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>
43 #include <linux/idr.h>
45 #include <linux/lru_cache.h>
46 #include <linux/prefetch.h>
47 #include <linux/drbd_genl_api.h>
48 #include <linux/drbd.h>
49 #include "drbd_state.h"
52 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
53 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
54 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
55 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
57 # define __protected_by(x)
58 # define __protected_read_by(x)
59 # define __protected_write_by(x)
60 # define __must_hold(x)
63 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
65 /* module parameter, defined in drbd_main.c */
66 extern unsigned int minor_count
;
67 extern int disable_sendpage
;
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 /* drbd_epoch flag bits */
593 DE_HAVE_BARRIER_NUMBER
,
600 EV_CLEANUP
= 32, /* used as flag */
603 struct drbd_wq_barrier
{
605 struct completion done
;
613 struct drbd_peer_request
{
615 struct drbd_epoch
*epoch
; /* for writes */
617 atomic_t pending_bios
;
618 struct drbd_interval i
;
619 /* see comments on ee flag bits below */
623 struct digest_info
*digest
;
628 * While corresponding bios are in flight, the only modification will be
629 * set_bit WAS_ERROR, which has to be atomic.
630 * If no bios are in flight yet, or all have been completed,
631 * non-atomic modification to ee->flags is ok.
634 __EE_CALL_AL_COMPLETE_IO
,
635 __EE_MAY_SET_IN_SYNC
,
637 /* In case a barrier failed,
638 * we need to resubmit without the barrier flag. */
641 /* we may have several bios per peer request.
642 * if any of those fail, we set this flag atomically
643 * from the endio callback */
646 /* This ee has a pointer to a digest instead of a block id */
649 /* Conflicting local requests need to be restarted after this request */
650 __EE_RESTART_REQUESTS
,
652 /* The peer wants a write ACK for this (wire proto C) */
655 /* Is set when net_conf had two_primaries set while creating this peer_req */
656 __EE_IN_INTERVAL_TREE
,
658 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
659 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
660 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
661 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
662 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
663 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
664 #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK)
665 #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE)
667 /* flag bits per mdev */
669 UNPLUG_REMOTE
, /* sending a "UnplugRemote" could help */
670 MD_DIRTY
, /* current uuids and flags not yet on disk */
671 USE_DEGR_WFC_T
, /* degr-wfc-timeout instead of wfc-timeout. */
674 CRASHED_PRIMARY
, /* This node was a crashed primary.
675 * Gets cleared when the state.conn
676 * goes into C_CONNECTED state. */
679 MD_NO_FUA
, /* Users wants us to not use FUA/FLUSH on meta data dev */
680 SUSPEND_IO
, /* suspend application io */
681 BITMAP_IO
, /* suspend application io;
682 once no more io in flight, start bitmap io */
683 BITMAP_IO_QUEUED
, /* Started bitmap IO */
684 GO_DISKLESS
, /* Disk is being detached, on io-error or admin request. */
685 WAS_IO_ERROR
, /* Local disk failed returned IO error */
686 FORCE_DETACH
, /* Force-detach from local disk, aborting any pending local IO */
687 RESYNC_AFTER_NEG
, /* Resync after online grow after the attach&negotiate finished. */
688 RESIZE_PENDING
, /* Size change detected locally, waiting for the response from
689 * the peer, if it changed there as well. */
690 NEW_CUR_UUID
, /* Create new current UUID when thawing IO */
691 AL_SUSPENDED
, /* Activity logging is currently suspended. */
692 AHEAD_TO_SYNC_SOURCE
, /* Ahead -> SyncSource queued */
693 B_RS_H_DONE
, /* Before resync handler done (already executed) */
694 DISCARD_MY_DATA
, /* discard_my_data flag per volume */
698 struct drbd_bitmap
; /* opaque for drbd_conf */
700 /* definition of bits in bm_flags to be used in drbd_bm_lock
701 * and drbd_bitmap_io and friends. */
703 /* do we need to kfree, or vfree bm_pages? */
704 BM_P_VMALLOCED
= 0x10000, /* internal use only, will be masked out */
706 /* currently locked for bulk operation */
707 BM_LOCKED_MASK
= 0xf,
709 /* in detail, that is: */
714 /* so we can mark it locked for bulk operation,
715 * and still allow all non-bulk operations */
718 /* (test bit, count bit) allowed (common case) */
719 BM_LOCKED_TEST_ALLOWED
= BM_DONT_CLEAR
| BM_DONT_SET
| BM_IS_LOCKED
,
721 /* testing bits, as well as setting new bits allowed, but clearing bits
722 * would be unexpected. Used during bitmap receive. Setting new bits
723 * requires sending of "out-of-sync" information, though. */
724 BM_LOCKED_SET_ALLOWED
= BM_DONT_CLEAR
| BM_IS_LOCKED
,
726 /* for drbd_bm_write_copy_pages, everything is allowed,
727 * only concurrent bulk operations are locked out. */
728 BM_LOCKED_CHANGE_ALLOWED
= BM_IS_LOCKED
,
731 struct drbd_work_queue
{
733 spinlock_t q_lock
; /* to protect the list. */
734 wait_queue_head_t q_wait
;
739 struct socket
*socket
;
740 /* this way we get our
741 * send/receive buffers off the stack */
747 u64 md_offset
; /* sector offset to 'super' block */
749 u64 la_size_sect
; /* last agreed size, unit sectors */
755 s32 al_offset
; /* signed relative sector offset to al area */
756 s32 bm_offset
; /* signed relative sector offset to bitmap */
758 /* u32 al_nr_extents; important for restoring the AL
759 * is stored into ldev->dc.al_extents, which in turn
760 * gets applied to act_log->nr_elements
764 struct drbd_backing_dev
{
765 struct block_device
*backing_bdev
;
766 struct block_device
*md_bdev
;
768 struct disk_conf
*disk_conf
; /* RCU, for updates: mdev->tconn->conf_update */
769 sector_t known_size
; /* last known size of that backing device */
781 int (*io_fn
)(struct drbd_conf
*mdev
);
782 void (*done
)(struct drbd_conf
*mdev
, int rv
);
785 enum write_ordering_e
{
792 unsigned int head_index
;
794 int total
; /* sum of all values */
797 extern struct fifo_buffer
*fifo_alloc(int fifo_size
);
799 /* flag bits per tconn */
801 NET_CONGESTED
, /* The data socket is congested */
802 RESOLVE_CONFLICTS
, /* Set on one node, cleared on the peer! */
803 SEND_PING
, /* whether asender should send a ping asap */
804 SIGNAL_ASENDER
, /* whether asender wants to be interrupted */
805 GOT_PING_ACK
, /* set when we receive a ping_ack packet, ping_wait gets woken */
806 CONN_WD_ST_CHG_REQ
, /* A cluster wide state change on the connection is active */
809 CONN_DRY_RUN
, /* Expect disconnect after resync handshake. */
810 CREATE_BARRIER
, /* next P_DATA is preceded by a P_BARRIER */
811 STATE_SENT
, /* Do not change state/UUIDs while this is set */
812 CALLBACK_PENDING
, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
813 * pending, from drbd worker context.
814 * If set, bdi_write_congested() returns true,
815 * so shrink_page_list() would not recurse into,
816 * and potentially deadlock on, this drbd worker.
