4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
6 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
7 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
8 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
10 drbd is free software; you can redistribute it and/or modify
11 it under the terms of the GNU General Public License as published by
12 the Free Software Foundation; either version 2, or (at your option)
15 drbd is distributed in the hope that it will be useful,
16 but WITHOUT ANY WARRANTY; without even the implied warranty of
17 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 GNU General Public License for more details.
20 You should have received a copy of the GNU General Public License
21 along with drbd; see the file COPYING. If not, write to
22 the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
29 #include <linux/compiler.h>
30 #include <linux/types.h>
31 #include <linux/version.h>
32 #include <linux/list.h>
33 #include <linux/sched.h>
34 #include <linux/bitops.h>
35 #include <linux/slab.h>
36 #include <linux/crypto.h>
37 #include <linux/ratelimit.h>
38 #include <linux/tcp.h>
39 #include <linux/mutex.h>
40 #include <linux/major.h>
41 #include <linux/blkdev.h>
42 #include <linux/genhd.h>
44 #include <linux/lru_cache.h>
47 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
48 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
49 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
50 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
52 # define __protected_by(x)
53 # define __protected_read_by(x)
54 # define __protected_write_by(x)
55 # define __must_hold(x)
58 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
60 /* module parameter, defined in drbd_main.c */
61 extern unsigned int minor_count
;
62 extern int disable_sendpage
;
64 extern unsigned int cn_idx
;
66 #ifdef CONFIG_DRBD_FAULT_INJECTION
67 extern int enable_faults
;
68 extern int fault_rate
;
69 extern int fault_devs
;
72 extern char usermode_helper
[];
82 /* I don't remember why XCPU ...
83 * This is used to wake the asender,
84 * and to interrupt sending the sending task
87 #define DRBD_SIG SIGXCPU
89 /* This is used to stop/restart our threads.
90 * Cannot use SIGTERM nor SIGKILL, since these
91 * are sent out by init on runlevel changes
92 * I choose SIGHUP for now.
94 #define DRBD_SIGKILL SIGHUP
96 /* All EEs on the free list should have ID_VACANT (== 0)
97 * freshly allocated EEs get !ID_VACANT (== 1)
98 * so if it says "cannot dereference null pointer at address 0x00000001",
99 * it is most likely one of these :( */
101 #define ID_IN_SYNC (4711ULL)
102 #define ID_OUT_OF_SYNC (4712ULL)
104 #define ID_SYNCER (-1ULL)
106 #define is_syncer_block_id(id) ((id) == ID_SYNCER)
111 /* to shorten dev_warn(DEV, "msg"); and relatives statements */
112 #define DEV (disk_to_dev(mdev->vdisk))
114 #define D_ASSERT(exp) if (!(exp)) \
115 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
117 #define ERR_IF(exp) if (({ \
118 int _b = (exp) != 0; \
119 if (_b) dev_err(DEV, "%s: (%s) in %s:%d\n", \
120 __func__, #exp, __FILE__, __LINE__); \
124 /* Defines to control fault insertion */
126 DRBD_FAULT_MD_WR
= 0, /* meta data write */
127 DRBD_FAULT_MD_RD
= 1, /* read */
128 DRBD_FAULT_RS_WR
= 2, /* resync */
129 DRBD_FAULT_RS_RD
= 3,
130 DRBD_FAULT_DT_WR
= 4, /* data */
131 DRBD_FAULT_DT_RD
= 5,
132 DRBD_FAULT_DT_RA
= 6, /* data read ahead */
133 DRBD_FAULT_BM_ALLOC
= 7, /* bitmap allocation */
134 DRBD_FAULT_AL_EE
= 8, /* alloc ee */
135 DRBD_FAULT_RECEIVE
= 9, /* Changes some bytes upon receiving a [rs]data block */
140 #ifdef CONFIG_DRBD_FAULT_INJECTION
142 _drbd_insert_fault(struct drbd_conf
*mdev
, unsigned int type
);
144 drbd_insert_fault(struct drbd_conf
*mdev
, unsigned int type
) {
146 (enable_faults
& (1<<type
)) &&
147 _drbd_insert_fault(mdev
, type
);
149 #define FAULT_ACTIVE(_m, _t) (drbd_insert_fault((_m), (_t)))
152 #define FAULT_ACTIVE(_m, _t) (0)
155 /* integer division, round _UP_ to the next integer */
156 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
157 /* usual integer division */
158 #define div_floor(A, B) ((A)/(B))
160 /* drbd_meta-data.c (still in drbd_main.c) */
161 /* 4th incarnation of the disk layout. */
162 #define DRBD_MD_MAGIC (DRBD_MAGIC+4)
164 extern struct drbd_conf
**minor_table
;
165 extern struct ratelimit_state drbd_ratelimit_state
;
169 /* receiver (data socket) */
171 P_DATA_REPLY
= 0x01, /* Response to P_DATA_REQUEST */
172 P_RS_DATA_REPLY
= 0x02, /* Response to P_RS_DATA_REQUEST */
175 P_BECOME_SYNC_TARGET
= 0x05,
176 P_BECOME_SYNC_SOURCE
= 0x06,
177 P_UNPLUG_REMOTE
= 0x07, /* Used at various times to hint the peer */
178 P_DATA_REQUEST
= 0x08, /* Used to ask for a data block */
179 P_RS_DATA_REQUEST
= 0x09, /* Used to ask for a data block for resync */
186 P_AUTH_CHALLENGE
= 0x10,
187 P_AUTH_RESPONSE
= 0x11,
188 P_STATE_CHG_REQ
= 0x12,
190 /* asender (meta socket */
193 P_RECV_ACK
= 0x15, /* Used in protocol B */
194 P_WRITE_ACK
= 0x16, /* Used in protocol C */
195 P_RS_WRITE_ACK
= 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
196 P_DISCARD_ACK
= 0x18, /* Used in proto C, two-primaries conflict detection */
197 P_NEG_ACK
= 0x19, /* Sent if local disk is unusable */
198 P_NEG_DREPLY
= 0x1a, /* Local disk is broken... */
199 P_NEG_RS_DREPLY
= 0x1b, /* Local disk is broken... */
200 P_BARRIER_ACK
= 0x1c,
201 P_STATE_CHG_REPLY
= 0x1d,
203 /* "new" commands, no longer fitting into the ordering scheme above */
205 P_OV_REQUEST
= 0x1e, /* data socket */
207 P_OV_RESULT
= 0x20, /* meta socket */
208 P_CSUM_RS_REQUEST
= 0x21, /* data socket */
209 P_RS_IS_IN_SYNC
= 0x22, /* meta socket */
210 P_SYNC_PARAM89
= 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
211 P_COMPRESSED_BITMAP
= 0x24, /* compressed or otherwise encoded bitmap transfer */
212 /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */
213 /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */
214 P_DELAY_PROBE
= 0x27, /* is used on BOTH sockets */
217 P_MAY_IGNORE
= 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
218 P_MAX_OPT_CMD
= 0x101,
220 /* special command ids for handshake */
222 P_HAND_SHAKE_M
= 0xfff1, /* First Packet on the MetaSock */
223 P_HAND_SHAKE_S
= 0xfff2, /* First Packet on the Socket */
225 P_HAND_SHAKE
= 0xfffe /* FIXED for the next century! */
228 static inline const char *cmdname(enum drbd_packets cmd
)
230 /* THINK may need to become several global tables
231 * when we want to support more than
233 static const char *cmdnames
[] = {
235 [P_DATA_REPLY
] = "DataReply",
236 [P_RS_DATA_REPLY
] = "RSDataReply",
237 [P_BARRIER
] = "Barrier",
238 [P_BITMAP
] = "ReportBitMap",
239 [P_BECOME_SYNC_TARGET
] = "BecomeSyncTarget",
240 [P_BECOME_SYNC_SOURCE
] = "BecomeSyncSource",
241 [P_UNPLUG_REMOTE
] = "UnplugRemote",
242 [P_DATA_REQUEST
] = "DataRequest",
243 [P_RS_DATA_REQUEST
] = "RSDataRequest",
244 [P_SYNC_PARAM
] = "SyncParam",
245 [P_SYNC_PARAM89
] = "SyncParam89",
246 [P_PROTOCOL
] = "ReportProtocol",
247 [P_UUIDS
] = "ReportUUIDs",
248 [P_SIZES
] = "ReportSizes",
249 [P_STATE
] = "ReportState",
250 [P_SYNC_UUID
] = "ReportSyncUUID",
251 [P_AUTH_CHALLENGE
] = "AuthChallenge",
252 [P_AUTH_RESPONSE
] = "AuthResponse",
254 [P_PING_ACK
] = "PingAck",
255 [P_RECV_ACK
] = "RecvAck",
256 [P_WRITE_ACK
] = "WriteAck",
257 [P_RS_WRITE_ACK
] = "RSWriteAck",
258 [P_DISCARD_ACK
] = "DiscardAck",
259 [P_NEG_ACK
] = "NegAck",
260 [P_NEG_DREPLY
] = "NegDReply",
261 [P_NEG_RS_DREPLY
] = "NegRSDReply",
262 [P_BARRIER_ACK
] = "BarrierAck",
263 [P_STATE_CHG_REQ
] = "StateChgRequest",
264 [P_STATE_CHG_REPLY
] = "StateChgReply",
265 [P_OV_REQUEST
] = "OVRequest",
266 [P_OV_REPLY
] = "OVReply",
267 [P_OV_RESULT
] = "OVResult",
268 [P_CSUM_RS_REQUEST
] = "CsumRSRequest",
269 [P_RS_IS_IN_SYNC
] = "CsumRSIsInSync",
270 [P_COMPRESSED_BITMAP
] = "CBitmap",
271 [P_DELAY_PROBE
] = "DelayProbe",
275 if (cmd
== P_HAND_SHAKE_M
)
277 if (cmd
== P_HAND_SHAKE_S
)
279 if (cmd
== P_HAND_SHAKE
)
281 if (cmd
>= P_MAX_CMD
)
283 return cmdnames
[cmd
];
286 /* for sending/receiving the bitmap,
287 * possibly in some encoding scheme */
290 * stores total bits and long words
291 * of the bitmap, so we don't need to
292 * call the accessor functions over and again. */
293 unsigned long bm_bits
;
294 unsigned long bm_words
;
295 /* during xfer, current position within the bitmap */
296 unsigned long bit_offset
;
297 unsigned long word_offset
;
299 /* statistics; index: (h->command == P_BITMAP) */
304 extern void INFO_bm_xfer_stats(struct drbd_conf
*mdev
,
305 const char *direction
, struct bm_xfer_ctx
*c
);
307 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx
*c
)
309 /* word_offset counts "native long words" (32 or 64 bit),
311 * Encoded packet may end at an unaligned bit offset.
