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/list.h>
32 #include <linux/sched.h>
33 #include <linux/bitops.h>
34 #include <linux/slab.h>
35 #include <linux/crypto.h>
36 #include <linux/ratelimit.h>
37 #include <linux/tcp.h>
38 #include <linux/mutex.h>
39 #include <linux/major.h>
40 #include <linux/blkdev.h>
41 #include <linux/genhd.h>
42 #include <linux/idr.h>
44 #include <linux/lru_cache.h>
45 #include <linux/prefetch.h>
46 #include <linux/drbd_genl_api.h>
47 #include <linux/drbd.h>
48 #include "drbd_state.h"
51 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
52 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
53 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
54 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
56 # define __protected_by(x)
57 # define __protected_read_by(x)
58 # define __protected_write_by(x)
59 # define __must_hold(x)
62 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
64 /* module parameter, defined in drbd_main.c */
65 extern unsigned int minor_count
;
66 extern bool disable_sendpage
;
67 extern bool allow_oos
;
68 void tl_abort_disk_io(struct drbd_conf
*mdev
);
70 #ifdef CONFIG_DRBD_FAULT_INJECTION
71 extern int enable_faults
;
72 extern int fault_rate
;
73 extern int fault_devs
;
76 extern char usermode_helper
[];
79 /* I don't remember why XCPU ...
80 * This is used to wake the asender,
81 * and to interrupt sending the sending task
84 #define DRBD_SIG SIGXCPU
86 /* This is used to stop/restart our threads.
87 * Cannot use SIGTERM nor SIGKILL, since these
88 * are sent out by init on runlevel changes
89 * I choose SIGHUP for now.
91 #define DRBD_SIGKILL SIGHUP
93 #define ID_IN_SYNC (4711ULL)
94 #define ID_OUT_OF_SYNC (4712ULL)
95 #define ID_SYNCER (-1ULL)
97 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
103 /* to shorten dev_warn(DEV, "msg"); and relatives statements */
104 #define DEV (disk_to_dev(mdev->vdisk))
106 #define conn_printk(LEVEL, TCONN, FMT, ARGS...) \
107 printk(LEVEL "d-con %s: " FMT, TCONN->name , ## ARGS)
108 #define conn_alert(TCONN, FMT, ARGS...) conn_printk(KERN_ALERT, TCONN, FMT, ## ARGS)
109 #define conn_crit(TCONN, FMT, ARGS...) conn_printk(KERN_CRIT, TCONN, FMT, ## ARGS)
110 #define conn_err(TCONN, FMT, ARGS...) conn_printk(KERN_ERR, TCONN, FMT, ## ARGS)
111 #define conn_warn(TCONN, FMT, ARGS...) conn_printk(KERN_WARNING, TCONN, FMT, ## ARGS)
112 #define conn_notice(TCONN, FMT, ARGS...) conn_printk(KERN_NOTICE, TCONN, FMT, ## ARGS)
113 #define conn_info(TCONN, FMT, ARGS...) conn_printk(KERN_INFO, TCONN, FMT, ## ARGS)
114 #define conn_dbg(TCONN, FMT, ARGS...) conn_printk(KERN_DEBUG, TCONN, FMT, ## ARGS)
116 #define D_ASSERT(exp) if (!(exp)) \
117 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
120 * expect - Make an assertion
122 * Unlike the assert macro, this macro returns a boolean result.
124 #define expect(exp) ({ \
125 bool _bool = (exp); \
127 dev_err(DEV, "ASSERTION %s FAILED in %s\n", \
132 /* Defines to control fault insertion */
134 DRBD_FAULT_MD_WR
= 0, /* meta data write */
135 DRBD_FAULT_MD_RD
= 1, /* read */
136 DRBD_FAULT_RS_WR
= 2, /* resync */
137 DRBD_FAULT_RS_RD
= 3,
138 DRBD_FAULT_DT_WR
= 4, /* data */
139 DRBD_FAULT_DT_RD
= 5,
140 DRBD_FAULT_DT_RA
= 6, /* data read ahead */
141 DRBD_FAULT_BM_ALLOC
= 7, /* bitmap allocation */
142 DRBD_FAULT_AL_EE
= 8, /* alloc ee */
143 DRBD_FAULT_RECEIVE
= 9, /* Changes some bytes upon receiving a [rs]data block */
149 _drbd_insert_fault(struct drbd_conf
*mdev
, unsigned int type
);
152 drbd_insert_fault(struct drbd_conf
*mdev
, unsigned int type
) {
153 #ifdef CONFIG_DRBD_FAULT_INJECTION
155 (enable_faults
& (1<<type
)) &&
156 _drbd_insert_fault(mdev
, type
);
162 /* integer division, round _UP_ to the next integer */
163 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
164 /* usual integer division */
165 #define div_floor(A, B) ((A)/(B))
167 extern struct ratelimit_state drbd_ratelimit_state
;
168 extern struct idr minors
; /* RCU, updates: genl_lock() */
169 extern struct list_head drbd_tconns
; /* RCU, updates: genl_lock() */
173 /* receiver (data socket) */
175 P_DATA_REPLY
= 0x01, /* Response to P_DATA_REQUEST */
176 P_RS_DATA_REPLY
= 0x02, /* Response to P_RS_DATA_REQUEST */
179 P_BECOME_SYNC_TARGET
= 0x05,
180 P_BECOME_SYNC_SOURCE
= 0x06,
181 P_UNPLUG_REMOTE
= 0x07, /* Used at various times to hint the peer */
182 P_DATA_REQUEST
= 0x08, /* Used to ask for a data block */
183 P_RS_DATA_REQUEST
= 0x09, /* Used to ask for a data block for resync */
190 P_AUTH_CHALLENGE
= 0x10,
191 P_AUTH_RESPONSE
= 0x11,
192 P_STATE_CHG_REQ
= 0x12,
194 /* asender (meta socket */
197 P_RECV_ACK
= 0x15, /* Used in protocol B */
198 P_WRITE_ACK
= 0x16, /* Used in protocol C */
199 P_RS_WRITE_ACK
= 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
200 P_SUPERSEDED
= 0x18, /* Used in proto C, two-primaries conflict detection */
201 P_NEG_ACK
= 0x19, /* Sent if local disk is unusable */
202 P_NEG_DREPLY
= 0x1a, /* Local disk is broken... */
203 P_NEG_RS_DREPLY
= 0x1b, /* Local disk is broken... */
204 P_BARRIER_ACK
= 0x1c,
205 P_STATE_CHG_REPLY
= 0x1d,
207 /* "new" commands, no longer fitting into the ordering scheme above */
209 P_OV_REQUEST
= 0x1e, /* data socket */
211 P_OV_RESULT
= 0x20, /* meta socket */
212 P_CSUM_RS_REQUEST
= 0x21, /* data socket */
213 P_RS_IS_IN_SYNC
= 0x22, /* meta socket */
214 P_SYNC_PARAM89
= 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
215 P_COMPRESSED_BITMAP
= 0x24, /* compressed or otherwise encoded bitmap transfer */
216 /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */
217 /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */
218 P_DELAY_PROBE
= 0x27, /* is used on BOTH sockets */
219 P_OUT_OF_SYNC
= 0x28, /* Mark as out of sync (Outrunning), data socket */
220 P_RS_CANCEL
= 0x29, /* meta: Used to cancel RS_DATA_REQUEST packet by SyncSource */
221 P_CONN_ST_CHG_REQ
= 0x2a, /* data sock: Connection wide state request */
222 P_CONN_ST_CHG_REPLY
= 0x2b, /* meta sock: Connection side state req reply */
223 P_RETRY_WRITE
= 0x2c, /* Protocol C: retry conflicting write request */
224 P_PROTOCOL_UPDATE
= 0x2d, /* data sock: is used in established connections */
226 P_MAY_IGNORE
= 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
227 P_MAX_OPT_CMD
= 0x101,
229 /* special command ids for handshake */
231 P_INITIAL_META
= 0xfff1, /* First Packet on the MetaSock */
232 P_INITIAL_DATA
= 0xfff2, /* First Packet on the Socket */
234 P_CONNECTION_FEATURES
= 0xfffe /* FIXED for the next century! */
237 extern const char *cmdname(enum drbd_packet cmd
);
239 /* for sending/receiving the bitmap,
240 * possibly in some encoding scheme */
243 * stores total bits and long words
244 * of the bitmap, so we don't need to
245 * call the accessor functions over and again. */
246 unsigned long bm_bits
;
247 unsigned long bm_words
;
248 /* during xfer, current position within the bitmap */
249 unsigned long bit_offset
;
250 unsigned long word_offset
;
252 /* statistics; index: (h->command == P_BITMAP) */
257 extern void INFO_bm_xfer_stats(struct drbd_conf
*mdev
,
258 const char *direction
, struct bm_xfer_ctx
*c
);
260 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx
*c
)
262 /* word_offset counts "native long words" (32 or 64 bit),
264 * Encoded packet may end at an unaligned bit offset.
265 * In case a fallback clear text packet is transmitted in
266 * between, we adjust this offset back to the last 64bit
267 * aligned "native long word", which makes coding and decoding
268 * the plain text bitmap much more convenient. */
269 #if BITS_PER_LONG == 64
270 c
->word_offset
= c
->bit_offset
>> 6;
271 #elif BITS_PER_LONG == 32
272 c
->word_offset
= c
->bit_offset
>> 5;
273 c
->word_offset
&= ~(1UL);
275 # error "unsupported BITS_PER_LONG"
280 #define __packed __attribute__((packed))
283 /* This is the layout for a packet on the wire.
284 * The byteorder is the network byte order.
285 * (except block_id and barrier fields.
286 * these are pointers to local structs
287 * and have no relevance for the partner,
288 * which just echoes them as received.)
290 * NOTE that the payload starts at a long aligned offset,
291 * regardless of 32 or 64 bit arch!
