4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
6 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
7 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
8 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
10 drbd is free software; you can redistribute it and/or modify
11 it under the terms of the GNU General Public License as published by
12 the Free Software Foundation; either version 2, or (at your option)
15 drbd is distributed in the hope that it will be useful,
16 but WITHOUT ANY WARRANTY; without even the implied warranty of
17 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 GNU General Public License for more details.
20 You should have received a copy of the GNU General Public License
21 along with drbd; see the file COPYING. If not, write to
22 the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
29 #include <linux/compiler.h>
30 #include <linux/types.h>
31 #include <linux/version.h>
32 #include <linux/list.h>
33 #include <linux/sched.h>
34 #include <linux/bitops.h>
35 #include <linux/slab.h>
36 #include <linux/crypto.h>
37 #include <linux/ratelimit.h>
38 #include <linux/tcp.h>
39 #include <linux/mutex.h>
40 #include <linux/major.h>
41 #include <linux/blkdev.h>
42 #include <linux/genhd.h>
44 #include <linux/lru_cache.h>
45 #include <linux/prefetch.h>
48 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
49 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
50 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
51 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
53 # define __protected_by(x)
54 # define __protected_read_by(x)
55 # define __protected_write_by(x)
56 # define __must_hold(x)
59 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
61 /* module parameter, defined in drbd_main.c */
62 extern unsigned int minor_count
;
63 extern int disable_sendpage
;
65 extern unsigned int cn_idx
;
67 #ifdef CONFIG_DRBD_FAULT_INJECTION
68 extern int enable_faults
;
69 extern int fault_rate
;
70 extern int fault_devs
;
73 extern char usermode_helper
[];
76 /* I don't remember why XCPU ...
77 * This is used to wake the asender,
78 * and to interrupt sending the sending task
81 #define DRBD_SIG SIGXCPU
83 /* This is used to stop/restart our threads.
84 * Cannot use SIGTERM nor SIGKILL, since these
85 * are sent out by init on runlevel changes
86 * I choose SIGHUP for now.
88 #define DRBD_SIGKILL SIGHUP
90 #define ID_IN_SYNC (4711ULL)
91 #define ID_OUT_OF_SYNC (4712ULL)
92 #define ID_SYNCER (-1ULL)
94 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
99 /* to shorten dev_warn(DEV, "msg"); and relatives statements */
100 #define DEV (disk_to_dev(mdev->vdisk))
102 #define D_ASSERT(exp) if (!(exp)) \
103 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
105 #define ERR_IF(exp) if (({ \
106 int _b = (exp) != 0; \
107 if (_b) dev_err(DEV, "ASSERT FAILED: %s: (%s) in %s:%d\n", \
108 __func__, #exp, __FILE__, __LINE__); \
112 /* Defines to control fault insertion */
114 DRBD_FAULT_MD_WR
= 0, /* meta data write */
115 DRBD_FAULT_MD_RD
= 1, /* read */
116 DRBD_FAULT_RS_WR
= 2, /* resync */
117 DRBD_FAULT_RS_RD
= 3,
118 DRBD_FAULT_DT_WR
= 4, /* data */
119 DRBD_FAULT_DT_RD
= 5,
120 DRBD_FAULT_DT_RA
= 6, /* data read ahead */
121 DRBD_FAULT_BM_ALLOC
= 7, /* bitmap allocation */
122 DRBD_FAULT_AL_EE
= 8, /* alloc ee */
123 DRBD_FAULT_RECEIVE
= 9, /* Changes some bytes upon receiving a [rs]data block */
129 _drbd_insert_fault(struct drbd_conf
*mdev
, unsigned int type
);
132 drbd_insert_fault(struct drbd_conf
*mdev
, unsigned int type
) {
133 #ifdef CONFIG_DRBD_FAULT_INJECTION
135 (enable_faults
& (1<<type
)) &&
136 _drbd_insert_fault(mdev
, type
);
142 /* integer division, round _UP_ to the next integer */
143 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
144 /* usual integer division */
145 #define div_floor(A, B) ((A)/(B))
147 /* drbd_meta-data.c (still in drbd_main.c) */
148 /* 4th incarnation of the disk layout. */
149 #define DRBD_MD_MAGIC (DRBD_MAGIC+4)
151 extern struct drbd_conf
**minor_table
;
152 extern struct ratelimit_state drbd_ratelimit_state
;
156 /* receiver (data socket) */
158 P_DATA_REPLY
= 0x01, /* Response to P_DATA_REQUEST */
159 P_RS_DATA_REPLY
= 0x02, /* Response to P_RS_DATA_REQUEST */
162 P_BECOME_SYNC_TARGET
= 0x05,
163 P_BECOME_SYNC_SOURCE
= 0x06,
164 P_UNPLUG_REMOTE
= 0x07, /* Used at various times to hint the peer */
165 P_DATA_REQUEST
= 0x08, /* Used to ask for a data block */
166 P_RS_DATA_REQUEST
= 0x09, /* Used to ask for a data block for resync */
173 P_AUTH_CHALLENGE
= 0x10,
174 P_AUTH_RESPONSE
= 0x11,
175 P_STATE_CHG_REQ
= 0x12,
177 /* asender (meta socket */
180 P_RECV_ACK
= 0x15, /* Used in protocol B */
181 P_WRITE_ACK
= 0x16, /* Used in protocol C */
182 P_RS_WRITE_ACK
= 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
183 P_DISCARD_ACK
= 0x18, /* Used in proto C, two-primaries conflict detection */
184 P_NEG_ACK
= 0x19, /* Sent if local disk is unusable */
185 P_NEG_DREPLY
= 0x1a, /* Local disk is broken... */
186 P_NEG_RS_DREPLY
= 0x1b, /* Local disk is broken... */
187 P_BARRIER_ACK
= 0x1c,
188 P_STATE_CHG_REPLY
= 0x1d,
190 /* "new" commands, no longer fitting into the ordering scheme above */
192 P_OV_REQUEST
= 0x1e, /* data socket */
194 P_OV_RESULT
= 0x20, /* meta socket */
195 P_CSUM_RS_REQUEST
= 0x21, /* data socket */
196 P_RS_IS_IN_SYNC
= 0x22, /* meta socket */
197 P_SYNC_PARAM89
= 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
198 P_COMPRESSED_BITMAP
= 0x24, /* compressed or otherwise encoded bitmap transfer */
199 /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */
200 /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */
201 P_DELAY_PROBE
= 0x27, /* is used on BOTH sockets */
202 P_OUT_OF_SYNC
= 0x28, /* Mark as out of sync (Outrunning), data socket */
203 P_RS_CANCEL
= 0x29, /* meta: Used to cancel RS_DATA_REQUEST packet by SyncSource */
206 P_MAY_IGNORE
= 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
207 P_MAX_OPT_CMD
= 0x101,
209 /* special command ids for handshake */
211 P_HAND_SHAKE_M
= 0xfff1, /* First Packet on the MetaSock */
212 P_HAND_SHAKE_S
= 0xfff2, /* First Packet on the Socket */
214 P_HAND_SHAKE
= 0xfffe /* FIXED for the next century! */
217 static inline const char *cmdname(enum drbd_packets cmd
)
219 /* THINK may need to become several global tables
220 * when we want to support more than
222 static const char *cmdnames
[] = {
224 [P_DATA_REPLY
] = "DataReply",
225 [P_RS_DATA_REPLY
] = "RSDataReply",
226 [P_BARRIER
] = "Barrier",
227 [P_BITMAP
] = "ReportBitMap",
228 [P_BECOME_SYNC_TARGET
] = "BecomeSyncTarget",
229 [P_BECOME_SYNC_SOURCE
] = "BecomeSyncSource",
230 [P_UNPLUG_REMOTE
] = "UnplugRemote",
231 [P_DATA_REQUEST
] = "DataRequest",
232 [P_RS_DATA_REQUEST
] = "RSDataRequest",
233 [P_SYNC_PARAM
] = "SyncParam",
234 [P_SYNC_PARAM89
] = "SyncParam89",
235 [P_PROTOCOL
] = "ReportProtocol",
236 [P_UUIDS
] = "ReportUUIDs",
237 [P_SIZES
] = "ReportSizes",
238 [P_STATE
] = "ReportState",
239 [P_SYNC_UUID
] = "ReportSyncUUID",
240 [P_AUTH_CHALLENGE
] = "AuthChallenge",
241 [P_AUTH_RESPONSE
] = "AuthResponse",
243 [P_PING_ACK
] = "PingAck",
244 [P_RECV_ACK
] = "RecvAck",
245 [P_WRITE_ACK
] = "WriteAck",
246 [P_RS_WRITE_ACK
] = "RSWriteAck",
247 [P_DISCARD_ACK
] = "DiscardAck",
248 [P_NEG_ACK
] = "NegAck",
249 [P_NEG_DREPLY
] = "NegDReply",
250 [P_NEG_RS_DREPLY
] = "NegRSDReply",
251 [P_BARRIER_ACK
] = "BarrierAck",
252 [P_STATE_CHG_REQ
] = "StateChgRequest",
253 [P_STATE_CHG_REPLY
] = "StateChgReply",
254 [P_OV_REQUEST
] = "OVRequest",
255 [P_OV_REPLY
] = "OVReply",
256 [P_OV_RESULT
] = "OVResult",
257 [P_CSUM_RS_REQUEST
] = "CsumRSRequest",
258 [P_RS_IS_IN_SYNC
] = "CsumRSIsInSync",
259 [P_COMPRESSED_BITMAP
] = "CBitmap",
260 [P_DELAY_PROBE
] = "DelayProbe",
261 [P_OUT_OF_SYNC
] = "OutOfSync",
265 if (cmd
== P_HAND_SHAKE_M
)
267 if (cmd
== P_HAND_SHAKE_S
)
269 if (cmd
== P_HAND_SHAKE
)
271 if (cmd
>= P_MAX_CMD
)
273 return cmdnames
[cmd
];
276 /* for sending/receiving the bitmap,
277 * possibly in some encoding scheme */
280 * stores total bits and long words
281 * of the bitmap, so we don't need to
282 * call the accessor functions over and again. */
283 unsigned long bm_bits
;
284 unsigned long bm_words
;
285 /* during xfer, current position within the bitmap */
286 unsigned long bit_offset
;
287 unsigned long word_offset
;
289 /* statistics; index: (h->command == P_BITMAP) */
294 extern void INFO_bm_xfer_stats(struct drbd_conf
*mdev
,
295 const char *direction
, struct bm_xfer_ctx
*c
);
297 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx
*c
)
299 /* word_offset counts "native long words" (32 or 64 bit),
301 * Encoded packet may end at an unaligned bit offset.