312 * In case a fallback clear text packet is transmitted in
313 * between, we adjust this offset back to the last 64bit
314 * aligned "native long word", which makes coding and decoding
315 * the plain text bitmap much more convenient. */
316 #if BITS_PER_LONG == 64
317 c
->word_offset
= c
->bit_offset
>> 6;
318 #elif BITS_PER_LONG == 32
319 c
->word_offset
= c
->bit_offset
>> 5;
320 c
->word_offset
&= ~(1UL);
322 # error "unsupported BITS_PER_LONG"
327 #define __packed __attribute__((packed))
330 /* This is the layout for a packet on the wire.
331 * The byteorder is the network byte order.
332 * (except block_id and barrier fields.
333 * these are pointers to local structs
334 * and have no relevance for the partner,
335 * which just echoes them as received.)
337 * NOTE that the payload starts at a long aligned offset,
338 * regardless of 32 or 64 bit arch!
343 u16 length
; /* bytes of data after this header */
346 /* 8 bytes. packet FIXED for the next century! */
349 * short commands, packets without payload, plain p_header:
352 * P_BECOME_SYNC_TARGET
353 * P_BECOME_SYNC_SOURCE
358 * commands with out-of-struct payload:
359 * P_BITMAP (no additional fields)
360 * P_DATA, P_DATA_REPLY (see p_data)
361 * P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
364 /* these defines must not be changed without changing the protocol version */
365 #define DP_HARDBARRIER 1
367 #define DP_MAY_SET_IN_SYNC 4
370 struct p_header head
;
371 u64 sector
; /* 64 bits sector number */
372 u64 block_id
; /* to identify the request in protocol B&C */
378 * commands which share a struct:
380 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
381 * P_DISCARD_ACK (proto C, two-primaries conflict detection)
383 * P_DATA_REQUEST, P_RS_DATA_REQUEST
386 struct p_header head
;
395 struct p_header head
;
399 u32 pad
; /* to multiple of 8 Byte */
403 * commands with their own struct for additional fields:
412 struct p_header head
; /* 8 bytes */
417 /* should be more than enough for future enhancements
418 * for now, feature_flags and the reserverd array shall be zero.
424 /* 80 bytes, FIXED for the next century */
427 struct p_header head
;
428 u32 barrier
; /* barrier number _handle_ only */
429 u32 pad
; /* to multiple of 8 Byte */
432 struct p_barrier_ack
{
433 struct p_header head
;
439 struct p_header head
;
442 /* Since protocol version 88 and higher. */
446 struct p_rs_param_89
{
447 struct p_header head
;
449 /* protocol version 89: */
450 char verify_alg
[SHARED_SECRET_MAX
];
451 char csums_alg
[SHARED_SECRET_MAX
];
454 enum drbd_conn_flags
{
460 struct p_header head
;
468 /* Since protocol version 87 and higher. */
469 char integrity_alg
[0];
474 struct p_header head
;
475 u64 uuid
[UI_EXTENDED_SIZE
];
479 struct p_header head
;
484 struct p_header head
;
485 u64 d_size
; /* size of disk */
486 u64 u_size
; /* user requested size */
487 u64 c_size
; /* current exported size */
488 u32 max_segment_size
; /* Maximal size of a BIO */
489 u16 queue_order_type
; /* not yet implemented in DRBD*/
490 u16 dds_flags
; /* use enum dds_flags here. */
494 struct p_header head
;
499 struct p_header head
;
504 struct p_req_state_reply
{
505 struct p_header head
;
509 struct p_drbd06_param
{
520 struct p_header head
;
526 /* Valid values for the encoding field.
527 * Bump proto version when changing this. */
528 enum drbd_bitmap_code
{
529 /* RLE_VLI_Bytes = 0,
530 * and other bit variants had been defined during
531 * algorithm evaluation. */
535 struct p_compressed_bm
{
536 struct p_header head
;
537 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
538 * (encoding & 0x80): polarity (set/unset) of first runlength
539 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
540 * used to pad up to head.length bytes
547 struct p_delay_probe
{
548 struct p_header head
;
549 u32 seq_num
; /* sequence number to match the two probe packets */
550 u32 offset
; /* usecs the probe got sent after the reference time point */
554 struct list_head list
;
555 unsigned int seq_num
;
559 /* DCBP: Drbd Compressed Bitmap Packet ... */
560 static inline enum drbd_bitmap_code
561 DCBP_get_code(struct p_compressed_bm
*p
)
563 return (enum drbd_bitmap_code
)(p
->encoding
& 0x0f);
567 DCBP_set_code(struct p_compressed_bm
*p
, enum drbd_bitmap_code code
)
570 p
->encoding
= (p
->encoding
& ~0xf) | code
;
574 DCBP_get_start(struct p_compressed_bm
*p
)
576 return (p
->encoding
& 0x80) != 0;
580 DCBP_set_start(struct p_compressed_bm
*p
, int set
)
582 p
->encoding
= (p
->encoding
& ~0x80) | (set
? 0x80 : 0);
586 DCBP_get_pad_bits(struct p_compressed_bm
*p
)
588 return (p
->encoding
>> 4) & 0x7;
592 DCBP_set_pad_bits(struct p_compressed_bm
*p
, int n
)
595 p
->encoding
= (p
->encoding
& (~0x7 << 4)) | (n
<< 4);
598 /* one bitmap packet, including the p_header,
599 * should fit within one _architecture independend_ page.
600 * so we need to use the fixed size 4KiB page size
601 * most architechtures have used for a long time.
603 #define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header))
604 #define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
605 #define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm))
606 #if (PAGE_SIZE < 4096)
607 /* drbd_send_bitmap / receive_bitmap would break horribly */
608 #error "PAGE_SIZE too small"
612 struct p_header header
;
613 struct p_handshake handshake
;
615 struct p_block_ack block_ack
;
616 struct p_barrier barrier
;
617 struct p_barrier_ack barrier_ack
;
618 struct p_rs_param_89 rs_param_89
;
619 struct p_protocol protocol
;
620 struct p_sizes sizes
;
621 struct p_uuids uuids
;
622 struct p_state state
;
623 struct p_req_state req_state
;
624 struct p_req_state_reply req_state_reply
;
625 struct p_block_req block_req
;
628 /**********************************************************************/
629 enum drbd_thread_state
{
638 struct task_struct
*task
;
639 struct completion stop
;
640 enum drbd_thread_state t_state
;
641 int (*function
) (struct drbd_thread
*);
642 struct drbd_conf
*mdev
;
646 static inline enum drbd_thread_state
get_t_state(struct drbd_thread
*thi
)
648 /* THINK testing the t_state seems to be uncritical in all cases
649 * (but thread_{start,stop}), so we can read it *without* the lock.
658 * Having this as the first member of a struct provides sort of "inheritance".
659 * "derived" structs can be "drbd_queue_work()"ed.
660 * The callback should know and cast back to the descendant struct.
661 * drbd_request and drbd_epoch_entry are descendants of drbd_work.
664 typedef int (*drbd_work_cb
)(struct drbd_conf
*, struct drbd_work
*, int cancel
);
666 struct list_head list
;
670 struct drbd_tl_epoch
;
671 struct drbd_request
{
673 struct drbd_conf
*mdev
;
675 /* if local IO is not allowed, will be NULL.
676 * if local IO _is_ allowed, holds the locally submitted bio clone,
677 * or, after local IO completion, the ERR_PTR(error).
678 * see drbd_endio_pri(). */
679 struct bio
*private_bio
;
681 struct hlist_node colision
;
684 unsigned int epoch
; /* barrier_nr */
686 /* barrier_nr: used to check on "completion" whether this req was in
687 * the current epoch, and we therefore have to close it,
688 * starting a new epoch...
691 /* up to here, the struct layout is identical to drbd_epoch_entry;
692 * we might be able to use that to our advantage... */
694 struct list_head tl_requests
; /* ring list in the transfer log */
695 struct bio
*master_bio
; /* master bio pointer */
696 unsigned long rq_state
; /* see comments above _req_mod() */
698 unsigned long start_time
;
701 struct drbd_tl_epoch
{
703 struct list_head requests
; /* requests before */
704 struct drbd_tl_epoch
*next
; /* pointer to the next barrier */
705 unsigned int br_number
; /* the barriers identifier. */
706 int n_req
; /* number of requests attached before this barrier */
711 /* These Tl_epoch_entries may be in one of 6 lists:
712 active_ee .. data packet being written
713 sync_ee .. syncer block being written
714 done_ee .. block written, need to send P_WRITE_ACK
715 read_ee .. [RS]P_DATA_REQUEST being read
719 struct list_head list
;
720 unsigned int barrier_nr
;
721 atomic_t epoch_size
; /* increased on every request added. */
722 atomic_t active
; /* increased on every req. added, and dec on every finished. */
726 /* drbd_epoch flag bits */
728 DE_BARRIER_IN_NEXT_EPOCH_ISSUED
,
729 DE_BARRIER_IN_NEXT_EPOCH_DONE
,
730 DE_CONTAINS_A_BARRIER
,
731 DE_HAVE_BARRIER_NUMBER
,
740 EV_CLEANUP
= 32, /* used as flag */
743 struct drbd_wq_barrier
{
745 struct completion done
;
753 struct drbd_epoch_entry
{
755 struct hlist_node colision
;
756 struct drbd_epoch
*epoch
;
757 struct drbd_conf
*mdev
;
759 atomic_t pending_bios
;
761 /* see comments on ee flag bits below */
768 * While corresponding bios are in flight, the only modification will be
769 * set_bit WAS_ERROR, which has to be atomic.