296 u16 length
; /* bytes of data after this header */
299 /* Header for big packets, Used for data packets exceeding 64kB */
301 u16 magic
; /* use DRBD_MAGIC_BIG here */
314 extern unsigned int drbd_header_size(struct drbd_tconn
*tconn
);
316 /* these defines must not be changed without changing the protocol version */
317 #define DP_HARDBARRIER 1 /* depricated */
318 #define DP_RW_SYNC 2 /* equals REQ_SYNC */
319 #define DP_MAY_SET_IN_SYNC 4
320 #define DP_UNPLUG 8 /* not used anymore */
321 #define DP_FUA 16 /* equals REQ_FUA */
322 #define DP_FLUSH 32 /* equals REQ_FLUSH */
323 #define DP_DISCARD 64 /* equals REQ_DISCARD */
324 #define DP_SEND_RECEIVE_ACK 128 /* This is a proto B write request */
325 #define DP_SEND_WRITE_ACK 256 /* This is a proto C write request */
328 u64 sector
; /* 64 bits sector number */
329 u64 block_id
; /* to identify the request in protocol B&C */
335 * commands which share a struct:
337 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
338 * P_SUPERSEDED (proto C, two-primaries conflict detection)
340 * P_DATA_REQUEST, P_RS_DATA_REQUEST
353 u32 pad
; /* to multiple of 8 Byte */
357 * commands with their own struct for additional fields:
358 * P_CONNECTION_FEATURES
365 struct p_connection_features
{
370 /* should be more than enough for future enhancements
371 * for now, feature_flags and the reserved array shall be zero.
379 u32 barrier
; /* barrier number _handle_ only */
380 u32 pad
; /* to multiple of 8 Byte */
383 struct p_barrier_ack
{
391 /* Since protocol version 88 and higher. */
395 struct p_rs_param_89
{
397 /* protocol version 89: */
398 char verify_alg
[SHARED_SECRET_MAX
];
399 char csums_alg
[SHARED_SECRET_MAX
];
402 struct p_rs_param_95
{
404 char verify_alg
[SHARED_SECRET_MAX
];
405 char csums_alg
[SHARED_SECRET_MAX
];
412 enum drbd_conn_flags
{
413 CF_DISCARD_MY_DATA
= 1,
425 /* Since protocol version 87 and higher. */
426 char integrity_alg
[0];
431 u64 uuid
[UI_EXTENDED_SIZE
];
439 u64 d_size
; /* size of disk */
440 u64 u_size
; /* user requested size */
441 u64 c_size
; /* current exported size */
442 u32 max_bio_size
; /* Maximal size of a BIO */
443 u16 queue_order_type
; /* not yet implemented in DRBD*/
444 u16 dds_flags
; /* use enum dds_flags here. */
456 struct p_req_state_reply
{
460 struct p_drbd06_param
{
470 struct p_block_desc
{
473 u32 pad
; /* to multiple of 8 Byte */
476 /* Valid values for the encoding field.
477 * Bump proto version when changing this. */
478 enum drbd_bitmap_code
{
479 /* RLE_VLI_Bytes = 0,
480 * and other bit variants had been defined during
481 * algorithm evaluation. */
485 struct p_compressed_bm
{
486 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
487 * (encoding & 0x80): polarity (set/unset) of first runlength
488 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
489 * used to pad up to head.length bytes
496 struct p_delay_probe93
{
497 u32 seq_num
; /* sequence number to match the two probe packets */
498 u32 offset
; /* usecs the probe got sent after the reference time point */
502 * Bitmap packets need to fit within a single page on the sender and receiver,
503 * so we are limited to 4 KiB (and not to PAGE_SIZE, which can be bigger).
505 #define DRBD_SOCKET_BUFFER_SIZE 4096
507 /**********************************************************************/
508 enum drbd_thread_state
{
517 struct task_struct
*task
;
518 struct completion stop
;
519 enum drbd_thread_state t_state
;
520 int (*function
) (struct drbd_thread
*);
521 struct drbd_tconn
*tconn
;
526 static inline enum drbd_thread_state
get_t_state(struct drbd_thread
*thi
)
528 /* THINK testing the t_state seems to be uncritical in all cases
529 * (but thread_{start,stop}), so we can read it *without* the lock.
537 struct list_head list
;
538 int (*cb
)(struct drbd_work
*, int cancel
);
540 struct drbd_conf
*mdev
;
541 struct drbd_tconn
*tconn
;
545 #include "drbd_interval.h"
547 extern int drbd_wait_misc(struct drbd_conf
*, struct drbd_interval
*);
549 struct drbd_request
{
552 /* if local IO is not allowed, will be NULL.
553 * if local IO _is_ allowed, holds the locally submitted bio clone,
554 * or, after local IO completion, the ERR_PTR(error).
555 * see drbd_request_endio(). */
556 struct bio
*private_bio
;
558 struct drbd_interval i
;
560 /* epoch: used to check on "completion" whether this req was in
561 * the current epoch, and we therefore have to close it,
562 * causing a p_barrier packet to be send, starting a new epoch.
564 * This corresponds to "barrier" in struct p_barrier[_ack],
565 * and to "barrier_nr" in struct drbd_epoch (and various
566 * comments/function parameters/local variable names).
570 struct list_head tl_requests
; /* ring list in the transfer log */
571 struct bio
*master_bio
; /* master bio pointer */
572 unsigned long start_time
;
574 /* once it hits 0, we may complete the master_bio */
575 atomic_t completion_ref
;
576 /* once it hits 0, we may destroy this drbd_request object */
579 unsigned rq_state
; /* see comments above _req_mod() */
583 struct drbd_tconn
*tconn
;
584 struct list_head list
;
585 unsigned int barrier_nr
;
586 atomic_t epoch_size
; /* increased on every request added. */
587 atomic_t active
; /* increased on every req. added, and dec on every finished. */
591 /* Prototype declaration of function defined in drbd_receiver.c */
592 int drbdd_init(struct drbd_thread
*);
593 int drbd_asender(struct drbd_thread
*);
595 /* drbd_epoch flag bits */
597 DE_HAVE_BARRIER_NUMBER
,
604 EV_CLEANUP
= 32, /* used as flag */
607 struct drbd_wq_barrier
{
609 struct completion done
;
617 struct drbd_peer_request
{
619 struct drbd_epoch
*epoch
; /* for writes */
621 atomic_t pending_bios
;
622 struct drbd_interval i
;
623 /* see comments on ee flag bits below */
627 struct digest_info
*digest
;
632 * While corresponding bios are in flight, the only modification will be
633 * set_bit WAS_ERROR, which has to be atomic.
634 * If no bios are in flight yet, or all have been completed,
635 * non-atomic modification to ee->flags is ok.
638 __EE_CALL_AL_COMPLETE_IO
,
639 __EE_MAY_SET_IN_SYNC
,
641 /* In case a barrier failed,
642 * we need to resubmit without the barrier flag. */
645 /* we may have several bios per peer request.
646 * if any of those fail, we set this flag atomically
647 * from the endio callback */
650 /* This ee has a pointer to a digest instead of a block id */
653 /* Conflicting local requests need to be restarted after this request */
654 __EE_RESTART_REQUESTS
,
656 /* The peer wants a write ACK for this (wire proto C) */
659 /* Is set when net_conf had two_primaries set while creating this peer_req */
660 __EE_IN_INTERVAL_TREE
,
662 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
663 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
664 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
665 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
666 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
667 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
668 #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK)
669 #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE)
671 /* flag bits per mdev */
673 UNPLUG_REMOTE
, /* sending a "UnplugRemote" could help */
674 MD_DIRTY
, /* current uuids and flags not yet on disk */
675 USE_DEGR_WFC_T
, /* degr-wfc-timeout instead of wfc-timeout. */
678 CRASHED_PRIMARY
, /* This node was a crashed primary.
679 * Gets cleared when the state.conn
680 * goes into C_CONNECTED state. */
683 MD_NO_FUA
, /* Users wants us to not use FUA/FLUSH on meta data dev */
684 SUSPEND_IO
, /* suspend application io */
685 BITMAP_IO
, /* suspend application io;
686 once no more io in flight, start bitmap io */
687 BITMAP_IO_QUEUED
, /* Started bitmap IO */
688 GO_DISKLESS
, /* Disk is being detached, on io-error or admin request. */
689 WAS_IO_ERROR
, /* Local disk failed, returned IO error */
690 WAS_READ_ERROR
, /* Local disk READ failed (set additionally to the above) */
691 FORCE_DETACH
, /* Force-detach from local disk, aborting any pending local IO */
692 RESYNC_AFTER_NEG
, /* Resync after online grow after the attach&negotiate finished. */
693 RESIZE_PENDING
, /* Size change detected locally, waiting for the response from
694 * the peer, if it changed there as well. */
695 NEW_CUR_UUID
, /* Create new current UUID when thawing IO */
696 AL_SUSPENDED
, /* Activity logging is currently suspended. */
697 AHEAD_TO_SYNC_SOURCE
, /* Ahead -> SyncSource queued */
698 B_RS_H_DONE
, /* Before resync handler done (already executed) */
699 DISCARD_MY_DATA
, /* discard_my_data flag per volume */
703 struct drbd_bitmap
; /* opaque for drbd_conf */
705 /* definition of bits in bm_flags to be used in drbd_bm_lock
706 * and drbd_bitmap_io and friends. */
708 /* do we need to kfree, or vfree bm_pages? */
709 BM_P_VMALLOCED
= 0x10000, /* internal use only, will be masked out */
711 /* currently locked for bulk operation */
712 BM_LOCKED_MASK
= 0xf,
714 /* in detail, that is: */
719 /* so we can mark it locked for bulk operation,
720 * and still allow all non-bulk operations */
723 /* (test bit, count bit) allowed (common case) */
724 BM_LOCKED_TEST_ALLOWED
= BM_DONT_CLEAR
| BM_DONT_SET
| BM_IS_LOCKED
,
726 /* testing bits, as well as setting new bits allowed, but clearing bits
727 * would be unexpected. Used during bitmap receive. Setting new bits
728 * requires sending of "out-of-sync" information, though. */
729 BM_LOCKED_SET_ALLOWED
= BM_DONT_CLEAR
| BM_IS_LOCKED
,
731 /* for drbd_bm_write_copy_pages, everything is allowed,
732 * only concurrent bulk operations are locked out. */
733 BM_LOCKED_CHANGE_ALLOWED
= BM_IS_LOCKED
,
736 struct drbd_work_queue
{
738 spinlock_t q_lock
; /* to protect the list. */
739 wait_queue_head_t q_wait
;
744 struct socket
*socket
;
745 /* this way we get our
746 * send/receive buffers off the stack */
752 u64 md_offset
; /* sector offset to 'super' block */
754 u64 la_size_sect
; /* last agreed size, unit sectors */
755 spinlock_t uuid_lock
;
761 s32 al_offset
; /* signed relative sector offset to activity log */
762 s32 bm_offset
; /* signed relative sector offset to bitmap */
764 /* cached value of bdev->disk_conf->meta_dev_idx (see below) */
767 /* see al_tr_number_to_on_disk_sector() */
769 u32 al_stripe_size_4k
;
770 u32 al_size_4k
; /* cached product of the above */
773 struct drbd_backing_dev
{
774 struct block_device
*backing_bdev
;
775 struct block_device
*md_bdev
;
777 struct disk_conf
*disk_conf
; /* RCU, for updates: mdev->tconn->conf_update */
778 sector_t known_size
; /* last known size of that backing device */
790 int (*io_fn
)(struct drbd_conf
*mdev
);
791 void (*done
)(struct drbd_conf
*mdev
, int rv
);
794 enum write_ordering_e
{
801 unsigned int head_index
;
803 int total
; /* sum of all values */
806 extern struct fifo_buffer
*fifo_alloc(int fifo_size
);
808 /* flag bits per tconn */
810 NET_CONGESTED
, /* The data socket is congested */
811 RESOLVE_CONFLICTS
, /* Set on one node, cleared on the peer! */
812 SEND_PING
, /* whether asender should send a ping asap */
813 SIGNAL_ASENDER
, /* whether asender wants to be interrupted */
814 GOT_PING_ACK
, /* set when we receive a ping_ack packet, ping_wait gets woken */
815 CONN_WD_ST_CHG_REQ
, /* A cluster wide state change on the connection is active */
818 CONN_DRY_RUN
, /* Expect disconnect after resync handshake. */
819 CREATE_BARRIER
, /* next P_DATA is preceded by a P_BARRIER */
820 STATE_SENT
, /* Do not change state/UUIDs while this is set */
821 CALLBACK_PENDING
, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
822 * pending, from drbd worker context.
