137
ERR("ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
139
#define ERR_IF(exp) if (({ \
140
int _b = (exp) != 0; \
141
if (_b) ERR("%s: (%s) in %s:%d\n", \
142
__func__, #exp, __FILE__, __LINE__); \
146
/* Defines to control fault insertion */
148
DRBD_FAULT_MD_WR = 0, /* meta data write */
149
DRBD_FAULT_MD_RD, /* read */
150
DRBD_FAULT_RS_WR, /* resync */
152
DRBD_FAULT_DT_WR, /* data */
154
DRBD_FAULT_DT_RA, /* data read ahead */
155
DRBD_FAULT_BM_ALLOC, /* bitmap allocation */
156
DRBD_FAULT_AL_EE, /* alloc ee */
161
#ifdef DRBD_ENABLE_FAULTS
163
_drbd_insert_fault(struct drbd_conf *mdev, unsigned int type);
165
drbd_insert_fault(struct drbd_conf *mdev, unsigned int type) {
167
(enable_faults & (1<<type)) &&
168
_drbd_insert_fault(mdev, type);
170
#define FAULT_ACTIVE(_m, _t) (drbd_insert_fault((_m), (_t)))
173
#define FAULT_ACTIVE(_m, _t) (0)
176
/* integer division, round _UP_ to the next integer */
177
#define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
178
/* usual integer division */
179
#define div_floor(A, B) ((A)/(B))
181
/* drbd_meta-data.c (still in drbd_main.c) */
182
/* 4th incarnation of the disk layout. */
183
#define DRBD_MD_MAGIC (DRBD_MAGIC+4)
185
extern struct drbd_conf **minor_table;
186
extern struct ratelimit_state drbd_ratelimit_state;
190
*********************************************************************/
192
enum Drbd_Packet_Cmd {
193
/* receiver (data socket) */
195
DataReply = 0x01, /* Response to DataRequest */
196
RSDataReply = 0x02, /* Response to RSDataRequest */
199
BecomeSyncTarget = 0x05,
200
BecomeSyncSource = 0x06,
201
UnplugRemote = 0x07, /* Used at various times to hint the peer */
202
DataRequest = 0x08, /* Used to ask for a data block */
203
RSDataRequest = 0x09, /* Used to ask for a data block for resync */
205
ReportProtocol = 0x0b,
209
ReportSyncUUID = 0x0f,
210
AuthChallenge = 0x10,
212
StateChgRequest = 0x12,
214
/* asender (meta socket */
217
RecvAck = 0x15, /* Used in protocol B */
218
WriteAck = 0x16, /* Used in protocol C */
219
RSWriteAck = 0x17, /* Is a WriteAck, additionally call set_in_sync(). */
220
DiscardAck = 0x18, /* Used in proto C, two-primaries conflict detection */
221
NegAck = 0x19, /* Sent if local disk is unusable */
222
NegDReply = 0x1a, /* Local disk is broken... */
223
NegRSDReply = 0x1b, /* Local disk is broken... */
225
StateChgReply = 0x1d,
227
/* "new" commands, no longer fitting into the ordering scheme above */
229
OVRequest = 0x1e, /* data socket */
231
OVResult = 0x20, /* meta socket */
232
CsumRSRequest = 0x21, /* data socket */
233
RSIsInSync = 0x22, /* meta socket */
234
SyncParam89 = 0x23, /* data socket, protocol version 89 replacement for SyncParam */
235
ReportCBitMap = 0x24, /* compressed or otherwise encoded bitmap transfer */
238
MayIgnore = 0x100, /* Flag to test if (cmd > MayIgnore) ... */
241
/* special command ids for handshake */
243
HandShakeM = 0xfff1, /* First Packet on the MetaSock */
244
HandShakeS = 0xfff2, /* First Packet on the Socket */
246
HandShake = 0xfffe /* FIXED for the next century! */
249
static inline const char *cmdname(enum Drbd_Packet_Cmd cmd)
251
/* THINK may need to become several global tables
252
* when we want to support more than
254
static const char *cmdnames[] = {
256
[DataReply] = "DataReply",
257
[RSDataReply] = "RSDataReply",
258
[Barrier] = "Barrier",
259
[ReportBitMap] = "ReportBitMap",
260
[BecomeSyncTarget] = "BecomeSyncTarget",
261
[BecomeSyncSource] = "BecomeSyncSource",
262
[UnplugRemote] = "UnplugRemote",
263
[DataRequest] = "DataRequest",
264
[RSDataRequest] = "RSDataRequest",
265
[SyncParam] = "SyncParam",
266
[SyncParam89] = "SyncParam89",
267
[ReportProtocol] = "ReportProtocol",
268
[ReportUUIDs] = "ReportUUIDs",
269
[ReportSizes] = "ReportSizes",
270
[ReportState] = "ReportState",
271
[ReportSyncUUID] = "ReportSyncUUID",
272
[AuthChallenge] = "AuthChallenge",
273
[AuthResponse] = "AuthResponse",
275
[PingAck] = "PingAck",
276
[RecvAck] = "RecvAck",
277
[WriteAck] = "WriteAck",
278
[RSWriteAck] = "RSWriteAck",
279
[DiscardAck] = "DiscardAck",
281
[NegDReply] = "NegDReply",
282
[NegRSDReply] = "NegRSDReply",
283
[BarrierAck] = "BarrierAck",
284
[StateChgRequest] = "StateChgRequest",
285
[StateChgReply] = "StateChgReply",
286
[OVRequest] = "OVRequest",
287
[OVReply] = "OVReply",
288
[OVResult] = "OVResult",
289
[CsumRSRequest] = "CsumRSRequest",
290
[RSIsInSync] = "RSIsInSync",
291
[ReportCBitMap] = "ReportCBitMap",
295
if (cmd == HandShakeM)
297
if (cmd == HandShakeS)
299
if (cmd == HandShake)
303
return cmdnames[cmd];
306
/* for sending/receiving the bitmap,
307
* possibly in some encoding scheme */
310
* stores total bits and long words
311
* of the bitmap, so we don't need to
312
* call the accessor functions over and again. */
313
unsigned long bm_bits;
314
unsigned long bm_words;
315
/* during xfer, current position within the bitmap */
316
unsigned long bit_offset;
317
unsigned long word_offset;
319
/* statistics; index: (h->command == ReportBitMap) */
324
extern void INFO_bm_xfer_stats(struct drbd_conf *mdev,
325
const char *direction, struct bm_xfer_ctx *c);
327
static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
329
/* word_offset counts "native long words" (32 or 64 bit),
331
* Encoded packet may end at an unaligned bit offset.
332
* In case a fallback clear text packet is transmitted in
333
* between, we adjust this offset back to the last 64bit
334
* aligned "native long word", which makes coding and decoding
335
* the plain text bitmap much more convenient. */
336
#if BITS_PER_LONG == 64
337
c->word_offset = c->bit_offset >> 6;
338
#elif BITS_PER_LONG == 32
339
c->word_offset = c->bit_offset >> 5;
340
c->word_offset &= ~(1UL);
342
# error "unsupported BITS_PER_LONG"
346
/* This is the layout for a packet on the wire.
347
* The byteorder is the network byte order.
348
* (except block_id and barrier fields.
349
* these are pointers to local structs
350
* and have no relevance for the partner,
351
* which just echoes them as received.)
353
* NOTE that the payload starts at a long aligned offset,
354
* regardless of 32 or 64 bit arch!
359
u16 length; /* bytes of data after this header */
361
} __attribute((packed));
362
/* 8 bytes. packet FIXED for the next century! */
365
* short commands, packets without payload, plain Drbd_Header:
374
* commands with out-of-struct payload:
375
* ReportBitMap (no additional fields)
376
* Data, DataReply (see Drbd_Data_Packet)
377
* ReportCBitMap (see receive_compressed_bitmap)
380
/* these defines must not be changed without changing the protocol version */
381
#define DP_HARDBARRIER 1
383
#define DP_MAY_SET_IN_SYNC 4
385
struct Drbd_Data_Packet {
386
struct Drbd_Header head;
387
u64 sector; /* 64 bits sector number */
388
u64 block_id; /* to identify the request in protocol B&C */
391
} __attribute((packed));
394
* commands which share a struct:
395
* Drbd_BlockAck_Packet:
396
* RecvAck (proto B), WriteAck (proto C),
397
* DiscardAck (proto C, two-primaries conflict detection)
398
* Drbd_BlockRequest_Packet:
399
* DataRequest, RSDataRequest
401
struct Drbd_BlockAck_Packet {
402
struct Drbd_Header head;
407
} __attribute((packed));
410
struct Drbd_BlockRequest_Packet {
411
struct Drbd_Header head;
415
u32 pad; /* to multiple of 8 Byte */
416
} __attribute((packed));
419
* commands with their own struct for additional fields:
427
struct Drbd_HandShake_Packet {
428
struct Drbd_Header head; /* 8 bytes */
433
/* should be more than enough for future enhancements
434
* for now, feature_flags and the reserverd array shall be zero.