770 * If no bios are in flight yet, or all have been completed,
771 * non-atomic modification to ee->flags is ok.
774 __EE_CALL_AL_COMPLETE_IO
,
775 __EE_MAY_SET_IN_SYNC
,
777 /* This epoch entry closes an epoch using a barrier.
778 * On sucessful completion, the epoch is released,
779 * and the P_BARRIER_ACK send. */
782 /* In case a barrier failed,
783 * we need to resubmit without the barrier flag. */
786 /* we may have several bios per epoch entry.
787 * if any of those fail, we set this flag atomically
788 * from the endio callback */
791 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
792 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
793 #define EE_IS_BARRIER (1<<__EE_IS_BARRIER)
794 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
795 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
797 /* global flag bits */
799 CREATE_BARRIER
, /* next P_DATA is preceeded by a P_BARRIER */
800 SIGNAL_ASENDER
, /* whether asender wants to be interrupted */
801 SEND_PING
, /* whether asender should send a ping asap */
803 STOP_SYNC_TIMER
, /* tell timer to cancel itself */
804 UNPLUG_QUEUED
, /* only relevant with kernel 2.4 */
805 UNPLUG_REMOTE
, /* sending a "UnplugRemote" could help */
806 MD_DIRTY
, /* current uuids and flags not yet on disk */
807 DISCARD_CONCURRENT
, /* Set on one node, cleared on the peer! */
808 USE_DEGR_WFC_T
, /* degr-wfc-timeout instead of wfc-timeout. */
809 CLUSTER_ST_CHANGE
, /* Cluster wide state change going on... */
812 CRASHED_PRIMARY
, /* This node was a crashed primary.
813 * Gets cleared when the state.conn
814 * goes into C_CONNECTED state. */
815 WRITE_BM_AFTER_RESYNC
, /* A kmalloc() during resync failed */
816 NO_BARRIER_SUPP
, /* underlying block device doesn't implement barriers */
819 MD_NO_BARRIER
, /* meta data device does not support barriers,
821 SUSPEND_IO
, /* suspend application io */
822 BITMAP_IO
, /* suspend application io;
823 once no more io in flight, start bitmap io */
824 BITMAP_IO_QUEUED
, /* Started bitmap IO */
825 RESYNC_AFTER_NEG
, /* Resync after online grow after the attach&negotiate finished. */
826 NET_CONGESTED
, /* The data socket is congested */
828 CONFIG_PENDING
, /* serialization of (re)configuration requests.
829 * if set, also prevents the device from dying */
830 DEVICE_DYING
, /* device became unconfigured,
831 * but worker thread is still handling the cleanup.
832 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
833 * while this is set. */
834 RESIZE_PENDING
, /* Size change detected locally, waiting for the response from
835 * the peer, if it changed there as well. */
836 CONN_DRY_RUN
, /* Expect disconnect after resync handshake. */
837 GOT_PING_ACK
, /* set when we receive a ping_ack packet, misc wait gets woken */
840 struct drbd_bitmap
; /* opaque for drbd_conf */
842 /* TODO sort members for performance
843 * MAYBE group them further */
845 /* THINK maybe we actually want to use the default "event/%s" worker threads
846 * or similar in linux 2.6, which uses per cpu data and threads.
848 * To be general, this might need a spin_lock member.
849 * For now, please use the mdev->req_lock to protect list_head,
850 * see drbd_queue_work below.
852 struct drbd_work_queue
{
854 struct semaphore s
; /* producers up it, worker down()s it */
855 spinlock_t q_lock
; /* to protect the list. */
859 struct drbd_work_queue work
;
861 struct socket
*socket
;
862 /* this way we get our
863 * send/receive buffers off the stack */
864 union p_polymorph sbuf
;
865 union p_polymorph rbuf
;
869 u64 md_offset
; /* sector offset to 'super' block */
871 u64 la_size_sect
; /* last agreed size, unit sectors */
877 s32 al_offset
; /* signed relative sector offset to al area */
878 s32 bm_offset
; /* signed relative sector offset to bitmap */
880 /* u32 al_nr_extents; important for restoring the AL
881 * is stored into sync_conf.al_extents, which in turn
882 * gets applied to act_log->nr_elements
886 /* for sync_conf and other types... */
887 #define NL_PACKET(name, number, fields) struct name { fields };
888 #define NL_INTEGER(pn,pr,member) int member;
889 #define NL_INT64(pn,pr,member) __u64 member;
890 #define NL_BIT(pn,pr,member) unsigned member:1;
891 #define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
892 #include "linux/drbd_nl.h"
894 struct drbd_backing_dev
{
895 struct block_device
*backing_bdev
;
896 struct block_device
*md_bdev
;
897 struct file
*lo_file
;
898 struct file
*md_file
;
900 struct disk_conf dc
; /* The user provided config... */
901 sector_t known_size
; /* last known size of that backing device */
905 struct drbd_conf
*mdev
;
906 struct completion event
;
913 int (*io_fn
)(struct drbd_conf
*mdev
);
914 void (*done
)(struct drbd_conf
*mdev
, int rv
);
917 enum write_ordering_e
{
925 /* things that are stored as / read from meta data on disk */
928 /* configured by drbdsetup */
929 struct net_conf
*net_conf
; /* protected by get_net_conf() and put_net_conf() */
930 struct syncer_conf sync_conf
;
931 struct drbd_backing_dev
*ldev
__protected_by(local
);
933 sector_t p_size
; /* partner's disk size */
934 struct request_queue
*rq_queue
;
935 struct block_device
*this_bdev
;
936 struct gendisk
*vdisk
;
938 struct drbd_socket data
; /* data/barrier/cstate/parameter packets */
939 struct drbd_socket meta
; /* ping/ack (metadata) packets */
940 int agreed_pro_version
; /* actually used protocol version */
941 unsigned long last_received
; /* in jiffies, either socket */
942 unsigned int ko_count
;
943 struct drbd_work resync_work
,
947 struct timer_list resync_timer
;
948 struct timer_list md_sync_timer
;
949 struct timer_list delay_probe_timer
;
951 /* Used after attach while negotiating new disk state. */
952 union drbd_state new_state_tmp
;
954 union drbd_state state
;
955 wait_queue_head_t misc_wait
;
956 wait_queue_head_t state_wait
; /* upon each state change. */
957 unsigned int send_cnt
;
958 unsigned int recv_cnt
;
959 unsigned int read_cnt
;
960 unsigned int writ_cnt
;
961 unsigned int al_writ_cnt
;
962 unsigned int bm_writ_cnt
;
963 atomic_t ap_bio_cnt
; /* Requests we need to complete */
964 atomic_t ap_pending_cnt
; /* AP data packets on the wire, ack expected */
965 atomic_t rs_pending_cnt
; /* RS request/data packets on the wire */
966 atomic_t unacked_cnt
; /* Need to send replys for */
967 atomic_t local_cnt
; /* Waiting for local completion */
968 atomic_t net_cnt
; /* Users of net_conf */
970 struct drbd_tl_epoch
*unused_spare_tle
; /* for pre-allocation */
971 struct drbd_tl_epoch
*newest_tle
;
972 struct drbd_tl_epoch
*oldest_tle
;
973 struct list_head out_of_sequence_requests
;
974 struct hlist_head
*tl_hash
;
975 unsigned int tl_hash_s
;
977 /* blocks to sync in this run [unit BM_BLOCK_SIZE] */
978 unsigned long rs_total
;
979 /* number of sync IOs that failed in this run */
980 unsigned long rs_failed
;
981 /* Syncer's start time [unit jiffies] */
982 unsigned long rs_start
;
983 /* cumulated time in PausedSyncX state [unit jiffies] */
984 unsigned long rs_paused
;
985 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
986 unsigned long rs_mark_left
;
987 /* marks's time [unit jiffies] */
988 unsigned long rs_mark_time
;
989 /* skipped because csum was equeal [unit BM_BLOCK_SIZE] */
990 unsigned long rs_same_csum
;
992 /* where does the admin want us to start? (sector) */
993 sector_t ov_start_sector
;
994 /* where are we now? (sector) */
995 sector_t ov_position
;
996 /* Start sector of out of sync range (to merge printk reporting). */
997 sector_t ov_last_oos_start
;
998 /* size of out-of-sync range in sectors. */
999 sector_t ov_last_oos_size
;
1000 unsigned long ov_left
; /* in bits */
1001 struct crypto_hash
*csums_tfm
;
1002 struct crypto_hash
*verify_tfm
;
1004 struct drbd_thread receiver
;
1005 struct drbd_thread worker
;
1006 struct drbd_thread asender
;
1007 struct drbd_bitmap
*bitmap
;
1008 unsigned long bm_resync_fo
; /* bit offset for drbd_bm_find_next */
1010 /* Used to track operations of resync... */
1011 struct lru_cache
*resync
;
1012 /* Number of locked elements in resync LRU */
1013 unsigned int resync_locked
;
1014 /* resync extent number waiting for application requests */
1015 unsigned int resync_wenr
;
1019 struct drbd_epoch
*current_epoch
;
1020 spinlock_t epoch_lock
;
1021 unsigned int epochs
;
1022 enum write_ordering_e write_ordering
;
1023 struct list_head active_ee
; /* IO in progress */
1024 struct list_head sync_ee
; /* IO in progress */
1025 struct list_head done_ee
; /* send ack */
1026 struct list_head read_ee
; /* IO in progress */
1027 struct list_head net_ee
; /* zero-copy network send in progress */
1028 struct hlist_head
*ee_hash
; /* is proteced by req_lock! */
1029 unsigned int ee_hash_s
;
1031 /* this one is protected by ee_lock, single thread */
1032 struct drbd_epoch_entry
*last_write_w_barrier
;
1034 int next_barrier_nr
;
1035 struct hlist_head
*app_reads_hash
; /* is proteced by req_lock */
1036 struct list_head resync_reads
;
1038 wait_queue_head_t ee_wait
;
1039 struct page
*md_io_page
; /* one page buffer for md_io */
1040 struct page
*md_io_tmpp
; /* for logical_block_size != 512 */
1041 struct mutex md_io_mutex
; /* protects the md_io_buffer */
1043 wait_queue_head_t al_wait
;
1044 struct lru_cache
*act_log
; /* activity log */
1045 unsigned int al_tr_number
;
1047 int al_tr_pos
; /* position of the next transaction in the journal */
1048 struct crypto_hash
*cram_hmac_tfm
;
1049 struct crypto_hash
*integrity_w_tfm
; /* to be used by the worker thread */
1050 struct crypto_hash
*integrity_r_tfm
; /* to be used by the receiver thread */
1054 wait_queue_head_t seq_wait
;
1055 atomic_t packet_seq
;
1056 unsigned int peer_seq
;
1057 spinlock_t peer_seq_lock
;
1059 unsigned long comm_bm_set
; /* communicated number of set bits. */
1060 cpumask_var_t cpu_mask
;
1061 struct bm_io_work bm_io_work
;
1062 u64 ed_uuid
; /* UUID of the exposed data */
1063 struct mutex state_mutex
;
1064 char congestion_reason
; /* Why we where congested... */
1065 struct list_head delay_probes
; /* protected by peer_seq_lock */
1066 int data_delay
; /* Delay of packets on the data-sock behind meta-sock */
1067 unsigned int delay_seq
; /* To generate sequence numbers of delay probes */
1068 struct timeval dps_time
; /* delay-probes-start-time */
1069 unsigned int dp_volume_last
; /* send_cnt of last delay probe */
1070 int c_sync_rate
; /* current resync rate after delay_probe magic */
1073 static inline struct drbd_conf
*minor_to_mdev(unsigned int minor
)
1075 struct drbd_conf
*mdev
;
1077 mdev
= minor
< minor_count
? minor_table
[minor
] : NULL
;
1082 static inline unsigned int mdev_to_minor(struct drbd_conf
*mdev
)
1087 /* returns 1 if it was successfull,
1088 * returns 0 if there was no data socket.