823 * If set, bdi_write_congested() returns true,
824 * so shrink_page_list() would not recurse into,
825 * and potentially deadlock on, this drbd worker.
830 struct drbd_tconn
{ /* is a resource from the config file */
831 char *name
; /* Resource name */
832 struct list_head all_tconn
; /* linked on global drbd_tconns */
834 struct idr volumes
; /* <tconn, vnr> to mdev mapping */
835 enum drbd_conns cstate
; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
836 unsigned susp
:1; /* IO suspended by user */
837 unsigned susp_nod
:1; /* IO suspended because no data */
838 unsigned susp_fen
:1; /* IO suspended because fence peer handler runs */
839 struct mutex cstate_mutex
; /* Protects graceful disconnects */
840 unsigned int connect_cnt
; /* Inc each time a connection is established */
843 struct net_conf
*net_conf
; /* content protected by rcu */
844 struct mutex conf_update
; /* mutex for ready-copy-update of net_conf and disk_conf */
845 wait_queue_head_t ping_wait
; /* Woken upon reception of a ping, and a state change */
846 struct res_opts res_opts
;
848 struct sockaddr_storage my_addr
;
850 struct sockaddr_storage peer_addr
;
853 struct drbd_socket data
; /* data/barrier/cstate/parameter packets */
854 struct drbd_socket meta
; /* ping/ack (metadata) packets */
855 int agreed_pro_version
; /* actually used protocol version */
856 unsigned long last_received
; /* in jiffies, either socket */
857 unsigned int ko_count
;
861 struct list_head transfer_log
; /* all requests not yet fully processed */
863 struct crypto_hash
*cram_hmac_tfm
;
864 struct crypto_hash
*integrity_tfm
; /* checksums we compute, updates protected by tconn->data->mutex */
865 struct crypto_hash
*peer_integrity_tfm
; /* checksums we verify, only accessed from receiver thread */
866 struct crypto_hash
*csums_tfm
;
867 struct crypto_hash
*verify_tfm
;
872 struct drbd_epoch
*current_epoch
;
873 spinlock_t epoch_lock
;
875 enum write_ordering_e write_ordering
;
876 atomic_t current_tle_nr
; /* transfer log epoch number */
877 unsigned current_tle_writes
; /* writes seen within this tl epoch */
879 unsigned long last_reconnect_jif
;
880 struct drbd_thread receiver
;
881 struct drbd_thread worker
;
882 struct drbd_thread asender
;
883 cpumask_var_t cpu_mask
;
886 struct drbd_work_queue sender_work
;
889 /* whether this sender thread
890 * has processed a single write yet. */
891 bool seen_any_write_yet
;
893 /* Which barrier number to send with the next P_BARRIER */
894 int current_epoch_nr
;
896 /* how many write requests have been sent
897 * with req->epoch == current_epoch_nr.
898 * If none, no P_BARRIER will be sent. */
899 unsigned current_epoch_writes
;
903 struct submit_worker
{
904 struct workqueue_struct
*wq
;
905 struct work_struct worker
;
908 struct list_head writes
;
912 struct drbd_tconn
*tconn
;
913 int vnr
; /* volume number within the connection */
916 /* things that are stored as / read from meta data on disk */
919 /* configured by drbdsetup */
920 struct drbd_backing_dev
*ldev
__protected_by(local
);
922 sector_t p_size
; /* partner's disk size */
923 struct request_queue
*rq_queue
;
924 struct block_device
*this_bdev
;
925 struct gendisk
*vdisk
;
927 unsigned long last_reattach_jif
;
928 struct drbd_work resync_work
,
933 struct timer_list resync_timer
;
934 struct timer_list md_sync_timer
;
935 struct timer_list start_resync_timer
;
936 struct timer_list request_timer
;
937 #ifdef DRBD_DEBUG_MD_SYNC
941 } last_md_mark_dirty
;
944 /* Used after attach while negotiating new disk state. */
945 union drbd_state new_state_tmp
;
947 union drbd_dev_state state
;
948 wait_queue_head_t misc_wait
;
949 wait_queue_head_t state_wait
; /* upon each state change. */
950 unsigned int send_cnt
;
951 unsigned int recv_cnt
;
952 unsigned int read_cnt
;
953 unsigned int writ_cnt
;
954 unsigned int al_writ_cnt
;
955 unsigned int bm_writ_cnt
;
956 atomic_t ap_bio_cnt
; /* Requests we need to complete */
957 atomic_t ap_pending_cnt
; /* AP data packets on the wire, ack expected */
958 atomic_t rs_pending_cnt
; /* RS request/data packets on the wire */
959 atomic_t unacked_cnt
; /* Need to send replies for */
960 atomic_t local_cnt
; /* Waiting for local completion */
962 /* Interval tree of pending local requests */
963 struct rb_root read_requests
;
964 struct rb_root write_requests
;
966 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
967 unsigned long rs_total
;
968 /* number of resync blocks that failed in this run */
969 unsigned long rs_failed
;
970 /* Syncer's start time [unit jiffies] */
971 unsigned long rs_start
;
972 /* cumulated time in PausedSyncX state [unit jiffies] */
973 unsigned long rs_paused
;
974 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
975 unsigned long rs_same_csum
;
976 #define DRBD_SYNC_MARKS 8
977 #define DRBD_SYNC_MARK_STEP (3*HZ)
978 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
979 unsigned long rs_mark_left
[DRBD_SYNC_MARKS
];
980 /* marks's time [unit jiffies] */
981 unsigned long rs_mark_time
[DRBD_SYNC_MARKS
];
982 /* current index into rs_mark_{left,time} */
984 unsigned long rs_last_bcast
; /* [unit jiffies] */
986 /* where does the admin want us to start? (sector) */
987 sector_t ov_start_sector
;
988 sector_t ov_stop_sector
;
989 /* where are we now? (sector) */
990 sector_t ov_position
;
991 /* Start sector of out of sync range (to merge printk reporting). */
992 sector_t ov_last_oos_start
;
993 /* size of out-of-sync range in sectors. */
994 sector_t ov_last_oos_size
;
995 unsigned long ov_left
; /* in bits */
997 struct drbd_bitmap
*bitmap
;
998 unsigned long bm_resync_fo
; /* bit offset for drbd_bm_find_next */
1000 /* Used to track operations of resync... */
1001 struct lru_cache
*resync
;
1002 /* Number of locked elements in resync LRU */
1003 unsigned int resync_locked
;
1004 /* resync extent number waiting for application requests */
1005 unsigned int resync_wenr
;
1010 struct list_head active_ee
; /* IO in progress (P_DATA gets written to disk) */
1011 struct list_head sync_ee
; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
1012 struct list_head done_ee
; /* need to send P_WRITE_ACK */
1013 struct list_head read_ee
; /* [RS]P_DATA_REQUEST being read */
1014 struct list_head net_ee
; /* zero-copy network send in progress */
1016 int next_barrier_nr
;
1017 struct list_head resync_reads
;
1018 atomic_t pp_in_use
; /* allocated from page pool */
1019 atomic_t pp_in_use_by_net
; /* sendpage()d, still referenced by tcp */
1020 wait_queue_head_t ee_wait
;
1021 struct page
*md_io_page
; /* one page buffer for md_io */
1022 struct drbd_md_io md_io
;
1023 atomic_t md_io_in_use
; /* protects the md_io, md_io_page and md_io_tmpp */
1025 wait_queue_head_t al_wait
;
1026 struct lru_cache
*act_log
; /* activity log */
1027 unsigned int al_tr_number
;
1029 wait_queue_head_t seq_wait
;
1030 atomic_t packet_seq
;
1031 unsigned int peer_seq
;
1032 spinlock_t peer_seq_lock
;
1034 unsigned long comm_bm_set
; /* communicated number of set bits. */
1035 struct bm_io_work bm_io_work
;
1036 u64 ed_uuid
; /* UUID of the exposed data */
1037 struct mutex own_state_mutex
;
1038 struct mutex
*state_mutex
; /* either own_state_mutex or mdev->tconn->cstate_mutex */
1039 char congestion_reason
; /* Why we where congested... */
1040 atomic_t rs_sect_in
; /* for incoming resync data rate, SyncTarget */
1041 atomic_t rs_sect_ev
; /* for submitted resync data rate, both */
1042 int rs_last_sect_ev
; /* counter to compare with */
1043 int rs_last_events
; /* counter of read or write "events" (unit sectors)
1044 * on the lower level device when we last looked. */
1045 int c_sync_rate
; /* current resync rate after syncer throttle magic */
1046 struct fifo_buffer
*rs_plan_s
; /* correction values of resync planer (RCU, tconn->conn_update) */
1047 int rs_in_flight
; /* resync sectors in flight (to proxy, in proxy and from proxy) */
1048 atomic_t ap_in_flight
; /* App sectors in flight (waiting for ack) */
1049 unsigned int peer_max_bio_size
;
1050 unsigned int local_max_bio_size
;
1052 /* any requests that would block in drbd_make_request()
1053 * are deferred to this single-threaded work queue */
1054 struct submit_worker submit
;
1057 static inline struct drbd_conf
*minor_to_mdev(unsigned int minor
)
1059 return (struct drbd_conf
*)idr_find(&minors
, minor
);
1062 static inline unsigned int mdev_to_minor(struct drbd_conf