439
} __attribute((packed));
440
/* 80 bytes, FIXED for the next century */
442
struct Drbd_Barrier_Packet {
443
struct Drbd_Header head;
444
u32 barrier; /* barrier number _handle_ only */
445
u32 pad; /* to multiple of 8 Byte */
446
} __attribute((packed));
448
struct Drbd_BarrierAck_Packet {
449
struct Drbd_Header head;
452
} __attribute((packed));
454
struct Drbd_SyncParam_Packet {
455
struct Drbd_Header head;
458
/* Since protocol version 88 and higher. */
460
} __attribute((packed));
462
struct Drbd_SyncParam89_Packet {
463
struct Drbd_Header head;
465
/* protocol version 89: */
466
char verify_alg[SHARED_SECRET_MAX];
467
char csums_alg[SHARED_SECRET_MAX];
468
} __attribute((packed));
470
struct Drbd_Protocol_Packet {
471
struct Drbd_Header head;
479
/* Since protocol version 87 and higher. */
480
char integrity_alg[0];
482
} __attribute((packed));
484
struct Drbd_GenCnt_Packet {
485
struct Drbd_Header head;
486
u64 uuid[EXT_UUID_SIZE];
487
} __attribute((packed));
489
struct Drbd_SyncUUID_Packet {
490
struct Drbd_Header head;
492
} __attribute((packed));
494
struct Drbd_Sizes_Packet {
495
struct Drbd_Header head;
496
u64 d_size; /* size of disk */
497
u64 u_size; /* user requested size */
498
u64 c_size; /* current exported size */
499
u32 max_segment_size; /* Maximal size of a BIO */
500
u32 queue_order_type;
501
} __attribute((packed));
503
struct Drbd_State_Packet {
504
struct Drbd_Header head;
506
} __attribute((packed));
508
struct Drbd_Req_State_Packet {
509
struct Drbd_Header head;
512
} __attribute((packed));
514
struct Drbd_RqS_Reply_Packet {
515
struct Drbd_Header head;
517
} __attribute((packed));
519
struct Drbd06_Parameter_P {
527
} __attribute((packed));
529
struct Drbd_Discard_Packet {
530
struct Drbd_Header head;
534
} __attribute((packed));
536
/* Valid values for the encoding field.
537
* Bump proto version when changing this. */
538
enum Drbd_bitmap_code {
540
RLE_VLI_BitsFibD_0_1 = 1,
541
RLE_VLI_BitsFibD_1_1 = 2,
542
RLE_VLI_BitsFibD_1_2 = 3,
543
RLE_VLI_BitsFibD_2_3 = 4,
544
RLE_VLI_BitsFibD_3_5 = 5,
547
struct Drbd_Compressed_Bitmap_Packet {
548
struct Drbd_Header head;
549
/* (encoding & 0x0f): actual encoding, see enum Drbd_bitmap_code
550
* (encoding & 0x80): polarity (set/unset) of first runlength
551
* ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
552
* used to pad up to head.length bytes
557
} __attribute((packed));
559
static inline enum Drbd_bitmap_code
560
DCBP_get_code(struct Drbd_Compressed_Bitmap_Packet *p)
562
return (enum Drbd_bitmap_code)(p->encoding & 0x0f);
566
DCBP_set_code(struct Drbd_Compressed_Bitmap_Packet *p, enum Drbd_bitmap_code code)
569
p->encoding = (p->encoding & ~0xf) | code;
573
DCBP_get_start(struct Drbd_Compressed_Bitmap_Packet *p)
575
return (p->encoding & 0x80) != 0;
579
DCBP_set_start(struct Drbd_Compressed_Bitmap_Packet *p, int set)
581
p->encoding = (p->encoding & ~0x80) | (set ? 0x80 : 0);
585
DCBP_get_pad_bits(struct Drbd_Compressed_Bitmap_Packet *p)
587
return (p->encoding >> 4) & 0x7;
591
DCBP_set_pad_bits(struct Drbd_Compressed_Bitmap_Packet *p, int n)
594
p->encoding = (p->encoding & (~0x7 << 4)) | (n << 4);
597
/* one bitmap packet, including the Drbd_Header,
598
* should fit within one _architecture independend_ page.
599
* so we need to use the fixed size 4KiB page size
600
* most architechtures have used for a long time.
602
#define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct Drbd_Header))
603
#define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
604
#define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct Drbd_Compressed_Bitmap_Packet))
605
#if (PAGE_SIZE < 4096)
606
/* drbd_send_bitmap / receive_bitmap would break horribly */
607
#error "PAGE_SIZE too small"
610
union Drbd_Polymorph_Packet {
611
struct Drbd_Header head;
612
struct Drbd_HandShake_Packet HandShake;
613
struct Drbd_Data_Packet Data;
614
struct Drbd_BlockAck_Packet BlockAck;
615
struct Drbd_Barrier_Packet Barrier;
616
struct Drbd_BarrierAck_Packet BarrierAck;
617
struct Drbd_SyncParam89_Packet SyncParam89;
618
struct Drbd_Protocol_Packet Protocol;
619
struct Drbd_Sizes_Packet Sizes;
620
struct Drbd_GenCnt_Packet GenCnt;
621
struct Drbd_State_Packet State;
622
struct Drbd_Req_State_Packet ReqState;
623
struct Drbd_RqS_Reply_Packet RqSReply;
624
struct Drbd_BlockRequest_Packet BlockRequest;
625
} __attribute((packed));
627
/**********************************************************************/
628
enum Drbd_thread_state {
637
struct task_struct *task;
638
struct completion stop;
639
enum Drbd_thread_state t_state;
640
int (*function) (struct Drbd_thread *);
641
struct drbd_conf *mdev;
645
static inline enum Drbd_thread_state get_t_state(struct Drbd_thread *thi)
647
/* THINK testing the t_state seems to be uncritical in all cases
648
* (but thread_{start,stop}), so we can read it *without* the lock.
657
* Having this as the first member of a struct provides sort of "inheritance".
658
* "derived" structs can be "drbd_queue_work()"ed.
659
* The callback should know and cast back to the descendant struct.
660
* drbd_request and Tl_epoch_entry are descendants of drbd_work.
663
typedef int (*drbd_work_cb)(struct drbd_conf *, struct drbd_work *, int cancel);
665
struct list_head list;
670
struct drbd_request {
672
struct drbd_conf *mdev;
673
struct bio *private_bio;
674
struct hlist_node colision;
677
unsigned int epoch; /* barrier_nr */
679
/* barrier_nr: used to check on "completion" whether this req was in
680
* the current epoch, and we therefore have to close it,
681
* starting a new epoch...
684
/* up to here, the struct layout is identical to Tl_epoch_entry;
685
* we might be able to use that to our advantage... */
687
struct list_head tl_requests; /* ring list in the transfer log */
688
struct bio *master_bio; /* master bio pointer */
689
unsigned long rq_state; /* see comments above _req_mod() */
691
unsigned long start_time;
694
struct drbd_barrier {
696
struct list_head requests; /* requests before */
697
struct drbd_barrier *next; /* pointer to the next barrier */
698
unsigned int br_number; /* the barriers identifier. */
699
int n_req; /* number of requests attached before this barrier */
704
/* These Tl_epoch_entries may be in one of 6 lists:
705
active_ee .. data packet being written
706
sync_ee .. syncer block being written
707
done_ee .. block written, need to send WriteAck
708
read_ee .. [RS]DataRequest being read
712
struct list_head list;
713
unsigned int barrier_nr;
714
atomic_t epoch_size; /* increased on every request added. */
715
atomic_t active; /* increased on every req. added, and dec on every finished. */
719
/* drbd_epoch flag bits */
721
DE_BARRIER_IN_NEXT_EPOCH_ISSUED,
722
DE_BARRIER_IN_NEXT_EPOCH_DONE,
723
DE_CONTAINS_A_BARRIER,
724
DE_HAVE_BARRIER_NUMBER,
728
struct Tl_epoch_entry {
730
struct drbd_conf *mdev;
731
struct bio *private_bio;
732
struct hlist_node colision;
735
struct drbd_epoch *epoch;
737
/* up to here, the struct layout is identical to drbd_request;
738
* we might be able to use that to our advantage... */
751
__EE_CALL_AL_COMPLETE_IO,
752
__EE_CONFLICT_PENDING,
753
__EE_MAY_SET_IN_SYNC,
756
#define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
757
#define EE_CONFLICT_PENDING (1<<__EE_CONFLICT_PENDING)
758
#define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
759
#define EE_IS_BARRIER (1<<__EE_IS_BARRIER)
761
/* global flag bits */
763
CREATE_BARRIER, /* next Data is preceeded by a Barrier */
764
SIGNAL_ASENDER, /* whether asender wants to be interrupted */
765
SEND_PING, /* whether asender should send a ping asap */
766
WORK_PENDING, /* completion flag for drbd_disconnect */
767
STOP_SYNC_TIMER, /* tell timer to cancel itself */
768
UNPLUG_QUEUED, /* only relevant with kernel 2.4 */
769
UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
770
MD_DIRTY, /* current uuids and flags not yet on disk */
771
DISCARD_CONCURRENT, /* Set on one node, cleared on the peer! */
772
USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
773
CLUSTER_ST_CHANGE, /* Cluster wide state change going on... */
776
CRASHED_PRIMARY, /* This node was a crashed primary.