302 * In case a fallback clear text packet is transmitted in
303 * between, we adjust this offset back to the last 64bit
304 * aligned "native long word", which makes coding and decoding
305 * the plain text bitmap much more convenient. */
306 #if BITS_PER_LONG == 64
307 c
->word_offset
= c
->bit_offset
>> 6;
308 #elif BITS_PER_LONG == 32
309 c
->word_offset
= c
->bit_offset
>> 5;
310 c
->word_offset
&= ~(1UL);
312 # error "unsupported BITS_PER_LONG"
317 #define __packed __attribute__((packed))
320 /* This is the layout for a packet on the wire.
321 * The byteorder is the network byte order.
322 * (except block_id and barrier fields.
323 * these are pointers to local structs
324 * and have no relevance for the partner,
325 * which just echoes them as received.)
327 * NOTE that the payload starts at a long aligned offset,
328 * regardless of 32 or 64 bit arch!
333 u16 length
; /* bytes of data after this header */
337 /* Header for big packets, Used for data packets exceeding 64kB */
339 u16 magic
; /* use DRBD_MAGIC_BIG here */
341 u32 length
; /* Use only 24 bits of that. Ignore the highest 8 bit. */
346 struct p_header80 h80
;
347 struct p_header95 h95
;
351 * short commands, packets without payload, plain p_header:
354 * P_BECOME_SYNC_TARGET
355 * P_BECOME_SYNC_SOURCE
360 * commands with out-of-struct payload:
361 * P_BITMAP (no additional fields)
362 * P_DATA, P_DATA_REPLY (see p_data)
363 * P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
366 /* these defines must not be changed without changing the protocol version */
367 #define DP_HARDBARRIER 1 /* depricated */
368 #define DP_RW_SYNC 2 /* equals REQ_SYNC */
369 #define DP_MAY_SET_IN_SYNC 4
370 #define DP_UNPLUG 8 /* not used anymore */
371 #define DP_FUA 16 /* equals REQ_FUA */
372 #define DP_FLUSH 32 /* equals REQ_FLUSH */
373 #define DP_DISCARD 64 /* equals REQ_DISCARD */
377 u64 sector
; /* 64 bits sector number */
378 u64 block_id
; /* to identify the request in protocol B&C */
384 * commands which share a struct:
386 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
387 * P_DISCARD_ACK (proto C, two-primaries conflict detection)
389 * P_DATA_REQUEST, P_RS_DATA_REQUEST
392 struct p_header80 head
;
401 struct p_header80 head
;
405 u32 pad
; /* to multiple of 8 Byte */
409 * commands with their own struct for additional fields:
418 struct p_header80 head
; /* 8 bytes */
423 /* should be more than enough for future enhancements
424 * for now, feature_flags and the reserverd array shall be zero.
430 /* 80 bytes, FIXED for the next century */
433 struct p_header80 head
;
434 u32 barrier
; /* barrier number _handle_ only */
435 u32 pad
; /* to multiple of 8 Byte */
438 struct p_barrier_ack
{
439 struct p_header80 head
;
445 struct p_header80 head
;
448 /* Since protocol version 88 and higher. */
452 struct p_rs_param_89
{
453 struct p_header80 head
;
455 /* protocol version 89: */
456 char verify_alg
[SHARED_SECRET_MAX
];
457 char csums_alg
[SHARED_SECRET_MAX
];
460 struct p_rs_param_95
{
461 struct p_header80 head
;
463 char verify_alg
[SHARED_SECRET_MAX
];
464 char csums_alg
[SHARED_SECRET_MAX
];
471 enum drbd_conn_flags
{
477 struct p_header80 head
;
485 /* Since protocol version 87 and higher. */
486 char integrity_alg
[0];
491 struct p_header80 head
;
492 u64 uuid
[UI_EXTENDED_SIZE
];
496 struct p_header80 head
;
501 struct p_header80 head
;
502 u64 d_size
; /* size of disk */
503 u64 u_size
; /* user requested size */
504 u64 c_size
; /* current exported size */
505 u32 max_bio_size
; /* Maximal size of a BIO */
506 u16 queue_order_type
; /* not yet implemented in DRBD*/
507 u16 dds_flags
; /* use enum dds_flags here. */
511 struct p_header80 head
;
516 struct p_header80 head
;
521 struct p_req_state_reply
{
522 struct p_header80 head
;
526 struct p_drbd06_param
{
537 struct p_header80 head
;
543 struct p_block_desc
{
544 struct p_header80 head
;
547 u32 pad
; /* to multiple of 8 Byte */
550 /* Valid values for the encoding field.
551 * Bump proto version when changing this. */
552 enum drbd_bitmap_code
{
553 /* RLE_VLI_Bytes = 0,
554 * and other bit variants had been defined during
555 * algorithm evaluation. */
559 struct p_compressed_bm
{
560 struct p_header80 head
;
561 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
562 * (encoding & 0x80): polarity (set/unset) of first runlength
563 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
564 * used to pad up to head.length bytes
571 struct p_delay_probe93
{
572 struct p_header80 head
;
573 u32 seq_num
; /* sequence number to match the two probe packets */
574 u32 offset
; /* usecs the probe got sent after the reference time point */
577 /* DCBP: Drbd Compressed Bitmap Packet ... */
578 static inline enum drbd_bitmap_code
579 DCBP_get_code(struct p_compressed_bm
*p
)
581 return (enum drbd_bitmap_code
)(p
->encoding
& 0x0f);
585 DCBP_set_code(struct p_compressed_bm
*p
, enum drbd_bitmap_code code
)
588 p
->encoding
= (p
->encoding
& ~0xf) | code
;
592 DCBP_get_start(struct p_compressed_bm
*p
)
594 return (p
->encoding
& 0x80) != 0;
598 DCBP_set_start(struct p_compressed_bm
*p
, int set
)
600 p
->encoding
= (p
->encoding
& ~0x80) | (set
? 0x80 : 0);
604 DCBP_get_pad_bits(struct p_compressed_bm
*p
)
606 return (p
->encoding
>> 4) & 0x7;
610 DCBP_set_pad_bits(struct p_compressed_bm
*p
, int n
)
613 p
->encoding
= (p
->encoding
& (~0x7 << 4)) | (n
<< 4);
616 /* one bitmap packet, including the p_header,
617 * should fit within one _architecture independend_ page.
618 * so we need to use the fixed size 4KiB page size
619 * most architectures have used for a long time.
621 #define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header80))
622 #define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
623 #define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm))
624 #if (PAGE_SIZE < 4096)
625 /* drbd_send_bitmap / receive_bitmap would break horribly */
626 #error "PAGE_SIZE too small"
630 union p_header header
;
631 struct p_handshake handshake
;
633 struct p_block_ack block_ack
;
634 struct p_barrier barrier
;
635 struct p_barrier_ack barrier_ack
;
636 struct p_rs_param_89 rs_param_89
;
637 struct p_rs_param_95 rs_param_95
;
638 struct p_protocol protocol
;
639 struct p_sizes sizes
;
640 struct p_uuids uuids
;
641 struct p_state state
;
642 struct p_req_state req_state
;
643 struct p_req_state_reply req_state_reply
;
644 struct p_block_req block_req
;
645 struct p_delay_probe93 delay_probe93
;
646 struct p_rs_uuid rs_uuid
;
647 struct p_block_desc block_desc
;
650 /**********************************************************************/
651 enum drbd_thread_state
{
660 struct task_struct
*task
;
661 struct completion stop
;
662 enum drbd_thread_state t_state
;
663 int (*function
) (struct drbd_thread
*);
664 struct drbd_conf
*mdev
;
668 static inline enum drbd_thread_state
get_t_state(struct drbd_thread
*thi
)
670 /* THINK testing the t_state seems to be uncritical in all cases
671 * (but thread_{start,stop}), so we can read it *without* the lock.
679 typedef int (*drbd_work_cb
)(struct drbd_conf
*, struct drbd_work
*, int cancel
);
681 struct list_head list
;
685 #include "drbd_interval.h"
687 struct drbd_request
{
689 struct drbd_conf
*mdev
;
691 /* if local IO is not allowed, will be NULL.
692 * if local IO _is_ allowed, holds the locally submitted bio clone,
693 * or, after local IO completion, the ERR_PTR(error).
694 * see drbd_endio_pri(). */
695 struct bio
*private_bio
;
697 struct hlist_node collision
;
698 struct drbd_interval i
;
699 unsigned int epoch
; /* barrier_nr */
701 /* barrier_nr: used to check on "completion" whether this req was in
702 * the current epoch, and we therefore have to close it,
703 * starting a new epoch...
706 struct list_head tl_requests
; /* ring list in the transfer log */
707 struct bio
*master_bio
; /* master bio pointer */
708 unsigned long rq_state
; /* see comments above _req_mod() */
710 unsigned long start_time
;
713 struct drbd_tl_epoch
{
715 struct list_head requests
; /* requests before */
716 struct drbd_tl_epoch
*next
; /* pointer to the next barrier */
717 unsigned int br_number
; /* the barriers identifier. */
718 int n_writes
; /* number of requests attached before this barrier */
723 /* These Tl_epoch_entries may be in one of 6 lists:
724 active_ee .. data packet being written
725 sync_ee .. syncer block being written
726 done_ee .. block written, need to send P_WRITE_ACK
727 read_ee .. [RS]P_DATA_REQUEST being read
731 struct list_head list
;
732 unsigned int barrier_nr
;
733 atomic_t epoch_size
; /* increased on every request added. */
734 atomic_t active
; /* increased on every req. added, and dec on every finished. */
738 /* drbd_epoch flag bits */
740 DE_HAVE_BARRIER_NUMBER
,
747 EV_CLEANUP
= 32, /* used as flag */
750 struct drbd_wq_barrier
{
752 struct completion done
;
760 struct drbd_epoch_entry
{
762 struct hlist_node collision
;
763 struct drbd_epoch
*epoch
; /* for writes */
764 struct drbd_conf
*mdev
;
766 atomic_t pending_bios
;
768 /* see comments on ee flag bits below */
773 struct digest_info
*digest
;
778 * While corresponding bios are in flight, the only modification will be
779 * set_bit WAS_ERROR, which has to be atomic.
780 * If no bios are in flight yet, or all have been completed,
781 * non-atomic modification to ee->flags is ok.
784 __EE_CALL_AL_COMPLETE_IO
,
785 __EE_MAY_SET_IN_SYNC
,
787 /* In case a barrier failed,
788 * we need to resubmit without the barrier flag. */
791 /* we may have several bios per epoch entry.