*mdev
)
1067 static inline struct drbd_conf
*vnr_to_mdev(struct drbd_tconn
*tconn
, int vnr
)
1069 return (struct drbd_conf
*)idr_find(&tconn
->volumes
, vnr
);
1073 * function declarations
1074 *************************/
1080 DDSF_NO_RESYNC
= 2, /* Do not run a resync for the new space */
1083 extern void drbd_init_set_defaults(struct drbd_conf
*mdev
);
1084 extern int drbd_thread_start(struct drbd_thread
*thi
);
1085 extern void _drbd_thread_stop(struct drbd_thread
*thi
, int restart
, int wait
);
1086 extern char *drbd_task_to_thread_name(struct drbd_tconn
*tconn
, struct task_struct
*task
);
1088 extern void drbd_thread_current_set_cpu(struct drbd_thread
*thi
);
1089 extern void drbd_calc_cpu_mask(struct drbd_tconn
*tconn
);
1091 #define drbd_thread_current_set_cpu(A) ({})
1092 #define drbd_calc_cpu_mask(A) ({})
1094 extern void tl_release(struct drbd_tconn
*, unsigned int barrier_nr
,
1095 unsigned int set_size
);
1096 extern void tl_clear(struct drbd_tconn
*);
1097 extern void drbd_free_sock(struct drbd_tconn
*tconn
);
1098 extern int drbd_send(struct drbd_tconn
*tconn
, struct socket
*sock
,
1099 void *buf
, size_t size
, unsigned msg_flags
);
1100 extern int drbd_send_all(struct drbd_tconn
*, struct socket
*, void *, size_t,
1103 extern int __drbd_send_protocol(struct drbd_tconn
*tconn
, enum drbd_packet cmd
);
1104 extern int drbd_send_protocol(struct drbd_tconn
*tconn
);
1105 extern int drbd_send_uuids(struct drbd_conf
*mdev
);
1106 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf
*mdev
);
1107 extern void drbd_gen_and_send_sync_uuid(struct drbd_conf
*mdev
);
1108 extern int drbd_send_sizes(struct drbd_conf
*mdev
, int trigger_reply
, enum dds_flags flags
);
1109 extern int drbd_send_state(struct drbd_conf
*mdev
, union drbd_state s
);
1110 extern int drbd_send_current_state(struct drbd_conf
*mdev
);
1111 extern int drbd_send_sync_param(struct drbd_conf
*mdev
);
1112 extern void drbd_send_b_ack(struct drbd_tconn
*tconn
, u32 barrier_nr
,
1114 extern int drbd_send_ack(struct drbd_conf
*, enum drbd_packet
,
1115 struct drbd_peer_request
*);
1116 extern void drbd_send_ack_rp(struct drbd_conf
*mdev
, enum drbd_packet cmd
,
1117 struct p_block_req
*rp
);
1118 extern void drbd_send_ack_dp(struct drbd_conf
*mdev
, enum drbd_packet cmd
,
1119 struct p_data
*dp
, int data_size
);
1120 extern int drbd_send_ack_ex(struct drbd_conf
*mdev
, enum drbd_packet cmd
,
1121 sector_t sector
, int blksize
, u64 block_id
);
1122 extern int drbd_send_out_of_sync(struct drbd_conf
*, struct drbd_request
*);
1123 extern int drbd_send_block(struct drbd_conf
*, enum drbd_packet
,
1124 struct drbd_peer_request
*);
1125 extern int drbd_send_dblock(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1126 extern int drbd_send_drequest(struct drbd_conf
*mdev
, int cmd
,
1127 sector_t sector
, int size
, u64 block_id
);
1128 extern int drbd_send_drequest_csum(struct drbd_conf
*mdev
, sector_t sector
,
1129 int size
, void *digest
, int digest_size
,
1130 enum drbd_packet cmd
);
1131 extern int drbd_send_ov_request(struct drbd_conf
*mdev
,sector_t sector
,int size
);
1133 extern int drbd_send_bitmap(struct drbd_conf
*mdev
);
1134 extern void drbd_send_sr_reply(struct drbd_conf
*mdev
, enum drbd_state_rv retcode
);
1135 extern void conn_send_sr_reply(struct drbd_tconn
*tconn
, enum drbd_state_rv retcode
);
1136 extern void drbd_free_bc(struct drbd_backing_dev
*ldev
);
1137 extern void drbd_mdev_cleanup(struct drbd_conf
*mdev
);
1138 void drbd_print_uuids(struct drbd_conf
*mdev
, const char *text
);
1140 extern void conn_md_sync(struct drbd_tconn
*tconn
);
1141 extern void drbd_md_write(struct drbd_conf
*mdev
, void *buffer
);
1142 extern void drbd_md_sync(struct drbd_conf
*mdev
);
1143 extern int drbd_md_read(struct drbd_conf
*mdev
, struct drbd_backing_dev
*bdev
);
1144 extern void drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1145 extern void _drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1146 extern void drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1147 extern void drbd_uuid_set_bm(struct drbd_conf
*mdev
, u64 val
) __must_hold(local
);
1148 extern void drbd_uuid_move_history(struct drbd_conf
*mdev
) __must_hold(local
);
1149 extern void __drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1150 extern void drbd_md_set_flag(struct drbd_conf
*mdev
, int flags
) __must_hold(local
);
1151 extern void drbd_md_clear_flag(struct drbd_conf
*mdev
, int flags
)__must_hold(local
);
1152 extern int drbd_md_test_flag(struct drbd_backing_dev
*, int);
1153 #ifndef DRBD_DEBUG_MD_SYNC
1154 extern void drbd_md_mark_dirty(struct drbd_conf
*mdev
);
1156 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1157 extern void drbd_md_mark_dirty_(struct drbd_conf
*mdev
,
1158 unsigned int line
, const char *func
);
1160 extern void drbd_queue_bitmap_io(struct drbd_conf
*mdev
,
1161 int (*io_fn
)(struct drbd_conf
*),
1162 void (*done
)(struct drbd_conf
*, int),
1163 char *why
, enum bm_flag flags
);
1164 extern int drbd_bitmap_io(struct drbd_conf
*mdev
,
1165 int (*io_fn
)(struct drbd_conf
*),
1166 char *why
, enum bm_flag flags
);
1167 extern int drbd_bitmap_io_from_worker(struct drbd_conf
*mdev
,
1168 int (*io_fn
)(struct drbd_conf
*),
1169 char *why
, enum bm_flag flags
);
1170 extern int drbd_bmio_set_n_write(struct drbd_conf
*mdev
);
1171 extern int drbd_bmio_clear_n_write(struct drbd_conf
*mdev
);
1172 extern void drbd_ldev_destroy(struct drbd_conf
*mdev
);
1176 * We currently have two possible layouts.
1177 * Offsets in (512 byte) sectors.
1179 * |----------- md_size_sect ------------------|
1180 * [ 4k superblock ][ activity log ][ Bitmap ]
1181 * | al_offset == 8 |
1182 * | bm_offset = al_offset + X |
1183 * ==> bitmap sectors = md_size_sect - bm_offset
1186 * old, indexed fixed size meta data:
1189 * |----------- md_size_sect ------------------|
1190 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*]
1192 * | bm_offset = al_offset - Y |
1193 * ==> bitmap sectors = Y = al_offset - bm_offset
1195 * [padding*] are zero or up to 7 unused 512 Byte sectors to the
1196 * end of the device, so that the [4k superblock] will be 4k aligned.
1198 * The activity log consists of 4k transaction blocks,
1199 * which are written in a ring-buffer, or striped ring-buffer like fashion,
1200 * which are writtensize used to be fixed 32kB,
1201 * but is about to become configurable.
1204 /* Our old fixed size meta data layout
1205 * allows up to about 3.8TB, so if you want more,
1206 * you need to use the "flexible" meta data format. */
1207 #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */
1208 #define MD_4kB_SECT 8
1209 #define MD_32kB_SECT 64
1211 /* One activity log extent represents 4M of storage */
1212 #define AL_EXTENT_SHIFT 22
1213 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1215 /* We could make these currently hardcoded constants configurable
1216 * variables at create-md time (or even re-configurable at runtime?).
1217 * Which will require some more changes to the DRBD "super block"
1220 * updates per transaction:
1221 * This many changes to the active set can be logged with one transaction.
1222 * This number is arbitrary.
1223 * context per transaction:
1224 * This many context extent numbers are logged with each transaction.
1225 * This number is resulting from the transaction block size (4k), the layout
1226 * of the transaction header, and the number of updates per transaction.
1227 * See drbd_actlog.c:struct al_transaction_on_disk
1229 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1230 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1232 #if BITS_PER_LONG == 32
1234 #define cpu_to_lel(A) cpu_to_le32(A)
1235 #define lel_to_cpu(A) le32_to_cpu(A)
1236 #elif BITS_PER_LONG == 64
1238 #define cpu_to_lel(A) cpu_to_le64(A)
1239 #define lel_to_cpu(A) le64_to_cpu(A)
1241 #error "LN2 of BITS_PER_LONG unknown!"
1245 /* 16MB sized 'bitmap extent' to track syncer usage */
1247 int rs_left
; /* number of bits set (out of sync) in this extent. */
1248 int rs_failed
; /* number of failed resync requests in this extent. */
1249 unsigned long flags
;
1250 struct lc_element lce
;
1253 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1254 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1255 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1259 * We need to store one bit for a block.
