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 /* drbd_meta-data.c (still in drbd_main.c) */
168 /* 4th incarnation of the disk layout. */
169 #define DRBD_MD_MAGIC (DRBD_MAGIC+4)
171 extern struct ratelimit_state drbd_ratelimit_state
;
172 extern struct idr minors
;
173 extern struct list_head drbd_tconns
;
174 extern struct mutex drbd_cfg_mutex
;
178 /* receiver (data socket) */
180 P_DATA_REPLY
= 0x01, /* Response to P_DATA_REQUEST */
181 P_RS_DATA_REPLY
= 0x02, /* Response to P_RS_DATA_REQUEST */
184 P_BECOME_SYNC_TARGET
= 0x05,
185 P_BECOME_SYNC_SOURCE
= 0x06,
186 P_UNPLUG_REMOTE
= 0x07, /* Used at various times to hint the peer */
187 P_DATA_REQUEST
= 0x08, /* Used to ask for a data block */
188 P_RS_DATA_REQUEST
= 0x09, /* Used to ask for a data block for resync */
195 P_AUTH_CHALLENGE
= 0x10,
196 P_AUTH_RESPONSE
= 0x11,
197 P_STATE_CHG_REQ
= 0x12,
199 /* asender (meta socket */
202 P_RECV_ACK
= 0x15, /* Used in protocol B */
203 P_WRITE_ACK
= 0x16, /* Used in protocol C */
204 P_RS_WRITE_ACK
= 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
205 P_DISCARD_WRITE
= 0x18, /* Used in proto C, two-primaries conflict detection */
206 P_NEG_ACK
= 0x19, /* Sent if local disk is unusable */
207 P_NEG_DREPLY
= 0x1a, /* Local disk is broken... */
208 P_NEG_RS_DREPLY
= 0x1b, /* Local disk is broken... */
209 P_BARRIER_ACK
= 0x1c,
210 P_STATE_CHG_REPLY
= 0x1d,
212 /* "new" commands, no longer fitting into the ordering scheme above */
214 P_OV_REQUEST
= 0x1e, /* data socket */
216 P_OV_RESULT
= 0x20, /* meta socket */
217 P_CSUM_RS_REQUEST
= 0x21, /* data socket */
218 P_RS_IS_IN_SYNC
= 0x22, /* meta socket */
219 P_SYNC_PARAM89
= 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
220 P_COMPRESSED_BITMAP
= 0x24, /* compressed or otherwise encoded bitmap transfer */
221 /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */
222 /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */
223 P_DELAY_PROBE
= 0x27, /* is used on BOTH sockets */
224 P_OUT_OF_SYNC
= 0x28, /* Mark as out of sync (Outrunning), data socket */
225 P_RS_CANCEL
= 0x29, /* meta: Used to cancel RS_DATA_REQUEST packet by SyncSource */
226 P_CONN_ST_CHG_REQ
= 0x2a, /* data sock: Connection wide state request */
227 P_CONN_ST_CHG_REPLY
= 0x2b, /* meta sock: Connection side state req reply */
228 P_RETRY_WRITE
= 0x2c, /* Protocol C: retry conflicting write request */
230 P_MAY_IGNORE
= 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
231 P_MAX_OPT_CMD
= 0x101,
233 /* special command ids for handshake */
235 P_INITIAL_META
= 0xfff1, /* First Packet on the MetaSock */
236 P_INITIAL_DATA
= 0xfff2, /* First Packet on the Socket */
238 P_CONNECTION_FEATURES
= 0xfffe /* FIXED for the next century! */
241 extern const char *cmdname(enum drbd_packet cmd
);
243 /* for sending/receiving the bitmap,
244 * possibly in some encoding scheme */
247 * stores total bits and long words
248 * of the bitmap, so we don't need to
249 * call the accessor functions over and again. */
250 unsigned long bm_bits
;
251 unsigned long bm_words
;
252 /* during xfer, current position within the bitmap */
253 unsigned long bit_offset
;
254 unsigned long word_offset
;
256 /* statistics; index: (h->command == P_BITMAP) */
261 extern void INFO_bm_xfer_stats(struct drbd_conf
*mdev
,
262 const char *direction
, struct bm_xfer_ctx
*c
);
264 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx
*c
)
266 /* word_offset counts "native long words" (32 or 64 bit),
268 * Encoded packet may end at an unaligned bit offset.
269 * In case a fallback clear text packet is transmitted in
270 * between, we adjust this offset back to the last 64bit
271 * aligned "native long word", which makes coding and decoding
272 * the plain text bitmap much more convenient. */
273 #if BITS_PER_LONG == 64
274 c
->word_offset
= c
->bit_offset
>> 6;
275 #elif BITS_PER_LONG == 32
276 c
->word_offset
= c
->bit_offset
>> 5;
277 c
->word_offset
&= ~(1UL);
279 # error "unsupported BITS_PER_LONG"
284 #define __packed __attribute__((packed))
287 /* This is the layout for a packet on the wire.