777
* Gets cleared when the state.conn
778
* goes into Connected state. */
779
WRITE_BM_AFTER_RESYNC, /* A kmalloc() during resync failed */
780
NO_BARRIER_SUPP, /* underlying block device doesn't implement barriers */
783
MD_NO_BARRIER, /* meta data device does not support barriers,
785
SUSPEND_IO, /* suspend application io */
786
BITMAP_IO, /* suspend application io;
787
once no more io in flight, start bitmap io */
788
BITMAP_IO_QUEUED, /* Started bitmap IO */
789
RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
790
NET_CONGESTED, /* The data socket is congested */
793
struct drbd_bitmap; /* opaque for drbd_conf */
795
/* TODO sort members for performance
796
* MAYBE group them further */
798
/* THINK maybe we actually want to use the default "event/%s" worker threads
799
* or similar in linux 2.6, which uses per cpu data and threads.
801
* To be general, this might need a spin_lock member.
802
* For now, please use the mdev->req_lock to protect list_head,
803
* see drbd_queue_work below.
805
struct drbd_work_queue {
807
struct semaphore s; /* producers up it, worker down()s it */
808
spinlock_t q_lock; /* to protect the list. */
812
struct drbd_work_queue work;
814
struct socket *socket;
815
/* this way we get our
816
* send/receive buffers off the stack */
817
union Drbd_Polymorph_Packet sbuf;
818
union Drbd_Polymorph_Packet rbuf;
822
u64 md_offset; /* sector offset to 'super' block */
824
u64 la_size_sect; /* last agreed size, unit sectors */
830
s32 al_offset; /* signed relative sector offset to al area */
831
s32 bm_offset; /* signed relative sector offset to bitmap */
833
/* u32 al_nr_extents; important for restoring the AL
834
* is stored into sync_conf.al_extents, which in turn
835
* gets applied to act_log->nr_elements
839
/* for sync_conf and other types... */
840
#define NL_PACKET(name, number, fields) struct name { fields };
841
#define NL_INTEGER(pn,pr,member) int member;
842
#define NL_INT64(pn,pr,member) __u64 member;
843
#define NL_BIT(pn,pr,member) unsigned member:1;
844
#define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
845
#include "linux/drbd_nl.h"
847
struct drbd_backing_dev {
848
struct block_device *backing_bdev;
849
struct block_device *md_bdev;
850
struct file *lo_file;
851
struct file *md_file;
853
struct disk_conf dc; /* The user provided config... */
854
sector_t known_size; /* last known size of that backing device */
858
struct drbd_conf *mdev;
859
struct completion event;
866
int (*io_fn)(struct drbd_conf *mdev);
867
void (*done)(struct drbd_conf *mdev, int rv);
870
enum write_ordering_e {
878
/* things that are stored as / read from meta data on disk */
881
/* configured by drbdsetup */
882
struct net_conf *net_conf; /* protected by inc_net() and dec_net() */
883
struct syncer_conf sync_conf;
884
struct drbd_backing_dev *bc __protected_by(local);
886
sector_t p_size; /* partner's disk size */
887
struct request_queue *rq_queue;
888
struct block_device *this_bdev;
889
struct gendisk *vdisk;
891
struct drbd_socket data; /* data/barrier/cstate/parameter packets */
892
struct drbd_socket meta; /* ping/ack (metadata) packets */
893
int agreed_pro_version; /* actually used protocol version */
894
unsigned long last_received; /* in jiffies, either socket */
895
unsigned int ko_count;
896
struct drbd_work resync_work,
899
struct timer_list resync_timer;
900
struct timer_list md_sync_timer;
902
/* Used after attach while negotiating new disk state. */
903
union drbd_state_t new_state_tmp;
905
union drbd_state_t state;
906
wait_queue_head_t misc_wait;
907
wait_queue_head_t state_wait; /* upon each state change. */
908
unsigned int send_cnt;
909
unsigned int recv_cnt;
910
unsigned int read_cnt;
911
unsigned int writ_cnt;
912
unsigned int al_writ_cnt;
913
unsigned int bm_writ_cnt;
914
atomic_t ap_bio_cnt; /* Requests we need to complete */
915
atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
916
atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
917
atomic_t unacked_cnt; /* Need to send replys for */
918
atomic_t local_cnt; /* Waiting for local completion */
919
atomic_t net_cnt; /* Users of net_conf */
921
struct drbd_barrier *unused_spare_barrier; /* for pre-allocation */
922
struct drbd_barrier *newest_barrier;
923
struct drbd_barrier *oldest_barrier;
924
struct list_head out_of_sequence_requests;
925
struct hlist_head *tl_hash;
926
unsigned int tl_hash_s;
928
/* blocks to sync in this run [unit BM_BLOCK_SIZE] */
929
unsigned long rs_total;
930
/* number of sync IOs that failed in this run */
931
unsigned long rs_failed;
932
/* Syncer's start time [unit jiffies] */
933
unsigned long rs_start;
934
/* cumulated time in PausedSyncX state [unit jiffies] */
935
unsigned long rs_paused;
936
/* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
937
unsigned long rs_mark_left;
938
/* marks's time [unit jiffies] */
939
unsigned long rs_mark_time;
940
/* skipped because csum was equeal [unit BM_BLOCK_SIZE] */
941
unsigned long rs_same_csum;
942
sector_t ov_position;
943
/* Start sector of out of sync range. */
944
sector_t ov_last_oos_start;
945
/* size of out-of-sync range in sectors. */
946
sector_t ov_last_oos_size;
947
unsigned long ov_left;
948
struct crypto_hash *csums_tfm;
949
struct crypto_hash *verify_tfm;
951
struct Drbd_thread receiver;
952
struct Drbd_thread worker;
953
struct Drbd_thread asender;
954
struct drbd_bitmap *bitmap;
955
unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
957
/* Used to track operations of resync... */
958
struct lru_cache *resync;
959
/* Number of locked elements in resync LRU */
960
unsigned int resync_locked;
961
/* resync extent number waiting for application requests */
962
unsigned int resync_wenr;
966
struct drbd_epoch *current_epoch;
967
spinlock_t epoch_lock;
969
enum write_ordering_e write_ordering;
970
struct list_head active_ee; /* IO in progress */
971
struct list_head sync_ee; /* IO in progress */
972
struct list_head done_ee; /* send ack */
973
struct list_head read_ee; /* IO in progress */
974
struct list_head net_ee; /* zero-copy network send in progress */
975
struct hlist_head *ee_hash; /* is proteced by req_lock! */
976
unsigned int ee_hash_s;
978
/* this one is protected by ee_lock, single thread */
979
struct Tl_epoch_entry *last_write_w_barrier;
982
struct hlist_head *app_reads_hash; /* is proteced by req_lock */
983
struct list_head resync_reads;
985
wait_queue_head_t ee_wait;
986
struct page *md_io_page; /* one page buffer for md_io */
987
struct page *md_io_tmpp; /* for hardsect != 512 [s390 only?] */
988
struct mutex md_io_mutex; /* protects the md_io_buffer */
990
wait_queue_head_t al_wait;
991
struct lru_cache *act_log; /* activity log */
992
unsigned int al_tr_number;
994
int al_tr_pos; /* position of the next transaction in the journal */
995
struct crypto_hash *cram_hmac_tfm;
996
struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */
997
struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */
1001
wait_queue_head_t seq_wait;
1002
atomic_t packet_seq;
1003
unsigned int peer_seq;
1004
spinlock_t peer_seq_lock;
1006
unsigned long comm_bm_set; /* communicated number of set bits. */
1008
struct bm_io_work bm_io_work;
1009
u64 ed_uuid; /* UUID of the exposed data */
1010
struct mutex state_mutex;
1011
char congestion_reason; /* Why we where congested... */
1014
static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
1016
struct drbd_conf *mdev;
1018
mdev = minor < minor_count ? minor_table[minor] : NULL;
1023
static inline unsigned int mdev_to_minor(struct drbd_conf *mdev)
1028
/* returns 1 if it was successfull,
1029
* returns 0 if there was no data socket.