1260 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1261 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1262 * Bit 1 ==> local node thinks this block needs to be synced.
1265 #define SLEEP_TIME (HZ/10)
1267 /* We do bitmap IO in units of 4k blocks.
1268 * We also still have a hardcoded 4k per bit relation. */
1269 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1270 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1271 /* mostly arbitrarily set the represented size of one bitmap extent,
1272 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1273 * at 4k per bit resolution) */
1274 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
1275 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1277 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1278 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1281 /* thus many _storage_ sectors are described by one bit */
1282 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1283 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1284 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1286 /* bit to represented kilo byte conversion */
1287 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1289 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1290 * _storage_ sector is located in */
1291 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1293 /* how much _storage_ sectors we have per bitmap sector */
1294 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1295 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1297 /* in one sector of the bitmap, we have this many activity_log extents. */
1298 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1300 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1301 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1303 /* the extent in "PER_EXTENT" below is an activity log extent
1304 * we need that many (long words/bytes) to store the bitmap
1305 * of one AL_EXTENT_SIZE chunk of storage.
1306 * we can store the bitmap for that many AL_EXTENTS within
1307 * one sector of the _on_disk_ bitmap:
1308 * bit 0 bit 37 bit 38 bit (512*8)-1
1309 * ...|........|........|.. // ..|........|
1310 * sect. 0 `296 `304 ^(512*8*8)-1
1312 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1313 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1314 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1317 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1318 /* we have a certain meta data variant that has a fixed on-disk size of 128
1319 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity
1320 * log, leaving this many sectors for the bitmap.
1323 #define DRBD_MAX_SECTORS_FIXED_BM \
1324 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9)))
1325 #if !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1326 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1327 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1329 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM
1330 /* 16 TB in units of sectors */
1331 #if BITS_PER_LONG == 32
1332 /* adjust by one page worth of bitmap,
1333 * so we won't wrap around in drbd_bm_find_next_bit.
1334 * you should use 64bit OS for that much storage, anyways. */
1335 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1337 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1338 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1339 /* corresponds to (1UL << 38) bits right now. */
1343 /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE,
1344 * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte.
1345 * Since we may live in a mixed-platform cluster,
1346 * we limit us to a platform agnostic constant here for now.
1347 * A followup commit may allow even bigger BIO sizes,
1348 * once we thought that through. */
1349 #define DRBD_MAX_BIO_SIZE (1U << 20)
1350 #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1351 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1353 #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */
1355 #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */
1356 #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */
1358 extern int drbd_bm_init(struct drbd_conf
*mdev
);
1359 extern int drbd_bm_resize(struct drbd_conf
*mdev
, sector_t sectors
, int set_new_bits
);
1360 extern void drbd_bm_cleanup(struct drbd_conf
*mdev
);
1361 extern void drbd_bm_set_all(struct drbd_conf
*mdev
);
1362 extern void drbd_bm_clear_all(struct drbd_conf
*mdev
);
1363 /* set/clear/test only a few bits at a time */
1364 extern int drbd_bm_set_bits(
1365 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1366 extern int drbd_bm_clear_bits(
1367 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1368 extern int drbd_bm_count_bits(
1369 struct drbd_conf
*mdev
, const unsigned long s
, const unsigned long e
);
1370 /* bm_set_bits variant for use while holding drbd_bm_lock,
1371 * may process the whole bitmap in one go */
1372 extern void _drbd_bm_set_bits(struct drbd_conf
*mdev
,
1373 const unsigned long s
, const unsigned long e
);
1374 extern int drbd_bm_test_bit(struct drbd_conf
*mdev
, unsigned long bitnr
);
1375 extern int drbd_bm_e_weight(struct drbd_conf
*mdev
, unsigned long enr
);
1376 extern int drbd_bm_write_page(struct drbd_conf
*mdev
, unsigned int idx
) __must_hold(local
);
1377 extern int drbd_bm_read(struct drbd_conf
*mdev
) __must_hold(local
);
1378 extern void drbd_bm_mark_for_writeout(struct drbd_conf
*mdev
, int page_nr
);
1379 extern int drbd_bm_write(struct drbd_conf
*mdev
) __must_hold(local
);
1380 extern int drbd_bm_write_hinted(struct drbd_conf
*mdev
) __must_hold(local
);
1381 extern int drbd_bm_write_all(struct drbd_conf
*mdev
) __must_hold(local
);
1382 extern int drbd_bm_write_copy_pages(struct drbd_conf
*mdev
) __must_hold(local
);
1383 extern size_t drbd_bm_words(struct drbd_conf
*mdev
);
1384 extern unsigned long drbd_bm_bits(struct drbd_conf
*mdev
);
1385 extern sector_t
drbd_bm_capacity(struct drbd_conf
*mdev
);
1387 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1388 extern unsigned long drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1389 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1390 extern unsigned long _drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1391 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1392 extern unsigned long _drbd_bm_total_weight(struct drbd_conf
*mdev
);
1393 extern unsigned long drbd_bm_total_weight(struct drbd_conf
*mdev
);
1394 extern int drbd_bm_rs_done(struct drbd_conf
*mdev
);
1395 /* for receive_bitmap */
1396 extern void drbd_bm_merge_lel(struct drbd_conf
*mdev
, size_t offset
,
1397 size_t number
, unsigned long *buffer
);
1398 /* for _drbd_send_bitmap */
1399 extern void drbd_bm_get_lel(struct drbd_conf
*mdev
, size_t offset
,
1400 size_t number
, unsigned long *buffer
);
1402 extern void drbd_bm_lock(struct drbd_conf
*mdev
, char *why
, enum bm_flag flags
);
1403 extern void drbd_bm_unlock(struct drbd_conf
*mdev
);
1406 extern struct kmem_cache
*drbd_request_cache
;
1407 extern struct kmem_cache
*drbd_ee_cache
; /* peer requests */
1408 extern struct kmem_cache
*drbd_bm_ext_cache
; /* bitmap extents */
1409 extern struct kmem_cache
*drbd_al_ext_cache
; /* activity log extents */
1410 extern mempool_t
*drbd_request_mempool
;
1411 extern mempool_t
*drbd_ee_mempool
;
1413 /* drbd's page pool, used to buffer data received from the peer,
1414 * or data requested by the peer.
1416 * This does not have an emergency reserve.
1418 * When allocating from this pool, it first takes pages from the pool.
1419 * Only if the pool is depleted will try to allocate from the system.
1421 * The assumption is that pages taken from this pool will be processed,
1422 * and given back, "quickly", and then can be recycled, so we can avoid
1423 * frequent calls to alloc_page(), and still will be able to make progress even
1424 * under memory pressure.
1426 extern struct page
*drbd_pp_pool
;
1427 extern spinlock_t drbd_pp_lock
;
1428 extern int drbd_pp_vacant
;
1429 extern wait_queue_head_t drbd_pp_wait
;
1431 /* We also need a standard (emergency-reserve backed) page pool
1432 * for meta data IO (activity log, bitmap).
1433 * We can keep it global, as long as it is used as "N pages at a time".
1434 * 128 should be plenty, currently we probably can get away with as few as 1.
1436 #define DRBD_MIN_POOL_PAGES 128
1437 extern mempool_t
*drbd_md_io_page_pool
;
1439 /* We also need to make sure we get a bio
1440 * when we need it for housekeeping purposes */
1441 extern struct bio_set
*drbd_md_io_bio_set
;
1442 /* to allocate from that set */
1443 extern struct bio
*bio_alloc_drbd(gfp_t gfp_mask
);
1445 extern rwlock_t global_state_lock
;
1447 extern int conn_lowest_minor(struct drbd_tconn
*tconn
);
1448 enum drbd_ret_code
conn_new_minor(struct drbd_tconn
*tconn
, unsigned int minor
, int vnr
);
1449 extern void drbd_minor_destroy(struct kref
*kref
);
1451 extern int set_resource_options(struct drbd_tconn
*tconn
, struct res_opts
*res_opts
);
1452 extern struct drbd_tconn
*conn_create(const char *name
, struct res_opts
*res_opts
);
1453 extern void conn_destroy(struct kref
*kref
);
1454 struct drbd_tconn
*conn_get_by_name(const char *name
);
1455 extern struct drbd_tconn
*conn_get_by_addrs(void *my_addr
, int my_addr_len
,
1456 void *peer_addr
, int peer_addr_len
);
1457 extern void conn_free_crypto(struct drbd_tconn
*tconn
);
1459 extern int proc_details
;
1462 extern void do_submit(struct work_struct
*ws
);
1463 extern void __drbd_make_request(struct drbd_conf
*, struct bio
*, unsigned long);
1464 extern void drbd_make_request(struct request_queue
*q
, struct bio
*bio
);
1465 extern int drbd_read_remote(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1466 extern int drbd_merge_bvec(struct request_queue
*q
, struct bvec_merge_data
*bvm
, struct bio_vec
*bvec
);
1467 extern int is_valid_ar_handle(struct drbd_request
*, sector_t
);
1471 extern int drbd_msg_put_info(const char *info
);
1472 extern void drbd_suspend_io(struct drbd_conf
*mdev
);
1473 extern void drbd_resume_io(struct drbd_conf
*mdev
);