820 struct drbd_tconn
{ /* is a resource from the config file */
821 char *name
; /* Resource name */
822 struct list_head all_tconn
; /* linked on global drbd_tconns */
824 struct idr volumes
; /* <tconn, vnr> to mdev mapping */
825 enum drbd_conns cstate
; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
826 unsigned susp
:1; /* IO suspended by user */
827 unsigned susp_nod
:1; /* IO suspended because no data */
828 unsigned susp_fen
:1; /* IO suspended because fence peer handler runs */
829 struct mutex cstate_mutex
; /* Protects graceful disconnects */
832 struct net_conf
*net_conf
; /* content protected by rcu */
833 struct mutex conf_update
; /* mutex for ready-copy-update of net_conf and disk_conf */
834 wait_queue_head_t ping_wait
; /* Woken upon reception of a ping, and a state change */
835 struct res_opts res_opts
;
837 struct sockaddr_storage my_addr
;
839 struct sockaddr_storage peer_addr
;
842 struct drbd_socket data
; /* data/barrier/cstate/parameter packets */
843 struct drbd_socket meta
; /* ping/ack (metadata) packets */
844 int agreed_pro_version
; /* actually used protocol version */
845 unsigned long last_received
; /* in jiffies, either socket */
846 unsigned int ko_count
;
850 struct list_head transfer_log
; /* all requests not yet fully processed */
852 struct crypto_hash
*cram_hmac_tfm
;
853 struct crypto_hash
*integrity_tfm
; /* checksums we compute, updates protected by tconn->data->mutex */
854 struct crypto_hash
*peer_integrity_tfm
; /* checksums we verify, only accessed from receiver thread */
855 struct crypto_hash
*csums_tfm
;
856 struct crypto_hash
*verify_tfm
;
861 struct drbd_epoch
*current_epoch
;
862 spinlock_t epoch_lock
;
864 enum write_ordering_e write_ordering
;
865 atomic_t current_tle_nr
; /* transfer log epoch number */
866 unsigned current_tle_writes
; /* writes seen within this tl epoch */
868 unsigned long last_reconnect_jif
;
869 struct drbd_thread receiver
;
870 struct drbd_thread worker
;
871 struct drbd_thread asender
;
872 cpumask_var_t cpu_mask
;
875 struct drbd_work_queue sender_work
;
878 /* whether this sender thread
879 * has processed a single write yet. */
880 bool seen_any_write_yet
;
882 /* Which barrier number to send with the next P_BARRIER */
883 int current_epoch_nr
;
885 /* how many write requests have been sent
886 * with req->epoch == current_epoch_nr.
887 * If none, no P_BARRIER will be sent. */
888 unsigned current_epoch_writes
;
893 struct drbd_tconn
*tconn
;
894 int vnr
; /* volume number within the connection */
897 /* things that are stored as / read from meta data on disk */
900 /* configured by drbdsetup */
901 struct drbd_backing_dev
*ldev
__protected_by(local
);
903 sector_t p_size
; /* partner's disk size */
904 struct request_queue
*rq_queue
;
905 struct block_device
*this_bdev
;
906 struct gendisk
*vdisk
;
908 unsigned long last_reattach_jif
;
909 struct drbd_work resync_work
,
914 struct timer_list resync_timer
;
915 struct timer_list md_sync_timer
;
916 struct timer_list start_resync_timer
;
917 struct timer_list request_timer
;
918 #ifdef DRBD_DEBUG_MD_SYNC
922 } last_md_mark_dirty
;
925 /* Used after attach while negotiating new disk state. */
926 union drbd_state new_state_tmp
;
928 union drbd_dev_state state
;
929 wait_queue_head_t misc_wait
;
930 wait_queue_head_t state_wait
; /* upon each state change. */
931 unsigned int send_cnt
;
932 unsigned int recv_cnt
;
933 unsigned int read_cnt
;
934 unsigned int writ_cnt
;
935 unsigned int al_writ_cnt
;
936 unsigned int bm_writ_cnt
;
937 atomic_t ap_bio_cnt
; /* Requests we need to complete */
938 atomic_t ap_pending_cnt
; /* AP data packets on the wire, ack expected */
939 atomic_t rs_pending_cnt
; /* RS request/data packets on the wire */
940 atomic_t unacked_cnt
; /* Need to send replies for */
941 atomic_t local_cnt
; /* Waiting for local completion */
943 /* Interval tree of pending local requests */
944 struct rb_root read_requests
;
945 struct rb_root write_requests
;
947 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
948 unsigned long rs_total
;
949 /* number of resync blocks that failed in this run */
950 unsigned long rs_failed
;
951 /* Syncer's start time [unit jiffies] */
952 unsigned long rs_start
;
953 /* cumulated time in PausedSyncX state [unit jiffies] */
954 unsigned long rs_paused
;
955 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
956 unsigned long rs_same_csum
;
957 #define DRBD_SYNC_MARKS 8
958 #define DRBD_SYNC_MARK_STEP (3*HZ)
959 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
960 unsigned long rs_mark_left
[DRBD_SYNC_MARKS
];
961 /* marks's time [unit jiffies] */
962 unsigned long rs_mark_time
[DRBD_SYNC_MARKS
];
963 /* current index into rs_mark_{left,time} */
966 /* where does the admin want us to start? (sector) */
967 sector_t ov_start_sector
;
968 sector_t ov_stop_sector
;
969 /* where are we now? (sector) */
970 sector_t ov_position
;
971 /* Start sector of out of sync range (to merge printk reporting). */
972 sector_t ov_last_oos_start
;
973 /* size of out-of-sync range in sectors. */
974 sector_t ov_last_oos_size
;
975 unsigned long ov_left
; /* in bits */
977 struct drbd_bitmap
*bitmap
;
978 unsigned long bm_resync_fo
; /* bit offset for drbd_bm_find_next */
980 /* Used to track operations of resync... */
981 struct lru_cache
*resync
;
982 /* Number of locked elements in resync LRU */
983 unsigned int resync_locked
;
984 /* resync extent number waiting for application requests */
985 unsigned int resync_wenr
;
990 struct list_head active_ee
; /* IO in progress (P_DATA gets written to disk) */
991 struct list_head sync_ee
; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
992 struct list_head done_ee
; /* need to send P_WRITE_ACK */
993 struct list_head read_ee
; /* [RS]P_DATA_REQUEST being read */
994 struct list_head net_ee
; /* zero-copy network send in progress */
997 struct list_head resync_reads
;
998 atomic_t pp_in_use
; /* allocated from page pool */
999 atomic_t pp_in_use_by_net
; /* sendpage()d, still referenced by tcp */
1000 wait_queue_head_t ee_wait
;
1001 struct page
*md_io_page
; /* one page buffer for md_io */
1002 struct drbd_md_io md_io
;
1003 atomic_t md_io_in_use
; /* protects the md_io, md_io_page and md_io_tmpp */
1005 wait_queue_head_t al_wait
;
1006 struct lru_cache
*act_log
; /* activity log */
1007 unsigned int al_tr_number
;
1009 int al_tr_pos
; /* position of the next transaction in the journal */
1010 wait_queue_head_t seq_wait
;
1011 atomic_t packet_seq
;
1012 unsigned int peer_seq
;
1013 spinlock_t peer_seq_lock
;
1015 unsigned long comm_bm_set
; /* communicated number of set bits. */
1016 struct bm_io_work bm_io_work
;
1017 u64 ed_uuid
; /* UUID of the exposed data */
1018 struct mutex own_state_mutex
;
1019 struct mutex
*state_mutex
; /* either own_state_mutex or mdev->tconn->cstate_mutex */
1020 char congestion_reason
; /* Why we where congested... */
1021 atomic_t rs_sect_in
; /* for incoming resync data rate, SyncTarget */
1022 atomic_t rs_sect_ev
; /* for submitted resync data rate, both */
1023 int rs_last_sect_ev
; /* counter to compare with */
1024 int rs_last_events
; /* counter of read or write "events" (unit sectors)
1025 * on the lower level device when we last looked. */
1026 int c_sync_rate
; /* current resync rate after syncer throttle magic */
1027 struct fifo_buffer
*rs_plan_s
; /* correction values of resync planer (RCU, tconn->conn_update) */
1028 int rs_in_flight
; /* resync sectors in flight (to proxy, in proxy and from proxy) */
1029 atomic_t ap_in_flight
; /* App sectors in flight (waiting for ack) */
1030 int peer_max_bio_size
;
1031 int local_max_bio_size
;
1034 static inline struct drbd_conf