1089 * so wherever you are going to use the data.socket, e.g. do
1090 * if (!drbd_get_data_sock(mdev))
1093 * drbd_put_data_sock(mdev);
1095 static inline int drbd_get_data_sock(struct drbd_conf
*mdev
)
1097 mutex_lock(&mdev
->data
.mutex
);
1098 /* drbd_disconnect() could have called drbd_free_sock()
1099 * while we were waiting in down()... */
1100 if (unlikely(mdev
->data
.socket
== NULL
)) {
1101 mutex_unlock(&mdev
->data
.mutex
);
1107 static inline void drbd_put_data_sock(struct drbd_conf
*mdev
)
1109 mutex_unlock(&mdev
->data
.mutex
);
1113 * function declarations
1114 *************************/
1118 enum chg_state_flags
{
1121 CS_WAIT_COMPLETE
= 4,
1123 CS_ORDERED
= CS_WAIT_COMPLETE
+ CS_SERIALIZE
,
1128 DDSF_NO_RESYNC
= 2, /* Do not run a resync for the new space */
1131 extern void drbd_init_set_defaults(struct drbd_conf
*mdev
);
1132 extern int drbd_change_state(struct drbd_conf
*mdev
, enum chg_state_flags f
,
1133 union drbd_state mask
, union drbd_state val
);
1134 extern void drbd_force_state(struct drbd_conf
*, union drbd_state
,
1136 extern int _drbd_request_state(struct drbd_conf
*, union drbd_state
,
1137 union drbd_state
, enum chg_state_flags
);
1138 extern int __drbd_set_state(struct drbd_conf
*, union drbd_state
,
1139 enum chg_state_flags
, struct completion
*done
);
1140 extern void print_st_err(struct drbd_conf
*, union drbd_state
,
1141 union drbd_state
, int);
1142 extern int drbd_thread_start(struct drbd_thread
*thi
);
1143 extern void _drbd_thread_stop(struct drbd_thread
*thi
, int restart
, int wait
);
1145 extern void drbd_thread_current_set_cpu(struct drbd_conf
*mdev
);
1146 extern void drbd_calc_cpu_mask(struct drbd_conf
*mdev
);
1148 #define drbd_thread_current_set_cpu(A) ({})
1149 #define drbd_calc_cpu_mask(A) ({})
1151 extern void drbd_free_resources(struct drbd_conf
*mdev
);
1152 extern void tl_release(struct drbd_conf
*mdev
, unsigned int barrier_nr
,
1153 unsigned int set_size
);
1154 extern void tl_clear(struct drbd_conf
*mdev
);
1155 extern void _tl_add_barrier(struct drbd_conf
*, struct drbd_tl_epoch
*);
1156 extern void drbd_free_sock(struct drbd_conf
*mdev
);
1157 extern int drbd_send(struct drbd_conf
*mdev
, struct socket
*sock
,
1158 void *buf
, size_t size
, unsigned msg_flags
);
1159 extern int drbd_send_protocol(struct drbd_conf
*mdev
);
1160 extern int drbd_send_uuids(struct drbd_conf
*mdev
);
1161 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf
*mdev
);
1162 extern int drbd_send_sync_uuid(struct drbd_conf
*mdev
, u64 val
);
1163 extern int drbd_send_sizes(struct drbd_conf
*mdev
, int trigger_reply
, enum dds_flags flags
);
1164 extern int _drbd_send_state(struct drbd_conf
*mdev
);
1165 extern int drbd_send_state(struct drbd_conf
*mdev
);
1166 extern int _drbd_send_cmd(struct drbd_conf
*mdev
, struct socket
*sock
,
1167 enum drbd_packets cmd
, struct p_header
*h
,
1168 size_t size
, unsigned msg_flags
);
1169 #define USE_DATA_SOCKET 1
1170 #define USE_META_SOCKET 0
1171 extern int drbd_send_cmd(struct drbd_conf
*mdev
, int use_data_socket
,
1172 enum drbd_packets cmd
, struct p_header
*h
,
1174 extern int drbd_send_cmd2(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1175 char *data
, size_t size
);
1176 extern int drbd_send_sync_param(struct drbd_conf
*mdev
, struct syncer_conf
*sc
);
1177 extern int drbd_send_b_ack(struct drbd_conf
*mdev
, u32 barrier_nr
,
1179 extern int drbd_send_ack(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1180 struct drbd_epoch_entry
*e
);
1181 extern int drbd_send_ack_rp(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1182 struct p_block_req
*rp
);
1183 extern int drbd_send_ack_dp(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1185 extern int drbd_send_ack_ex(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1186 sector_t sector
, int blksize
, u64 block_id
);
1187 extern int drbd_send_block(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1188 struct drbd_epoch_entry
*e
);
1189 extern int drbd_send_dblock(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1190 extern int _drbd_send_barrier(struct drbd_conf
*mdev
,
1191 struct drbd_tl_epoch
*barrier
);
1192 extern int drbd_send_drequest(struct drbd_conf
*mdev
, int cmd
,
1193 sector_t sector
, int size
, u64 block_id
);
1194 extern int drbd_send_drequest_csum(struct drbd_conf
*mdev
,
1195 sector_t sector
,int size
,
1196 void *digest
, int digest_size
,
1197 enum drbd_packets cmd
);
1198 extern int drbd_send_ov_request(struct drbd_conf
*mdev
,sector_t sector
,int size
);
1200 extern int drbd_send_bitmap(struct drbd_conf
*mdev
);
1201 extern int _drbd_send_bitmap(struct drbd_conf
*mdev
);
1202 extern int drbd_send_sr_reply(struct drbd_conf
*mdev
, int retcode
);
1203 extern void drbd_free_bc(struct drbd_backing_dev
*ldev
);
1204 extern void drbd_mdev_cleanup(struct drbd_conf
*mdev
);
1206 /* drbd_meta-data.c (still in drbd_main.c) */
1207 extern void drbd_md_sync(struct drbd_conf
*mdev
);
1208 extern int drbd_md_read(struct drbd_conf
*mdev
, struct drbd_backing_dev
*bdev
);
1209 /* maybe define them below as inline? */
1210 extern void drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1211 extern void _drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1212 extern void drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1213 extern void _drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1214 extern void drbd_uuid_set_bm(struct drbd_conf
*mdev
, u64 val
) __must_hold(local
);
1215 extern void drbd_md_set_flag(struct drbd_conf
*mdev
, int flags
) __must_hold(local
);
1216 extern void drbd_md_clear_flag(struct drbd_conf
*mdev
, int flags
)__must_hold(local
);
1217 extern int drbd_md_test_flag(struct drbd_backing_dev
*, int);
1218 extern void drbd_md_mark_dirty(struct drbd_conf
*mdev
);
1219 extern void drbd_queue_bitmap_io(struct drbd_conf
*mdev
,
1220 int (*io_fn
)(struct drbd_conf
*),
1221 void (*done
)(struct drbd_conf
*, int),
1223 extern int drbd_bmio_set_n_write(struct drbd_conf
*mdev
);
1224 extern int drbd_bmio_clear_n_write(struct drbd_conf
*mdev
);
1225 extern int drbd_bitmap_io(struct drbd_conf
*mdev
, int (*io_fn
)(struct drbd_conf
*), char *why
);
1229 We reserve a 128MB Block (4k aligned)
1230 * either at the end of the backing device
1231 * or on a separate meta data device. */
1233 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1234 /* The following numbers are sectors */
1235 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1236 #define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */
1237 /* Allows up to about 3.8TB */
1238 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1240 /* Since the smalles IO unit is usually 512 byte */
1241 #define MD_SECTOR_SHIFT 9
1242 #define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT)
1245 #define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1246 #define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */
1247 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1249 #if BITS_PER_LONG == 32
1251 #define cpu_to_lel(A) cpu_to_le32(A)
1252 #define lel_to_cpu(A) le32_to_cpu(A)
1253 #elif BITS_PER_LONG == 64
1255 #define cpu_to_lel(A) cpu_to_le64(A)
1256 #define lel_to_cpu(A) le64_to_cpu(A)
1258 #error "LN2 of BITS_PER_LONG unknown!"