1474 extern char *ppsize(char *buf
, unsigned long long size
);
1475 extern sector_t
drbd_new_dev_size(struct drbd_conf
*, struct drbd_backing_dev
*, sector_t
, int);
1476 enum determine_dev_size
{
1477 DS_ERROR_SHRINK
= -3,
1478 DS_ERROR_SPACE_MD
= -2,
1483 DS_GREW_FROM_ZERO
= 3,
1485 extern enum determine_dev_size
1486 drbd_determine_dev_size(struct drbd_conf
*, enum dds_flags
, struct resize_parms
*) __must_hold(local
);
1487 extern void resync_after_online_grow(struct drbd_conf
*);
1488 extern void drbd_reconsider_max_bio_size(struct drbd_conf
*mdev
);
1489 extern enum drbd_state_rv
drbd_set_role(struct drbd_conf
*mdev
,
1490 enum drbd_role new_role
,
1492 extern bool conn_try_outdate_peer(struct drbd_tconn
*tconn
);
1493 extern void conn_try_outdate_peer_async(struct drbd_tconn
*tconn
);
1494 extern int drbd_khelper(struct drbd_conf
*mdev
, char *cmd
);
1497 extern int drbd_worker(struct drbd_thread
*thi
);
1498 enum drbd_ret_code
drbd_resync_after_valid(struct drbd_conf
*mdev
, int o_minor
);
1499 void drbd_resync_after_changed(struct drbd_conf
*mdev
);
1500 extern void drbd_start_resync(struct drbd_conf
*mdev
, enum drbd_conns side
);
1501 extern void resume_next_sg(struct drbd_conf
*mdev
);
1502 extern void suspend_other_sg(struct drbd_conf
*mdev
);
1503 extern int drbd_resync_finished(struct drbd_conf
*mdev
);
1504 /* maybe rather drbd_main.c ? */
1505 extern void *drbd_md_get_buffer(struct drbd_conf
*mdev
);
1506 extern void drbd_md_put_buffer(struct drbd_conf
*mdev
);
1507 extern int drbd_md_sync_page_io(struct drbd_conf
*mdev
,
1508 struct drbd_backing_dev
*bdev
, sector_t sector
, int rw
);
1509 extern void drbd_ov_out_of_sync_found(struct drbd_conf
*, sector_t
, int);
1510 extern void wait_until_done_or_force_detached(struct drbd_conf
*mdev
,
1511 struct drbd_backing_dev
*bdev
, unsigned int *done
);
1512 extern void drbd_rs_controller_reset(struct drbd_conf
*mdev
);
1514 static inline void ov_out_of_sync_print(struct drbd_conf
*mdev
)
1516 if (mdev
->ov_last_oos_size
) {
1517 dev_err(DEV
, "Out of sync: start=%llu, size=%lu (sectors)\n",
1518 (unsigned long long)mdev
->ov_last_oos_start
,
1519 (unsigned long)mdev
->ov_last_oos_size
);
1521 mdev
->ov_last_oos_size
=0;
1525 extern void drbd_csum_bio(struct drbd_conf
*, struct crypto_hash
*, struct bio
*, void *);
1526 extern void drbd_csum_ee(struct drbd_conf
*, struct crypto_hash
*,
1527 struct drbd_peer_request
*, void *);
1528 /* worker callbacks */
1529 extern int w_e_end_data_req(struct drbd_work
*, int);
1530 extern int w_e_end_rsdata_req(struct drbd_work
*, int);
1531 extern int w_e_end_csum_rs_req(struct drbd_work
*, int);
1532 extern int w_e_end_ov_reply(struct drbd_work
*, int);
1533 extern int w_e_end_ov_req(struct drbd_work
*, int);
1534 extern int w_ov_finished(struct drbd_work
*, int);
1535 extern int w_resync_timer(struct drbd_work
*, int);
1536 extern int w_send_write_hint(struct drbd_work
*, int);
1537 extern int w_make_resync_request(struct drbd_work
*, int);
1538 extern int w_send_dblock(struct drbd_work
*, int);
1539 extern int w_send_read_req(struct drbd_work
*, int);
1540 extern int w_prev_work_done(struct drbd_work
*, int);
1541 extern int w_e_reissue(struct drbd_work
*, int);
1542 extern int w_restart_disk_io(struct drbd_work
*, int);
1543 extern int w_send_out_of_sync(struct drbd_work
*, int);
1544 extern int w_start_resync(struct drbd_work
*, int);
1546 extern void resync_timer_fn(unsigned long data
);
1547 extern void start_resync_timer_fn(unsigned long data
);
1549 /* drbd_receiver.c */
1550 extern int drbd_rs_should_slow_down(struct drbd_conf
*mdev
, sector_t sector
);
1551 extern int drbd_submit_peer_request(struct drbd_conf
*,
1552 struct drbd_peer_request
*, const unsigned,
1554 extern int drbd_free_peer_reqs(struct drbd_conf
*, struct list_head
*);
1555 extern struct drbd_peer_request
*drbd_alloc_peer_req(struct drbd_conf
*, u64
,
1556 sector_t
, unsigned int,
1557 gfp_t
) __must_hold(local
);
1558 extern void __drbd_free_peer_req(struct drbd_conf
*, struct drbd_peer_request
*,
1560 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1561 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
1562 extern struct page
*drbd_alloc_pages(struct drbd_conf
*, unsigned int, bool);
1563 extern void drbd_set_recv_tcq(struct drbd_conf
*mdev
, int tcq_enabled
);
1564 extern void _drbd_clear_done_ee(struct drbd_conf
*mdev
, struct list_head
*to_be_freed
);
1565 extern void conn_flush_workqueue(struct drbd_tconn
*tconn
);
1566 extern int drbd_connected(struct drbd_conf
*mdev
);
1567 static inline void drbd_flush_workqueue(struct drbd_conf
*mdev
)
1569 conn_flush_workqueue(mdev
->tconn
);
1572 /* Yes, there is kernel_setsockopt, but only since 2.6.18.
1573 * So we have our own copy of it here. */
1574 static inline int drbd_setsockopt(struct socket
*sock
, int level
, int optname
,
1575 char *optval
, int optlen
)
1577 mm_segment_t oldfs
= get_fs();
1578 char __user
*uoptval
;
1581 uoptval
= (char __user __force
*)optval
;
1584 if (level
== SOL_SOCKET
)
1585 err
= sock_setsockopt(sock
, level
, optname
, uoptval
, optlen
);
1587 err
= sock
->ops
->setsockopt(sock
, level
, optname
, uoptval
,
1593 static inline void drbd_tcp_cork(struct socket
*sock
)
1596 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1597 (char*)&val
, sizeof(val
));
1600 static inline void drbd_tcp_uncork(struct socket
*sock
)
1603 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1604 (char*)&val
, sizeof(val
));
1607 static inline void drbd_tcp_nodelay(struct socket
*sock
)
1610 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_NODELAY
,
1611 (char*)&val
, sizeof(val
));
1614 static inline void drbd_tcp_quickack(struct socket
*sock
)
1617 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_QUICKACK
,
1618 (char*)&val
, sizeof(val
));
1621 void drbd_bump_write_ordering(struct drbd_tconn
*tconn
, enum write_ordering_e wo
);
1624 extern struct proc_dir_entry
*drbd_proc
;
1625 extern const struct file_operations drbd_proc_fops
;
1626 extern const char *drbd_conn_str(enum drbd_conns s
);
1627 extern const char *drbd_role_str(enum drbd_role s
);
1630 extern int drbd_al_begin_io_nonblock(struct drbd_conf
*mdev
, struct drbd_interval
*i
);
1631 extern void drbd_al_begin_io_commit(struct drbd_conf
*mdev
, bool delegate
);
1632 extern bool drbd_al_begin_io_fastpath(struct drbd_conf
*mdev
, struct drbd_interval
*i
);
1633 extern void drbd_al_begin_io(struct drbd_conf
*mdev
, struct drbd_interval
*i
, bool delegate
);
1634 extern void drbd_al_complete_io(struct drbd_conf
*mdev
, struct drbd_interval
*i
);
1635 extern void drbd_rs_complete_io(struct drbd_conf
*mdev
, sector_t sector
);
1636 extern int drbd_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1637 extern int drbd_try_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1638 extern void drbd_rs_cancel_all(struct drbd_conf
*mdev
);
1639 extern int drbd_rs_del_all(struct drbd_conf
*mdev
);
1640 extern void drbd_rs_failed_io(struct drbd_conf
*mdev
,
1641 sector_t sector
, int size
);
1642 extern void drbd_advance_rs_marks(struct drbd_conf
*mdev
, unsigned long still_to_go
);
1643 extern void __drbd_set_in_sync(struct drbd_conf
*mdev
, sector_t sector
,
1644 int size
, const char *file
, const unsigned int line
);
1645 #define drbd_set_in_sync(mdev, sector, size) \
1646 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1647 extern int __drbd_set_out_of_sync(struct drbd_conf
*mdev
, sector_t sector
,
1648 int size
, const char *file
, const unsigned int line
);
1649 #define drbd_set_out_of_sync(mdev, sector, size) \
1650 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1651 extern void drbd_al_shrink(struct drbd_conf
*mdev
);
1652 extern int drbd_initialize_al(struct drbd_conf
*, void *);
1655 /* state info broadcast */
1657 enum drbd_state_info_bcast_reason sib_reason
;
1661 unsigned helper_exit_code
;
1664 union drbd_state os
;
1665 union drbd_state ns
;
1669 void drbd_bcast_event(struct drbd_conf
*mdev
, const struct sib_info
*sib
);
1672 * inline helper functions
1673 *************************/
1675 /* see also page_chain_add and friends in drbd_receiver.c */
1676 static inline struct page
*page_chain_next(struct page
*page
)
1678 return (struct page
*)page_private(page
);
1680 #define page_chain_for_each(page) \
1681 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1682 page = page_chain_next(page))
1683 #define page_chain_for_each_safe(page, n) \
1684 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1687 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request
*peer_req
)
1689 struct page
*page
= peer_req
->pages
;
1690 page_chain_for_each(page
) {
1691 if (page_count(page
) > 1)
1697 static inline enum drbd_state_rv
1698 _drbd_set_state(struct drbd_conf
*mdev
, union drbd_state ns
,
1699 enum chg_state_flags flags
, struct completion
*done
)
1701 enum drbd_state_rv rv
;
1703 read_lock(&global_state_lock
);
1704 rv
= __drbd_set_state(mdev
, ns
, flags
, done
);
1705 read_unlock(&global_state_lock
);
1710 static inline union drbd_state
drbd_read_state(struct drbd_conf
*mdev
)
1712 union drbd_state rv
;
1714 rv
.i
= mdev
->state
.i
;
1715 rv
.susp
= mdev
->tconn
->susp
;
1716 rv
.susp_nod
= mdev
->tconn
->susp_nod
;
1717 rv
.susp_fen
= mdev
->tconn
->susp_fen
;
1722 enum drbd_force_detach_flags
{
1729 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1730 static inline void __drbd_chk_io_error_(struct drbd_conf
*mdev
,
1731 enum drbd_force_detach_flags df
,
1734 enum drbd_io_error_p ep
;
1737 ep
= rcu_dereference(mdev
->ldev
->disk_conf
)->on_io_error
;
1740 case EP_PASS_ON
: /* FIXME would this be better named "Ignore"? */
1741 if (df
== DRBD_READ_ERROR
|| df
== DRBD_WRITE_ERROR
) {
1742 if (__ratelimit(&drbd_ratelimit_state
))
1743 dev_err(DEV
, "Local IO failed in %s.\n", where
);
1744 if (mdev
->state
.disk
> D_INCONSISTENT
)
1745 _drbd_set_state(_NS(mdev
, disk
, D_INCONSISTENT
), CS_HARD
, NULL
);
1748 /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */
1750 case EP_CALL_HELPER
:
1751 /* Remember whether we saw a READ or WRITE error.