*minor_to_mdev(unsigned int minor
)
1036 return (struct drbd_conf
*)idr_find(&minors
, minor
);
1039 static inline unsigned int mdev_to_minor(struct drbd_conf
*mdev
)
1044 static inline struct drbd_conf
*vnr_to_mdev(struct drbd_tconn
*tconn
, int vnr
)
1046 return (struct drbd_conf
*)idr_find(&tconn
->volumes
, vnr
);
1050 * function declarations
1051 *************************/
1057 DDSF_NO_RESYNC
= 2, /* Do not run a resync for the new space */
1060 extern void drbd_init_set_defaults(struct drbd_conf
*mdev
);
1061 extern int drbd_thread_start(struct drbd_thread
*thi
);
1062 extern void _drbd_thread_stop(struct drbd_thread
*thi
, int restart
, int wait
);
1063 extern char *drbd_task_to_thread_name(struct drbd_tconn
*tconn
, struct task_struct
*task
);
1065 extern void drbd_thread_current_set_cpu(struct drbd_thread
*thi
);
1066 extern void drbd_calc_cpu_mask(struct drbd_tconn
*tconn
);
1068 #define drbd_thread_current_set_cpu(A) ({})
1069 #define drbd_calc_cpu_mask(A) ({})
1071 extern void tl_release(struct drbd_tconn
*, unsigned int barrier_nr
,
1072 unsigned int set_size
);
1073 extern void tl_clear(struct drbd_tconn
*);
1074 extern void drbd_free_sock(struct drbd_tconn
*tconn
);
1075 extern int drbd_send(struct drbd_tconn
*tconn
, struct socket
*sock
,
1076 void *buf
, size_t size
, unsigned msg_flags
);
1077 extern int drbd_send_all(struct drbd_tconn
*, struct socket
*, void *, size_t,
1080 extern int __drbd_send_protocol(struct drbd_tconn
*tconn
, enum drbd_packet cmd
);
1081 extern int drbd_send_protocol(struct drbd_tconn
*tconn
);
1082 extern int drbd_send_uuids(struct drbd_conf
*mdev
);
1083 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf
*mdev
);
1084 extern void drbd_gen_and_send_sync_uuid(struct drbd_conf
*mdev
);
1085 extern int drbd_send_sizes(struct drbd_conf
*mdev
, int trigger_reply
, enum dds_flags flags
);
1086 extern int drbd_send_state(struct drbd_conf
*mdev
, union drbd_state s
);
1087 extern int drbd_send_current_state(struct drbd_conf
*mdev
);
1088 extern int drbd_send_sync_param(struct drbd_conf
*mdev
);
1089 extern void drbd_send_b_ack(struct drbd_tconn
*tconn
, u32 barrier_nr
,
1091 extern int drbd_send_ack(struct drbd_conf
*, enum drbd_packet
,
1092 struct drbd_peer_request
*);
1093 extern void drbd_send_ack_rp(struct drbd_conf
*mdev
, enum drbd_packet cmd
,
1094 struct p_block_req
*rp
);
1095 extern void drbd_send_ack_dp(struct drbd_conf
*mdev
, enum drbd_packet cmd
,
1096 struct p_data
*dp
, int data_size
);
1097 extern int drbd_send_ack_ex(struct drbd_conf
*mdev
, enum drbd_packet cmd
,
1098 sector_t sector
, int blksize
, u64 block_id
);
1099 extern int drbd_send_out_of_sync(struct drbd_conf
*, struct drbd_request
*);
1100 extern int drbd_send_block(struct drbd_conf
*, enum drbd_packet
,
1101 struct drbd_peer_request
*);
1102 extern int drbd_send_dblock(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1103 extern int drbd_send_drequest(struct drbd_conf
*mdev
, int cmd
,
1104 sector_t sector
, int size
, u64 block_id
);
1105 extern int drbd_send_drequest_csum(struct drbd_conf
*mdev
, sector_t sector
,
1106 int size
, void *digest
, int digest_size
,
1107 enum drbd_packet cmd
);
1108 extern int drbd_send_ov_request(struct drbd_conf
*mdev
,sector_t sector
,int size
);
1110 extern int drbd_send_bitmap(struct drbd_conf
*mdev
);
1111 extern void drbd_send_sr_reply(struct drbd_conf
*mdev
, enum drbd_state_rv retcode
);
1112 extern void conn_send_sr_reply(struct drbd_tconn
*tconn
, enum drbd_state_rv retcode
);
1113 extern void drbd_free_bc(struct drbd_backing_dev
*ldev
);
1114 extern void drbd_mdev_cleanup(struct drbd_conf
*mdev
);
1115 void drbd_print_uuids(struct drbd_conf
*mdev
, const char *text
);
1117 extern void drbd_md_sync(struct drbd_conf
*mdev
);
1118 extern int drbd_md_read(struct drbd_conf
*mdev
, struct drbd_backing_dev
*bdev
);
1119 extern void drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1120 extern void _drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1121 extern void drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1122 extern void _drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1123 extern void drbd_uuid_set_bm(struct drbd_conf
*mdev
, u64 val
) __must_hold(local
);
1124 extern void drbd_md_set_flag(struct drbd_conf
*mdev
, int flags
) __must_hold(local
);
1125 extern void drbd_md_clear_flag(struct drbd_conf
*mdev
, int flags
)__must_hold(local
);
1126 extern int drbd_md_test_flag(struct drbd_backing_dev
*, int);
1127 #ifndef DRBD_DEBUG_MD_SYNC
1128 extern void drbd_md_mark_dirty(struct drbd_conf
*mdev
);
1130 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1131 extern void drbd_md_mark_dirty_(struct drbd_conf
*mdev
,
1132 unsigned int line
, const char *func
);
1134 extern void drbd_queue_bitmap_io(struct drbd_conf
*mdev
,
1135 int (*io_fn
)(struct drbd_conf
*),
1136 void (*done
)(struct drbd_conf
*, int),
1137 char *why
, enum bm_flag flags
);
1138 extern int drbd_bitmap_io(struct drbd_conf
*mdev
,
1139 int (*io_fn
)(struct drbd_conf
*),
1140 char *why
, enum bm_flag flags
);
1141 extern int drbd_bmio_set_n_write(struct drbd_conf
*mdev
);
1142 extern int drbd_bmio_clear_n_write(struct drbd_conf
*mdev
);
1143 extern void drbd_go_diskless(struct drbd_conf
*mdev
);
1144 extern void drbd_ldev_destroy(struct drbd_conf
*mdev
);
1147 We reserve a 128MB Block (4k aligned)
1148 * either at the end of the backing device
1149 * or on a separate meta data device. */
1151 /* The following numbers are sectors */
1152 /* Allows up to about 3.8TB, so if you want more,
1153 * you need to use the "flexible" meta data format. */
1154 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1155 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1156 #define MD_AL_SECTORS 64 /* = 32 kB on disk activity log ring buffer */
1157 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_SECTORS)
1159 /* we do all meta data IO in 4k blocks */
1160 #define MD_BLOCK_SHIFT 12
1161 #define MD_BLOCK_SIZE (1<<MD_BLOCK_SHIFT)
1163 /* One activity log extent represents 4M of storage */
1164 #define AL_EXTENT_SHIFT 22
1165 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1167 /* We could make these currently hardcoded constants configurable
1168 * variables at create-md time (or even re-configurable at runtime?).
1169 * Which will require some more changes to the DRBD "super block"
1172 * updates per transaction:
1173 * This many changes to the active set can be logged with one transaction.
1174 * This number is arbitrary.
1175 * context per transaction:
1176 * This many context extent numbers are logged with each transaction.
1177 * This number is resulting from the transaction block size (4k), the layout
1178 * of the transaction header, and the number of updates per transaction.
1179 * See drbd_actlog.c:struct al_transaction_on_disk
1181 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1182 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1184 #if BITS_PER_LONG == 32
1186 #define cpu_to_lel(A) cpu_to_le32(A)
1187 #define lel_to_cpu(A) le32_to_cpu(A)
1188 #elif BITS_PER_LONG == 64
1190 #define cpu_to_lel(A) cpu_to_le64(A)
1191 #define lel_to_cpu(A) le64_to_cpu(A)
1193 #error "LN2 of BITS_PER_LONG unknown!"
1197 /* 16MB sized 'bitmap extent' to track syncer usage */
1199 int rs_left
; /* number of bits set (out of sync) in this extent. */
1200 int rs_failed
; /* number of failed resync requests in this extent. */
1201 unsigned long flags
;
1202 struct lc_element lce
;
1205 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1206 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1207 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1211 * We need to store one bit for a block.