288 * The byteorder is the network byte order.
289 * (except block_id and barrier fields.
290 * these are pointers to local structs
291 * and have no relevance for the partner,
292 * which just echoes them as received.)
294 * NOTE that the payload starts at a long aligned offset,
295 * regardless of 32 or 64 bit arch!
300 u16 length
; /* bytes of data after this header */
303 /* Header for big packets, Used for data packets exceeding 64kB */
305 u16 magic
; /* use DRBD_MAGIC_BIG here */
307 u32 length
; /* Use only 24 bits of that. Ignore the highest 8 bit. */
312 struct p_header80 h80
;
313 struct p_header95 h95
;
318 extern unsigned int drbd_header_size(struct drbd_tconn
*tconn
);
321 * short commands, packets without payload, plain p_header:
324 * P_BECOME_SYNC_TARGET
325 * P_BECOME_SYNC_SOURCE
330 * commands with out-of-struct payload:
331 * P_BITMAP (no additional fields)
332 * P_DATA, P_DATA_REPLY (see p_data)
333 * P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
336 /* these defines must not be changed without changing the protocol version */
337 #define DP_HARDBARRIER 1 /* depricated */
338 #define DP_RW_SYNC 2 /* equals REQ_SYNC */
339 #define DP_MAY_SET_IN_SYNC 4
340 #define DP_UNPLUG 8 /* not used anymore */
341 #define DP_FUA 16 /* equals REQ_FUA */
342 #define DP_FLUSH 32 /* equals REQ_FLUSH */
343 #define DP_DISCARD 64 /* equals REQ_DISCARD */
346 struct p_header head
;
347 u64 sector
; /* 64 bits sector number */
348 u64 block_id
; /* to identify the request in protocol B&C */
354 * commands which share a struct:
356 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
357 * P_DISCARD_WRITE (proto C, two-primaries conflict detection)
359 * P_DATA_REQUEST, P_RS_DATA_REQUEST
362 struct p_header head
;
370 struct p_header head
;
374 u32 pad
; /* to multiple of 8 Byte */
378 * commands with their own struct for additional fields:
379 * P_CONNECTION_FEATURES
386 struct p_connection_features
{
387 struct p_header head
; /* Note: vnr will be ignored */
392 /* should be more than enough for future enhancements
393 * for now, feature_flags and the reserverd array shall be zero.
399 /* 80 bytes, FIXED for the next century */
402 struct p_header head
;
403 u32 barrier
; /* barrier number _handle_ only */
404 u32 pad
; /* to multiple of 8 Byte */
407 struct p_barrier_ack
{
408 struct p_header head
;
414 struct p_header head
;
417 /* Since protocol version 88 and higher. */
421 struct p_rs_param_89
{
422 struct p_header head
;
424 /* protocol version 89: */
425 char verify_alg
[SHARED_SECRET_MAX
];
426 char csums_alg
[SHARED_SECRET_MAX
];
429 struct p_rs_param_95
{
430 struct p_header head
;
432 char verify_alg
[SHARED_SECRET_MAX
];
433 char csums_alg
[SHARED_SECRET_MAX
];
440 enum drbd_conn_flags
{
446 struct p_header head
;
454 /* Since protocol version 87 and higher. */
455 char integrity_alg
[0];
460 struct p_header head
;
461 u64 uuid
[UI_EXTENDED_SIZE
];
465 struct p_header head
;
470 struct p_header head
;
471 u64 d_size
; /* size of disk */
472 u64 u_size
; /* user requested size */
473 u64 c_size
; /* current exported size */
474 u32 max_bio_size
; /* Maximal size of a BIO */
475 u16 queue_order_type
; /* not yet implemented in DRBD*/
476 u16 dds_flags
; /* use enum dds_flags here. */
480 struct p_header head
;
485 struct p_header head
;
490 struct p_req_state_reply
{
491 struct p_header head
;
495 struct p_drbd06_param
{
506 struct p_header head
;
512 struct p_block_desc
{
513 struct p_header head
;
516 u32 pad
; /* to multiple of 8 Byte */
519 /* Valid values for the encoding field.
520 * Bump proto version when changing this. */
521 enum drbd_bitmap_code
{
522 /* RLE_VLI_Bytes = 0,
523 * and other bit variants had been defined during
524 * algorithm evaluation. */
528 struct p_compressed_bm
{
529 struct p_header head
;
530 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
531 * (encoding & 0x80): polarity (set/unset) of first runlength
532 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
533 * used to pad up to head.length bytes
540 struct p_delay_probe93
{
541 struct p_header head
;
542 u32 seq_num
; /* sequence number to match the two probe packets */
543 u32 offset
; /* usecs the probe got sent after the reference time point */
547 * Bitmap packets need to fit within a single page on the sender and receiver,
548 * so we are limited to 4 KiB (and not to PAGE_SIZE, which can be bigger).