1030
* so wherever you are going to use the data.socket, e.g. do
1031
* if (!drbd_get_data_sock(mdev))
1034
* drbd_put_data_sock(mdev);
1036
static inline int drbd_get_data_sock(struct drbd_conf *mdev)
1038
mutex_lock(&mdev->data.mutex);
1039
/* drbd_disconnect() could have called drbd_free_sock()
1040
* while we were waiting in down()... */
1041
if (unlikely(mdev->data.socket == NULL)) {
1042
mutex_unlock(&mdev->data.mutex);
1048
static inline void drbd_put_data_sock(struct drbd_conf *mdev)
1050
mutex_unlock(&mdev->data.mutex);
1054
* function declarations
1055
*************************/
1059
enum chg_state_flags {
1061
ChgStateVerbose = 2,
1062
ChgWaitComplete = 4,
1064
ChgOrdered = ChgWaitComplete + ChgSerialize,
1067
extern void drbd_init_set_defaults(struct drbd_conf *mdev);
1068
extern int drbd_change_state(struct drbd_conf *mdev, enum chg_state_flags f,
1069
union drbd_state_t mask, union drbd_state_t val);
1070
extern void drbd_force_state(struct drbd_conf *, union drbd_state_t,
1071
union drbd_state_t);
1072
extern int _drbd_request_state(struct drbd_conf *, union drbd_state_t,
1073
union drbd_state_t, enum chg_state_flags);
1074
extern int __drbd_set_state(struct drbd_conf *, union drbd_state_t,
1075
enum chg_state_flags, struct completion *done);
1076
extern void print_st_err(struct drbd_conf *, union drbd_state_t,
1077
union drbd_state_t, int);
1078
extern int drbd_thread_start(struct Drbd_thread *thi);
1079
extern void _drbd_thread_stop(struct Drbd_thread *thi, int restart, int wait);
1081
extern void drbd_thread_current_set_cpu(struct drbd_conf *mdev);
1082
extern cpumask_t drbd_calc_cpu_mask(struct drbd_conf *mdev);
1084
#define drbd_thread_current_set_cpu(A) ({})
1085
#define drbd_calc_cpu_mask(A) CPU_MASK_ALL
1087
extern void drbd_free_resources(struct drbd_conf *mdev);
1088
extern void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr,
1089
unsigned int set_size);
1090
extern void tl_clear(struct drbd_conf *mdev);
1091
extern void _tl_add_barrier(struct drbd_conf *, struct drbd_barrier *);
1092
extern void drbd_free_sock(struct drbd_conf *mdev);
1093
extern int drbd_send(struct drbd_conf *mdev, struct socket *sock,
1094
void *buf, size_t size, unsigned msg_flags);
1095
extern int drbd_send_protocol(struct drbd_conf *mdev);
1096
extern int _drbd_send_uuids(struct drbd_conf *mdev);
1097
extern int drbd_send_uuids(struct drbd_conf *mdev);
1098
extern int drbd_send_sync_uuid(struct drbd_conf *mdev, u64 val);
1099
extern int drbd_send_sizes(struct drbd_conf *mdev);
1100
extern int _drbd_send_state(struct drbd_conf *mdev);
1101
extern int drbd_send_state(struct drbd_conf *mdev);
1102
extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock,
1103
enum Drbd_Packet_Cmd cmd, struct Drbd_Header *h,
1104
size_t size, unsigned msg_flags);
1105
#define USE_DATA_SOCKET 1
1106
#define USE_META_SOCKET 0
1107
extern int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket,
1108
enum Drbd_Packet_Cmd cmd, struct Drbd_Header *h,
1110
extern int drbd_send_cmd2(struct drbd_conf *mdev, enum Drbd_Packet_Cmd cmd,
1111
char *data, size_t size);
1112
extern int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc);
1113
extern int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr,
1115
extern int drbd_send_ack(struct drbd_conf *mdev, enum Drbd_Packet_Cmd cmd,
1116
struct Tl_epoch_entry *e);
1117
extern int drbd_send_ack_rp(struct drbd_conf *mdev, enum Drbd_Packet_Cmd cmd,
1118
struct Drbd_BlockRequest_Packet *rp);
1119
extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum Drbd_Packet_Cmd cmd,
1120
struct Drbd_Data_Packet *dp);
1121
extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum Drbd_Packet_Cmd cmd,
1122
sector_t sector, int blksize, u64 block_id);
1123
extern int _drbd_send_page(struct drbd_conf *mdev, struct page *page,
1124
int offset, size_t size);
1125
extern int drbd_send_block(struct drbd_conf *mdev, enum Drbd_Packet_Cmd cmd,
1126
struct Tl_epoch_entry *e);
1127
extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
1128
extern int _drbd_send_barrier(struct drbd_conf *mdev,
1129
struct drbd_barrier *barrier);
1130
extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1131
sector_t sector, int size, u64 block_id);
1132
extern int drbd_send_drequest_csum(struct drbd_conf *mdev,
1133
sector_t sector,int size,
1134
void *digest, int digest_size,
1135
enum Drbd_Packet_Cmd cmd);
1136
extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size);
1138
extern int drbd_send_bitmap(struct drbd_conf *mdev);
1139
extern int _drbd_send_bitmap(struct drbd_conf *mdev);
1140
extern int drbd_send_sr_reply(struct drbd_conf *mdev, int retcode);
1141
extern void drbd_free_bc(struct drbd_backing_dev *bc);
1142
extern int drbd_io_error(struct drbd_conf *mdev, int forcedetach);
1143
extern void drbd_mdev_cleanup(struct drbd_conf *mdev);
1145
/* drbd_meta-data.c (still in drbd_main.c) */
1146
extern void drbd_md_sync(struct drbd_conf *mdev);
1147
extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
1148
/* maybe define them below as inline? */
1149
extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1150
extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1151
extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1152
extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1153
extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
1154
extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
1155
extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
1156
extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1157
extern void drbd_md_mark_dirty(struct drbd_conf *mdev);
1158
extern void drbd_queue_bitmap_io(struct drbd_conf *mdev,
1159
int (*io_fn)(struct drbd_conf *),
1160
void (*done)(struct drbd_conf *, int),
1162
extern int drbd_bmio_set_n_write(struct drbd_conf *mdev);
1163
extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev);
1164
extern int drbd_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), char *why);
1168
We reserve a 128MB Block (4k aligned)
1169
* either at the end of the backing device
1170
* or on a seperate meta data device. */
1172
#define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1173
/* The following numbers are sectors */
1174
#define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1175
#define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */
1176
/* Allows up to about 3.8TB */
1177
#define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1179
/* Since the smalles IO unit is usually 512 byte */
1180
#define MD_HARDSECT_B 9
1181
#define MD_HARDSECT (1<<MD_HARDSECT_B)
1184
#define AL_EXTENTS_PT ((MD_HARDSECT-12)/8-1) /* 61 ; Extents per 512B sector */
1185
#define AL_EXTENT_SIZE_B 22 /* One extent represents 4M Storage */
1186
#define AL_EXTENT_SIZE (1<<AL_EXTENT_SIZE_B)
1188
#if BITS_PER_LONG == 32
1190
#define cpu_to_lel(A) cpu_to_le32(A)
1191
#define lel_to_cpu(A) le32_to_cpu(A)
1192
#elif BITS_PER_LONG == 64
1194
#define cpu_to_lel(A) cpu_to_le64(A)
1195
#define lel_to_cpu(A) le64_to_cpu(A)
1197
#error "LN2 of BITS_PER_LONG unknown!"
1201
/* 16MB sized 'bitmap extent' to track syncer usage */
1203
struct lc_element lce;
1204
int rs_left; /* number of bits set (out of sync) in this extent. */
1205
int rs_failed; /* number of failed resync requests in this extent. */
1206
unsigned long flags;
1209
#define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1210
#define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1214
* We need to store one bit for a block.
1215
* Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1216
* Bit 0 ==> local node thinks this block is binary identical on both nodes
1217
* Bit 1 ==> local node thinks this block needs to be synced.
1220
#define BM_BLOCK_SIZE_B 12 /* 4k per bit */
1221
#define BM_BLOCK_SIZE (1<<BM_BLOCK_SIZE_B)
1222
/* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1223
* per sector of on disk bitmap */
1224
#define BM_EXT_SIZE_B (BM_BLOCK_SIZE_B + MD_HARDSECT_B + 3) /* = 24 */
1225
#define BM_EXT_SIZE (1<<BM_EXT_SIZE_B)
1227
#if (BM_EXT_SIZE_B != 24) || (BM_BLOCK_SIZE_B != 12)
1228
#error "HAVE YOU FIXED drbdmeta AS WELL??"
1231
/* thus many _storage_ sectors are described by one bit */
1232
#define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SIZE_B-9))
1233
#define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SIZE_B-9))
1234
#define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1236
/* bit to represented kilo byte conversion */
1237
#define Bit2KB(bits) ((bits)<<(BM_BLOCK_SIZE_B-10))
1239
/* in which _bitmap_ extent (resp. sector) the bit for a certain
1240
* _storage_ sector is located in */
1241
#define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SIZE_B-9))
1243
/* how much _storage_ sectors we have per bitmap sector */
1244
#define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SIZE_B-9))
1245
#define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1247
/* in one sector of the bitmap, we have this many activity_log extents. */
1248
#define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SIZE_B - AL_EXTENT_SIZE_B))
1249
#define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SIZE_B-BM_BLOCK_SIZE_B-LN2_BPL))
1251
#define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SIZE_B - BM_BLOCK_SIZE_B)
1252
#define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1254
/* the extent in "PER_EXTENT" below is an activity log extent
1255
* we need that many (long words/bytes) to store the bitmap
1256
* of one AL_EXTENT_SIZE chunk of storage.