1262 /* 16MB sized 'bitmap extent' to track syncer usage */
1264 int rs_left
; /* number of bits set (out of sync) in this extent. */
1265 int rs_failed
; /* number of failed resync requests in this extent. */
1266 unsigned long flags
;
1267 struct lc_element lce
;
1270 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1271 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1275 * We need to store one bit for a block.
1276 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1277 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1278 * Bit 1 ==> local node thinks this block needs to be synced.
1281 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1282 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1283 /* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1284 * per sector of on disk bitmap */
1285 #define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */
1286 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1288 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1289 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1292 /* thus many _storage_ sectors are described by one bit */
1293 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1294 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1295 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1297 /* bit to represented kilo byte conversion */
1298 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1300 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1301 * _storage_ sector is located in */
1302 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1304 /* how much _storage_ sectors we have per bitmap sector */
1305 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1306 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1308 /* in one sector of the bitmap, we have this many activity_log extents. */
1309 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1310 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1312 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1313 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1315 /* the extent in "PER_EXTENT" below is an activity log extent
1316 * we need that many (long words/bytes) to store the bitmap
1317 * of one AL_EXTENT_SIZE chunk of storage.
1318 * we can store the bitmap for that many AL_EXTENTS within
1319 * one sector of the _on_disk_ bitmap:
1320 * bit 0 bit 37 bit 38 bit (512*8)-1
1321 * ...|........|........|.. // ..|........|
1322 * sect. 0 `296 `304 ^(512*8*8)-1
1324 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1325 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1326 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1329 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1330 #define DRBD_MAX_SECTORS_BM \
1331 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1332 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1333 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1334 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1335 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1336 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1337 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1339 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1340 /* 16 TB in units of sectors */
1341 #if BITS_PER_LONG == 32
1342 /* adjust by one page worth of bitmap,
1343 * so we won't wrap around in drbd_bm_find_next_bit.
1344 * you should use 64bit OS for that much storage, anyways. */
1345 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1347 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0x1LU << 32)
1351 /* Sector shift value for the "hash" functions of tl_hash and ee_hash tables.
1352 * With a value of 6 all IO in one 32K block make it to the same slot of the
1355 #define DRBD_MAX_SEGMENT_SIZE (1U<<(9+HT_SHIFT))
1357 /* Number of elements in the app_reads_hash */
1358 #define APP_R_HSIZE 15
1360 extern int drbd_bm_init(struct drbd_conf
*mdev
);
1361 extern int drbd_bm_resize(struct drbd_conf
*mdev
, sector_t sectors
, int set_new_bits
);
1362 extern void drbd_bm_cleanup(struct drbd_conf
*mdev
);
1363 extern void drbd_bm_set_all(struct drbd_conf
*mdev
);
1364 extern void drbd_bm_clear_all(struct drbd_conf
*mdev
);
1365 extern int drbd_bm_set_bits(
1366 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1367 extern int drbd_bm_clear_bits(
1368 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1369 /* bm_set_bits variant for use while holding drbd_bm_lock */
1370 extern void _drbd_bm_set_bits(struct drbd_conf
*mdev
,
1371 const unsigned long s
, const unsigned long e
);
1372 extern int drbd_bm_test_bit(struct drbd_conf
*mdev
, unsigned long bitnr
);
1373 extern int drbd_bm_e_weight(struct drbd_conf
*mdev
, unsigned long enr
);
1374 extern int drbd_bm_write_sect(struct drbd_conf
*mdev
, unsigned long enr
) __must_hold(local
);
1375 extern int drbd_bm_read(struct drbd_conf
*mdev
) __must_hold(local
);
1376 extern int drbd_bm_write(struct drbd_conf
*mdev
) __must_hold(local
);
1377 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf
*mdev
,
1378 unsigned long al_enr
);
1379 extern size_t drbd_bm_words(struct drbd_conf
*mdev
);
1380 extern unsigned long drbd_bm_bits(struct drbd_conf
*mdev
);
1381 extern sector_t
drbd_bm_capacity(struct drbd_conf
*mdev
);
1382 extern unsigned long drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1383 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1384 extern unsigned long _drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1385 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1386 extern unsigned long drbd_bm_total_weight(struct drbd_conf
*mdev
);
1387 extern int drbd_bm_rs_done(struct drbd_conf
*mdev
);
1388 /* for receive_bitmap */
1389 extern void drbd_bm_merge_lel(struct drbd_conf
*mdev
, size_t offset
,
1390 size_t number
, unsigned long *buffer
);
1391 /* for _drbd_send_bitmap and drbd_bm_write_sect */
1392 extern void drbd_bm_get_lel(struct drbd_conf
*mdev
, size_t offset
,
1393 size_t number
, unsigned long *buffer
);
1395 extern void drbd_bm_lock(struct drbd_conf
*mdev
, char *why
);
1396 extern void drbd_bm_unlock(struct drbd_conf
*mdev
);
1398 extern int drbd_bm_count_bits(struct drbd_conf
*mdev
, const unsigned long s
, const unsigned long e
);
1401 extern struct kmem_cache
*drbd_request_cache
;
1402 extern struct kmem_cache
*drbd_ee_cache
; /* epoch entries */
1403 extern struct kmem_cache
*drbd_bm_ext_cache
; /* bitmap extents */
1404 extern struct kmem_cache
*drbd_al_ext_cache
; /* activity log extents */
1405 extern mempool_t
*drbd_request_mempool
;
1406 extern mempool_t
*drbd_ee_mempool
;
1408 extern struct page
*drbd_pp_pool
; /* drbd's page pool */
1409 extern spinlock_t drbd_pp_lock
;
1410 extern int drbd_pp_vacant
;
1411 extern wait_queue_head_t drbd_pp_wait
;
1413 extern rwlock_t global_state_lock
;
1415 extern struct drbd_conf
*drbd_new_device(unsigned int minor
);
1416 extern void drbd_free_mdev(struct drbd_conf
*mdev
);
1418 extern int proc_details
;
1421 extern int drbd_make_request_26(struct request_queue
*q
, struct bio
*bio
);
1422 extern int drbd_read_remote(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1423 extern int drbd_merge_bvec(struct request_queue
*q
, struct bvec_merge_data
*bvm
, struct bio_vec
*bvec
);
1424 extern int is_valid_ar_handle(struct drbd_request
*, sector_t
);
1428 extern void drbd_suspend_io(struct drbd_conf
*mdev
);
1429 extern void drbd_resume_io(struct drbd_conf
*mdev
);
1430 extern char *ppsize(char *buf
, unsigned long long size
);
1431 extern sector_t
drbd_new_dev_size(struct drbd_conf
*, struct drbd_backing_dev
*, int);
1432 enum determine_dev_size
{ dev_size_error
= -1, unchanged
= 0, shrunk
= 1, grew
= 2 };
1433 extern enum determine_dev_size
drbd_determin_dev_size(struct drbd_conf
*, enum dds_flags
) __must_hold(local
);
1434 extern void resync_after_online_grow(struct drbd_conf
*);
1435 extern void drbd_setup_queue_param(struct drbd_conf
*mdev
, unsigned int) __must_hold(local
);
1436 extern int drbd_set_role(struct drbd_conf
*mdev
, enum drbd_role new_role
,
1438 enum drbd_disk_state
drbd_try_outdate_peer(struct drbd_conf
*mdev
);
1439 extern int drbd_khelper(struct drbd_conf
*mdev
, char *cmd
);
1442 extern int drbd_worker(struct drbd_thread
*thi
);
1443 extern int drbd_alter_sa(struct drbd_conf
*mdev
, int na
);
1444 extern void drbd_start_resync(struct drbd_conf
*mdev
, enum drbd_conns side
);
1445 extern void resume_next_sg(struct drbd_conf
*mdev
);
1446 extern void suspend_other_sg(struct drbd_conf
*mdev
);
1447 extern int drbd_resync_finished(struct drbd_conf
*mdev
);
1448 /* maybe rather drbd_main.c ? */
1449 extern int drbd_md_sync_page_io(struct drbd_conf
*mdev
,
1450 struct drbd_backing_dev
*bdev
, sector_t sector
, int rw
);
1451 extern void drbd_ov_oos_found(struct drbd_conf
*, sector_t
, int);
1453 static inline void ov_oos_print(struct drbd_conf
*mdev
)
1455 if (mdev
->ov_last_oos_size
) {
1456 dev_err(DEV
, "Out of sync: start=%llu, size=%lu (sectors)\n",
1457 (unsigned long long)mdev
->ov_last_oos_start
,
1458 (unsigned long)mdev
->ov_last_oos_size
);
1460 mdev
->ov_last_oos_size
=0;
1464 extern void drbd_csum_bio(struct drbd_conf
*, struct crypto_hash
*, struct bio
*, void *);
1465 extern void drbd_csum_ee(struct drbd_conf
*, struct crypto_hash