792 * if any of those fail, we set this flag atomically
793 * from the endio callback */
796 /* This ee has a pointer to a digest instead of a block id */
799 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
800 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
801 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
802 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
803 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
805 /* global flag bits */
807 CREATE_BARRIER
, /* next P_DATA is preceded by a P_BARRIER */
808 SIGNAL_ASENDER
, /* whether asender wants to be interrupted */
809 SEND_PING
, /* whether asender should send a ping asap */
811 UNPLUG_QUEUED
, /* only relevant with kernel 2.4 */
812 UNPLUG_REMOTE
, /* sending a "UnplugRemote" could help */
813 MD_DIRTY
, /* current uuids and flags not yet on disk */
814 DISCARD_CONCURRENT
, /* Set on one node, cleared on the peer! */
815 USE_DEGR_WFC_T
, /* degr-wfc-timeout instead of wfc-timeout. */
816 CLUSTER_ST_CHANGE
, /* Cluster wide state change going on... */
819 CRASHED_PRIMARY
, /* This node was a crashed primary.
820 * Gets cleared when the state.conn
821 * goes into C_CONNECTED state. */
822 NO_BARRIER_SUPP
, /* underlying block device doesn't implement barriers */
825 MD_NO_FUA
, /* Users wants us to not use FUA/FLUSH on meta data dev */
826 SUSPEND_IO
, /* suspend application io */
827 BITMAP_IO
, /* suspend application io;
828 once no more io in flight, start bitmap io */
829 BITMAP_IO_QUEUED
, /* Started bitmap IO */
830 GO_DISKLESS
, /* Disk is being detached, on io-error or admin request. */
831 WAS_IO_ERROR
, /* Local disk failed returned IO error */
832 RESYNC_AFTER_NEG
, /* Resync after online grow after the attach&negotiate finished. */
833 NET_CONGESTED
, /* The data socket is congested */
835 CONFIG_PENDING
, /* serialization of (re)configuration requests.
836 * if set, also prevents the device from dying */
837 DEVICE_DYING
, /* device became unconfigured,
838 * but worker thread is still handling the cleanup.
839 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
840 * while this is set. */
841 RESIZE_PENDING
, /* Size change detected locally, waiting for the response from
842 * the peer, if it changed there as well. */
843 CONN_DRY_RUN
, /* Expect disconnect after resync handshake. */
844 GOT_PING_ACK
, /* set when we receive a ping_ack packet, misc wait gets woken */
845 NEW_CUR_UUID
, /* Create new current UUID when thawing IO */
846 AL_SUSPENDED
, /* Activity logging is currently suspended. */
847 AHEAD_TO_SYNC_SOURCE
, /* Ahead -> SyncSource queued */
850 struct drbd_bitmap
; /* opaque for drbd_conf */
852 /* definition of bits in bm_flags to be used in drbd_bm_lock
853 * and drbd_bitmap_io and friends. */
855 /* do we need to kfree, or vfree bm_pages? */
856 BM_P_VMALLOCED
= 0x10000, /* internal use only, will be masked out */
858 /* currently locked for bulk operation */
859 BM_LOCKED_MASK
= 0x7,
861 /* in detail, that is: */
866 /* (test bit, count bit) allowed (common case) */
867 BM_LOCKED_TEST_ALLOWED
= 0x3,
869 /* testing bits, as well as setting new bits allowed, but clearing bits
870 * would be unexpected. Used during bitmap receive. Setting new bits
871 * requires sending of "out-of-sync" information, though. */
872 BM_LOCKED_SET_ALLOWED
= 0x1,
874 /* clear is not expected while bitmap is locked for bulk operation */
878 /* TODO sort members for performance
879 * MAYBE group them further */
881 /* THINK maybe we actually want to use the default "event/%s" worker threads
882 * or similar in linux 2.6, which uses per cpu data and threads.
884 struct drbd_work_queue
{
886 struct semaphore s
; /* producers up it, worker down()s it */
887 spinlock_t q_lock
; /* to protect the list. */
891 struct drbd_work_queue work
;
893 struct socket
*socket
;
894 /* this way we get our
895 * send/receive buffers off the stack */
896 union p_polymorph sbuf
;
897 union p_polymorph rbuf
;
901 u64 md_offset
; /* sector offset to 'super' block */
903 u64 la_size_sect
; /* last agreed size, unit sectors */
909 s32 al_offset
; /* signed relative sector offset to al area */
910 s32 bm_offset
; /* signed relative sector offset to bitmap */
912 /* u32 al_nr_extents; important for restoring the AL
913 * is stored into sync_conf.al_extents, which in turn
914 * gets applied to act_log->nr_elements
918 /* for sync_conf and other types... */
919 #define NL_PACKET(name, number, fields) struct name { fields };
920 #define NL_INTEGER(pn,pr,member) int member;
921 #define NL_INT64(pn,pr,member) __u64 member;
922 #define NL_BIT(pn,pr,member) unsigned member:1;
923 #define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
924 #include "linux/drbd_nl.h"
926 struct drbd_backing_dev
{
927 struct block_device
*backing_bdev
;
928 struct block_device
*md_bdev
;
930 struct disk_conf dc
; /* The user provided config... */
931 sector_t known_size
; /* last known size of that backing device */
935 struct drbd_conf
*mdev
;
936 struct completion event
;
944 int (*io_fn
)(struct drbd_conf
*mdev
);
945 void (*done
)(struct drbd_conf
*mdev
, int rv
);
948 enum write_ordering_e
{
956 unsigned int head_index
;
961 /* things that are stored as / read from meta data on disk */
964 /* configured by drbdsetup */
965 struct net_conf
*net_conf
; /* protected by get_net_conf() and put_net_conf() */
966 struct syncer_conf sync_conf
;
967 struct drbd_backing_dev
*ldev
__protected_by(local
);
969 sector_t p_size
; /* partner's disk size */
970 struct request_queue
*rq_queue
;
971 struct block_device
*this_bdev
;
972 struct gendisk
*vdisk
;
974 struct drbd_socket data
; /* data/barrier/cstate/parameter packets */
975 struct drbd_socket meta
; /* ping/ack (metadata) packets */
976 int agreed_pro_version
; /* actually used protocol version */
977 unsigned long last_received
; /* in jiffies, either socket */
978 unsigned int ko_count
;
979 struct drbd_work resync_work
,
984 struct timer_list resync_timer
;
985 struct timer_list md_sync_timer
;
986 struct timer_list start_resync_timer
;
987 struct timer_list request_timer
;
988 #ifdef DRBD_DEBUG_MD_SYNC
992 } last_md_mark_dirty
;
995 /* Used after attach while negotiating new disk state. */
996 union drbd_state new_state_tmp
;
998 union drbd_state state
;
999 wait_queue_head_t misc_wait
;
1000 wait_queue_head_t state_wait
; /* upon each state change. */
1001 wait_queue_head_t net_cnt_wait
;
1002 unsigned int send_cnt
;
1003 unsigned int recv_cnt
;
1004 unsigned int read_cnt
;
1005 unsigned int writ_cnt
;
1006 unsigned int al_writ_cnt
;
1007 unsigned int bm_writ_cnt
;
1008 atomic_t ap_bio_cnt
; /* Requests we need to complete */
1009 atomic_t ap_pending_cnt
; /* AP data packets on the wire, ack expected */
1010 atomic_t rs_pending_cnt
; /* RS request/data packets on the wire */
1011 atomic_t unacked_cnt
; /* Need to send replys for */
1012 atomic_t local_cnt
; /* Waiting for local completion */
1013 atomic_t net_cnt
; /* Users of net_conf */
1014 spinlock_t req_lock
;
1015 struct drbd_tl_epoch
*unused_spare_tle
; /* for pre-allocation */
1016 struct drbd_tl_epoch
*newest_tle
;
1017 struct drbd_tl_epoch
*oldest_tle
;
1018 struct list_head out_of_sequence_requests
;
1019 struct hlist_head
*tl_hash
;
1020 unsigned int tl_hash_s
;
1022 /* Interval tree of pending local write requests */
1023 struct rb_root write_requests
;
1025 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
1026 unsigned long rs_total
;
1027 /* number of resync blocks that failed in this run */
1028 unsigned long rs_failed
;
1029 /* Syncer's start time [unit jiffies] */
1030 unsigned long rs_start
;
1031 /* cumulated time in PausedSyncX state [unit jiffies] */
1032 unsigned long rs_paused
;
1033 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
1034 unsigned long rs_same_csum
;
1035 #define DRBD_SYNC_MARKS 8
1036 #define DRBD_SYNC_MARK_STEP (3*HZ)
1037 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
1038 unsigned long rs_mark_left
[DRBD_SYNC_MARKS
];
1039 /* marks's time [unit jiffies] */
1040 unsigned long rs_mark_time
[DRBD_SYNC_MARKS
];
1041 /* current index into rs_mark_{left,time} */
1044 /* where does the admin want us to start? (sector) */
1045 sector_t ov_start_sector
;
1046 /* where are we now? (sector) */
1047 sector_t ov_position
;
1048 /* Start sector of out of sync range (to merge printk reporting). */
1049 sector_t ov_last_oos_start
;
1050 /* size of out-of-sync range in sectors. */
1051 sector_t ov_last_oos_size
;
1052 unsigned long ov_left
; /* in bits */
1053 struct crypto_hash
*csums_tfm
;
1054 struct crypto_hash
*verify_tfm
;
1056 struct drbd_thread receiver
;
1057 struct drbd_thread worker
;
1058 struct drbd_thread asender
;
1059 struct drbd_bitmap
*bitmap
;
1060 unsigned long bm_resync_fo
; /* bit offset for drbd_bm_find_next */
1062 /* Used to track operations of resync... */
1063 struct lru_cache
*resync
;
1064 /* Number of locked elements in resync LRU */
1065 unsigned int resync_locked
;
1066 /* resync extent number waiting for application requests */
1067 unsigned int resync_wenr
;
1071 struct drbd_epoch
*current_epoch
;
1072 spinlock_t epoch_lock
;
1073 unsigned int epochs
;
1074 enum write_ordering_e write_ordering
;
1075 struct list_head active_ee
; /* IO in progress (P_DATA gets written to disk) */
1076 struct list_head sync_ee
; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
1077 struct list_head done_ee
; /* send ack */
1078 struct list_head read_ee
; /* IO in progress (any read) */
1079 struct list_head net_ee
; /* zero-copy network send in progress */
1080 struct hlist_head
*ee_hash
; /* is proteced by req_lock! */
1081 unsigned int ee_hash_s
;
1083 /* this one is protected by ee_lock, single thread */
1084 struct drbd_epoch_entry
*last_write_w_barrier
;
1086 int next_barrier_nr
;
1087 struct hlist_head
*app_reads_hash
; /* is proteced by req_lock */
1088 struct list_head resync_reads
;
1089 atomic_t pp_in_use
; /* allocated from page pool */
1090 atomic_t pp_in_use_by_net
; /* sendpage()d, still referenced by tcp */
1091 wait_queue_head_t ee_wait
;
1092 struct page
*md_io_page
; /* one page buffer for md_io */
1093 struct page
*md_io_tmpp
; /* for logical_block_size != 512 */
1094 struct mutex md_io_mutex
; /* protects the md_io_buffer */
1096 wait_queue_head_t al_wait
;
1097 struct lru_cache
*act_log
; /* activity log */
1098 unsigned int al_tr_number
;
1100 int al_tr_pos
; /* position of the next transaction in the journal */
1101 struct crypto_hash
*cram_hmac_tfm
;
1102 struct crypto_hash
*integrity_w_tfm
; /* to be used by the worker thread */
1103 struct crypto_hash
*integrity_r_tfm
; /* to be used by the receiver thread */
1107 wait_queue_head_t seq_wait
;
1108 atomic_t packet_seq
;
1109 unsigned int peer_seq
;
1110 spinlock_t peer_seq_lock
;
1112 unsigned long comm_bm_set
; /* communicated number of set bits. */
1113 cpumask_var_t cpu_mask
;
1114 struct bm_io_work bm_io_work
;
1115 u64 ed_uuid
; /* UUID of the exposed data */
1116 struct mutex state_mutex
;
1117 char congestion_reason
; /* Why we where congested... */
1118 atomic_t rs_sect_in
; /* for incoming resync data rate, SyncTarget */
1119 atomic_t rs_sect_ev
; /* for submitted resync data rate, both */
1120 int rs_last_sect_ev
; /* counter to compare with */
1121 int rs_last_events
; /* counter of read or write "events" (unit sectors)
1122 * on the lower level device when we last looked. */
1123 int c_sync_rate
; /* current resync rate after syncer throttle magic */
1124 struct fifo_buffer rs_plan_s
; /* correction values of resync planer */
1125 int rs_in_flight
; /* resync sectors in flight (to proxy, in proxy and from proxy) */
1126 int rs_planed
; /* resync sectors already planned */
1127 atomic_t ap_in_flight
; /* App sectors in flight (waiting for ack) */
1128 int peer_max_bio_size
;
1129 int local_max_bio_size
;
1132 static inline struct drbd_conf
*minor_to_mdev(unsigned int minor
)
1134 struct drbd_conf
*mdev
;
1136 mdev
= minor
< minor_count
? minor_table
[minor
] : NULL
;
1141 static inline unsigned int mdev_to_minor(struct drbd_conf
*mdev
)
1146 /* returns 1 if it was successful,
1147 * returns 0 if there was no data socket.