1257
* we can store the bitmap for that many AL_EXTENTS within
1258
* one sector of the _on_disk_ bitmap:
1259
* bit 0 bit 37 bit 38 bit (512*8)-1
1260
* ...|........|........|.. // ..|........|
1261
* sect. 0 `296 `304 ^(512*8*8)-1
1263
#define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1264
#define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1265
#define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1268
#define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1269
#define DRBD_MAX_SECTORS_BM \
1270
((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SIZE_B-9)))
1271
#if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1272
#define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1273
#define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1274
#elif !defined(CONFIG_LBD) && BITS_PER_LONG == 32
1275
#define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1276
#define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1278
#define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1279
/* 16 TB in units of sectors */
1280
#if BITS_PER_LONG == 32
1281
/* adjust by one page worth of bitmap,
1282
* so we won't wrap around in drbd_bm_find_next_bit.
1283
* you should use 64bit OS for that much storage, anyways. */
1284
#define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1286
#define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0x1LU << 32)
1290
/* Sector shift value for the "hash" functions of tl_hash and ee_hash tables.
1291
* With a value of 6 all IO in one 32K block make it to the same slot of the
1294
#define DRBD_MAX_SEGMENT_SIZE (1U<<(9+HT_SHIFT))
1296
/* Number of elements in the app_reads_hash */
1297
#define APP_R_HSIZE 15
1299
extern int drbd_bm_init(struct drbd_conf *mdev);
1300
extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors);
1301
extern void drbd_bm_cleanup(struct drbd_conf *mdev);
1302
extern void drbd_bm_set_all(struct drbd_conf *mdev);
1303
extern void drbd_bm_clear_all(struct drbd_conf *mdev);
1304
extern int drbd_bm_set_bits(
1305
struct drbd_conf *mdev, unsigned long s, unsigned long e);
1306
extern int drbd_bm_clear_bits(
1307
struct drbd_conf *mdev, unsigned long s, unsigned long e);
1308
/* bm_set_bits variant for use while holding drbd_bm_lock */
1309
extern int _drbd_bm_set_bits(struct drbd_conf *mdev,
1310
const unsigned long s, const unsigned long e);
1311
extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr);
1312
extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr);
1313
extern int drbd_bm_write_sect(struct drbd_conf *mdev, unsigned long enr) __must_hold(local);
1314
extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
1315
extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
1316
extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev,
1317
unsigned long al_enr);
1318
extern size_t drbd_bm_words(struct drbd_conf *mdev);
1319
extern unsigned long drbd_bm_bits(struct drbd_conf *mdev);
1320
extern sector_t drbd_bm_capacity(struct drbd_conf *mdev);
1321
extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1322
/* bm_find_next variants for use while you hold drbd_bm_lock() */
1323
extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1324
extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo);
1325
extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev);
1326
extern int drbd_bm_rs_done(struct drbd_conf *mdev);
1327
/* for receive_bitmap */
1328
extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset,
1329
size_t number, unsigned long *buffer);
1330
/* for _drbd_send_bitmap and drbd_bm_write_sect */
1331
extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset,
1332
size_t number, unsigned long *buffer);
1334
extern void drbd_bm_lock(struct drbd_conf *mdev, char *why);
1335
extern void drbd_bm_unlock(struct drbd_conf *mdev);
1337
extern void _drbd_bm_recount_bits(struct drbd_conf *mdev, char *file, int line);
1338
#define drbd_bm_recount_bits(mdev) \
1339
_drbd_bm_recount_bits(mdev, __FILE__, __LINE__)
1340
extern int drbd_bm_count_bits(struct drbd_conf *mdev, const unsigned long s, const unsigned long e);
1343
extern struct kmem_cache *drbd_request_cache;
1344
extern struct kmem_cache *drbd_ee_cache;
1345
extern mempool_t *drbd_request_mempool;
1346
extern mempool_t *drbd_ee_mempool;
1348
extern struct page *drbd_pp_pool; /* drbd's page pool */
1349
extern spinlock_t drbd_pp_lock;
1350
extern int drbd_pp_vacant;
1351
extern wait_queue_head_t drbd_pp_wait;
1353
extern rwlock_t global_state_lock;
1355
extern struct drbd_conf *drbd_new_device(unsigned int minor);
1356
extern void drbd_free_mdev(struct drbd_conf *mdev);
1358
/* Dynamic tracing framework */
1359
#ifdef ENABLE_DYNAMIC_TRACE
1361
extern int proc_details;
1362
extern int trace_type;
1363
extern int trace_devs;
1364
extern int trace_level;
1375
TraceTypePacket = 0x00000001,
1376
TraceTypeRq = 0x00000002,
1377
TraceTypeUuid = 0x00000004,
1378
TraceTypeResync = 0x00000008,
1379
TraceTypeEE = 0x00000010,
1380
TraceTypeUnplug = 0x00000020,
1381
TraceTypeNl = 0x00000040,
1382
TraceTypeALExts = 0x00000080,
1383
TraceTypeIntRq = 0x00000100,
1384
TraceTypeMDIO = 0x00000200,
1385
TraceTypeEpochs = 0x00000400,
1389
is_trace(unsigned int type, unsigned int level) {
1390
return (trace_level >= level) && (type & trace_type);
1393
is_mdev_trace(struct drbd_conf *mdev, unsigned int type, unsigned int level) {
1394
return is_trace(type, level) &&
1395
((1 << mdev_to_minor(mdev)) & trace_devs);
1398
#define MTRACE(type, lvl, code...) \
1400
if (unlikely(is_mdev_trace(mdev, type, lvl))) { \
1405
#define TRACE(type, lvl, code...) \
1407
if (unlikely(is_trace(type, lvl))) { \
1412
/* Buffer printing support
1413
* dbg_print_flags: used for Flags arg to drbd_print_buffer
1414
* - DBGPRINT_BUFFADDR; if set, each line starts with the
1415
* virtual address of the line being output. If clear,
1416
* each line starts with the offset from the beginning
1418
enum dbg_print_flags {
1419
DBGPRINT_BUFFADDR = 0x0001,
1422
extern void drbd_print_uuid(struct drbd_conf *mdev, unsigned int idx);
1424
extern void drbd_print_buffer(const char *prefix, unsigned int flags, int size,
1425
const void *buffer, const void *buffer_va,
1426
unsigned int length);
1428
/* Bio printing support */
1429
extern void _dump_bio(const char *pfx, struct drbd_conf *mdev, struct bio *bio, int complete, struct drbd_request *r);
1431
static inline void dump_bio(struct drbd_conf *mdev,
1432
struct bio *bio, int complete, struct drbd_request *r)
1434
MTRACE(TraceTypeRq, TraceLvlSummary,
1435
_dump_bio("Rq", mdev, bio, complete, r);
1439
static inline void dump_internal_bio(const char *pfx, struct drbd_conf *mdev, struct bio *bio, int complete)
1441
MTRACE(TraceTypeIntRq, TraceLvlSummary,
1442
_dump_bio(pfx, mdev, bio, complete, NULL);
1446
/* Packet dumping support */
1447
extern void _dump_packet(struct drbd_conf *mdev, struct socket *sock,
1448
int recv, union Drbd_Polymorph_Packet *p,
1449
char *file, int line);
1452
dump_packet(struct drbd_conf *mdev, struct socket *sock,
1453
int recv, union Drbd_Polymorph_Packet *p, char *file, int line)
1455
MTRACE(TraceTypePacket, TraceLvlSummary,
1456
_dump_packet(mdev, sock, recv, p, file, line);
1462
#define MTRACE(ignored...) ((void)0)
1463
#define TRACE(ignored...) ((void)0)
1465
#define dump_bio(ignored...) ((void)0)
1466
#define dump_internal_bio(ignored...) ((void)0)
1467
#define dump_packet(ignored...) ((void)0)
1471
extern int drbd_make_request_26(struct request_queue *q, struct bio *bio);
1472
extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req);
1473
extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1474
extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1478
extern void drbd_suspend_io(struct drbd_conf *mdev);
1479
extern void drbd_resume_io(struct drbd_conf *mdev);
1480
extern char *ppsize(char *buf, unsigned long long size);
1481
extern sector_t drbd_new_dev_size(struct drbd_conf *,
1482
struct drbd_backing_dev *);
1483
enum determin_dev_size_enum { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
1484
extern enum determin_dev_size_enum drbd_determin_dev_size(struct drbd_conf *) __must_hold(local);
1485
extern void resync_after_online_grow(struct drbd_conf *);
1486
extern void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int) __must_hold(local);
1487
extern int drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role,
1489
enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev);
1490
extern int drbd_khelper(struct drbd_conf *mdev, char *cmd);
1493
extern int drbd_worker(struct Drbd_thread *thi);
1494
extern void drbd_alter_sa(struct drbd_conf *mdev, int na);
1495
extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side);
1496
extern void resume_next_sg(struct drbd_conf *mdev);