1753 * Recovery of the affected area for WRITE failure is covered
1754 * by the activity log.
1755 * READ errors may fall outside that area though. Certain READ
1756 * errors can be "healed" by writing good data to the affected
1757 * blocks, which triggers block re-allocation in lower layers.
1759 * If we can not write the bitmap after a READ error,
1760 * we may need to trigger a full sync (see w_go_diskless()).
1762 * Force-detach is not really an IO error, but rather a
1763 * desperate measure to try to deal with a completely
1764 * unresponsive lower level IO stack.
1765 * Still it should be treated as a WRITE error.
1767 * Meta IO error is always WRITE error:
1768 * we read meta data only once during attach,
1769 * which will fail in case of errors.
1771 set_bit(WAS_IO_ERROR
, &mdev
->flags
);
1772 if (df
== DRBD_READ_ERROR
)
1773 set_bit(WAS_READ_ERROR
, &mdev
->flags
);
1774 if (df
== DRBD_FORCE_DETACH
)
1775 set_bit(FORCE_DETACH
, &mdev
->flags
);
1776 if (mdev
->state
.disk
> D_FAILED
) {
1777 _drbd_set_state(_NS(mdev
, disk
, D_FAILED
), CS_HARD
, NULL
);
1779 "Local IO failed in %s. Detaching...\n", where
);
1786 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1787 * @mdev: DRBD device.
1788 * @error: Error code passed to the IO completion callback
1789 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1791 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1793 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1794 static inline void drbd_chk_io_error_(struct drbd_conf
*mdev
,
1795 int error
, enum drbd_force_detach_flags forcedetach
, const char *where
)
1798 unsigned long flags
;
1799 spin_lock_irqsave(&mdev
->tconn
->req_lock
, flags
);
1800 __drbd_chk_io_error_(mdev
, forcedetach
, where
);
1801 spin_unlock_irqrestore(&mdev
->tconn
->req_lock
, flags
);
1807 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1808 * @bdev: Meta data block device.
1810 * BTW, for internal meta data, this happens to be the maximum capacity
1811 * we could agree upon with our peer node.
1813 static inline sector_t
drbd_md_first_sector(struct drbd_backing_dev
*bdev
)
1815 switch (bdev
->md
.meta_dev_idx
) {
1816 case DRBD_MD_INDEX_INTERNAL
:
1817 case DRBD_MD_INDEX_FLEX_INT
:
1818 return bdev
->md
.md_offset
+ bdev
->md
.bm_offset
;
1819 case DRBD_MD_INDEX_FLEX_EXT
:
1821 return bdev
->md
.md_offset
;
1826 * drbd_md_last_sector() - Return the last sector number of the meta data area
1827 * @bdev: Meta data block device.
1829 static inline sector_t
drbd_md_last_sector(struct drbd_backing_dev
*bdev
)
1831 switch (bdev
->md
.meta_dev_idx
) {
1832 case DRBD_MD_INDEX_INTERNAL
:
1833 case DRBD_MD_INDEX_FLEX_INT
:
1834 return bdev
->md
.md_offset
+ MD_4kB_SECT
-1;
1835 case DRBD_MD_INDEX_FLEX_EXT
:
1837 return bdev
->md
.md_offset
+ bdev
->md
.md_size_sect
-1;
1841 /* Returns the number of 512 byte sectors of the device */
1842 static inline sector_t
drbd_get_capacity(struct block_device
*bdev
)
1844 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1845 return bdev
? i_size_read(bdev
->bd_inode
) >> 9 : 0;
1849 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1850 * @bdev: Meta data block device.
1852 * returns the capacity we announce to out peer. we clip ourselves at the
1853 * various MAX_SECTORS, because if we don't, current implementation will
1854 * oops sooner or later
1856 static inline sector_t
drbd_get_max_capacity(struct drbd_backing_dev
*bdev
)
1860 switch (bdev
->md
.meta_dev_idx
) {
1861 case DRBD_MD_INDEX_INTERNAL
:
1862 case DRBD_MD_INDEX_FLEX_INT
:
1863 s
= drbd_get_capacity(bdev
->backing_bdev
)
1864 ? min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1865 drbd_md_first_sector(bdev
))
1868 case DRBD_MD_INDEX_FLEX_EXT
:
1869 s
= min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1870 drbd_get_capacity(bdev
->backing_bdev
));
1871 /* clip at maximum size the meta device can support */
1872 s
= min_t(sector_t
, s
,
1873 BM_EXT_TO_SECT(bdev
->md
.md_size_sect
1874 - bdev
->md
.bm_offset
));
1877 s
= min_t(sector_t
, DRBD_MAX_SECTORS
,
1878 drbd_get_capacity(bdev
->backing_bdev
));
1884 * drbd_md_ss() - Return the sector number of our meta data super block
1885 * @bdev: Meta data block device.
1887 static inline sector_t
drbd_md_ss(struct drbd_backing_dev
*bdev
)
1889 const int meta_dev_idx
= bdev
->md
.meta_dev_idx
;
1891 if (meta_dev_idx
== DRBD_MD_INDEX_FLEX_EXT
)
1894 /* Since drbd08, internal meta data is always "flexible".
1895 * position: last 4k aligned block of 4k size */
1896 if (meta_dev_idx
== DRBD_MD_INDEX_INTERNAL
||
1897 meta_dev_idx
== DRBD_MD_INDEX_FLEX_INT
)
1898 return (drbd_get_capacity(bdev
->backing_bdev
) & ~7ULL) - 8;
1900 /* external, some index; this is the old fixed size layout */
1901 return MD_128MB_SECT
* bdev
->md
.meta_dev_idx
;
1905 drbd_queue_work_front(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1907 unsigned long flags
;
1908 spin_lock_irqsave(&q
->q_lock
, flags
);
1909 list_add(&w
->list
, &q
->q
);
1910 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1911 wake_up(&q
->q_wait
);
1915 drbd_queue_work(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1917 unsigned long flags
;
1918 spin_lock_irqsave(&q
->q_lock
, flags
);
1919 list_add_tail(&w
->list
, &q
->q
);
1920 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1921 wake_up(&q
->q_wait
);
1924 static inline void wake_asender(struct drbd_tconn
*tconn
)
1926 if (test_bit(SIGNAL_ASENDER
, &tconn
->flags
))
1927 force_sig(DRBD_SIG
, tconn
->asender
.task
);
1930 static inline void request_ping(struct drbd_tconn
*tconn
)
1932 set_bit(SEND_PING
, &tconn
->flags
);
1933 wake_asender(tconn
);
1936 extern void *conn_prepare_command(struct drbd_tconn
*, struct drbd_socket
*);
1937 extern void *drbd_prepare_command(struct drbd_conf
*, struct drbd_socket
*);
1938 extern int conn_send_command(struct drbd_tconn
*, struct drbd_socket
*,
1939 enum drbd_packet
, unsigned int, void *,
1941 extern int drbd_send_command(struct drbd_conf
*, struct drbd_socket
*,
1942 enum drbd_packet
, unsigned int, void *,
1945 extern int drbd_send_ping(struct drbd_tconn
*tconn
);
1946 extern int drbd_send_ping_ack(struct drbd_tconn
*tconn
);
1947 extern int drbd_send_state_req(struct drbd_conf
*, union drbd_state
, union drbd_state
);
1948 extern int conn_send_state_req(struct drbd_tconn
*, union drbd_state
, union drbd_state
);
1950 static inline void drbd_thread_stop(struct drbd_thread
*thi
)
1952 _drbd_thread_stop(thi
, false, true);
1955 static inline void drbd_thread_stop_nowait(struct drbd_thread
*thi
)
1957 _drbd_thread_stop(thi
, false, false);
1960 static inline void drbd_thread_restart_nowait(struct drbd_thread
*thi
)
1962 _drbd_thread_stop(thi
, true, false);
1965 /* counts how many answer packets packets we expect from our peer,
1966 * for either explicit application requests,
1967 * or implicit barrier packets as necessary.
1970 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1971 * it is much easier and equally valid to count what we queue for the
1972 * worker, even before it actually was queued or send.
1973 * (drbd_make_request_common; recovery path on read io-error)
1975 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1976 * _req_mod(req, DATA_RECEIVED)
1977 * [from receive_DataReply]
1978 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1979 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1980 * for some reason it is NOT decreased in got_NegAck,
1981 * but in the resulting cleanup code from report_params.
1982 * we should try to remember the reason for that...
1983 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
1984 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
1985 * [from tl_clear_barrier]
1987 static inline void inc_ap_pending(struct drbd_conf
*mdev
)
1989 atomic_inc(&mdev
->ap_pending_cnt
);
1992 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \
1993 if (atomic_read(&mdev->which) < 0) \
1994 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
1996 atomic_read(&mdev->which))
1998 #define dec_ap_pending(mdev) _dec_ap_pending(mdev, __FUNCTION__, __LINE__)
1999 static inline void _dec_ap_pending(struct drbd_conf
*mdev
, const char *func
, int line
)
2001 if (atomic_dec_and_test(&mdev
->ap_pending_cnt
))
2002 wake_up(&mdev
->misc_wait
);
2003 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt
, func
, line
);
2006 /* counts how many resync-related answers we still expect from the peer
2008 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
2009 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
2010 * (or P_NEG_ACK with ID_SYNCER)
2012 static inline void inc_rs_pending(struct drbd_conf
*mdev
)
2014 atomic_inc(&mdev
->rs_pending_cnt
);
2017 #define dec_rs_pending(mdev) _dec_rs_pending(mdev, __FUNCTION__, __LINE__)
2018 static inline void _dec_rs_pending(struct drbd_conf
*mdev
, const char *func
, int line
)
2020 atomic_dec(&mdev
->rs_pending_cnt
);
2021 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt
, func
, line
);
2024 /* counts how many answers we still need to send to the peer.