1148 * so wherever you are going to use the data.socket, e.g. do
1149 * if (!drbd_get_data_sock(mdev))
1152 * drbd_put_data_sock(mdev);
1154 static inline int drbd_get_data_sock(struct drbd_conf
*mdev
)
1156 mutex_lock(&mdev
->data
.mutex
);
1157 /* drbd_disconnect() could have called drbd_free_sock()
1158 * while we were waiting in down()... */
1159 if (unlikely(mdev
->data
.socket
== NULL
)) {
1160 mutex_unlock(&mdev
->data
.mutex
);
1166 static inline void drbd_put_data_sock(struct drbd_conf
*mdev
)
1168 mutex_unlock(&mdev
->data
.mutex
);
1172 * function declarations
1173 *************************/
1177 enum chg_state_flags
{
1180 CS_WAIT_COMPLETE
= 4,
1182 CS_ORDERED
= CS_WAIT_COMPLETE
+ CS_SERIALIZE
,
1187 DDSF_NO_RESYNC
= 2, /* Do not run a resync for the new space */
1190 extern void drbd_init_set_defaults(struct drbd_conf
*mdev
);
1191 extern enum drbd_state_rv
drbd_change_state(struct drbd_conf
*mdev
,
1192 enum chg_state_flags f
,
1193 union drbd_state mask
,
1194 union drbd_state val
);
1195 extern void drbd_force_state(struct drbd_conf
*, union drbd_state
,
1197 extern enum drbd_state_rv
_drbd_request_state(struct drbd_conf
*,
1200 enum chg_state_flags
);
1201 extern enum drbd_state_rv
__drbd_set_state(struct drbd_conf
*, union drbd_state
,
1202 enum chg_state_flags
,
1203 struct completion
*done
);
1204 extern void print_st_err(struct drbd_conf
*, union drbd_state
,
1205 union drbd_state
, int);
1206 extern int drbd_thread_start(struct drbd_thread
*thi
);
1207 extern void _drbd_thread_stop(struct drbd_thread
*thi
, int restart
, int wait
);
1209 extern void drbd_thread_current_set_cpu(struct drbd_conf
*mdev
);
1210 extern void drbd_calc_cpu_mask(struct drbd_conf
*mdev
);
1212 #define drbd_thread_current_set_cpu(A) ({})
1213 #define drbd_calc_cpu_mask(A) ({})
1215 extern void drbd_free_resources(struct drbd_conf
*mdev
);
1216 extern void tl_release(struct drbd_conf
*mdev
, unsigned int barrier_nr
,
1217 unsigned int set_size
);
1218 extern void tl_clear(struct drbd_conf
*mdev
);
1219 extern void _tl_add_barrier(struct drbd_conf
*, struct drbd_tl_epoch
*);
1220 extern void drbd_free_sock(struct drbd_conf
*mdev
);
1221 extern int drbd_send(struct drbd_conf
*mdev
, struct socket
*sock
,
1222 void *buf
, size_t size
, unsigned msg_flags
);
1223 extern int drbd_send_protocol(struct drbd_conf
*mdev
);
1224 extern int drbd_send_uuids(struct drbd_conf
*mdev
);
1225 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf
*mdev
);
1226 extern int drbd_gen_and_send_sync_uuid(struct drbd_conf
*mdev
);
1227 extern int drbd_send_sizes(struct drbd_conf
*mdev
, int trigger_reply
, enum dds_flags flags
);
1228 extern int _drbd_send_state(struct drbd_conf
*mdev
);
1229 extern int drbd_send_state(struct drbd_conf
*mdev
);
1230 extern int _drbd_send_cmd(struct drbd_conf
*mdev
, struct socket
*sock
,
1231 enum drbd_packets cmd
, struct p_header80
*h
,
1232 size_t size
, unsigned msg_flags
);
1233 #define USE_DATA_SOCKET 1
1234 #define USE_META_SOCKET 0
1235 extern int drbd_send_cmd(struct drbd_conf
*mdev
, int use_data_socket
,
1236 enum drbd_packets cmd
, struct p_header80
*h
,
1238 extern int drbd_send_cmd2(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1239 char *data
, size_t size
);
1240 extern int drbd_send_sync_param(struct drbd_conf
*mdev
, struct syncer_conf
*sc
);
1241 extern int drbd_send_b_ack(struct drbd_conf
*mdev
, u32 barrier_nr
,
1243 extern int drbd_send_ack(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1244 struct drbd_epoch_entry
*e
);
1245 extern int drbd_send_ack_rp(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1246 struct p_block_req
*rp
);
1247 extern int drbd_send_ack_dp(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1248 struct p_data
*dp
, int data_size
);
1249 extern int drbd_send_ack_ex(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1250 sector_t sector
, int blksize
, u64 block_id
);
1251 extern int drbd_send_oos(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1252 extern int drbd_send_block(struct drbd_conf
*mdev
, enum drbd_packets cmd
,
1253 struct drbd_epoch_entry
*e
);
1254 extern int drbd_send_dblock(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1255 extern int drbd_send_drequest(struct drbd_conf
*mdev
, int cmd
,
1256 sector_t sector
, int size
, u64 block_id
);
1257 extern int drbd_send_drequest_csum(struct drbd_conf
*mdev
,
1258 sector_t sector
,int size
,
1259 void *digest
, int digest_size
,
1260 enum drbd_packets cmd
);
1261 extern int drbd_send_ov_request(struct drbd_conf
*mdev
,sector_t sector
,int size
);
1263 extern int drbd_send_bitmap(struct drbd_conf
*mdev
);
1264 extern int _drbd_send_bitmap(struct drbd_conf
*mdev
);
1265 extern int drbd_send_sr_reply(struct drbd_conf
*mdev
, enum drbd_state_rv retcode
);
1266 extern void drbd_free_bc(struct drbd_backing_dev
*ldev
);
1267 extern void drbd_mdev_cleanup(struct drbd_conf
*mdev
);
1268 void drbd_print_uuids(struct drbd_conf
*mdev
, const char *text
);
1270 extern void drbd_md_sync(struct drbd_conf
*mdev
);
1271 extern int drbd_md_read(struct drbd_conf
*mdev
, struct drbd_backing_dev
*bdev
);
1272 extern void drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1273 extern void _drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1274 extern void drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1275 extern void _drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1276 extern void drbd_uuid_set_bm(struct drbd_conf
*mdev
, u64 val
) __must_hold(local
);
1277 extern void drbd_md_set_flag(struct drbd_conf
*mdev
, int flags
) __must_hold(local
);
1278 extern void drbd_md_clear_flag(struct drbd_conf
*mdev
, int flags
)__must_hold(local
);
1279 extern int drbd_md_test_flag(struct drbd_backing_dev
*, int);
1280 #ifndef DRBD_DEBUG_MD_SYNC
1281 extern void drbd_md_mark_dirty(struct drbd_conf
*mdev
);
1283 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1284 extern void drbd_md_mark_dirty_(struct drbd_conf
*mdev
,
1285 unsigned int line
, const char *func
);
1287 extern void drbd_queue_bitmap_io(struct drbd_conf
*mdev
,
1288 int (*io_fn
)(struct drbd_conf
*),
1289 void (*done
)(struct drbd_conf
*, int),
1290 char *why
, enum bm_flag flags
);
1291 extern int drbd_bitmap_io(struct drbd_conf
*mdev
,
1292 int (*io_fn
)(struct drbd_conf
*),
1293 char *why
, enum bm_flag flags
);
1294 extern int drbd_bmio_set_n_write(struct drbd_conf
*mdev
);
1295 extern int drbd_bmio_clear_n_write(struct drbd_conf
*mdev
);
1296 extern void drbd_go_diskless(struct drbd_conf
*mdev
);
1297 extern void drbd_ldev_destroy(struct drbd_conf
*mdev
);
1301 We reserve a 128MB Block (4k aligned)
1302 * either at the end of the backing device
1303 * or on a separate meta data device. */
1305 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1306 /* The following numbers are sectors */
1307 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1308 #define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */
1309 /* Allows up to about 3.8TB */
1310 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1312 /* Since the smalles IO unit is usually 512 byte */
1313 #define MD_SECTOR_SHIFT 9
1314 #define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT)
1317 #define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1318 #define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */
1319 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1321 #if BITS_PER_LONG == 32
1323 #define cpu_to_lel(A) cpu_to_le32(A)
1324 #define lel_to_cpu(A) le32_to_cpu(A)
1325 #elif BITS_PER_LONG == 64
1327 #define cpu_to_lel(A) cpu_to_le64(A)
1328 #define lel_to_cpu(A) le64_to_cpu(A)
1330 #error "LN2 of BITS_PER_LONG unknown!"