1497
extern void suspend_other_sg(struct drbd_conf *mdev);
1498
extern int drbd_resync_finished(struct drbd_conf *mdev);
1499
/* maybe rather drbd_main.c ? */
1500
extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
1501
struct drbd_backing_dev *bdev, sector_t sector, int rw);
1502
extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int);
1504
static inline void ov_oos_print(struct drbd_conf *mdev)
1506
if (mdev->ov_last_oos_size) {
1507
ERR("Out of sync: start=%llu, size=%lu (sectors)\n",
1508
(unsigned long long)mdev->ov_last_oos_start,
1509
(unsigned long)mdev->ov_last_oos_size);
1511
mdev->ov_last_oos_size=0;
1515
void drbd_csum(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
1516
/* worker callbacks */
1517
extern int w_req_cancel_conflict(struct drbd_conf *, struct drbd_work *, int);
1518
extern int w_read_retry_remote(struct drbd_conf *, struct drbd_work *, int);
1519
extern int w_e_end_data_req(struct drbd_conf *, struct drbd_work *, int);
1520
extern int w_e_end_rsdata_req(struct drbd_conf *, struct drbd_work *, int);
1521
extern int w_e_end_csum_rs_req(struct drbd_conf *, struct drbd_work *, int);
1522
extern int w_e_end_ov_reply(struct drbd_conf *, struct drbd_work *, int);
1523
extern int w_e_end_ov_req(struct drbd_conf *, struct drbd_work *, int);
1524
extern int w_ov_finished(struct drbd_conf *, struct drbd_work *, int);
1525
extern int w_resync_inactive(struct drbd_conf *, struct drbd_work *, int);
1526
extern int w_resume_next_sg(struct drbd_conf *, struct drbd_work *, int);
1527
extern int w_io_error(struct drbd_conf *, struct drbd_work *, int);
1528
extern int w_send_write_hint(struct drbd_conf *, struct drbd_work *, int);
1529
extern int w_make_resync_request(struct drbd_conf *, struct drbd_work *, int);
1530
extern int w_send_dblock(struct drbd_conf *, struct drbd_work *, int);
1531
extern int w_send_barrier(struct drbd_conf *, struct drbd_work *, int);
1532
extern int w_send_read_req(struct drbd_conf *, struct drbd_work *, int);
1533
extern int w_prev_work_done(struct drbd_conf *, struct drbd_work *, int);
1534
extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int);
1536
extern void resync_timer_fn(unsigned long data);
1538
/* drbd_receiver.c */
1539
extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list);
1540
extern struct Tl_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev,
1543
unsigned int data_size,
1544
gfp_t gfp_mask) __must_hold(local);
1545
extern void drbd_free_ee(struct drbd_conf *mdev, struct Tl_epoch_entry *e);
1546
extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1547
struct list_head *head);
1548
extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1549
struct list_head *head);
1550
extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled);
1551
extern void _drbd_clear_done_ee(struct drbd_conf *mdev);
1553
/* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1554
* mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1555
static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1556
char __user *optval, int optlen)
1559
if (level == SOL_SOCKET)
1560
err = sock_setsockopt(sock, level, optname, optval, optlen);
1562
err = sock->ops->setsockopt(sock, level, optname, optval,
1567
static inline void drbd_tcp_cork(struct socket *sock)
1570
(void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1571
(char __user *)&val, sizeof(val));
1574
static inline void drbd_tcp_uncork(struct socket *sock)
1577
(void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1578
(char __user *)&val, sizeof(val));
1581
static inline void drbd_tcp_nodelay(struct socket *sock)
1584
(void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1585
(char __user *)&val, sizeof(val));
1588
static inline void drbd_tcp_quickack(struct socket *sock)
1591
(void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1592
(char __user *)&val, sizeof(val));
1595
void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo);
1598
extern struct proc_dir_entry *drbd_proc;
1599
extern struct file_operations drbd_proc_fops;
1600
extern const char *conns_to_name(enum drbd_conns s);
1601
extern const char *roles_to_name(enum drbd_role s);
1604
extern void drbd_al_begin_io(struct drbd_conf *mdev, sector_t sector);
1605
extern void drbd_al_complete_io(struct drbd_conf *mdev, sector_t sector);
1606
extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector);
1607
extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1608
extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1609
extern void drbd_rs_cancel_all(struct drbd_conf *mdev);
1610
extern int drbd_rs_del_all(struct drbd_conf *mdev);
1611
extern void drbd_rs_failed_io(struct drbd_conf *mdev,
1612
sector_t sector, int size);
1613
extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *);
1614
extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
1615
int size, const char *file, const unsigned int line);
1616
#define drbd_set_in_sync(mdev, sector, size) \
1617
__drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1618
extern void __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
1619
int size, const char *file, const unsigned int line);
1620
#define drbd_set_out_of_sync(mdev, sector, size) \
1621
__drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1622
extern void drbd_al_apply_to_bm(struct drbd_conf *mdev);
1623
extern void drbd_al_to_on_disk_bm(struct drbd_conf *mdev);
1624
extern void drbd_al_shrink(struct drbd_conf *mdev);
1629
void drbd_nl_cleanup(void);
1630
int __init drbd_nl_init(void);
1631
void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state_t);
1632
void drbd_bcast_sync_progress(struct drbd_conf *mdev);
1633
void drbd_bcast_ee(struct drbd_conf *mdev,
1634
const char *reason, const int dgs,
1635
const char* seen_hash, const char* calc_hash,
1636
const struct Tl_epoch_entry* e);
1639
/** DRBD State macros:
1640
* These macros are used to express state changes in easily readable form.
1642
* The NS macros expand to a mask and a value, that can be bit ored onto the
1643
* current state as soon as the spinlock (req_lock) was taken.
1645
* The _NS macros are used for state functions that get called with the
1646
* spinlock. These macros expand directly to the new state value.
1648
* Besides the basic forms NS() and _NS() additional _?NS[23] are defined
1649
* to express state changes that affect more than one aspect of the state.
1651
* E.g. NS2(conn, Connected, peer, Secondary)
1652
* Means that the network connection was established and that the peer
1653
* is in secondary role.
1655
#define peer_mask role_mask
1656
#define pdsk_mask disk_mask
1658
#define user_isp_mask 1
1659
#define aftr_isp_mask 1
1662
({ union drbd_state_t mask; mask.i = 0; mask.T = T##_mask; mask; }), \
1663
({ union drbd_state_t val; val.i = 0; val.T = (S); val; })
1664
#define NS2(T1, S1, T2, S2) \
1665
({ union drbd_state_t mask; mask.i = 0; mask.T1 = T1##_mask; \
1666
mask.T2 = T2##_mask; mask; }), \
1667
({ union drbd_state_t val; val.i = 0; val.T1 = (S1); \
1668
val.T2 = (S2); val; })
1669
#define NS3(T1, S1, T2, S2, T3, S3) \
1670
({ union drbd_state_t mask; mask.i = 0; mask.T1 = T1##_mask; \
1671
mask.T2 = T2##_mask; mask.T3 = T3##_mask; mask; }), \
1672
({ union drbd_state_t val; val.i = 0; val.T1 = (S1); \
1673
val.T2 = (S2); val.T3 = (S3); val; })
1675
#define _NS(D, T, S) \
1676
D, ({ union drbd_state_t __ns; __ns.i = D->state.i; __ns.T = (S); __ns; })
1677
#define _NS2(D, T1, S1, T2, S2) \
1678
D, ({ union drbd_state_t __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1679
__ns.T2 = (S2); __ns; })
1680
#define _NS3(D, T1, S1, T2, S2, T3, S3) \
1681
D, ({ union drbd_state_t __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1682
__ns.T2 = (S2); __ns.T3 = (S3); __ns; })
1685
* inline helper functions
1686
*************************/
1688
static inline void drbd_state_lock(struct drbd_conf *mdev)
1690
wait_event(mdev->misc_wait,
1691
!test_and_set_bit(CLUSTER_ST_CHANGE, &mdev->flags));
1694
static inline void drbd_state_unlock(struct drbd_conf *mdev)
1696
clear_bit(CLUSTER_ST_CHANGE, &mdev->flags);
1697
wake_up(&mdev->misc_wait);
1700
static inline int _drbd_set_state(struct drbd_conf *mdev,
1701
union drbd_state_t ns, enum chg_state_flags flags,
1702
struct completion *done)
1706
read_lock(&global_state_lock);
1707
rv = __drbd_set_state(mdev, ns, flags, done);
1708
read_unlock(&global_state_lock);
1713
static inline int drbd_request_state(struct drbd_conf *mdev,
1714
union drbd_state_t mask,
1715
union drbd_state_t val)
1717
return _drbd_request_state(mdev, mask, val, ChgStateVerbose + ChgOrdered);
1721
* drbd_chk_io_error: Handles the on_io_error setting, should be called from
1722
* all io completion handlers. See also drbd_io_error().