*, struct drbd_epoch_entry
*, void *);
1466 /* worker callbacks */
1467 extern int w_req_cancel_conflict(struct drbd_conf
*, struct drbd_work
*, int);
1468 extern int w_read_retry_remote(struct drbd_conf
*, struct drbd_work
*, int);
1469 extern int w_e_end_data_req(struct drbd_conf
*, struct drbd_work
*, int);
1470 extern int w_e_end_rsdata_req(struct drbd_conf
*, struct drbd_work
*, int);
1471 extern int w_e_end_csum_rs_req(struct drbd_conf
*, struct drbd_work
*, int);
1472 extern int w_e_end_ov_reply(struct drbd_conf
*, struct drbd_work
*, int);
1473 extern int w_e_end_ov_req(struct drbd_conf
*, struct drbd_work
*, int);
1474 extern int w_ov_finished(struct drbd_conf
*, struct drbd_work
*, int);
1475 extern int w_resync_inactive(struct drbd_conf
*, struct drbd_work
*, int);
1476 extern int w_resume_next_sg(struct drbd_conf
*, struct drbd_work
*, int);
1477 extern int w_send_write_hint(struct drbd_conf
*, struct drbd_work
*, int);
1478 extern int w_make_resync_request(struct drbd_conf
*, struct drbd_work
*, int);
1479 extern int w_send_dblock(struct drbd_conf
*, struct drbd_work
*, int);
1480 extern int w_send_barrier(struct drbd_conf
*, struct drbd_work
*, int);
1481 extern int w_send_read_req(struct drbd_conf
*, struct drbd_work
*, int);
1482 extern int w_prev_work_done(struct drbd_conf
*, struct drbd_work
*, int);
1483 extern int w_e_reissue(struct drbd_conf
*, struct drbd_work
*, int);
1485 extern void resync_timer_fn(unsigned long data
);
1487 /* drbd_receiver.c */
1488 extern int drbd_submit_ee(struct drbd_conf
*mdev
, struct drbd_epoch_entry
*e
,
1489 const unsigned rw
, const int fault_type
);
1490 extern int drbd_release_ee(struct drbd_conf
*mdev
, struct list_head
*list
);
1491 extern struct drbd_epoch_entry
*drbd_alloc_ee(struct drbd_conf
*mdev
,
1494 unsigned int data_size
,
1495 gfp_t gfp_mask
) __must_hold(local
);
1496 extern void drbd_free_ee(struct drbd_conf
*mdev
, struct drbd_epoch_entry
*e
);
1497 extern void drbd_wait_ee_list_empty(struct drbd_conf
*mdev
,
1498 struct list_head
*head
);
1499 extern void _drbd_wait_ee_list_empty(struct drbd_conf
*mdev
,
1500 struct list_head
*head
);
1501 extern void drbd_set_recv_tcq(struct drbd_conf
*mdev
, int tcq_enabled
);
1502 extern void _drbd_clear_done_ee(struct drbd_conf
*mdev
, struct list_head
*to_be_freed
);
1503 extern void drbd_flush_workqueue(struct drbd_conf
*mdev
);
1505 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1506 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1507 static inline int drbd_setsockopt(struct socket
*sock
, int level
, int optname
,
1508 char __user
*optval
, int optlen
)
1511 if (level
== SOL_SOCKET
)
1512 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
1514 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1519 static inline void drbd_tcp_cork(struct socket
*sock
)
1522 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1523 (char __user
*)&val
, sizeof(val
));
1526 static inline void drbd_tcp_uncork(struct socket
*sock
)
1529 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1530 (char __user
*)&val
, sizeof(val
));
1533 static inline void drbd_tcp_nodelay(struct socket
*sock
)
1536 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_NODELAY
,
1537 (char __user
*)&val
, sizeof(val
));
1540 static inline void drbd_tcp_quickack(struct socket
*sock
)
1543 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_QUICKACK
,
1544 (char __user
*)&val
, sizeof(val
));
1547 void drbd_bump_write_ordering(struct drbd_conf
*mdev
, enum write_ordering_e wo
);
1550 extern struct proc_dir_entry
*drbd_proc
;
1551 extern const struct file_operations drbd_proc_fops
;
1552 extern const char *drbd_conn_str(enum drbd_conns s
);
1553 extern const char *drbd_role_str(enum drbd_role s
);
1556 extern void drbd_al_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1557 extern void drbd_al_complete_io(struct drbd_conf
*mdev
, sector_t sector
);
1558 extern void drbd_rs_complete_io(struct drbd_conf
*mdev
, sector_t sector
);
1559 extern int drbd_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1560 extern int drbd_try_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1561 extern void drbd_rs_cancel_all(struct drbd_conf
*mdev
);
1562 extern int drbd_rs_del_all(struct drbd_conf
*mdev
);
1563 extern void drbd_rs_failed_io(struct drbd_conf
*mdev
,
1564 sector_t sector
, int size
);
1565 extern int drbd_al_read_log(struct drbd_conf
*mdev
, struct drbd_backing_dev
*);
1566 extern void __drbd_set_in_sync(struct drbd_conf
*mdev
, sector_t sector
,
1567 int size
, const char *file
, const unsigned int line
);
1568 #define drbd_set_in_sync(mdev, sector, size) \
1569 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1570 extern void __drbd_set_out_of_sync(struct drbd_conf
*mdev
, sector_t sector
,
1571 int size
, const char *file
, const unsigned int line
);
1572 #define drbd_set_out_of_sync(mdev, sector, size) \
1573 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1574 extern void drbd_al_apply_to_bm(struct drbd_conf
*mdev
);
1575 extern void drbd_al_to_on_disk_bm(struct drbd_conf
*mdev
);
1576 extern void drbd_al_shrink(struct drbd_conf
*mdev
);
1581 void drbd_nl_cleanup(void);
1582 int __init
drbd_nl_init(void);
1583 void drbd_bcast_state(struct drbd_conf
*mdev
, union drbd_state
);
1584 void drbd_bcast_sync_progress(struct drbd_conf
*mdev
);
1585 void drbd_bcast_ee(struct drbd_conf
*mdev
,
1586 const char *reason
, const int dgs
,
1587 const char* seen_hash
, const char* calc_hash
,
1588 const struct drbd_epoch_entry
* e
);
1592 * DOC: DRBD State macros
1594 * These macros are used to express state changes in easily readable form.
1596 * The NS macros expand to a mask and a value, that can be bit ored onto the
1597 * current state as soon as the spinlock (req_lock) was taken.
1599 * The _NS macros are used for state functions that get called with the
1600 * spinlock. These macros expand directly to the new state value.
1602 * Besides the basic forms NS() and _NS() additional _?NS[23] are defined
1603 * to express state changes that affect more than one aspect of the state.
1605 * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY)
1606 * Means that the network connection was established and that the peer
1607 * is in secondary role.
1609 #define role_MASK R_MASK
1610 #define peer_MASK R_MASK
1611 #define disk_MASK D_MASK
1612 #define pdsk_MASK D_MASK
1613 #define conn_MASK C_MASK
1615 #define user_isp_MASK 1
1616 #define aftr_isp_MASK 1
1619 ({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \
1620 ({ union drbd_state val; val.i = 0; val.T = (S); val; })
1621 #define NS2(T1, S1, T2, S2) \
1622 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1623 mask.T2 = T2##_MASK; mask; }), \
1624 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1625 val.T2 = (S2); val; })
1626 #define NS3(T1, S1, T2, S2, T3, S3) \
1627 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1628 mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \
1629 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1630 val.T2 = (S2); val.T3 = (S3); val; })
1632 #define _NS(D, T, S) \
1633 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; })
1634 #define _NS2(D, T1, S1, T2, S2) \
1635 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1636 __ns.T2 = (S2); __ns; })
1637 #define _NS3(D, T1, S1, T2, S2, T3, S3) \
1638 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1639 __ns.T2 = (S2); __ns.T3 = (S3); __ns; })
1642 * inline helper functions
1643 *************************/
1645 /* see also page_chain_add and friends in drbd_receiver.c */
1646 static inline struct page
*page_chain_next(struct page
*page
)
1648 return (struct page
*)page_private(page
);
1650 #define page_chain_for_each(page) \
1651 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1652 page = page_chain_next(page))
1653 #define page_chain_for_each_safe(page, n) \
1654 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1656 static inline int drbd_bio_has_active_page(struct bio
*bio
)
1658 struct bio_vec
*bvec
;
1661 __bio_for_each_segment(bvec
, bio
, i
, 0) {
1662 if (page_count(bvec
->bv_page
) > 1)
1669 static inline int drbd_ee_has_active_page(struct drbd_epoch_entry
*e
)
1671 struct page
*page
= e
->pages
;
1672 page_chain_for_each(page
) {
1673 if (page_count(page
) > 1)
1680 static inline void drbd_state_lock(struct drbd_conf
*mdev
)
1682 wait_event(mdev
->misc_wait
,
1683 !test_and_set_bit(CLUSTER_ST_CHANGE
, &mdev
->flags
));
1686 static inline void drbd_state_unlock(struct drbd_conf
*mdev
)
1688 clear_bit(CLUSTER_ST_CHANGE
, &mdev
->flags
);
1689 wake_up(&mdev
->misc_wait
);
1692 static inline int _drbd_set_state(struct drbd_conf
*mdev
,
1693 union drbd_state ns
, enum chg_state_flags flags
,
1694 struct completion
*done
)
1698 read_lock(&global_state_lock
);
1699 rv
= __drbd_set_state(mdev
, ns
, flags
, done
);
1700 read_unlock(&global_state_lock
);
1706 * drbd_request_state() - Reqest a state change
1707 * @mdev: DRBD device.
1708 * @mask: mask of state bits to change.
1709 * @val: value of new state bits.
1711 * This is the most graceful way of requesting a state change. It is verbose
1712 * quite verbose in case the state change is not possible, and all those
1713 * state changes are globally serialized.