1334 /* 16MB sized 'bitmap extent' to track syncer usage */
1336 int rs_left
; /* number of bits set (out of sync) in this extent. */
1337 int rs_failed
; /* number of failed resync requests in this extent. */
1338 unsigned long flags
;
1339 struct lc_element lce
;
1342 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1343 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1344 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1348 * We need to store one bit for a block.
1349 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1350 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1351 * Bit 1 ==> local node thinks this block needs to be synced.
1354 #define SLEEP_TIME (HZ/10)
1356 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1357 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1358 /* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1359 * per sector of on disk bitmap */
1360 #define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */
1361 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1363 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1364 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1367 /* thus many _storage_ sectors are described by one bit */
1368 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1369 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1370 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1372 /* bit to represented kilo byte conversion */
1373 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1375 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1376 * _storage_ sector is located in */
1377 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1379 /* how much _storage_ sectors we have per bitmap sector */
1380 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1381 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1383 /* in one sector of the bitmap, we have this many activity_log extents. */
1384 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1385 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1387 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1388 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1390 /* the extent in "PER_EXTENT" below is an activity log extent
1391 * we need that many (long words/bytes) to store the bitmap
1392 * of one AL_EXTENT_SIZE chunk of storage.
1393 * we can store the bitmap for that many AL_EXTENTS within
1394 * one sector of the _on_disk_ bitmap:
1395 * bit 0 bit 37 bit 38 bit (512*8)-1
1396 * ...|........|........|.. // ..|........|
1397 * sect. 0 `296 `304 ^(512*8*8)-1
1399 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1400 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1401 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1404 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1405 #define DRBD_MAX_SECTORS_BM \
1406 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1407 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1408 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1409 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1410 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1411 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1412 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1414 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1415 /* 16 TB in units of sectors */
1416 #if BITS_PER_LONG == 32
1417 /* adjust by one page worth of bitmap,
1418 * so we won't wrap around in drbd_bm_find_next_bit.
1419 * you should use 64bit OS for that much storage, anyways. */
1420 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1422 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1423 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1424 /* corresponds to (1UL << 38) bits right now. */
1428 /* Sector shift value for the "hash" functions of tl_hash and ee_hash tables.
1429 * With a value of 8 all IO in one 128K block make it to the same slot of the
1432 #define DRBD_MAX_BIO_SIZE (1U<<(9+HT_SHIFT))
1433 #define DRBD_MAX_BIO_SIZE_SAFE (1 << 12) /* Works always = 4k */
1435 #define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */
1437 /* Number of elements in the app_reads_hash */
1438 #define APP_R_HSIZE 15
1440 extern int drbd_bm_init(struct drbd_conf
*mdev
);
1441 extern int drbd_bm_resize(struct drbd_conf
*mdev
, sector_t sectors
, int set_new_bits
);
1442 extern void drbd_bm_cleanup(struct drbd_conf
*mdev
);
1443 extern void drbd_bm_set_all(struct drbd_conf
*mdev
);
1444 extern void drbd_bm_clear_all(struct drbd_conf
*mdev
);
1445 /* set/clear/test only a few bits at a time */
1446 extern int drbd_bm_set_bits(
1447 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1448 extern int drbd_bm_clear_bits(
1449 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1450 extern int drbd_bm_count_bits(
1451 struct drbd_conf
*mdev
, const unsigned long s
, const unsigned long e
);
1452 /* bm_set_bits variant for use while holding drbd_bm_lock,
1453 * may process the whole bitmap in one go */
1454 extern void _drbd_bm_set_bits(struct drbd_conf
*mdev
,
1455 const unsigned long s
, const unsigned long e
);
1456 extern int drbd_bm_test_bit(struct drbd_conf
*mdev
, unsigned long bitnr
);
1457 extern int drbd_bm_e_weight(struct drbd_conf
*mdev
, unsigned long enr
);
1458 extern int drbd_bm_write_page(struct drbd_conf
*mdev
, unsigned int idx
) __must_hold(local
);
1459 extern int drbd_bm_read(struct drbd_conf
*mdev
) __must_hold(local
);
1460 extern int drbd_bm_write(struct drbd_conf
*mdev
) __must_hold(local
);
1461 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf
*mdev
,
1462 unsigned long al_enr
);
1463 extern size_t drbd_bm_words(struct drbd_conf
*mdev
);
1464 extern unsigned long drbd_bm_bits(struct drbd_conf
*mdev
);
1465 extern sector_t
drbd_bm_capacity(struct drbd_conf
*mdev
);
1467 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1468 extern unsigned long drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1469 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1470 extern unsigned long _drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1471 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1472 extern unsigned long _drbd_bm_total_weight(struct drbd_conf
*mdev
);
1473 extern unsigned long drbd_bm_total_weight(struct drbd_conf
*mdev
);
1474 extern int drbd_bm_rs_done(struct drbd_conf
*mdev
);
1475 /* for receive_bitmap */
1476 extern void drbd_bm_merge_lel(struct drbd_conf
*mdev
, size_t offset
,
1477 size_t number
, unsigned long *buffer
);
1478 /* for _drbd_send_bitmap */
1479 extern void drbd_bm_get_lel(struct drbd_conf
*mdev
, size_t offset
,
1480 size_t number
, unsigned long *buffer
);
1482 extern void drbd_bm_lock(struct drbd_conf
*mdev
, char *why
, enum bm_flag flags
);
1483 extern void drbd_bm_unlock(struct drbd_conf
*mdev
);
1486 extern struct kmem_cache
*drbd_request_cache
;
1487 extern struct kmem_cache
*drbd_ee_cache
; /* epoch entries */
1488 extern struct kmem_cache
*drbd_bm_ext_cache
; /* bitmap extents */
1489 extern struct kmem_cache
*drbd_al_ext_cache
; /* activity log extents */
1490 extern mempool_t
*drbd_request_mempool
;
1491 extern mempool_t
*drbd_ee_mempool
;
1493 extern struct page
*drbd_pp_pool
; /* drbd's page pool */
1494 extern spinlock_t drbd_pp_lock
;
1495 extern int drbd_pp_vacant
;
1496 extern wait_queue_head_t drbd_pp_wait
;
1498 extern rwlock_t global_state_lock
;
1500 extern struct drbd_conf
*drbd_new_device(unsigned int minor
);
1501 extern void drbd_free_mdev(struct drbd_conf
*mdev
);
1503 extern int proc_details
;
1506 extern int drbd_make_request(struct request_queue
*q
, struct bio
*bio
);
1507 extern int drbd_read_remote(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1508 extern int drbd_merge_bvec(struct request_queue
*q
, struct bvec_merge_data
*bvm
, struct bio_vec
*bvec
);
1509 extern int is_valid_ar_handle(struct drbd_request
*, sector_t
);
1513 extern void drbd_suspend_io(struct drbd_conf
*mdev
);
1514 extern void drbd_resume_io(struct drbd_conf
*mdev
);
1515 extern char *ppsize(char *buf
, unsigned long long size
);
1516 extern sector_t
drbd_new_dev_size(struct drbd_conf
*, struct drbd_backing_dev
*, int);
1517 enum determine_dev_size
{ dev_size_error
= -1, unchanged
= 0, shrunk
= 1, grew
= 2 };
1518 extern enum determine_dev_size
drbd_determine_dev_size(struct drbd_conf
*, enum dds_flags
) __must_hold(local
);
1519 extern void resync_after_online_grow(struct drbd_conf
*);
1520 extern void drbd_reconsider_max_bio_size(struct drbd_conf
*mdev
);
1521 extern enum drbd_state_rv
drbd_set_role(struct drbd_conf
*mdev
,
1522 enum drbd_role new_role
,
1524 extern enum drbd_disk_state
drbd_try_outdate_peer(struct drbd_conf
*mdev
);
1525 extern void drbd_try_outdate_peer_async(struct drbd_conf
*mdev
);
1526 extern int drbd_khelper(struct drbd_conf
*mdev
, char *cmd
);
1529 extern int drbd_worker(struct drbd_thread
*thi
);
1530 extern int drbd_alter_sa(struct drbd_conf
*mdev
, int na
);
1531 extern void drbd_start_resync(struct drbd_conf
*mdev
, enum drbd_conns side
);
1532 extern void resume_next_sg(struct drbd_conf
*mdev
);
1533 extern void suspend_other_sg(struct drbd_conf
*mdev
);
1534 extern int drbd_resync_finished(struct drbd_conf
*mdev
);
1535 /* maybe rather drbd_main.c ? */
1536 extern int drbd_md_sync_page_io(struct drbd_conf
*mdev
,
1537 struct drbd_backing_dev
*bdev
, sector_t sector
, int rw
);
1538 extern void drbd_ov_oos_found(struct drbd_conf
*, sector_t
, int);
1539 extern void drbd_rs_controller_reset(struct drbd_conf
*mdev
);
1541 static inline void ov_oos_print(struct drbd_conf
*mdev
)
1543 if (mdev
->ov_last_oos_size
) {
1544 dev_err(DEV
, "Out of sync: start=%llu, size=%lu (sectors)\n",
1545 (unsigned long long)mdev
->ov_last_oos_start
,
1546 (unsigned long)mdev
->ov_last_oos_size
);
1548 mdev
->ov_last_oos_size
=0;
1552 extern void drbd_csum_bio(struct drbd_conf
*, struct crypto_hash
*, struct bio