1212 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1213 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1214 * Bit 1 ==> local node thinks this block needs to be synced.
1217 #define SLEEP_TIME (HZ/10)
1219 /* We do bitmap IO in units of 4k blocks.
1220 * We also still have a hardcoded 4k per bit relation. */
1221 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1222 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1223 /* mostly arbitrarily set the represented size of one bitmap extent,
1224 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1225 * at 4k per bit resolution) */
1226 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
1227 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1229 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1230 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1233 /* thus many _storage_ sectors are described by one bit */
1234 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1235 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1236 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1238 /* bit to represented kilo byte conversion */
1239 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1241 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1242 * _storage_ sector is located in */
1243 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1245 /* how much _storage_ sectors we have per bitmap sector */
1246 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1247 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1249 /* in one sector of the bitmap, we have this many activity_log extents. */
1250 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1251 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1253 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1254 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1256 /* the extent in "PER_EXTENT" below is an activity log extent
1257 * we need that many (long words/bytes) to store the bitmap
1258 * of one AL_EXTENT_SIZE chunk of storage.
1259 * we can store the bitmap for that many AL_EXTENTS within
1260 * one sector of the _on_disk_ bitmap:
1261 * bit 0 bit 37 bit 38 bit (512*8)-1
1262 * ...|........|........|.. // ..|........|
1263 * sect. 0 `296 `304 ^(512*8*8)-1
1265 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1266 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1267 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1270 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1271 #define DRBD_MAX_SECTORS_BM \
1272 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1273 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1274 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1275 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1276 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1277 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1278 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1280 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1281 /* 16 TB in units of sectors */
1282 #if BITS_PER_LONG == 32
1283 /* adjust by one page worth of bitmap,
1284 * so we won't wrap around in drbd_bm_find_next_bit.
1285 * you should use 64bit OS for that much storage, anyways. */
1286 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1288 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1289 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1290 /* corresponds to (1UL << 38) bits right now. */
1294 /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE,
1295 * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte.
1296 * Since we may live in a mixed-platform cluster,
1297 * we limit us to a platform agnostic constant here for now.
1298 * A followup commit may allow even bigger BIO sizes,
1299 * once we thought that through. */
1300 #define DRBD_MAX_BIO_SIZE (1 << 20)
1301 #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1302 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1304 #define DRBD_MAX_BIO_SIZE_SAFE (1 << 12) /* Works always = 4k */
1306 #define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* Header 80 only allows packets up to 32KiB data */
1307 #define DRBD_MAX_BIO_SIZE_P95 (1 << 17) /* Protocol 95 to 99 allows bios up to 128KiB */
1309 extern int drbd_bm_init(struct drbd_conf
*mdev
);
1310 extern int drbd_bm_resize(struct drbd_conf
*mdev
, sector_t sectors
, int set_new_bits
);
1311 extern void drbd_bm_cleanup(struct drbd_conf
*mdev
);
1312 extern void drbd_bm_set_all(struct drbd_conf
*mdev
);
1313 extern void drbd_bm_clear_all(struct drbd_conf
*mdev
);
1314 /* set/clear/test only a few bits at a time */
1315 extern int drbd_bm_set_bits(
1316 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1317 extern int drbd_bm_clear_bits(
1318 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1319 extern int drbd_bm_count_bits(
1320 struct drbd_conf
*mdev
, const unsigned long s
, const unsigned long e
);
1321 /* bm_set_bits variant for use while holding drbd_bm_lock,
1322 * may process the whole bitmap in one go */
1323 extern void _drbd_bm_set_bits(struct drbd_conf
*mdev
,
1324 const unsigned long s
, const unsigned long e
);
1325 extern int drbd_bm_test_bit(struct drbd_conf
*mdev
, unsigned long bitnr
);
1326 extern int drbd_bm_e_weight(struct drbd_conf
*mdev
, unsigned long enr
);
1327 extern int drbd_bm_write_page(struct drbd_conf
*mdev
, unsigned int idx
) __must_hold(local
);
1328 extern int drbd_bm_read(struct drbd_conf
*mdev
) __must_hold(local
);
1329 extern void drbd_bm_mark_for_writeout(struct drbd_conf
*mdev
, int page_nr
);
1330 extern int drbd_bm_write(struct drbd_conf
*mdev
) __must_hold(local
);
1331 extern int drbd_bm_write_hinted(struct drbd_conf
*mdev
) __must_hold(local
);
1332 extern int drbd_bm_write_all(struct drbd_conf
*mdev
) __must_hold(local
);
1333 extern int drbd_bm_write_copy_pages(struct drbd_conf
*mdev
) __must_hold(local
);
1334 extern size_t drbd_bm_words(struct drbd_conf
*mdev
);
1335 extern unsigned long drbd_bm_bits(struct drbd_conf
*mdev
);
1336 extern sector_t
drbd_bm_capacity(struct drbd_conf
*mdev
);
1338 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1339 extern unsigned long drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1340 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1341 extern unsigned long _drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1342 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1343 extern unsigned long _drbd_bm_total_weight(struct drbd_conf
*mdev
);
1344 extern unsigned long drbd_bm_total_weight(struct drbd_conf
*mdev
);
1345 extern int drbd_bm_rs_done(struct drbd_conf
*mdev
);
1346 /* for receive_bitmap */
1347 extern void drbd_bm_merge_lel(struct drbd_conf
*mdev
, size_t offset
,
1348 size_t number
, unsigned long *buffer
);
1349 /* for _drbd_send_bitmap */
1350 extern void drbd_bm_get_lel(struct drbd_conf
*mdev
, size_t offset
,
1351 size_t number
, unsigned long *buffer
);
1353 extern void drbd_bm_lock(struct drbd_conf
*mdev
, char *why
, enum bm_flag flags
);
1354 extern void drbd_bm_unlock(struct drbd_conf
*mdev
);
1357 extern struct kmem_cache
*drbd_request_cache
;
1358 extern struct kmem_cache
*drbd_ee_cache
; /* peer requests */
1359 extern struct kmem_cache
*drbd_bm_ext_cache
; /* bitmap extents */
1360 extern struct kmem_cache
*drbd_al_ext_cache
; /* activity log extents */
1361 extern mempool_t
*drbd_request_mempool
;
1362 extern mempool_t
*drbd_ee_mempool
;
1364 /* drbd's page pool, used to buffer data received from the peer,
1365 * or data requested by the peer.
1367 * This does not have an emergency reserve.
1369 * When allocating from this pool, it first takes pages from the pool.
1370 * Only if the pool is depleted will try to allocate from the system.
1372 * The assumption is that pages taken from this pool will be processed,
1373 * and given back, "quickly", and then can be recycled, so we can avoid
1374 * frequent calls to alloc_page(), and still will be able to make progress even
1375 * under memory pressure.
1377 extern struct page
*drbd_pp_pool
;
1378 extern spinlock_t drbd_pp_lock
;
1379 extern int drbd_pp_vacant
;
1380 extern wait_queue_head_t drbd_pp_wait
;
1382 /* We also need a standard (emergency-reserve backed) page pool
1383 * for meta data IO (activity log, bitmap).
1384 * We can keep it global, as long as it is used as "N pages at a time".
1385 * 128 should be plenty, currently we probably can get away with as few as 1.