1715 static inline int drbd_request_state(struct drbd_conf
*mdev
,
1716 union drbd_state mask
,
1717 union drbd_state val
)
1719 return _drbd_request_state(mdev
, mask
, val
, CS_VERBOSE
+ CS_ORDERED
);
1722 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1723 static inline void __drbd_chk_io_error_(struct drbd_conf
*mdev
, int forcedetach
, const char *where
)
1725 switch (mdev
->ldev
->dc
.on_io_error
) {
1728 if (__ratelimit(&drbd_ratelimit_state
))
1729 dev_err(DEV
, "Local IO failed in %s."
1730 "Passing error on...\n", where
);
1733 /* NOTE fall through to detach case if forcedetach set */
1735 case EP_CALL_HELPER
:
1736 if (mdev
->state
.disk
> D_FAILED
) {
1737 _drbd_set_state(_NS(mdev
, disk
, D_FAILED
), CS_HARD
, NULL
);
1738 dev_err(DEV
, "Local IO failed in %s."
1739 "Detaching...\n", where
);
1746 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1747 * @mdev: DRBD device.
1748 * @error: Error code passed to the IO completion callback
1749 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1751 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1753 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1754 static inline void drbd_chk_io_error_(struct drbd_conf
*mdev
,
1755 int error
, int forcedetach
, const char *where
)
1758 unsigned long flags
;
1759 spin_lock_irqsave(&mdev
->req_lock
, flags
);
1760 __drbd_chk_io_error_(mdev
, forcedetach
, where
);
1761 spin_unlock_irqrestore(&mdev
->req_lock
, flags
);
1767 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1768 * @bdev: Meta data block device.
1770 * BTW, for internal meta data, this happens to be the maximum capacity
1771 * we could agree upon with our peer node.
1773 static inline sector_t
drbd_md_first_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
+ bdev
->md
.bm_offset
;
1779 case DRBD_MD_INDEX_FLEX_EXT
:
1781 return bdev
->md
.md_offset
;
1786 * drbd_md_last_sector() - Return the last sector number of the meta data area
1787 * @bdev: Meta data block device.
1789 static inline sector_t
drbd_md_last_sector(struct drbd_backing_dev
*bdev
)
1791 switch (bdev
->dc
.meta_dev_idx
) {
1792 case DRBD_MD_INDEX_INTERNAL
:
1793 case DRBD_MD_INDEX_FLEX_INT
:
1794 return bdev
->md
.md_offset
+ MD_AL_OFFSET
- 1;
1795 case DRBD_MD_INDEX_FLEX_EXT
:
1797 return bdev
->md
.md_offset
+ bdev
->md
.md_size_sect
;
1801 /* Returns the number of 512 byte sectors of the device */
1802 static inline sector_t
drbd_get_capacity(struct block_device
*bdev
)
1804 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1805 return bdev
? bdev
->bd_inode
->i_size
>> 9 : 0;
1809 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1810 * @bdev: Meta data block device.
1812 * returns the capacity we announce to out peer. we clip ourselves at the
1813 * various MAX_SECTORS, because if we don't, current implementation will
1814 * oops sooner or later
1816 static inline sector_t
drbd_get_max_capacity(struct drbd_backing_dev
*bdev
)
1819 switch (bdev
->dc
.meta_dev_idx
) {
1820 case DRBD_MD_INDEX_INTERNAL
:
1821 case DRBD_MD_INDEX_FLEX_INT
:
1822 s
= drbd_get_capacity(bdev
->backing_bdev
)
1823 ? min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1824 drbd_md_first_sector(bdev
))
1827 case DRBD_MD_INDEX_FLEX_EXT
:
1828 s
= min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1829 drbd_get_capacity(bdev
->backing_bdev
));
1830 /* clip at maximum size the meta device can support */
1831 s
= min_t(sector_t
, s
,
1832 BM_EXT_TO_SECT(bdev
->md
.md_size_sect
1833 - bdev
->md
.bm_offset
));
1836 s
= min_t(sector_t
, DRBD_MAX_SECTORS
,
1837 drbd_get_capacity(bdev
->backing_bdev
));
1843 * drbd_md_ss__() - Return the sector number of our meta data super block
1844 * @mdev: DRBD device.
1845 * @bdev: Meta data block device.
1847 static inline sector_t
drbd_md_ss__(struct drbd_conf
*mdev
,
1848 struct drbd_backing_dev
*bdev
)
1850 switch (bdev
->dc
.meta_dev_idx
) {
1851 default: /* external, some index */
1852 return MD_RESERVED_SECT
* bdev
->dc
.meta_dev_idx
;
1853 case DRBD_MD_INDEX_INTERNAL
:
1854 /* with drbd08, internal meta data is always "flexible" */
1855 case DRBD_MD_INDEX_FLEX_INT
:
1856 /* sizeof(struct md_on_disk_07) == 4k
1857 * position: last 4k aligned block of 4k size */
1858 if (!bdev
->backing_bdev
) {
1859 if (__ratelimit(&drbd_ratelimit_state
)) {
1860 dev_err(DEV
, "bdev->backing_bdev==NULL\n");
1865 return (drbd_get_capacity(bdev
->backing_bdev
) & ~7ULL)
1867 case DRBD_MD_INDEX_FLEX_EXT
:
1873 _drbd_queue_work(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1875 list_add_tail(&w
->list
, &q
->q
);
1880 drbd_queue_work_front(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1882 unsigned long flags
;
1883 spin_lock_irqsave(&q
->q_lock
, flags
);
1884 list_add(&w
->list
, &q
->q
);
1885 up(&q
->s
); /* within the spinlock,
1886 see comment near end of drbd_worker() */
1887 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1891 drbd_queue_work(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1893 unsigned long flags
;
1894 spin_lock_irqsave(&q
->q_lock
, flags
);
1895 list_add_tail(&w
->list
, &q
->q
);
1896 up(&q
->s
); /* within the spinlock,
1897 see comment near end of drbd_worker() */
1898 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1901 static inline void wake_asender(struct drbd_conf
*mdev
)
1903 if (test_bit(SIGNAL_ASENDER
, &mdev
->flags
))
1904 force_sig(DRBD_SIG
, mdev
->asender
.task
);
1907 static inline void request_ping(struct drbd_conf
*mdev
)
1909 set_bit(SEND_PING
, &mdev
->flags
);
1913 static inline int drbd_send_short_cmd(struct drbd_conf
*mdev
,
1914 enum drbd_packets cmd
)
1917 return drbd_send_cmd(mdev
, USE_DATA_SOCKET
, cmd
, &h
, sizeof(h
));
1920 static inline int drbd_send_ping(struct drbd_conf
*mdev
)
1923 return drbd_send_cmd(mdev
, USE_META_SOCKET
, P_PING
, &h
, sizeof(h
));
1926 static inline int drbd_send_ping_ack(struct drbd_conf
*mdev
)
1929 return drbd_send_cmd(mdev
, USE_META_SOCKET
, P_PING_ACK
, &h
, sizeof(h
));
1932 static inline void drbd_thread_stop(struct drbd_thread
*thi
)
1934 _drbd_thread_stop(thi
, FALSE
, TRUE
);
1937 static inline void drbd_thread_stop_nowait(struct drbd_thread
*thi
)
1939 _drbd_thread_stop(thi
, FALSE
, FALSE
);
1942 static inline void drbd_thread_restart_nowait(struct drbd_thread
*thi
)
1944 _drbd_thread_stop(thi
, TRUE
, FALSE
);
1947 /* counts how many answer packets packets we expect from our peer,
1948 * for either explicit application requests,
1949 * or implicit barrier packets as necessary.
1952 * _req_mod(req, queue_for_net_write or queue_for_net_read);
1953 * it is much easier and equally valid to count what we queue for the
1954 * worker, even before it actually was queued or send.
1955 * (drbd_make_request_common; recovery path on read io-error)
1957 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1958 * _req_mod(req, data_received)
1959 * [from receive_DataReply]
1960 * _req_mod(req, write_acked_by_peer or recv_acked_by_peer or neg_acked)
1961 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1962 * for some reason it is NOT decreased in got_NegAck,
1963 * but in the resulting cleanup code from report_params.
1964 * we should try to remember the reason for that...
1965 * _req_mod(req, send_failed or send_canceled)
1966 * _req_mod(req, connection_lost_while_pending)
1967 * [from tl_clear_barrier]
1969 static inline void inc_ap_pending(struct drbd_conf
*mdev
)
1971 atomic_inc(&mdev
->ap_pending_cnt
);
1974 #define ERR_IF_CNT_IS_NEGATIVE(which) \
1975 if (atomic_read(&mdev->which) < 0) \
1976 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
1977 __func__ , __LINE__ , \
1978 atomic_read(&mdev->which))
1980 #define dec_ap_pending(mdev) do { \
1981 typecheck(struct drbd_conf *, mdev); \
1982 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \
1983 wake_up(&mdev->misc_wait); \
1984 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
1986 /* counts how many resync-related answers we still expect from the peer
1988 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1989 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK whith ID_SYNCER)
1990 * (or P_NEG_ACK with ID_SYNCER)
1992 static inline void inc_rs_pending(struct drbd_conf
*mdev
)
1994 atomic_inc(&mdev
->rs_pending_cnt
);
1997 #define dec_rs_pending(mdev) do { \
1998 typecheck(struct drbd_conf *, mdev); \
1999 atomic_dec(&mdev->rs_pending_cnt); \
2000 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
2002 /* counts how many answers we still need to send to the peer.
2004 * receive_Data unless protocol A;
2005 * we need to send a P_RECV_ACK (proto B)
2006 * or P_WRITE_ACK (proto C)
2007 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2008 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2009 * receive_Barrier_* we need to send a P_BARRIER_ACK
2011 static inline void inc_unacked(struct drbd_conf
*mdev
)
2013 atomic_inc(&mdev
->unacked_cnt
);
2016 #define dec_unacked(mdev) do { \
2017 typecheck(struct drbd_conf *, mdev); \
2018 atomic_dec(&mdev->unacked_cnt); \
2019 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2021 #define sub_unacked(mdev, n) do { \
2022 typecheck(struct drbd_conf *, mdev); \
2023 atomic_sub(n, &mdev->unacked_cnt); \
2024 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2027 static inline void put_net_conf(struct drbd_conf
*mdev
)
2029 if (atomic_dec_and_test(&mdev
->net_cnt
))
2030 wake_up(&mdev
->misc_wait
);
2034 * get_net_conf() - Increase ref count on mdev->net_conf; Returns 0 if nothing there
2035 * @mdev: DRBD device.