550 #define DRBD_SOCKET_BUFFER_SIZE 4096
552 /**********************************************************************/
553 enum drbd_thread_state
{
562 struct task_struct
*task
;
563 struct completion stop
;
564 enum drbd_thread_state t_state
;
565 int (*function
) (struct drbd_thread
*);
566 struct drbd_tconn
*tconn
;
571 static inline enum drbd_thread_state
get_t_state(struct drbd_thread
*thi
)
573 /* THINK testing the t_state seems to be uncritical in all cases
574 * (but thread_{start,stop}), so we can read it *without* the lock.
582 struct list_head list
;
583 int (*cb
)(struct drbd_work
*, int cancel
);
585 struct drbd_conf
*mdev
;
586 struct drbd_tconn
*tconn
;
590 #include "drbd_interval.h"
592 extern int drbd_wait_misc(struct drbd_conf
*, struct drbd_interval
*);
594 struct drbd_request
{
597 /* if local IO is not allowed, will be NULL.
598 * if local IO _is_ allowed, holds the locally submitted bio clone,
599 * or, after local IO completion, the ERR_PTR(error).
600 * see drbd_request_endio(). */
601 struct bio
*private_bio
;
603 struct drbd_interval i
;
604 unsigned int epoch
; /* barrier_nr */
606 /* barrier_nr: used to check on "completion" whether this req was in
607 * the current epoch, and we therefore have to close it,
608 * starting a new epoch...
611 struct list_head tl_requests
; /* ring list in the transfer log */
612 struct bio
*master_bio
; /* master bio pointer */
613 unsigned long rq_state
; /* see comments above _req_mod() */
615 unsigned long start_time
;
618 struct drbd_tl_epoch
{
620 struct list_head requests
; /* requests before */
621 struct drbd_tl_epoch
*next
; /* pointer to the next barrier */
622 unsigned int br_number
; /* the barriers identifier. */
623 int n_writes
; /* number of requests attached before this barrier */
627 struct list_head list
;
628 unsigned int barrier_nr
;
629 atomic_t epoch_size
; /* increased on every request added. */
630 atomic_t active
; /* increased on every req. added, and dec on every finished. */
634 /* drbd_epoch flag bits */
636 DE_HAVE_BARRIER_NUMBER
,
643 EV_CLEANUP
= 32, /* used as flag */
646 struct drbd_wq_barrier
{
648 struct completion done
;
656 struct drbd_peer_request
{
658 struct drbd_epoch
*epoch
; /* for writes */
660 atomic_t pending_bios
;
661 struct drbd_interval i
;
662 /* see comments on ee flag bits below */
666 struct digest_info
*digest
;
671 * While corresponding bios are in flight, the only modification will be
672 * set_bit WAS_ERROR, which has to be atomic.
673 * If no bios are in flight yet, or all have been completed,
674 * non-atomic modification to ee->flags is ok.
677 __EE_CALL_AL_COMPLETE_IO
,
678 __EE_MAY_SET_IN_SYNC
,
680 /* In case a barrier failed,
681 * we need to resubmit without the barrier flag. */
684 /* we may have several bios per peer request.
685 * if any of those fail, we set this flag atomically
686 * from the endio callback */
689 /* This ee has a pointer to a digest instead of a block id */
692 /* Conflicting local requests need to be restarted after this request */
693 __EE_RESTART_REQUESTS
,
695 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
696 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
697 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
698 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
699 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
700 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
702 /* flag bits per mdev */
704 CREATE_BARRIER
, /* next P_DATA is preceded by a P_BARRIER */
705 UNPLUG_QUEUED
, /* only relevant with kernel 2.4 */
706 UNPLUG_REMOTE
, /* sending a "UnplugRemote" could help */
707 MD_DIRTY
, /* current uuids and flags not yet on disk */
708 USE_DEGR_WFC_T
, /* degr-wfc-timeout instead of wfc-timeout. */
711 CRASHED_PRIMARY
, /* This node was a crashed primary.