1724
static inline void __drbd_chk_io_error(struct drbd_conf *mdev, int forcedetach)
1726
switch (mdev->bc->dc.on_io_error) {
1729
if (printk_ratelimit())
1730
ERR("Local IO failed. Passing error on...\n");
1733
/* NOTE fall through to detach case if forcedetach set */
1736
if (mdev->state.disk > Failed) {
1737
_drbd_set_state(_NS(mdev, disk, Failed), ChgStateHard, NULL);
1738
ERR("Local IO failed. Detaching...\n");
1744
static inline void drbd_chk_io_error(struct drbd_conf *mdev,
1745
int error, int forcedetach)
1748
unsigned long flags;
1749
spin_lock_irqsave(&mdev->req_lock, flags);
1750
__drbd_chk_io_error(mdev, forcedetach);
1751
spin_unlock_irqrestore(&mdev->req_lock, flags);
1755
/* Returns the first sector number of our meta data,
1756
* which, for internal meta data, happens to be the maximum capacity
1757
* we could agree upon with our peer
1759
static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1761
switch (bdev->dc.meta_dev_idx) {
1762
case DRBD_MD_INDEX_INTERNAL:
1763
case DRBD_MD_INDEX_FLEX_INT:
1764
return bdev->md.md_offset + bdev->md.bm_offset;
1765
case DRBD_MD_INDEX_FLEX_EXT:
1767
return bdev->md.md_offset;
1771
/* returns the last sector number of our meta data,
1772
* to be able to catch out of band md access */
1773
static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1775
switch (bdev->dc.meta_dev_idx) {
1776
case DRBD_MD_INDEX_INTERNAL:
1777
case DRBD_MD_INDEX_FLEX_INT:
1778
return bdev->md.md_offset + MD_AL_OFFSET - 1;
1779
case DRBD_MD_INDEX_FLEX_EXT:
1781
return bdev->md.md_offset + bdev->md.md_size_sect;
1785
/* Returns the number of 512 byte sectors of the device */
1786
static inline sector_t drbd_get_capacity(struct block_device *bdev)
1788
/* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1789
return bdev ? bdev->bd_inode->i_size >> 9 : 0;
1792
/* returns the capacity we announce to out peer.
1793
* we clip ourselves at the various MAX_SECTORS, because if we don't,
1794
* current implementation will oops sooner or later */
1795
static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1798
switch (bdev->dc.meta_dev_idx) {
1799
case DRBD_MD_INDEX_INTERNAL:
1800
case DRBD_MD_INDEX_FLEX_INT:
1801
s = drbd_get_capacity(bdev->backing_bdev)
1802
? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1803
drbd_md_first_sector(bdev))
1806
case DRBD_MD_INDEX_FLEX_EXT:
1807
s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1808
drbd_get_capacity(bdev->backing_bdev));
1809
/* clip at maximum size the meta device can support */
1810
s = min_t(sector_t, s,
1811
BM_EXT_TO_SECT(bdev->md.md_size_sect
1812
- bdev->md.bm_offset));
1815
s = min_t(sector_t, DRBD_MAX_SECTORS,
1816
drbd_get_capacity(bdev->backing_bdev));
1821
/* returns the sector number of our meta data 'super' block */
1822
static inline sector_t drbd_md_ss__(struct drbd_conf *mdev,
1823
struct drbd_backing_dev *bdev)
1825
switch (bdev->dc.meta_dev_idx) {
1826
default: /* external, some index */
1827
return MD_RESERVED_SECT * bdev->dc.meta_dev_idx;
1828
case DRBD_MD_INDEX_INTERNAL:
1829
/* with drbd08, internal meta data is always "flexible" */
1830
case DRBD_MD_INDEX_FLEX_INT:
1831
/* sizeof(struct md_on_disk_07) == 4k
1832
* position: last 4k aligned block of 4k size */
1833
if (!bdev->backing_bdev) {
1834
if (__ratelimit(&drbd_ratelimit_state)) {
1835
ERR("bdev->backing_bdev==NULL\n");
1840
return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL)
1842
case DRBD_MD_INDEX_FLEX_EXT:
1848
_drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1850
list_add_tail(&w->list, &q->q);
1855
drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w)
1857
unsigned long flags;
1858
spin_lock_irqsave(&q->q_lock, flags);
1859
list_add(&w->list, &q->q);
1860
up(&q->s); /* within the spinlock,
1861
see comment near end of drbd_worker() */
1862
spin_unlock_irqrestore(&q->q_lock, flags);
1866
drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1868
unsigned long flags;
1869
spin_lock_irqsave(&q->q_lock, flags);
1870
list_add_tail(&w->list, &q->q);
1871
up(&q->s); /* within the spinlock,
1872
see comment near end of drbd_worker() */
1873
spin_unlock_irqrestore(&q->q_lock, flags);
1876
static inline void wake_asender(struct drbd_conf *mdev)
1878
if (test_bit(SIGNAL_ASENDER, &mdev->flags))
1879
force_sig(DRBD_SIG, mdev->asender.task);
1882
static inline void request_ping(struct drbd_conf *mdev)
1884
set_bit(SEND_PING, &mdev->flags);
1888
static inline int drbd_send_short_cmd(struct drbd_conf *mdev,
1889
enum Drbd_Packet_Cmd cmd)
1891
struct Drbd_Header h;
1892
return drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, &h, sizeof(h));
1895
static inline int drbd_send_ping(struct drbd_conf *mdev)
1897
struct Drbd_Header h;
1898
return drbd_send_cmd(mdev, USE_META_SOCKET, Ping, &h, sizeof(h));
1901
static inline int drbd_send_ping_ack(struct drbd_conf *mdev)
1903
struct Drbd_Header h;
1904
return drbd_send_cmd(mdev, USE_META_SOCKET, PingAck, &h, sizeof(h));
1907
static inline void drbd_thread_stop(struct Drbd_thread *thi)
1909
_drbd_thread_stop(thi, FALSE, TRUE);
1912
static inline void drbd_thread_stop_nowait(struct Drbd_thread *thi)
1914
_drbd_thread_stop(thi, FALSE, FALSE);
1917
static inline void drbd_thread_restart_nowait(struct Drbd_thread *thi)
1919
_drbd_thread_stop(thi, TRUE, FALSE);
1922
/* counts how many answer packets packets we expect from our peer,
1923
* for either explicit application requests,
1924
* or implicit barrier packets as necessary.
1927
* _req_mod(req, queue_for_net_write or queue_for_net_read);
1928
* it is much easier and equally valid to count what we queue for the
1929
* worker, even before it actually was queued or send.
1930
* (drbd_make_request_common; recovery path on read io-error)
1932
* got_BarrierAck (respective tl_clear, tl_clear_barrier)
1933
* _req_mod(req, data_received)
1934
* [from receive_DataReply]
1935
* _req_mod(req, write_acked_by_peer or recv_acked_by_peer or neg_acked)
1936
* [from got_BlockAck (WriteAck, RecvAck)]
1937
* for some reason it is NOT decreased in got_NegAck,
1938
* but in the resulting cleanup code from report_params.
1939
* we should try to remember the reason for that...
1940
* _req_mod(req, send_failed or send_canceled)
1941
* _req_mod(req, connection_lost_while_pending)
1942
* [from tl_clear_barrier]
1944
static inline void inc_ap_pending(struct drbd_conf *mdev)
1946
atomic_inc(&mdev->ap_pending_cnt);
1949
#define ERR_IF_CNT_IS_NEGATIVE(which) \
1950
if (atomic_read(&mdev->which) < 0) \
1951
ERR("in %s:%d: " #which " = %d < 0 !\n", \
1952
__func__ , __LINE__ , \
1953
atomic_read(&mdev->which))
1955
#define dec_ap_pending(mdev) do { \
1956
typecheck(struct drbd_conf *, mdev); \
1957
if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \
1958
wake_up(&mdev->misc_wait); \
1959
ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
1961
/* counts how many resync-related answers we still expect from the peer
1963
* SyncTarget sends RSDataRequest (and expects RSDataReply)
1964
* SyncSource sends RSDataReply (and expects WriteAck whith ID_SYNCER)
1965
* (or NegAck with ID_SYNCER)
1967
static inline void inc_rs_pending(struct drbd_conf *mdev)
1969
atomic_inc(&mdev->rs_pending_cnt);
1972
#define dec_rs_pending(mdev) do { \
1973
typecheck(struct drbd_conf *, mdev); \
1974
atomic_dec(&mdev->rs_pending_cnt); \
1975
ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
1977
/* counts how many answers we still need to send to the peer.
1979
* receive_Data unless protocol A;
1980
* we need to send a RecvAck (proto B)
1981
* or WriteAck (proto C)
1982
* receive_RSDataReply (recv_resync_read) we need to send a WriteAck
1983
* receive_DataRequest (receive_RSDataRequest) we need to send back Data
1984
* receive_Barrier_* we need to send a BarrierAck
1986
static inline void inc_unacked(struct drbd_conf *mdev)
1988
atomic_inc(&mdev->unacked_cnt);
1991
#define dec_unacked(mdev) do { \
1992
typecheck(struct drbd_conf *, mdev); \
1993
atomic_dec(&mdev->unacked_cnt); \
1994
ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
1996
#define sub_unacked(mdev, n) do { \
1997
typecheck(struct drbd_conf *, mdev); \
1998
atomic_sub(n, &mdev->unacked_cnt); \
1999
ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2002
static inline void dec_net(struct drbd_conf *mdev)
2004
if (atomic_dec_and_test(&mdev->net_cnt))
2005
wake_up(&mdev->misc_wait);
2009
* inc_net: Returns TRUE when it is ok to access mdev->net_conf. You
2010
* should call dec_net() when finished looking at mdev->net_conf.