*, void *);
1553 extern void drbd_csum_ee(struct drbd_conf
*, struct crypto_hash
*, struct drbd_epoch_entry
*, void *);
1554 /* worker callbacks */
1555 extern int w_req_cancel_conflict(struct drbd_conf
*, struct drbd_work
*, int);
1556 extern int w_read_retry_remote(struct drbd_conf
*, struct drbd_work
*, int);
1557 extern int w_e_end_data_req(struct drbd_conf
*, struct drbd_work
*, int);
1558 extern int w_e_end_rsdata_req(struct drbd_conf
*, struct drbd_work
*, int);
1559 extern int w_e_end_csum_rs_req(struct drbd_conf
*, struct drbd_work
*, int);
1560 extern int w_e_end_ov_reply(struct drbd_conf
*, struct drbd_work
*, int);
1561 extern int w_e_end_ov_req(struct drbd_conf
*, struct drbd_work
*, int);
1562 extern int w_ov_finished(struct drbd_conf
*, struct drbd_work
*, int);
1563 extern int w_resync_timer(struct drbd_conf
*, struct drbd_work
*, int);
1564 extern int w_resume_next_sg(struct drbd_conf
*, struct drbd_work
*, int);
1565 extern int w_send_write_hint(struct drbd_conf
*, struct drbd_work
*, int);
1566 extern int w_send_dblock(struct drbd_conf
*, struct drbd_work
*, int);
1567 extern int w_send_barrier(struct drbd_conf
*, struct drbd_work
*, int);
1568 extern int w_send_read_req(struct drbd_conf
*, struct drbd_work
*, int);
1569 extern int w_prev_work_done(struct drbd_conf
*, struct drbd_work
*, int);
1570 extern int w_e_reissue(struct drbd_conf
*, struct drbd_work
*, int);
1571 extern int w_restart_disk_io(struct drbd_conf
*, struct drbd_work
*, int);
1572 extern int w_send_oos(struct drbd_conf
*, struct drbd_work
*, int);
1573 extern int w_start_resync(struct drbd_conf
*, struct drbd_work
*, int);
1575 extern void resync_timer_fn(unsigned long data
);
1576 extern void start_resync_timer_fn(unsigned long data
);
1578 /* drbd_receiver.c */
1579 extern int drbd_rs_should_slow_down(struct drbd_conf
*mdev
, sector_t sector
);
1580 extern int drbd_submit_ee(struct drbd_conf
*mdev
, struct drbd_epoch_entry
*e
,
1581 const unsigned rw
, const int fault_type
);
1582 extern int drbd_release_ee(struct drbd_conf
*mdev
, struct list_head
*list
);
1583 extern struct drbd_epoch_entry
*drbd_alloc_ee(struct drbd_conf
*mdev
,
1586 unsigned int data_size
,
1587 gfp_t gfp_mask
) __must_hold(local
);
1588 extern void drbd_free_some_ee(struct drbd_conf
*mdev
, struct drbd_epoch_entry
*e
,
1590 #define drbd_free_ee(m,e) drbd_free_some_ee(m, e, 0)
1591 #define drbd_free_net_ee(m,e) drbd_free_some_ee(m, e, 1)
1592 extern void drbd_wait_ee_list_empty(struct drbd_conf
*mdev
,
1593 struct list_head
*head
);
1594 extern void _drbd_wait_ee_list_empty(struct drbd_conf
*mdev
,
1595 struct list_head
*head
);
1596 extern void drbd_set_recv_tcq(struct drbd_conf
*mdev
, int tcq_enabled
);
1597 extern void _drbd_clear_done_ee(struct drbd_conf
*mdev
, struct list_head
*to_be_freed
);
1598 extern void drbd_flush_workqueue(struct drbd_conf
*mdev
);
1600 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1601 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1602 static inline int drbd_setsockopt(struct socket
*sock
, int level
, int optname
,
1603 char __user
*optval
, int optlen
)
1606 if (level
== SOL_SOCKET
)
1607 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
1609 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1614 static inline void drbd_tcp_cork(struct socket
*sock
)
1617 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1618 (char __user
*)&val
, sizeof(val
));
1621 static inline void drbd_tcp_uncork(struct socket
*sock
)
1624 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1625 (char __user
*)&val
, sizeof(val
));
1628 static inline void drbd_tcp_nodelay(struct socket
*sock
)
1631 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_NODELAY
,
1632 (char __user
*)&val
, sizeof(val
));
1635 static inline void drbd_tcp_quickack(struct socket
*sock
)
1638 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_QUICKACK
,
1639 (char __user
*)&val
, sizeof(val
));
1642 void drbd_bump_write_ordering(struct drbd_conf
*mdev
, enum write_ordering_e wo
);
1645 extern struct proc_dir_entry
*drbd_proc
;
1646 extern const struct file_operations drbd_proc_fops
;
1647 extern const char *drbd_conn_str(enum drbd_conns s
);
1648 extern const char *drbd_role_str(enum drbd_role s
);
1651 extern void drbd_al_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1652 extern void drbd_al_complete_io(struct drbd_conf
*mdev
, sector_t sector
);
1653 extern void drbd_rs_complete_io(struct drbd_conf
*mdev
, sector_t sector
);
1654 extern int drbd_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1655 extern int drbd_try_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1656 extern void drbd_rs_cancel_all(struct drbd_conf
*mdev
);
1657 extern int drbd_rs_del_all(struct drbd_conf
*mdev
);
1658 extern void drbd_rs_failed_io(struct drbd_conf
*mdev
,
1659 sector_t sector
, int size
);
1660 extern int drbd_al_read_log(struct drbd_conf
*mdev
, struct drbd_backing_dev
*);
1661 extern void drbd_advance_rs_marks(struct drbd_conf
*mdev
, unsigned long still_to_go
);
1662 extern void __drbd_set_in_sync(struct drbd_conf
*mdev
, sector_t sector
,
1663 int size
, const char *file
, const unsigned int line
);
1664 #define drbd_set_in_sync(mdev, sector, size) \
1665 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1666 extern int __drbd_set_out_of_sync(struct drbd_conf
*mdev
, sector_t sector
,
1667 int size
, const char *file
, const unsigned int line
);
1668 #define drbd_set_out_of_sync(mdev, sector, size) \
1669 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1670 extern void drbd_al_apply_to_bm(struct drbd_conf
*mdev
);
1671 extern void drbd_al_shrink(struct drbd_conf
*mdev
);
1676 void drbd_nl_cleanup(void);
1677 int __init
drbd_nl_init(void);
1678 void drbd_bcast_state(struct drbd_conf
*mdev
, union drbd_state
);
1679 void drbd_bcast_sync_progress(struct drbd_conf
*mdev
);
1680 void drbd_bcast_ee(struct drbd_conf
*mdev
,
1681 const char *reason
, const int dgs
,
1682 const char* seen_hash
, const char* calc_hash
,
1683 const struct drbd_epoch_entry
* e
);
1687 * DOC: DRBD State macros
1689 * These macros are used to express state changes in easily readable form.
1691 * The NS macros expand to a mask and a value, that can be bit ored onto the
1692 * current state as soon as the spinlock (req_lock) was taken.
1694 * The _NS macros are used for state functions that get called with the
1695 * spinlock. These macros expand directly to the new state value.
1697 * Besides the basic forms NS() and _NS() additional _?NS[23] are defined
1698 * to express state changes that affect more than one aspect of the state.
1700 * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY)
1701 * Means that the network connection was established and that the peer
1702 * is in secondary role.
1704 #define role_MASK R_MASK
1705 #define peer_MASK R_MASK
1706 #define disk_MASK D_MASK
1707 #define pdsk_MASK D_MASK
1708 #define conn_MASK C_MASK
1710 #define user_isp_MASK 1
1711 #define aftr_isp_MASK 1
1712 #define susp_nod_MASK 1
1713 #define susp_fen_MASK 1
1716 ({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \
1717 ({ union drbd_state val; val.i = 0; val.T = (S); val; })
1718 #define NS2(T1, S1, T2, S2) \
1719 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1720 mask.T2 = T2##_MASK; mask; }), \
1721 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1722 val.T2 = (S2); val; })
1723 #define NS3(T1, S1, T2, S2, T3, S3) \
1724 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1725 mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \
1726 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1727 val.T2 = (S2); val.T3 = (S3); val; })
1729 #define _NS(D, T, S) \
1730 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; })
1731 #define _NS2(D, T1, S1, T2, S2) \
1732 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1733 __ns.T2 = (S2); __ns; })
1734 #define _NS3(D, T1, S1, T2, S2, T3, S3) \
1735 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1736 __ns.T2 = (S2); __ns.T3 = (S3); __ns; })
1739 * inline helper functions
1740 *************************/
1742 /* see also page_chain_add and friends in drbd_receiver.c */
1743 static inline struct page
*page_chain_next(struct page
*page
)
1745 return (struct page
*)page_private(page
);
1747 #define page_chain_for_each(page) \
1748 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1749 page = page_chain_next(page))
1750 #define page_chain_for_each_safe(page, n) \
1751 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1753 static inline int drbd_bio_has_active_page(struct bio
*bio
)
1755 struct bio_vec
*bvec
;
1758 __bio_for_each_segment(bvec
, bio
, i
, 0) {
1759 if (page_count(bvec
->bv_page
) > 1)
1766 static inline int drbd_ee_has_active_page(struct drbd_epoch_entry
*e
)
1768 struct page
*page
= e
->pages
;
1769 page_chain_for_each(page
) {
1770 if (page_count(page
) > 1)
1777 static inline void drbd_state_lock(struct drbd_conf
*mdev
)
1779 wait_event(mdev
->misc_wait
,
1780 !test_and_set_bit(CLUSTER_ST_CHANGE
, &mdev
->flags
));
1783 static inline void drbd_state_unlock(struct drbd_conf
*mdev
)
1785 clear_bit(CLUSTER_ST_CHANGE
, &mdev
->flags
);
1786 wake_up(&mdev
->misc_wait
);
1789 static inline enum drbd_state_rv
1790 _drbd_set_state(struct drbd_conf
*mdev
, union drbd_state ns
,
1791 enum chg_state_flags flags
, struct completion
*done
)
1793 enum drbd_state_rv rv
;
1795 read_lock(&global_state_lock
);
1796 rv
= __drbd_set_state(mdev
, ns
, flags
, done
);
1797 read_unlock(&global_state_lock
);
1803 * drbd_request_state() - Reqest a state change
1804 * @mdev: DRBD device.
1805 * @mask: mask of state bits to change.
1806 * @val: value of new state bits.
1808 * This is the most graceful way of requesting a state change. It is verbose
1809 * quite verbose in case the state change is not possible, and all those
1810 * state changes are globally serialized.