1387 #define DRBD_MIN_POOL_PAGES 128
1388 extern mempool_t
*drbd_md_io_page_pool
;
1390 /* We also need to make sure we get a bio
1391 * when we need it for housekeeping purposes */
1392 extern struct bio_set
*drbd_md_io_bio_set
;
1393 /* to allocate from that set */
1394 extern struct bio
*bio_alloc_drbd(gfp_t gfp_mask
);
1396 extern rwlock_t global_state_lock
;
1398 extern int conn_lowest_minor(struct drbd_tconn
*tconn
);
1399 enum drbd_ret_code
conn_new_minor(struct drbd_tconn
*tconn
, unsigned int minor
, int vnr
);
1400 extern void drbd_minor_destroy(struct kref
*kref
);
1402 extern int set_resource_options(struct drbd_tconn
*tconn
, struct res_opts
*res_opts
);
1403 extern struct drbd_tconn
*conn_create(const char *name
, struct res_opts
*res_opts
);
1404 extern void conn_destroy(struct kref
*kref
);
1405 struct drbd_tconn
*conn_get_by_name(const char *name
);
1406 extern struct drbd_tconn
*conn_get_by_addrs(void *my_addr
, int my_addr_len
,
1407 void *peer_addr
, int peer_addr_len
);
1408 extern void conn_free_crypto(struct drbd_tconn
*tconn
);
1410 extern int proc_details
;
1413 extern void __drbd_make_request(struct drbd_conf
*, struct bio
*, unsigned long);
1414 extern int drbd_make_request(struct request_queue
*q
, struct bio
*bio
);
1415 extern int drbd_read_remote(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1416 extern int drbd_merge_bvec(struct request_queue
*q
, struct bvec_merge_data
*bvm
, struct bio_vec
*bvec
);
1417 extern int is_valid_ar_handle(struct drbd_request
*, sector_t
);
1421 extern int drbd_msg_put_info(const char *info
);
1422 extern void drbd_suspend_io(struct drbd_conf
*mdev
);
1423 extern void drbd_resume_io(struct drbd_conf
*mdev
);
1424 extern char *ppsize(char *buf
, unsigned long long size
);
1425 extern sector_t
drbd_new_dev_size(struct drbd_conf
*, struct drbd_backing_dev
*, sector_t
, int);
1426 enum determine_dev_size
{ dev_size_error
= -1, unchanged
= 0, shrunk
= 1, grew
= 2 };
1427 extern enum determine_dev_size
drbd_determine_dev_size(struct drbd_conf
*, enum dds_flags
) __must_hold(local
);
1428 extern void resync_after_online_grow(struct drbd_conf
*);
1429 extern void drbd_reconsider_max_bio_size(struct drbd_conf
*mdev
);
1430 extern enum drbd_state_rv
drbd_set_role(struct drbd_conf
*mdev
,
1431 enum drbd_role new_role
,
1433 extern bool conn_try_outdate_peer(struct drbd_tconn
*tconn
);
1434 extern void conn_try_outdate_peer_async(struct drbd_tconn
*tconn
);
1435 extern int drbd_khelper(struct drbd_conf
*mdev
, char *cmd
);
1438 extern int drbd_worker(struct drbd_thread
*thi
);
1439 enum drbd_ret_code
drbd_resync_after_valid(struct drbd_conf
*mdev
, int o_minor
);
1440 void drbd_resync_after_changed(struct drbd_conf
*mdev
);
1441 extern void drbd_start_resync(struct drbd_conf
*mdev
, enum drbd_conns side
);
1442 extern void resume_next_sg(struct drbd_conf
*mdev
);
1443 extern void suspend_other_sg(struct drbd_conf
*mdev
);
1444 extern int drbd_resync_finished(struct drbd_conf
*mdev
);
1445 /* maybe rather drbd_main.c ? */
1446 extern void *drbd_md_get_buffer(struct drbd_conf
*mdev
);
1447 extern void drbd_md_put_buffer(struct drbd_conf
*mdev
);
1448 extern int drbd_md_sync_page_io(struct drbd_conf
*mdev
,
1449 struct drbd_backing_dev
*bdev
, sector_t sector
, int rw
);
1450 extern void drbd_ov_out_of_sync_found(struct drbd_conf
*, sector_t
, int);
1451 extern void wait_until_done_or_disk_failure(struct drbd_conf
*mdev
, struct drbd_backing_dev
*bdev
,
1452 unsigned int *done
);
1453 extern void drbd_rs_controller_reset(struct drbd_conf
*mdev
);
1455 static inline void ov_out_of_sync_print(struct drbd_conf
*mdev
)
1457 if (mdev
->ov_last_oos_size
) {
1458 dev_err(DEV
, "Out of sync: start=%llu, size=%lu (sectors)\n",
1459 (unsigned long long)mdev
->ov_last_oos_start
,
1460 (unsigned long)mdev
->ov_last_oos_size
);
1462 mdev
->ov_last_oos_size
=0;
1466 extern void drbd_csum_bio(struct drbd_conf
*, struct crypto_hash
*, struct bio
*, void *);
1467 extern void drbd_csum_ee(struct drbd_conf
*, struct crypto_hash
*,
1468 struct drbd_peer_request
*, void *);
1469 /* worker callbacks */
1470 extern int w_e_end_data_req(struct drbd_work
*, int);
1471 extern int w_e_end_rsdata_req(struct drbd_work
*, int);
1472 extern int w_e_end_csum_rs_req(struct drbd_work
*, int);
1473 extern int w_e_end_ov_reply(struct drbd_work
*, int);
1474 extern int w_e_end_ov_req(struct drbd_work
*, int);
1475 extern int w_ov_finished(struct drbd_work
*, int);
1476 extern int w_resync_timer(struct drbd_work
*, int);
1477 extern int w_send_write_hint(struct drbd_work
*, int);
1478 extern int w_make_resync_request(struct drbd_work
*, int);
1479 extern int w_send_dblock(struct drbd_work
*, int);
1480 extern int w_send_read_req(struct drbd_work
*, int);
1481 extern int w_prev_work_done(struct drbd_work
*, int);
1482 extern int w_e_reissue(struct drbd_work
*, int);
1483 extern int w_restart_disk_io(struct drbd_work
*, int);
1484 extern int w_send_out_of_sync(struct drbd_work
*, int);
1485 extern int w_start_resync(struct drbd_work
*, int);
1487 extern void resync_timer_fn(unsigned long data
);
1488 extern void start_resync_timer_fn(unsigned long data
);
1490 /* drbd_receiver.c */
1491 extern int drbd_rs_should_slow_down(struct drbd_conf
*mdev
, sector_t sector
);
1492 extern int drbd_submit_peer_request(struct drbd_conf
*,
1493 struct drbd_peer_request
*, const unsigned,
1495 extern int drbd_free_peer_reqs(struct drbd_conf
*, struct list_head
*);
1496 extern struct drbd_peer_request
*drbd_alloc_peer_req(struct drbd_conf
*, u64
,
1497 sector_t
, unsigned int,
1498 gfp_t
) __must_hold(local
);
1499 extern void __drbd_free_peer_req(struct drbd_conf
*, struct drbd_peer_request
*,
1501 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1502 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
1503 extern struct page
*drbd_alloc_pages(struct drbd_conf
*, unsigned int, bool);
1504 extern void drbd_set_recv_tcq(struct drbd_conf
*mdev
, int tcq_enabled
);
1505 extern void _drbd_clear_done_ee(struct drbd_conf
*mdev
, struct list_head
*to_be_freed
);
1506 extern void conn_flush_workqueue(struct drbd_tconn
*tconn
);
1507 extern int drbd_connected(struct drbd_conf
*mdev
);
1508 static inline void drbd_flush_workqueue(struct drbd_conf
*mdev
)
1510 conn_flush_workqueue(mdev
->tconn
);
1513 /* Yes, there is kernel_setsockopt, but only since 2.6.18.