2012
static inline int inc_net(struct drbd_conf *mdev)
2016
atomic_inc(&mdev->net_cnt);
2017
have_net_conf = mdev->state.conn >= Unconnected;
2020
return have_net_conf;
2024
* inc_local: Returns TRUE when local IO is possible. If it returns
2025
* TRUE you should call dec_local() after IO is completed.
2027
#define inc_local_if_state(M,MINS) __cond_lock(local, _inc_local_if_state(M,MINS))
2028
#define inc_local(M) __cond_lock(local, _inc_local_if_state(M,Inconsistent))
2030
static inline void dec_local(struct drbd_conf *mdev)
2033
if (atomic_dec_and_test(&mdev->local_cnt))
2034
wake_up(&mdev->misc_wait);
2035
D_ASSERT(atomic_read(&mdev->local_cnt) >= 0);
2039
static inline int _inc_local_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
2043
atomic_inc(&mdev->local_cnt);
2044
io_allowed = (mdev->state.disk >= mins);
2050
extern int _inc_local_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins);
2053
/* you must have an "inc_local" reference */
2054
static inline void drbd_get_syncer_progress(struct drbd_conf *mdev,
2055
unsigned long *bits_left, unsigned int *per_mil_done)
2058
* this is to break it at compile time when we change that
2059
* (we may feel 4TB maximum storage per drbd is not enough)
2061
typecheck(unsigned long, mdev->rs_total);
2063
/* note: both rs_total and rs_left are in bits, i.e. in
2064
* units of BM_BLOCK_SIZE.
2065
* for the percentage, we don't care. */
2067
*bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
2068
/* >> 10 to prevent overflow,
2069
* +1 to prevent division by zero */
2070
if (*bits_left > mdev->rs_total) {
2071
/* doh. maybe a logic bug somewhere.
2072
* may also be just a race condition
2073
* between this and a disconnect during sync.
2074
* for now, just prevent in-kernel buffer overflow.
2077
drbd_WARN("cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2078
conns_to_name(mdev->state.conn),
2079
*bits_left, mdev->rs_total, mdev->rs_failed);
2082
/* make sure the calculation happens in long context */
2083
unsigned long tmp = 1000UL -
2084
(*bits_left >> 10)*1000UL
2085
/ ((mdev->rs_total >> 10) + 1UL);
2086
*per_mil_done = tmp;
2091
/* this throttles on-the-fly application requests
2092
* according to max_buffers settings;
2093
* maybe re-implement using semaphores? */
2094
static inline int drbd_get_max_buffers(struct drbd_conf *mdev)
2096
int mxb = 1000000; /* arbitrary limit on open requests */
2097
if (inc_net(mdev)) {
2098
mxb = mdev->net_conf->max_buffers;
2104
static inline int drbd_state_is_stable(union drbd_state_t s)
2107
/* DO NOT add a default clause, we want the compiler to warn us
2108
* for any newly introduced state we may have forgotten to add here */
2110
switch ((enum drbd_conns)s.conn) {
2111
/* new io only accepted when there is no connection, ... */
2114
/* ... or there is a well established connection. */
2122
/* maybe stable, look at the disk state */
2125
/* no new io accepted during tansitional states
2126
* like handshake or teardown */
2131
case NetworkFailure:
2134
case WFReportParams:
2145
switch ((enum drbd_disk_state)s.disk) {
2151
/* disk state is stable as well. */
2154
/* no new io accepted during tansitional states */
2167
static inline int __inc_ap_bio_cond(struct drbd_conf *mdev)
2169
int mxb = drbd_get_max_buffers(mdev);
2171
if (mdev->state.susp)
2173
if (test_bit(SUSPEND_IO, &mdev->flags))
2176
/* to avoid potential deadlock or bitmap corruption,
2177
* in various places, we only allow new application io
2178
* to start during "stable" states. */
2180
/* no new io accepted when attaching or detaching the disk */
2181
if (!drbd_state_is_stable(mdev->state))
2184
/* since some older kernels don't have atomic_add_unless,
2185
* and we are within the spinlock anyways, we have this workaround. */
2186
if (atomic_read(&mdev->ap_bio_cnt) > mxb)
2188
if (test_bit(BITMAP_IO, &mdev->flags))
2193
/* I'd like to use wait_event_lock_irq,
2194
* but I'm not sure when it got introduced,
2195
* and not sure when it has 3 or 4 arguments */
2196
static inline void inc_ap_bio(struct drbd_conf *mdev, int one_or_two)
2198
/* compare with after_state_ch,
2199
* os.conn != WFBitMapS && ns.conn == WFBitMapS */
2203
* as long as the device is suspended
2204
* until the bitmap is no longer on the fly during connection
2205
* handshake as long as we would exeed the max_buffer limit.
2207
* to avoid races with the reconnect code,
2208
* we need to atomic_inc within the spinlock. */
2210
spin_lock_irq(&mdev->req_lock);
2211
while (!__inc_ap_bio_cond(mdev)) {
2212
prepare_to_wait(&mdev->misc_wait, &wait, TASK_UNINTERRUPTIBLE);
2213
spin_unlock_irq(&mdev->req_lock);
2215
finish_wait(&mdev->misc_wait, &wait);
2216
spin_lock_irq(&mdev->req_lock);
2218
atomic_add(one_or_two, &mdev->ap_bio_cnt);
2219
spin_unlock_irq(&mdev->req_lock);
2222
static inline void dec_ap_bio(struct drbd_conf *mdev)
2224
int mxb = drbd_get_max_buffers(mdev);
2225
int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt);
2227
D_ASSERT(ap_bio >= 0);
2228
/* this currently does wake_up for every dec_ap_bio!
2229
* maybe rather introduce some type of hysteresis?
2230
* e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2232
wake_up(&mdev->misc_wait);
2233
if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) {
2234
if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags))
2235
drbd_queue_work(&mdev->data.work, &mdev->bm_io_work.w);
2239
static inline void drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val)
2241
mdev->ed_uuid = val;
2243
MTRACE(TraceTypeUuid, TraceLvlMetrics,
2244
INFO(" exposed data uuid now %016llX\n",
2245
(unsigned long long)val);
2249
static inline int seq_cmp(u32 a, u32 b)
2251
/* we assume wrap around at 32bit.
2252
* for wrap around at 24bit (old atomic_t),
2256
return (s32)(a) - (s32)(b);
2258
#define seq_lt(a, b) (seq_cmp((a), (b)) < 0)
2259
#define seq_gt(a, b) (seq_cmp((a), (b)) > 0)
2260
#define seq_ge(a, b) (seq_cmp((a), (b)) >= 0)
2261
#define seq_le(a, b) (seq_cmp((a), (b)) <= 0)
2262
/* CAUTION: please no side effects in arguments! */
2263
#define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b)))
2265
static inline void update_peer_seq(struct drbd_conf *mdev, unsigned int new_seq)
2268
spin_lock(&mdev->peer_seq_lock);
2269
m = seq_max(mdev->peer_seq, new_seq);
2271
spin_unlock(&mdev->peer_seq_lock);
2273
wake_up(&mdev->seq_wait);
2276
static inline void drbd_update_congested(struct drbd_conf *mdev)
2278
struct sock *sk = mdev->data.socket->sk;
2279
if (sk->sk_wmem_queued > sk->sk_sndbuf * 4 / 5)
2280
set_bit(NET_CONGESTED, &mdev->flags);
2283
static inline int drbd_queue_order_type(struct drbd_conf *mdev)
2285
/* sorry, we currently have no working implementation
2286
* of distributed TCQ stuff */
2287
#ifndef QUEUE_ORDERED_NONE
2288
#define QUEUE_ORDERED_NONE 0
2290
return QUEUE_ORDERED_NONE;
2293
static inline void drbd_blk_run_queue(struct request_queue *q)
2295
if (q && q->unplug_fn)
2299
static inline void drbd_kick_lo(struct drbd_conf *mdev)
2301
if (inc_local(mdev)) {
2302
drbd_blk_run_queue(bdev_get_queue(mdev->bc->backing_bdev));
2307
static inline void drbd_md_flush(struct drbd_conf *mdev)
2311
if (test_bit(MD_NO_BARRIER, &mdev->flags))
2314
r = blkdev_issue_flush(mdev->bc->md_bdev, NULL);
2316
set_bit(MD_NO_BARRIER, &mdev->flags);
2317
ERR("meta data flush failed with status %d, disabling md-flushes\n", r);
added
removed
ubuntu/drbd/drbd_int.h