1812 static inline int drbd_request_state(struct drbd_conf
*mdev
,
1813 union drbd_state mask
,
1814 union drbd_state val
)
1816 return _drbd_request_state(mdev
, mask
, val
, CS_VERBOSE
+ CS_ORDERED
);
1819 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1820 static inline void __drbd_chk_io_error_(struct drbd_conf
*mdev
, int forcedetach
, const char *where
)
1822 switch (mdev
->ldev
->dc
.on_io_error
) {
1825 if (__ratelimit(&drbd_ratelimit_state
))
1826 dev_err(DEV
, "Local IO failed in %s.\n", where
);
1827 if (mdev
->state
.disk
> D_INCONSISTENT
)
1828 _drbd_set_state(_NS(mdev
, disk
, D_INCONSISTENT
), CS_HARD
, NULL
);
1831 /* NOTE fall through to detach case if forcedetach set */
1833 case EP_CALL_HELPER
:
1834 set_bit(WAS_IO_ERROR
, &mdev
->flags
);
1835 if (mdev
->state
.disk
> D_FAILED
) {
1836 _drbd_set_state(_NS(mdev
, disk
, D_FAILED
), CS_HARD
, NULL
);
1838 "Local IO failed in %s. Detaching...\n", where
);
1845 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1846 * @mdev: DRBD device.
1847 * @error: Error code passed to the IO completion callback
1848 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1850 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1852 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1853 static inline void drbd_chk_io_error_(struct drbd_conf
*mdev
,
1854 int error
, int forcedetach
, const char *where
)
1857 unsigned long flags
;
1858 spin_lock_irqsave(&mdev
->req_lock
, flags
);
1859 __drbd_chk_io_error_(mdev
, forcedetach
, where
);
1860 spin_unlock_irqrestore(&mdev
->req_lock
, flags
);
1866 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1867 * @bdev: Meta data block device.
1869 * BTW, for internal meta data, this happens to be the maximum capacity
1870 * we could agree upon with our peer node.
1872 static inline sector_t
drbd_md_first_sector(struct drbd_backing_dev
*bdev
)
1874 switch (bdev
->dc
.meta_dev_idx
) {
1875 case DRBD_MD_INDEX_INTERNAL
:
1876 case DRBD_MD_INDEX_FLEX_INT
:
1877 return bdev
->md
.md_offset
+ bdev
->md
.bm_offset
;
1878 case DRBD_MD_INDEX_FLEX_EXT
:
1880 return bdev
->md
.md_offset
;
1885 * drbd_md_last_sector() - Return the last sector number of the meta data area
1886 * @bdev: Meta data block device.
1888 static inline sector_t
drbd_md_last_sector(struct drbd_backing_dev
*bdev
)
1890 switch (bdev
->dc
.meta_dev_idx
) {
1891 case DRBD_MD_INDEX_INTERNAL
:
1892 case DRBD_MD_INDEX_FLEX_INT
:
1893 return bdev
->md
.md_offset
+ MD_AL_OFFSET
- 1;
1894 case DRBD_MD_INDEX_FLEX_EXT
:
1896 return bdev
->md
.md_offset
+ bdev
->md
.md_size_sect
;
1900 /* Returns the number of 512 byte sectors of the device */
1901 static inline sector_t
drbd_get_capacity(struct block_device
*bdev
)
1903 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1904 return bdev
? i_size_read(bdev
->bd_inode
) >> 9 : 0;
1908 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1909 * @bdev: Meta data block device.
1911 * returns the capacity we announce to out peer. we clip ourselves at the
1912 * various MAX_SECTORS, because if we don't, current implementation will
1913 * oops sooner or later
1915 static inline sector_t
drbd_get_max_capacity(struct drbd_backing_dev
*bdev
)
1918 switch (bdev
->dc
.meta_dev_idx
) {
1919 case DRBD_MD_INDEX_INTERNAL
:
1920 case DRBD_MD_INDEX_FLEX_INT
:
1921 s
= drbd_get_capacity(bdev
->backing_bdev
)
1922 ? min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1923 drbd_md_first_sector(bdev
))
1926 case DRBD_MD_INDEX_FLEX_EXT
:
1927 s
= min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1928 drbd_get_capacity(bdev
->backing_bdev
));
1929 /* clip at maximum size the meta device can support */
1930 s
= min_t(sector_t
, s
,
1931 BM_EXT_TO_SECT(bdev
->md
.md_size_sect
1932 - bdev
->md
.bm_offset
));
1935 s
= min_t(sector_t
, DRBD_MAX_SECTORS
,
1936 drbd_get_capacity(bdev
->backing_bdev
));
1942 * drbd_md_ss__() - Return the sector number of our meta data super block
1943 * @mdev: DRBD device.
1944 * @bdev: Meta data block device.
1946 static inline sector_t
drbd_md_ss__(struct drbd_conf
*mdev
,
1947 struct drbd_backing_dev
*bdev
)
1949 switch (bdev
->dc
.meta_dev_idx
) {
1950 default: /* external, some index */
1951 return MD_RESERVED_SECT
* bdev
->dc
.meta_dev_idx
;
1952 case DRBD_MD_INDEX_INTERNAL
:
1953 /* with drbd08, internal meta data is always "flexible" */
1954 case DRBD_MD_INDEX_FLEX_INT
:
1955 /* sizeof(struct md_on_disk_07) == 4k
1956 * position: last 4k aligned block of 4k size */
1957 if (!bdev
->backing_bdev
) {
1958 if (__ratelimit(&drbd_ratelimit_state
)) {
1959 dev_err(DEV
, "bdev->backing_bdev==NULL\n");
1964 return (drbd_get_capacity(bdev
->backing_bdev
) & ~7ULL)
1966 case DRBD_MD_INDEX_FLEX_EXT
:
1972 drbd_queue_work_front(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1974 unsigned long flags
;
1975 spin_lock_irqsave(&q
->q_lock
, flags
);
1976 list_add(&w
->list
, &q
->q
);
1977 up(&q
->s
); /* within the spinlock,
1978 see comment near end of drbd_worker() */
1979 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1983 drbd_queue_work(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1985 unsigned long flags
;
1986 spin_lock_irqsave(&q
->q_lock
, flags
);
1987 list_add_tail(&w
->list
, &q
->q
);
1988 up(&q
->s
); /* within the spinlock,
1989 see comment near end of drbd_worker() */
1990 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1993 static inline void wake_asender(struct drbd_conf
*mdev
)
1995 if (test_bit(SIGNAL_ASENDER
, &mdev
->flags
))
1996 force_sig(DRBD_SIG
, mdev
->asender
.task
);
1999 static inline void request_ping(struct drbd_conf
*mdev
)
2001 set_bit(SEND_PING
, &mdev
->flags
);
2005 static inline int drbd_send_short_cmd(struct drbd_conf
*mdev
,
2006 enum drbd_packets cmd
)
2008 struct p_header80 h
;
2009 return drbd_send_cmd(mdev
, USE_DATA_SOCKET
, cmd
, &h
, sizeof(h
));
2012 static inline int drbd_send_ping(struct drbd_conf
*mdev
)
2014 struct p_header80 h
;
2015 return drbd_send_cmd(mdev
, USE_META_SOCKET
, P_PING
, &h
, sizeof(h
));
2018 static inline int drbd_send_ping_ack(struct drbd_conf
*mdev
)
2020 struct p_header80 h
;
2021 return drbd_send_cmd(mdev
, USE_META_SOCKET
, P_PING_ACK
, &h
, sizeof(h
));
2024 static inline void drbd_thread_stop(struct drbd_thread
*thi
)
2026 _drbd_thread_stop(thi
, false, true);
2029 static inline void drbd_thread_stop_nowait(struct drbd_thread
*thi
)
2031 _drbd_thread_stop(thi
, false, false);
2034 static inline void drbd_thread_restart_nowait(struct drbd_thread
*thi
)
2036 _drbd_thread_stop(thi
, true, false);
2039 /* counts how many answer packets packets we expect from our peer,
2040 * for either explicit application requests,
2041 * or implicit barrier packets as necessary.
2044 * _req_mod(req, queue_for_net_write or queue_for_net_read);
2045 * it is much easier and equally valid to count what we queue for the
2046 * worker, even before it actually was queued or send.
2047 * (drbd_make_request_common; recovery path on read io-error)
2049 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
2050 * _req_mod(req, data_received)
2051 * [from receive_DataReply]
2052 * _req_mod(req, write_acked_by_peer or recv_acked_by_peer or neg_acked)
2053 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
2054 * for some reason it is NOT decreased in got_NegAck,
2055 * but in the resulting cleanup code from report_params.
2056 * we should try to remember the reason for that...
2057 * _req_mod(req, send_failed or send_canceled)
2058 * _req_mod(req, connection_lost_while_pending)
2059 * [from tl_clear_barrier]
2061 static inline void inc_ap_pending(struct drbd_conf
*mdev
)
2063 atomic_inc(&mdev
->ap_pending_cnt
);
2066 #define ERR_IF_CNT_IS_NEGATIVE(which) \
2067 if (atomic_read(&mdev->which) < 0) \
2068 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
2069 __func__ , __LINE__ , \
2070 atomic_read(&mdev->which))
2072 #define dec_ap_pending(mdev) do { \
2073 typecheck(struct drbd_conf *, mdev); \
2074 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \
2075 wake_up(&mdev->misc_wait); \
2076 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
2078 /* counts how many resync-related answers we still expect from the peer
2080 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
2081 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
2082 * (or P_NEG_ACK with ID_SYNCER)
2084 static inline void inc_rs_pending(struct drbd_conf
*mdev
)
2086 atomic_inc(&mdev
->rs_pending_cnt
);
2089 #define dec_rs_pending(mdev) do { \
2090 typecheck(struct drbd_conf *, mdev); \
2091 atomic_dec(&mdev->rs_pending_cnt); \
2092 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
2094 /* counts how many answers we still need to send to the peer.
2096 * receive_Data unless protocol A;
2097 * we need to send a P_RECV_ACK (proto B)
2098 * or P_WRITE_ACK (proto C)
2099 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2100 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2101 * receive_Barrier_* we need to send a P_BARRIER_ACK
2103 static inline void inc_unacked(struct drbd_conf
*mdev
)
2105 atomic_inc(&mdev
->unacked_cnt
);
2108 #define dec_unacked(mdev) do { \
2109 typecheck(struct drbd_conf *, mdev); \
2110 atomic_dec(&mdev->unacked_cnt); \
2111 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2113 #define sub_unacked(mdev, n) do { \
2114 typecheck(struct drbd_conf *, mdev); \
2115 atomic_sub(n, &mdev->unacked_cnt); \
2116 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2119 static inline void put_net_conf(struct drbd_conf
*mdev
)
2121 if (atomic_dec_and_test(&mdev
->net_cnt
))
2122 wake_up(&mdev
->net_cnt_wait
);
2126 * get_net_conf() - Increase ref count on mdev->net_conf; Returns 0 if nothing there
2127 * @mdev: DRBD device.