2037 * You have to call put_net_conf() when finished working with mdev->net_conf.
2039 static inline int get_net_conf(struct drbd_conf
*mdev
)
2043 atomic_inc(&mdev
->net_cnt
);
2044 have_net_conf
= mdev
->state
.conn
>= C_UNCONNECTED
;
2047 return have_net_conf
;
2051 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2054 * You have to call put_ldev() when finished working with mdev->ldev.
2056 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2057 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2059 static inline void put_ldev(struct drbd_conf
*mdev
)
2062 if (atomic_dec_and_test(&mdev
->local_cnt
))
2063 wake_up(&mdev
->misc_wait
);
2064 D_ASSERT(atomic_read(&mdev
->local_cnt
) >= 0);
2068 static inline int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
)
2072 atomic_inc(&mdev
->local_cnt
);
2073 io_allowed
= (mdev
->state
.disk
>= mins
);
2079 extern int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
);
2082 /* you must have an "get_ldev" reference */
2083 static inline void drbd_get_syncer_progress(struct drbd_conf
*mdev
,
2084 unsigned long *bits_left
, unsigned int *per_mil_done
)
2087 * this is to break it at compile time when we change that
2088 * (we may feel 4TB maximum storage per drbd is not enough)
2090 typecheck(unsigned long, mdev
->rs_total
);
2092 /* note: both rs_total and rs_left are in bits, i.e. in
2093 * units of BM_BLOCK_SIZE.
2094 * for the percentage, we don't care. */
2096 *bits_left
= drbd_bm_total_weight(mdev
) - mdev
->rs_failed
;
2097 /* >> 10 to prevent overflow,
2098 * +1 to prevent division by zero */
2099 if (*bits_left
> mdev
->rs_total
) {
2100 /* doh. maybe a logic bug somewhere.
2101 * may also be just a race condition
2102 * between this and a disconnect during sync.
2103 * for now, just prevent in-kernel buffer overflow.
2106 dev_warn(DEV
, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2107 drbd_conn_str(mdev
->state
.conn
),
2108 *bits_left
, mdev
->rs_total
, mdev
->rs_failed
);
2111 /* make sure the calculation happens in long context */
2112 unsigned long tmp
= 1000UL -
2113 (*bits_left
>> 10)*1000UL
2114 / ((mdev
->rs_total
>> 10) + 1UL);
2115 *per_mil_done
= tmp
;
2120 /* this throttles on-the-fly application requests
2121 * according to max_buffers settings;
2122 * maybe re-implement using semaphores? */
2123 static inline int drbd_get_max_buffers(struct drbd_conf
*mdev
)
2125 int mxb
= 1000000; /* arbitrary limit on open requests */
2126 if (get_net_conf(mdev
)) {
2127 mxb
= mdev
->net_conf
->max_buffers
;
2133 static inline int drbd_state_is_stable(union drbd_state s
)
2136 /* DO NOT add a default clause, we want the compiler to warn us
2137 * for any newly introduced state we may have forgotten to add here */
2139 switch ((enum drbd_conns
)s
.conn
) {
2140 /* new io only accepted when there is no connection, ... */
2142 case C_WF_CONNECTION
:
2143 /* ... or there is a well established connection. */
2149 case C_PAUSED_SYNC_S
:
2150 case C_PAUSED_SYNC_T
:
2151 /* maybe stable, look at the disk state */
2154 /* no new io accepted during tansitional states
2155 * like handshake or teardown */
2156 case C_DISCONNECTING
:
2160 case C_NETWORK_FAILURE
:
2161 case C_PROTOCOL_ERROR
:
2163 case C_WF_REPORT_PARAMS
:
2164 case C_STARTING_SYNC_S
:
2165 case C_STARTING_SYNC_T
:
2168 case C_WF_SYNC_UUID
:
2174 switch ((enum drbd_disk_state
)s
.disk
) {
2176 case D_INCONSISTENT
:
2180 /* disk state is stable as well. */
2183 /* no new io accepted during tansitional states */
2196 static inline int __inc_ap_bio_cond(struct drbd_conf
*mdev
)
2198 int mxb
= drbd_get_max_buffers(mdev
);
2200 if (mdev
->state
.susp
)
2202 if (test_bit(SUSPEND_IO
, &mdev
->flags
))
2205 /* to avoid potential deadlock or bitmap corruption,
2206 * in various places, we only allow new application io
2207 * to start during "stable" states. */
2209 /* no new io accepted when attaching or detaching the disk */
2210 if (!drbd_state_is_stable(mdev
->state
))
2213 /* since some older kernels don't have atomic_add_unless,
2214 * and we are within the spinlock anyways, we have this workaround. */
2215 if (atomic_read(&mdev
->ap_bio_cnt
) > mxb
)
2217 if (test_bit(BITMAP_IO
, &mdev
->flags
))
2222 /* I'd like to use wait_event_lock_irq,
2223 * but I'm not sure when it got introduced,
2224 * and not sure when it has 3 or 4 arguments */
2225 static inline void inc_ap_bio(struct drbd_conf
*mdev
, int count
)
2227 /* compare with after_state_ch,
2228 * os.conn != C_WF_BITMAP_S && ns.conn == C_WF_BITMAP_S */
2232 * as long as the device is suspended
2233 * until the bitmap is no longer on the fly during connection
2234 * handshake as long as we would exeed the max_buffer limit.
2236 * to avoid races with the reconnect code,
2237 * we need to atomic_inc within the spinlock. */
2239 spin_lock_irq(&mdev
->req_lock
);
2240 while (!__inc_ap_bio_cond(mdev
)) {
2241 prepare_to_wait(&mdev
->misc_wait
, &wait
, TASK_UNINTERRUPTIBLE
);
2242 spin_unlock_irq(&mdev
->req_lock
);
2244 finish_wait(&mdev
->misc_wait
, &wait
);
2245 spin_lock_irq(&mdev
->req_lock
);
2247 atomic_add(count
, &mdev
->ap_bio_cnt
);
2248 spin_unlock_irq(&mdev
->req_lock
);
2251 static inline void dec_ap_bio(struct drbd_conf
*mdev
)
2253 int mxb
= drbd_get_max_buffers(mdev
);
2254 int ap_bio
= atomic_dec_return(&mdev
->ap_bio_cnt
);
2256 D_ASSERT(ap_bio
>= 0);
2257 /* this currently does wake_up for every dec_ap_bio!
2258 * maybe rather introduce some type of hysteresis?
2259 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2261 wake_up(&mdev
->misc_wait
);
2262 if (ap_bio
== 0 && test_bit(BITMAP_IO
, &mdev
->flags
)) {
2263 if (!test_and_set_bit(BITMAP_IO_QUEUED
, &mdev
->flags
))
2264 drbd_queue_work(&mdev
->data
.work
, &mdev
->bm_io_work
.w
);
2268 static inline void drbd_set_ed_uuid(struct drbd_conf
*mdev
, u64 val
)
2270 mdev
->ed_uuid
= val
;
2273 static inline int seq_cmp(u32 a
, u32 b
)
2275 /* we assume wrap around at 32bit.
2276 * for wrap around at 24bit (old atomic_t),
2280 return (s32
)(a
) - (s32
)(b
);
2282 #define seq_lt(a, b) (seq_cmp((a), (b)) < 0)
2283 #define seq_gt(a, b) (seq_cmp((a), (b)) > 0)
2284 #define seq_ge(a, b) (seq_cmp((a), (b)) >= 0)
2285 #define seq_le(a, b) (seq_cmp((a), (b)) <= 0)
2286 /* CAUTION: please no side effects in arguments! */
2287 #define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b)))
2289 static inline void update_peer_seq(struct drbd_conf
*mdev
, unsigned int new_seq
)
2292 spin_lock(&mdev
->peer_seq_lock
);
2293 m
= seq_max(mdev
->peer_seq
, new_seq
);
2295 spin_unlock(&mdev
->peer_seq_lock
);
2297 wake_up(&mdev
->seq_wait
);
2300 static inline void drbd_update_congested(struct drbd_conf
*mdev
)
2302 struct sock
*sk
= mdev
->data
.socket
->sk
;
2303 if (sk
->sk_wmem_queued
> sk
->sk_sndbuf
* 4 / 5)
2304 set_bit(NET_CONGESTED
, &mdev
->flags
);
2307 static inline int drbd_queue_order_type(struct drbd_conf
*mdev
)
2309 /* sorry, we currently have no working implementation
2310 * of distributed TCQ stuff */
2311 #ifndef QUEUE_ORDERED_NONE
2312 #define QUEUE_ORDERED_NONE 0
2314 return QUEUE_ORDERED_NONE
;
2317 static inline void drbd_blk_run_queue(struct request_queue
*q
)
2319 if (q
&& q
->unplug_fn
)
2323 static inline void drbd_kick_lo(struct drbd_conf
*mdev
)
2325 if (get_ldev(mdev
)) {
2326 drbd_blk_run_queue(bdev_get_queue(mdev
->ldev
->backing_bdev
));
2331 static inline void drbd_md_flush(struct drbd_conf
*mdev
)
2335 if (test_bit(MD_NO_BARRIER
, &mdev
->flags
))
2338 r
= blkdev_issue_flush(mdev
->ldev
->md_bdev
, GFP_KERNEL
, NULL
,
2341 set_bit(MD_NO_BARRIER
, &mdev
->flags
);
2342 dev_err(DEV
, "meta data flush failed with status %d, disabling md-flushes\n", r
);