1514 * So we have our own copy of it here. */
1515 static inline int drbd_setsockopt(struct socket
*sock
, int level
, int optname
,
1516 char *optval
, int optlen
)
1518 mm_segment_t oldfs
= get_fs();
1519 char __user
*uoptval
;
1522 uoptval
= (char __user __force
*)optval
;
1525 if (level
== SOL_SOCKET
)
1526 err
= sock_setsockopt(sock
, level
, optname
, uoptval
, optlen
);
1528 err
= sock
->ops
->setsockopt(sock
, level
, optname
, uoptval
,
1534 static inline void drbd_tcp_cork(struct socket
*sock
)
1537 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1538 (char*)&val
, sizeof(val
));
1541 static inline void drbd_tcp_uncork(struct socket
*sock
)
1544 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1545 (char*)&val
, sizeof(val
));
1548 static inline void drbd_tcp_nodelay(struct socket
*sock
)
1551 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_NODELAY
,
1552 (char*)&val
, sizeof(val
));
1555 static inline void drbd_tcp_quickack(struct socket
*sock
)
1558 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_QUICKACK
,
1559 (char*)&val
, sizeof(val
));
1562 void drbd_bump_write_ordering(struct drbd_tconn
*tconn
, enum write_ordering_e wo
);
1565 extern struct proc_dir_entry
*drbd_proc
;
1566 extern const struct file_operations drbd_proc_fops
;
1567 extern const char *drbd_conn_str(enum drbd_conns s
);
1568 extern const char *drbd_role_str(enum drbd_role s
);
1571 extern void drbd_al_begin_io(struct drbd_conf
*mdev
, struct drbd_interval
*i
);
1572 extern void drbd_al_complete_io(struct drbd_conf
*mdev
, struct drbd_interval
*i
);
1573 extern void drbd_rs_complete_io(struct drbd_conf
*mdev
, sector_t sector
);
1574 extern int drbd_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1575 extern int drbd_try_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1576 extern void drbd_rs_cancel_all(struct drbd_conf
*mdev
);
1577 extern int drbd_rs_del_all(struct drbd_conf
*mdev
);
1578 extern void drbd_rs_failed_io(struct drbd_conf
*mdev
,
1579 sector_t sector
, int size
);
1580 extern void drbd_advance_rs_marks(struct drbd_conf
*mdev
, unsigned long still_to_go
);
1581 extern void __drbd_set_in_sync(struct drbd_conf
*mdev
, sector_t sector
,
1582 int size
, const char *file
, const unsigned int line
);
1583 #define drbd_set_in_sync(mdev, sector, size) \
1584 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1585 extern int __drbd_set_out_of_sync(struct drbd_conf
*mdev
, sector_t sector
,
1586 int size
, const char *file
, const unsigned int line
);
1587 #define drbd_set_out_of_sync(mdev, sector, size) \
1588 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1589 extern void drbd_al_shrink(struct drbd_conf
*mdev
);
1592 /* state info broadcast */
1594 enum drbd_state_info_bcast_reason sib_reason
;
1598 unsigned helper_exit_code
;
1601 union drbd_state os
;
1602 union drbd_state ns
;
1606 void drbd_bcast_event(struct drbd_conf
*mdev
, const struct sib_info
*sib
);
1609 * inline helper functions
1610 *************************/
1612 /* see also page_chain_add and friends in drbd_receiver.c */
1613 static inline struct page
*page_chain_next(struct page
*page
)
1615 return (struct page
*)page_private(page
);
1617 #define page_chain_for_each(page) \
1618 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1619 page = page_chain_next(page))
1620 #define page_chain_for_each_safe(page, n) \
1621 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1624 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request
*peer_req
)
1626 struct page
*page
= peer_req
->pages
;
1627 page_chain_for_each(page
) {
1628 if (page_count(page
) > 1)
1634 static inline enum drbd_state_rv
1635 _drbd_set_state(struct drbd_conf
*mdev
, union drbd_state ns
,
1636 enum chg_state_flags flags
, struct completion
*done
)
1638 enum drbd_state_rv rv
;
1640 read_lock(&global_state_lock
);
1641 rv
= __drbd_set_state(mdev
, ns
, flags
, done
);
1642 read_unlock(&global_state_lock
);
1647 static inline union drbd_state
drbd_read_state(struct drbd_conf
*mdev
)
1649 union drbd_state rv
;
1651 rv
.i
= mdev
->state
.i
;
1652 rv
.susp
= mdev
->tconn
->susp
;
1653 rv
.susp_nod
= mdev
->tconn
->susp_nod
;
1654 rv
.susp_fen
= mdev
->tconn
->susp_fen
;
1659 enum drbd_force_detach_flags
{
1665 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1666 static inline void __drbd_chk_io_error_(struct drbd_conf
*mdev
,
1667 enum drbd_force_detach_flags forcedetach
,
1670 enum drbd_io_error_p ep
;
1673 ep
= rcu_dereference(mdev
->ldev
->disk_conf
)->on_io_error
;
1676 case EP_PASS_ON
: /* FIXME would this be better named "Ignore"? */
1677 if (forcedetach
== DRBD_IO_ERROR
) {
1678 if (__ratelimit(&drbd_ratelimit_state
))
1679 dev_err(DEV
, "Local IO failed in %s.\n", where
);
1680 if (mdev
->state
.disk
> D_INCONSISTENT
)
1681 _drbd_set_state(_NS(mdev
, disk
, D_INCONSISTENT
), CS_HARD
, NULL
);
1684 /* NOTE fall through to detach case if forcedetach set */
1686 case EP_CALL_HELPER
:
1687 set_bit(WAS_IO_ERROR
, &mdev
->flags
);
1688 if (forcedetach
== DRBD_FORCE_DETACH
)
1689 set_bit(FORCE_DETACH
, &mdev
->flags
);
1690 if (mdev
->state
.disk
> D_FAILED
) {
1691 _drbd_set_state(_NS(mdev
, disk
, D_FAILED
), CS_HARD
, NULL
);
1693 "Local IO failed in %s. Detaching...\n", where
);
1700 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1701 * @mdev: DRBD device.
1702 * @error: Error code passed to the IO completion callback
1703 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1705 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1707 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1708 static inline void drbd_chk_io_error_(struct drbd_conf
*mdev
,
1709 int error
, enum drbd_force_detach_flags forcedetach
, const char *where
)
1712 unsigned long flags
;
1713 spin_lock_irqsave(&mdev
->tconn
->req_lock
, flags
);
1714 __drbd_chk_io_error_(mdev
, forcedetach
, where
);
1715 spin_unlock_irqrestore(&mdev
->tconn
->req_lock
, flags
);
1721 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1722 * @bdev: Meta data block device.
1724 * BTW, for internal meta data, this happens to be the maximum capacity
1725 * we could agree upon with our peer node.
1727 static inline sector_t
_drbd_md_first_sector(int meta_dev_idx
, struct drbd_backing_dev
*bdev
)
1729 switch (meta_dev_idx
) {
1730 case DRBD_MD_INDEX_INTERNAL
:
1731 case DRBD_MD_INDEX_FLEX_INT
:
1732 return bdev
->md
.md_offset
+ bdev
->md
.bm_offset
;
1733 case DRBD_MD_INDEX_FLEX_EXT
:
1735 return bdev
->md
.md_offset
;
1739 static inline sector_t
drbd_md_first_sector(struct drbd_backing_dev
*bdev
)
1744 meta_dev_idx
= rcu_dereference(bdev
->disk_conf
)->meta_dev_idx
;
1747 return _drbd_md_first_sector(meta_dev_idx
, bdev
);
1751 * drbd_md_last_sector() - Return the last sector number of the meta data area
1752 * @bdev: Meta data block device.
1754 static inline sector_t
drbd_md_last_sector(struct drbd_backing_dev
*bdev
)
1759 meta_dev_idx
= rcu_dereference(bdev
->disk_conf
)->meta_dev_idx
;
1762 switch (meta_dev_idx
) {
1763 case DRBD_MD_INDEX_INTERNAL
:
1764 case DRBD_MD_INDEX_FLEX_INT
:
1765 return bdev
->md
.md_offset
+ MD_AL_OFFSET
- 1;
1766 case DRBD_MD_INDEX_FLEX_EXT
:
1768 return bdev
->md
.md_offset
+ bdev
->md
.md_size_sect
;
1772 /* Returns the number of 512 byte sectors of the device */
1773 static inline sector_t
drbd_get_capacity(struct block_device
*bdev
)
1775 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1776 return bdev
? i_size_read(bdev
->bd_inode
) >> 9 : 0;
1780 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1781 * @bdev: Meta data block device.