2129 * You have to call put_net_conf() when finished working with mdev->net_conf.
2131 static inline int get_net_conf(struct drbd_conf
*mdev
)
2135 atomic_inc(&mdev
->net_cnt
);
2136 have_net_conf
= mdev
->state
.conn
>= C_UNCONNECTED
;
2139 return have_net_conf
;
2143 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2146 * You have to call put_ldev() when finished working with mdev->ldev.
2148 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2149 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2151 static inline void put_ldev(struct drbd_conf
*mdev
)
2153 int i
= atomic_dec_return(&mdev
->local_cnt
);
2155 /* This may be called from some endio handler,
2156 * so we must not sleep here. */
2161 if (mdev
->state
.disk
== D_DISKLESS
)
2162 /* even internal references gone, safe to destroy */
2163 drbd_ldev_destroy(mdev
);
2164 if (mdev
->state
.disk
== D_FAILED
)
2165 /* all application IO references gone. */
2166 drbd_go_diskless(mdev
);
2167 wake_up(&mdev
->misc_wait
);
2172 static inline int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
)
2176 /* never get a reference while D_DISKLESS */
2177 if (mdev
->state
.disk
== D_DISKLESS
)
2180 atomic_inc(&mdev
->local_cnt
);
2181 io_allowed
= (mdev
->state
.disk
>= mins
);
2187 extern int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
);
2190 /* you must have an "get_ldev" reference */
2191 static inline void drbd_get_syncer_progress(struct drbd_conf
*mdev
,
2192 unsigned long *bits_left
, unsigned int *per_mil_done
)
2194 /* this is to break it at compile time when we change that, in case we
2195 * want to support more than (1<<32) bits on a 32bit arch. */
2196 typecheck(unsigned long, mdev
->rs_total
);
2198 /* note: both rs_total and rs_left are in bits, i.e. in
2199 * units of BM_BLOCK_SIZE.
2200 * for the percentage, we don't care. */
2202 if (mdev
->state
.conn
== C_VERIFY_S
|| mdev
->state
.conn
== C_VERIFY_T
)
2203 *bits_left
= mdev
->ov_left
;
2205 *bits_left
= drbd_bm_total_weight(mdev
) - mdev
->rs_failed
;
2206 /* >> 10 to prevent overflow,
2207 * +1 to prevent division by zero */
2208 if (*bits_left
> mdev
->rs_total
) {
2209 /* doh. maybe a logic bug somewhere.
2210 * may also be just a race condition
2211 * between this and a disconnect during sync.
2212 * for now, just prevent in-kernel buffer overflow.
2215 dev_warn(DEV
, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2216 drbd_conn_str(mdev
->state
.conn
),
2217 *bits_left
, mdev
->rs_total
, mdev
->rs_failed
);
2220 /* Make sure the division happens in long context.
2221 * We allow up to one petabyte storage right now,
2222 * at a granularity of 4k per bit that is 2**38 bits.
2223 * After shift right and multiplication by 1000,
2224 * this should still fit easily into a 32bit long,
2225 * so we don't need a 64bit division on 32bit arch.
2226 * Note: currently we don't support such large bitmaps on 32bit
2227 * arch anyways, but no harm done to be prepared for it here.
2229 unsigned int shift
= mdev
->rs_total
>= (1ULL << 32) ? 16 : 10;
2230 unsigned long left
= *bits_left
>> shift
;
2231 unsigned long total
= 1UL + (mdev
->rs_total
>> shift
);
2232 unsigned long tmp
= 1000UL - left
* 1000UL/total
;
2233 *per_mil_done
= tmp
;
2238 /* this throttles on-the-fly application requests
2239 * according to max_buffers settings;
2240 * maybe re-implement using semaphores? */
2241 static inline int drbd_get_max_buffers(struct drbd_conf
*mdev
)
2243 int mxb
= 1000000; /* arbitrary limit on open requests */
2244 if (get_net_conf(mdev
)) {
2245 mxb
= mdev
->net_conf
->max_buffers
;
2251 static inline int drbd_state_is_stable(struct drbd_conf
*mdev
)
2253 union drbd_state s
= mdev
->state
;
2255 /* DO NOT add a default clause, we want the compiler to warn us
2256 * for any newly introduced state we may have forgotten to add here */
2258 switch ((enum drbd_conns
)s
.conn
) {
2259 /* new io only accepted when there is no connection, ... */
2261 case C_WF_CONNECTION
:
2262 /* ... or there is a well established connection. */
2268 case C_PAUSED_SYNC_S
:
2269 case C_PAUSED_SYNC_T
:
2272 /* transitional states, IO allowed */
2273 case C_DISCONNECTING
:
2277 case C_NETWORK_FAILURE
:
2278 case C_PROTOCOL_ERROR
:
2280 case C_WF_REPORT_PARAMS
:
2281 case C_STARTING_SYNC_S
:
2282 case C_STARTING_SYNC_T
:
2285 /* Allow IO in BM exchange states with new protocols */
2287 if (mdev
->agreed_pro_version
< 96)
2291 /* no new io accepted in these states */
2293 case C_WF_SYNC_UUID
:
2299 switch ((enum drbd_disk_state
)s
.disk
) {
2301 case D_INCONSISTENT
:
2305 /* disk state is stable as well. */
2308 /* no new io accepted during tansitional states */
2321 static inline int is_susp(union drbd_state s
)
2323 return s
.susp
|| s
.susp_nod
|| s
.susp_fen
;
2326 static inline bool may_inc_ap_bio(struct drbd_conf
*mdev
)
2328 int mxb
= drbd_get_max_buffers(mdev
);
2330 if (is_susp(mdev
->state
))
2332 if (test_bit(SUSPEND_IO
, &mdev
->flags
))
2335 /* to avoid potential deadlock or bitmap corruption,
2336 * in various places, we only allow new application io
2337 * to start during "stable" states. */
2339 /* no new io accepted when attaching or detaching the disk */
2340 if (!drbd_state_is_stable(mdev
))
2343 /* since some older kernels don't have atomic_add_unless,
2344 * and we are within the spinlock anyways, we have this workaround. */
2345 if (atomic_read(&mdev
->ap_bio_cnt
) > mxb
)
2347 if (test_bit(BITMAP_IO
, &mdev
->flags
))
2352 static inline bool inc_ap_bio_cond(struct drbd_conf
*mdev
, int count
)
2356 spin_lock_irq(&mdev
->req_lock
);
2357 rv
= may_inc_ap_bio(mdev
);
2359 atomic_add(count
, &mdev
->ap_bio_cnt
);
2360 spin_unlock_irq(&mdev
->req_lock
);
2365 static inline void inc_ap_bio(struct drbd_conf
*mdev
, int count
)
2368 * as long as the device is suspended
2369 * until the bitmap is no longer on the fly during connection
2370 * handshake as long as we would exeed the max_buffer limit.
2372 * to avoid races with the reconnect code,
2373 * we need to atomic_inc within the spinlock. */
2375 wait_event(mdev
->misc_wait
, inc_ap_bio_cond(mdev
, count
));
2378 static inline void dec_ap_bio(struct drbd_conf
*mdev
)
2380 int mxb
= drbd_get_max_buffers(mdev
);
2381 int ap_bio
= atomic_dec_return(&mdev
->ap_bio_cnt
);
2383 D_ASSERT(ap_bio
>= 0);
2384 /* this currently does wake_up for every dec_ap_bio!
2385 * maybe rather introduce some type of hysteresis?
2386 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2388 wake_up(&mdev
->misc_wait
);
2389 if (ap_bio
== 0 && test_bit(BITMAP_IO
, &mdev
->flags
)) {
2390 if (!test_and_set_bit(BITMAP_IO_QUEUED
, &mdev
->flags
))
2391 drbd_queue_work(&mdev
->data
.work
, &mdev
->bm_io_work
.w
);
2395 static inline int drbd_set_ed_uuid(struct drbd_conf
*mdev
, u64 val
)
2397 int changed
= mdev
->ed_uuid
!= val
;
2398 mdev
->ed_uuid
= val
;
2402 static inline int seq_cmp(u32 a
, u32 b
)
2404 /* we assume wrap around at 32bit.
2405 * for wrap around at 24bit (old atomic_t),
2409 return (s32
)(a
) - (s32
)(b
);
2411 #define seq_lt(a, b) (seq_cmp((a), (b)) < 0)
2412 #define seq_gt(a, b) (seq_cmp((a), (b)) > 0)
2413 #define seq_ge(a, b) (seq_cmp((a), (b)) >= 0)
2414 #define seq_le(a, b) (seq_cmp((a), (b)) <= 0)
2415 /* CAUTION: please no side effects in arguments! */
2416 #define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b)))
2418 static inline void update_peer_seq(struct drbd_conf
*mdev
, unsigned int new_seq
)
2421 spin_lock(&mdev
->peer_seq_lock
);
2422 m
= seq_max(mdev
->peer_seq
, new_seq
);
2424 spin_unlock(&mdev
->peer_seq_lock
);
2426 wake_up(&mdev
->seq_wait
);
2429 static inline void drbd_update_congested(struct drbd_conf
*mdev
)
2431 struct sock
*sk
= mdev
->data
.socket
->sk
;
2432 if (sk
->sk_wmem_queued
> sk
->sk_sndbuf
* 4 / 5)
2433 set_bit(NET_CONGESTED
, &mdev
->flags
);
2436 static inline int drbd_queue_order_type(struct drbd_conf
*mdev
)
2438 /* sorry, we currently have no working implementation
2439 * of distributed TCQ stuff */
2440 #ifndef QUEUE_ORDERED_NONE
2441 #define QUEUE_ORDERED_NONE 0
2443 return QUEUE_ORDERED_NONE
;
2446 static inline void drbd_md_flush(struct drbd_conf
*mdev
)
2450 if (test_bit(MD_NO_FUA
, &mdev
->flags
))
2453 r
= blkdev_issue_flush(mdev
->ldev
->md_bdev
, GFP_KERNEL
, NULL
);
2455 set_bit(MD_NO_FUA
, &mdev
->flags
);
2456 dev_err(DEV
, "meta data flush failed with status %d, disabling md-flushes\n", r
);