2026 * receive_Data unless protocol A;
2027 * we need to send a P_RECV_ACK (proto B)
2028 * or P_WRITE_ACK (proto C)
2029 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2030 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2031 * receive_Barrier_* we need to send a P_BARRIER_ACK
2033 static inline void inc_unacked(struct drbd_conf
*mdev
)
2035 atomic_inc(&mdev
->unacked_cnt
);
2038 #define dec_unacked(mdev) _dec_unacked(mdev, __FUNCTION__, __LINE__)
2039 static inline void _dec_unacked(struct drbd_conf
*mdev
, const char *func
, int line
)
2041 atomic_dec(&mdev
->unacked_cnt
);
2042 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt
, func
, line
);
2045 #define sub_unacked(mdev, n) _sub_unacked(mdev, n, __FUNCTION__, __LINE__)
2046 static inline void _sub_unacked(struct drbd_conf
*mdev
, int n
, const char *func
, int line
)
2048 atomic_sub(n
, &mdev
->unacked_cnt
);
2049 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt
, func
, line
);
2053 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2056 * You have to call put_ldev() when finished working with mdev->ldev.
2058 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2059 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2061 static inline void put_ldev(struct drbd_conf
*mdev
)
2063 int i
= atomic_dec_return(&mdev
->local_cnt
);
2065 /* This may be called from some endio handler,
2066 * so we must not sleep here. */
2071 if (mdev
->state
.disk
== D_DISKLESS
)
2072 /* even internal references gone, safe to destroy */
2073 drbd_ldev_destroy(mdev
);
2074 if (mdev
->state
.disk
== D_FAILED
) {
2075 /* all application IO references gone. */
2076 if (!test_and_set_bit(GO_DISKLESS
, &mdev
->flags
))
2077 drbd_queue_work(&mdev
->tconn
->sender_work
, &mdev
->go_diskless
);
2079 wake_up(&mdev
->misc_wait
);
2084 static inline int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
)
2088 /* never get a reference while D_DISKLESS */
2089 if (mdev
->state
.disk
== D_DISKLESS
)
2092 atomic_inc(&mdev
->local_cnt
);
2093 io_allowed
= (mdev
->state
.disk
>= mins
);
2099 extern int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
);
2102 /* you must have an "get_ldev" reference */
2103 static inline void drbd_get_syncer_progress(struct drbd_conf
*mdev
,
2104 unsigned long *bits_left
, unsigned int *per_mil_done
)
2106 /* this is to break it at compile time when we change that, in case we
2107 * want to support more than (1<<32) bits on a 32bit arch. */
2108 typecheck(unsigned long, mdev
->rs_total
);
2110 /* note: both rs_total and rs_left are in bits, i.e. in
2111 * units of BM_BLOCK_SIZE.
2112 * for the percentage, we don't care. */
2114 if (mdev
->state
.conn
== C_VERIFY_S
|| mdev
->state
.conn
== C_VERIFY_T
)
2115 *bits_left
= mdev
->ov_left
;
2117 *bits_left
= drbd_bm_total_weight(mdev
) - mdev
->rs_failed
;
2118 /* >> 10 to prevent overflow,
2119 * +1 to prevent division by zero */
2120 if (*bits_left
> mdev
->rs_total
) {
2121 /* doh. maybe a logic bug somewhere.
2122 * may also be just a race condition
2123 * between this and a disconnect during sync.
2124 * for now, just prevent in-kernel buffer overflow.
2127 dev_warn(DEV
, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2128 drbd_conn_str(mdev
->state
.conn
),
2129 *bits_left
, mdev
->rs_total
, mdev
->rs_failed
);
2132 /* Make sure the division happens in long context.
2133 * We allow up to one petabyte storage right now,
2134 * at a granularity of 4k per bit that is 2**38 bits.
2135 * After shift right and multiplication by 1000,
2136 * this should still fit easily into a 32bit long,
2137 * so we don't need a 64bit division on 32bit arch.
2138 * Note: currently we don't support such large bitmaps on 32bit
2139 * arch anyways, but no harm done to be prepared for it here.
2141 unsigned int shift
= mdev
->rs_total
> UINT_MAX
? 16 : 10;
2142 unsigned long left
= *bits_left
>> shift
;
2143 unsigned long total
= 1UL + (mdev
->rs_total
>> shift
);
2144 unsigned long tmp
= 1000UL - left
* 1000UL/total
;
2145 *per_mil_done
= tmp
;
2150 /* this throttles on-the-fly application requests
2151 * according to max_buffers settings;
2152 * maybe re-implement using semaphores? */
2153 static inline int drbd_get_max_buffers(struct drbd_conf
*mdev
)
2155 struct net_conf
*nc
;
2159 nc
= rcu_dereference(mdev
->tconn
->net_conf
);
2160 mxb
= nc
? nc
->max_buffers
: 1000000; /* arbitrary limit on open requests */
2166 static inline int drbd_state_is_stable(struct drbd_conf
*mdev
)
2168 union drbd_dev_state s
= mdev
->state
;
2170 /* DO NOT add a default clause, we want the compiler to warn us
2171 * for any newly introduced state we may have forgotten to add here */
2173 switch ((enum drbd_conns
)s
.conn
) {
2174 /* new io only accepted when there is no connection, ... */
2176 case C_WF_CONNECTION
:
2177 /* ... or there is a well established connection. */
2183 case C_PAUSED_SYNC_S
:
2184 case C_PAUSED_SYNC_T
:
2187 /* transitional states, IO allowed */
2188 case C_DISCONNECTING
:
2192 case C_NETWORK_FAILURE
:
2193 case C_PROTOCOL_ERROR
:
2195 case C_WF_REPORT_PARAMS
:
2196 case C_STARTING_SYNC_S
:
2197 case C_STARTING_SYNC_T
:
2200 /* Allow IO in BM exchange states with new protocols */
2202 if (mdev
->tconn
->agreed_pro_version
< 96)
2206 /* no new io accepted in these states */
2208 case C_WF_SYNC_UUID
:
2214 switch ((enum drbd_disk_state
)s
.disk
) {
2216 case D_INCONSISTENT
:
2221 /* disk state is stable as well. */
2224 /* no new io accepted during transitional states */
2236 static inline int drbd_suspended(struct drbd_conf
*mdev
)
2238 struct drbd_tconn
*tconn
= mdev
->tconn
;
2240 return tconn
->susp
|| tconn
->susp_fen
|| tconn
->susp_nod
;
2243 static inline bool may_inc_ap_bio(struct drbd_conf
*mdev
)
2245 int mxb
= drbd_get_max_buffers(mdev
);
2247 if (drbd_suspended(mdev
))
2249 if (test_bit(SUSPEND_IO
, &mdev
->flags
))
2252 /* to avoid potential deadlock or bitmap corruption,
2253 * in various places, we only allow new application io
2254 * to start during "stable" states. */
2256 /* no new io accepted when attaching or detaching the disk */
2257 if (!drbd_state_is_stable(mdev
))
2260 /* since some older kernels don't have atomic_add_unless,
2261 * and we are within the spinlock anyways, we have this workaround. */
2262 if (atomic_read(&mdev
->ap_bio_cnt
) > mxb
)
2264 if (test_bit(BITMAP_IO
, &mdev
->flags
))
2269 static inline bool inc_ap_bio_cond(struct drbd_conf
*mdev
)
2273 spin_lock_irq(&mdev
->tconn
->req_lock
);
2274 rv
= may_inc_ap_bio(mdev
);
2276 atomic_inc(&mdev
->ap_bio_cnt
);
2277 spin_unlock_irq(&mdev
->tconn
->req_lock
);
2282 static inline void inc_ap_bio(struct drbd_conf
*mdev
)
2285 * as long as the device is suspended
2286 * until the bitmap is no longer on the fly during connection
2287 * handshake as long as we would exceed the max_buffer limit.
2289 * to avoid races with the reconnect code,
2290 * we need to atomic_inc within the spinlock. */
2292 wait_event(mdev
->misc_wait
, inc_ap_bio_cond(mdev
));
2295 static inline void dec_ap_bio(struct drbd_conf
*mdev
)
2297 int mxb
= drbd_get_max_buffers(mdev
);
2298 int ap_bio
= atomic_dec_return(&mdev
->ap_bio_cnt
);
2300 D_ASSERT(ap_bio
>= 0);
2302 if (ap_bio
== 0 && test_bit(BITMAP_IO
, &mdev
->flags
)) {
2303 if (!test_and_set_bit(BITMAP_IO_QUEUED
, &mdev
->flags
))
2304 drbd_queue_work(&mdev
->tconn
->sender_work
, &mdev
->bm_io_work
.w
);
2307 /* this currently does wake_up for every dec_ap_bio!
2308 * maybe rather introduce some type of hysteresis?
2309 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2311 wake_up(&mdev
->misc_wait
);
2314 static inline bool verify_can_do_stop_sector(struct drbd_conf
*mdev
)
2316 return mdev
->tconn
->agreed_pro_version
>= 97 &&
2317 mdev
->tconn
->agreed_pro_version
!= 100;
2320 static inline int drbd_set_ed_uuid(struct drbd_conf
*mdev
, u64 val
)
2322 int changed
= mdev
->ed_uuid
!= val
;
2323 mdev
->ed_uuid
= val
;
2327 static inline int drbd_queue_order_type(struct drbd_conf
*mdev
)
2329 /* sorry, we currently have no working implementation
2330 * of distributed TCQ stuff */
2331 #ifndef QUEUE_ORDERED_NONE
2332 #define QUEUE_ORDERED_NONE 0
2334 return QUEUE_ORDERED_NONE
;
2337 static inline void drbd_md_flush(struct drbd_conf
*mdev
)
2341 if (mdev
->ldev
== NULL
) {
2342 dev_warn(DEV
, "mdev->ldev == NULL in drbd_md_flush\n");
2346 if (test_bit(MD_NO_FUA
, &mdev
->flags
))
2349 r
= blkdev_issue_flush(mdev
->ldev
->md_bdev
, GFP_NOIO
, NULL
);
2351 set_bit(MD_NO_FUA
, &mdev
->flags
);
2352 dev_err(DEV
, "meta data flush failed with status %d, disabling md-flushes\n", r
);