1783 * returns the capacity we announce to out peer. we clip ourselves at the
1784 * various MAX_SECTORS, because if we don't, current implementation will
1785 * oops sooner or later
1787 static inline sector_t
drbd_get_max_capacity(struct drbd_backing_dev
*bdev
)
1793 meta_dev_idx
= rcu_dereference(bdev
->disk_conf
)->meta_dev_idx
;
1796 switch (meta_dev_idx
) {
1797 case DRBD_MD_INDEX_INTERNAL
:
1798 case DRBD_MD_INDEX_FLEX_INT
:
1799 s
= drbd_get_capacity(bdev
->backing_bdev
)
1800 ? min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1801 _drbd_md_first_sector(meta_dev_idx
, bdev
))
1804 case DRBD_MD_INDEX_FLEX_EXT
:
1805 s
= min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1806 drbd_get_capacity(bdev
->backing_bdev
));
1807 /* clip at maximum size the meta device can support */
1808 s
= min_t(sector_t
, s
,
1809 BM_EXT_TO_SECT(bdev
->md
.md_size_sect
1810 - bdev
->md
.bm_offset
));
1813 s
= min_t(sector_t
, DRBD_MAX_SECTORS
,
1814 drbd_get_capacity(bdev
->backing_bdev
));
1820 * drbd_md_ss__() - Return the sector number of our meta data super block
1821 * @mdev: DRBD device.
1822 * @bdev: Meta data block device.
1824 static inline sector_t
drbd_md_ss__(struct drbd_conf
*mdev
,
1825 struct drbd_backing_dev
*bdev
)
1830 meta_dev_idx
= rcu_dereference(bdev
->disk_conf
)->meta_dev_idx
;
1833 switch (meta_dev_idx
) {
1834 default: /* external, some index */
1835 return MD_RESERVED_SECT
* meta_dev_idx
;
1836 case DRBD_MD_INDEX_INTERNAL
:
1837 /* with drbd08, internal meta data is always "flexible" */
1838 case DRBD_MD_INDEX_FLEX_INT
:
1839 /* sizeof(struct md_on_disk_07) == 4k
1840 * position: last 4k aligned block of 4k size */
1841 if (!bdev
->backing_bdev
) {
1842 if (__ratelimit(&drbd_ratelimit_state
)) {
1843 dev_err(DEV
, "bdev->backing_bdev==NULL\n");
1848 return (drbd_get_capacity(bdev
->backing_bdev
) & ~7ULL)
1850 case DRBD_MD_INDEX_FLEX_EXT
:
1856 drbd_queue_work_front(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1858 unsigned long flags
;
1859 spin_lock_irqsave(&q
->q_lock
, flags
);
1860 list_add(&w
->list
, &q
->q
);
1861 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1862 wake_up(&q
->q_wait
);
1866 drbd_queue_work(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1868 unsigned long flags
;
1869 spin_lock_irqsave(&q
->q_lock
, flags
);
1870 list_add_tail(&w
->list
, &q
->q
);
1871 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1872 wake_up(&q
->q_wait
);
1875 static inline void wake_asender(struct drbd_tconn
*tconn
)
1877 if (test_bit(SIGNAL_ASENDER
, &tconn
->flags
))
1878 force_sig(DRBD_SIG
, tconn
->asender
.task
);
1881 static inline void request_ping(struct drbd_tconn
*tconn
)
1883 set_bit(SEND_PING
, &tconn
->flags
);
1884 wake_asender(tconn
);
1887 extern void *conn_prepare_command(struct drbd_tconn
*, struct drbd_socket
*);
1888 extern void *drbd_prepare_command(struct drbd_conf
*, struct drbd_socket
*);
1889 extern int conn_send_command(struct drbd_tconn
*, struct drbd_socket
*,
1890 enum drbd_packet
, unsigned int, void *,
1892 extern int drbd_send_command(struct drbd_conf
*, struct drbd_socket
*,
1893 enum drbd_packet
, unsigned int, void *,
1896 extern int drbd_send_ping(struct drbd_tconn
*tconn
);
1897 extern int drbd_send_ping_ack(struct drbd_tconn
*tconn
);
1898 extern int drbd_send_state_req(struct drbd_conf
*, union drbd_state
, union drbd_state
);
1899 extern int conn_send_state_req(struct drbd_tconn
*, union drbd_state
, union drbd_state
);
1901 static inline void drbd_thread_stop(struct drbd_thread
*thi
)
1903 _drbd_thread_stop(thi
, false, true);
1906 static inline void drbd_thread_stop_nowait(struct drbd_thread
*thi
)
1908 _drbd_thread_stop(thi
, false, false);
1911 static inline void drbd_thread_restart_nowait(struct drbd_thread
*thi
)
1913 _drbd_thread_stop(thi
, true, false);
1916 /* counts how many answer packets packets we expect from our peer,
1917 * for either explicit application requests,
1918 * or implicit barrier packets as necessary.
1921 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1922 * it is much easier and equally valid to count what we queue for the
1923 * worker, even before it actually was queued or send.
1924 * (drbd_make_request_common; recovery path on read io-error)
1926 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1927 * _req_mod(req, DATA_RECEIVED)
1928 * [from receive_DataReply]
1929 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1930 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1931 * for some reason it is NOT decreased in got_NegAck,
1932 * but in the resulting cleanup code from report_params.
1933 * we should try to remember the reason for that...
1934 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
1935 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
1936 * [from tl_clear_barrier]
1938 static inline void inc_ap_pending(struct drbd_conf
*mdev
)
1940 atomic_inc(&mdev
->ap_pending_cnt
);
1943 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \
1944 if (atomic_read(&mdev->which) < 0) \
1945 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
1947 atomic_read(&mdev->which))
1949 #define dec_ap_pending(mdev) _dec_ap_pending(mdev, __FUNCTION__, __LINE__)
1950 static inline void _dec_ap_pending(struct drbd_conf
*mdev
, const char *func
, int line
)
1952 if (atomic_dec_and_test(&mdev
->ap_pending_cnt
))
1953 wake_up(&mdev
->misc_wait
);
1954 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt
, func
, line
);
1957 /* counts how many resync-related answers we still expect from the peer
1959 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1960 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
1961 * (or P_NEG_ACK with ID_SYNCER)
1963 static inline void inc_rs_pending(struct drbd_conf
*mdev
)
1965 atomic_inc(&mdev
->rs_pending_cnt
);
1968 #define dec_rs_pending(mdev) _dec_rs_pending(mdev, __FUNCTION__, __LINE__)
1969 static inline void _dec_rs_pending(struct drbd_conf
*mdev
, const char *func
, int line
)
1971 atomic_dec(&mdev
->rs_pending_cnt
);
1972 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt
, func
, line
);
1975 /* counts how many answers we still need to send to the peer.
1977 * receive_Data unless protocol A;
1978 * we need to send a P_RECV_ACK (proto B)
1979 * or P_WRITE_ACK (proto C)
1980 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1981 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1982 * receive_Barrier_* we need to send a P_BARRIER_ACK
1984 static inline void inc_unacked(struct drbd_conf
*mdev
)
1986 atomic_inc(&mdev
->unacked_cnt
);
1989 #define dec_unacked(mdev) _dec_unacked(mdev, __FUNCTION__, __LINE__)
1990 static inline void _dec_unacked(struct drbd_conf
*mdev
, const char *func
, int line
)
1992 atomic_dec(&mdev
->unacked_cnt
);
1993 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt
, func
, line
);
1996 #define sub_unacked(mdev, n) _sub_unacked(mdev, n, __FUNCTION__, __LINE__)
1997 static inline void _sub_unacked(struct drbd_conf
*mdev
, int n
, const char *func
, int line
)
1999 atomic_sub(n
, &mdev
->unacked_cnt
);
2000 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt
, func
, line
);
2004 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2007 * You have to call put_ldev() when finished working with mdev->ldev.
2009 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2010 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2012 static inline void put_ldev(struct drbd_conf
*mdev
)
2014 int i
= atomic_dec_return(&mdev
->local_cnt
);
2016 /* This may be called from some endio handler,
2017 * so we must not sleep here. */
2022 if (mdev
->state
.disk
== D_DISKLESS
)
2023 /* even internal references gone, safe to destroy */
2024 drbd_ldev_destroy(mdev
);
2025 if (mdev
->state
.disk
== D_FAILED
)
2026 /* all application IO references gone. */
2027 drbd_go_diskless(mdev
);
2028 wake_up(&mdev
->misc_wait
);
2033 static inline int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
)
2037 /* never get a reference while D_DISKLESS */
2038 if (mdev
->state
.disk
== D_DISKLESS
)
2041 atomic_inc(&mdev
->local_cnt
);
2042 io_allowed
= (mdev
->state
.disk
>= mins
);
2048 extern int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
);
2051 /* you must have an "get_ldev" reference */
2052 static inline void drbd_get_syncer_progress(struct drbd_conf
*mdev
,
2053 unsigned long *bits_left
, unsigned int *per_mil_done
)
2055 /* this is to break it at compile time when we change that, in case we
2056 * want to support more than (1<<32) bits on a 32bit arch. */
2057 typecheck(unsigned long, mdev
->rs_total
);
2059 /* note: both rs_total and rs_left are in bits, i.e. in
2060 * units of BM_BLOCK_SIZE.
2061 * for the percentage, we don't care. */
2063 if (mdev
->state
.conn
== C_VERIFY_S
|| mdev
->state
.conn
== C_VERIFY_T
)
2064 *bits_left
= mdev
->ov_left
;
2066 *bits_left
= drbd_bm_total_weight(mdev
) - mdev
->rs_failed
;
2067 /* >> 10 to prevent overflow,
2068 * +1 to prevent division by zero */
2069 if (*bits_left
> mdev
->rs_total
) {
2070 /* doh. maybe a logic bug somewhere.
2071 * may also be just a race condition
2072 * between this and a disconnect during sync.
2073 * for now, just prevent in-kernel buffer overflow.
2076 dev_warn(DEV
, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2077 drbd_conn_str(mdev
->state
.conn
),
2078 *bits_left
, mdev
->rs_total
, mdev
->rs_failed
);
2081 /* Make sure the division happens in long context.
2082 * We allow up to one petabyte storage right now,
2083 * at a granularity of 4k per bit that is 2**38 bits.
2084 * After shift right and multiplication by 1000,
2085 * this should still fit easily into a 32bit long,
2086 * so we don't need a 64bit division on 32bit arch.
2087 * Note: currently we don't support such large bitmaps on 32bit
2088 * arch anyways, but no harm done to be prepared for it here.
2090 unsigned int shift
= mdev
->rs_total
> UINT_MAX
? 16 : 10;
2091 unsigned long left
= *bits_left
>> shift
;
2092 unsigned long total
= 1UL + (mdev
->rs_total
>> shift
);
2093 unsigned long tmp
= 1000UL - left
* 1000UL/total
;
2094 *per_mil_done
= tmp
;
2099 /* this throttles on-the-fly application requests
2100 * according to max_buffers settings;
2101 * maybe re-implement using semaphores? */
2102 static inline int drbd_get_max_buffers(struct drbd_conf
*mdev
)
2104 struct net_conf
*nc
;
2108 nc
= rcu_dereference(mdev
->tconn
->net_conf
);
2109 mxb
= nc
? nc
->max_buffers
: 1000000; /* arbitrary limit on open requests */
2115 static inline int drbd_state_is_stable(struct drbd_conf
*mdev
)
2117 union drbd_dev_state s
= mdev
->state
;
2119 /* DO NOT add a default clause, we want the compiler to warn us
2120 * for any newly introduced state we may have forgotten to add here */
2122 switch ((enum drbd_conns
)s
.conn
) {
2123 /* new io only accepted when there is no connection, ... */
2125 case C_WF_CONNECTION
:
2126 /* ... or there is a well established connection. */
2132 case C_PAUSED_SYNC_S
:
2133 case C_PAUSED_SYNC_T
:
2136 /* transitional states, IO allowed */
2137 case C_DISCONNECTING
:
2141 case C_NETWORK_FAILURE
:
2142 case C_PROTOCOL_ERROR
:
2144 case C_WF_REPORT_PARAMS
:
2145 case C_STARTING_SYNC_S
:
2146 case C_STARTING_SYNC_T
:
2149 /* Allow IO in BM exchange states with new protocols */
2151 if (mdev
->tconn
->agreed_pro_version
< 96)
2155 /* no new io accepted in these states */
2157 case C_WF_SYNC_UUID
:
2163 switch ((enum drbd_disk_state
)s
.disk
) {
2165 case D_INCONSISTENT
:
2170 /* disk state is stable as well. */
2173 /* no new io accepted during transitional states */
2185 static inline int drbd_suspended(struct drbd_conf
*mdev
)
2187 struct drbd_tconn
*tconn
= mdev
->tconn
;
2189 return tconn
->susp
|| tconn
->susp_fen
|| tconn
->susp_nod
;
2192 static inline bool may_inc_ap_bio(struct drbd_conf
*mdev
)
2194 int mxb
= drbd_get_max_buffers(mdev
);
2196 if (drbd_suspended(mdev
))
2198 if (test_bit(SUSPEND_IO
, &mdev
->flags
))
2201 /* to avoid potential deadlock or bitmap corruption,
2202 * in various places, we only allow new application io
2203 * to start during "stable" states. */
2205 /* no new io accepted when attaching or detaching the disk */
2206 if (!drbd_state_is_stable(mdev
))
2209 /* since some older kernels don't have atomic_add_unless,
2210 * and we are within the spinlock anyways, we have this workaround. */
2211 if (atomic_read(&mdev
->ap_bio_cnt
) > mxb
)
2213 if (test_bit(BITMAP_IO
, &mdev
->flags
))
2218 static inline bool inc_ap_bio_cond(struct drbd_conf
*mdev
)
2222 spin_lock_irq(&mdev
->tconn
->req_lock
);
2223 rv
= may_inc_ap_bio(mdev
);
2225 atomic_inc(&mdev
->ap_bio_cnt
);
2226 spin_unlock_irq(&mdev
->tconn
->req_lock
);
2231 static inline void inc_ap_bio(struct drbd_conf
*mdev
)
2234 * as long as the device is suspended
2235 * until the bitmap is no longer on the fly during connection
2236 * handshake as long as we would exceed the max_buffer limit.
2238 * to avoid races with the reconnect code,
2239 * we need to atomic_inc within the spinlock. */
2241 wait_event(mdev
->misc_wait
, inc_ap_bio_cond(mdev
));
2244 static inline void dec_ap_bio(struct drbd_conf
*mdev
)
2246 int mxb
= drbd_get_max_buffers(mdev
);
2247 int ap_bio
= atomic_dec_return(&mdev
->ap_bio_cnt
);
2249 D_ASSERT(ap_bio
>= 0);
2251 if (ap_bio
== 0 && test_bit(BITMAP_IO
, &mdev
->flags
)) {
2252 if (!test_and_set_bit(BITMAP_IO_QUEUED
, &mdev
->flags
))
2253 drbd_queue_work(&mdev
->tconn
->sender_work
, &mdev
->bm_io_work
.w
);
2256 /* this currently does wake_up for every dec_ap_bio!
2257 * maybe rather introduce some type of hysteresis?
2258 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2260 wake_up(&mdev
->misc_wait
);
2263 static inline bool verify_can_do_stop_sector(struct drbd_conf
*mdev
)
2265 return mdev
->tconn
->agreed_pro_version
>= 97 &&
2266 mdev
->tconn
->agreed_pro_version
!= 100;
2269 static inline int drbd_set_ed_uuid(struct drbd_conf
*mdev
, u64 val
)
2271 int changed
= mdev
->ed_uuid
!= val
;
2272 mdev
->ed_uuid
= val
;
2276 static inline int drbd_queue_order_type(struct drbd_conf
*mdev
)
2278 /* sorry, we currently have no working implementation
2279 * of distributed TCQ stuff */
2280 #ifndef QUEUE_ORDERED_NONE
2281 #define QUEUE_ORDERED_NONE 0
2283 return QUEUE_ORDERED_NONE
;
2286 static inline void drbd_md_flush(struct drbd_conf
*mdev
)
2290 if (test_bit(MD_NO_FUA
, &mdev
->flags
))
2293 r
= blkdev_issue_flush(mdev
->ldev
->md_bdev
, GFP_KERNEL
, NULL
);
2295 set_bit(MD_NO_FUA
, &mdev
->flags
);
2296 dev_err(DEV
, "meta data flush failed with status %d, disabling md-flushes\n", r
);