712 * Gets cleared when the state.conn
713 * goes into C_CONNECTED state. */
714 NO_BARRIER_SUPP
, /* underlying block device doesn't implement barriers */
717 MD_NO_FUA
, /* Users wants us to not use FUA/FLUSH on meta data dev */
718 SUSPEND_IO
, /* suspend application io */
719 BITMAP_IO
, /* suspend application io;
720 once no more io in flight, start bitmap io */
721 BITMAP_IO_QUEUED
, /* Started bitmap IO */
722 GO_DISKLESS
, /* Disk is being detached, on io-error or admin request. */
723 WAS_IO_ERROR
, /* Local disk failed returned IO error */
724 RESYNC_AFTER_NEG
, /* Resync after online grow after the attach&negotiate finished. */
725 RESIZE_PENDING
, /* Size change detected locally, waiting for the response from
726 * the peer, if it changed there as well. */
727 NEW_CUR_UUID
, /* Create new current UUID when thawing IO */
728 AL_SUSPENDED
, /* Activity logging is currently suspended. */
729 AHEAD_TO_SYNC_SOURCE
, /* Ahead -> SyncSource queued */
730 B_RS_H_DONE
, /* Before resync handler done (already executed) */
733 struct drbd_bitmap
; /* opaque for drbd_conf */
735 /* definition of bits in bm_flags to be used in drbd_bm_lock
736 * and drbd_bitmap_io and friends. */
738 /* do we need to kfree, or vfree bm_pages? */
739 BM_P_VMALLOCED
= 0x10000, /* internal use only, will be masked out */
741 /* currently locked for bulk operation */
742 BM_LOCKED_MASK
= 0x7,
744 /* in detail, that is: */
749 /* (test bit, count bit) allowed (common case) */
750 BM_LOCKED_TEST_ALLOWED
= 0x3,
752 /* testing bits, as well as setting new bits allowed, but clearing bits
753 * would be unexpected. Used during bitmap receive. Setting new bits
754 * requires sending of "out-of-sync" information, though. */
755 BM_LOCKED_SET_ALLOWED
= 0x1,
757 /* clear is not expected while bitmap is locked for bulk operation */
761 /* TODO sort members for performance
762 * MAYBE group them further */
764 /* THINK maybe we actually want to use the default "event/%s" worker threads
765 * or similar in linux 2.6, which uses per cpu data and threads.
767 struct drbd_work_queue
{
769 struct semaphore s
; /* producers up it, worker down()s it */
770 spinlock_t q_lock
; /* to protect the list. */
774 struct drbd_work_queue work
;
776 struct socket
*socket
;
777 /* this way we get our
778 * send/receive buffers off the stack */
784 u64 md_offset
; /* sector offset to 'super' block */
786 u64 la_size_sect
; /* last agreed size, unit sectors */
792 s32 al_offset
; /* signed relative sector offset to al area */
793 s32 bm_offset
; /* signed relative sector offset to bitmap */
795 /* u32 al_nr_extents; important for restoring the AL
796 * is stored into ldev->dc.al_extents, which in turn
797 * gets applied to act_log->nr_elements
801 struct drbd_backing_dev
{
802 struct block_device
*backing_bdev
;
803 struct block_device
*md_bdev
;
805 struct disk_conf dc
; /* The user provided config... */
806 sector_t known_size
; /* last known size of that backing device */
810 struct drbd_conf
*mdev
;
811 struct completion event
;
819 int (*io_fn
)(struct drbd_conf
*mdev
);
820 void (*done
)(struct drbd_conf
*mdev
, int rv
);
823 enum write_ordering_e
{
831 unsigned int head_index
;
835 /* flag bits per tconn */
837 NET_CONGESTED
, /* The data socket is congested */
838 DISCARD_CONCURRENT
, /* Set on one node, cleared on the peer! */
839 SEND_PING
, /* whether asender should send a ping asap */
840 SIGNAL_ASENDER
, /* whether asender wants to be interrupted */
841 GOT_PING_ACK
, /* set when we receive a ping_ack packet, ping_wait gets woken */
844 CONFIG_PENDING
, /* serialization of (re)configuration requests.
845 * if set, also prevents the device from dying */
846 OBJECT_DYING
, /* device became unconfigured,
847 * but worker thread is still handling the cleanup.
848 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
849 * while this is set. */
850 CONN_DRY_RUN
, /* Expect disconnect after resync handshake. */
853 struct drbd_tconn
{ /* is a resource from the config file */
854 char *name
; /* Resource name */
855 struct list_head all_tconn
; /* linked on global drbd_tconns */
856 struct idr volumes
; /* <tconn, vnr> to mdev mapping */
857 enum drbd_conns cstate
; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
858 unsigned susp
:1; /* IO suspended by user */
859 unsigned susp_nod
:1; /* IO suspended because no data */
860 unsigned susp_fen
:1; /* IO suspended because fence peer handler runs */
861 struct mutex cstate_mutex
; /* Protects graceful disconnects */
864 struct net_conf
*net_conf
; /* protected by get_net_conf() and put_net_conf() */
865 atomic_t net_cnt
; /* Users of net_conf */
866 wait_queue_head_t net_cnt_wait
;
867 wait_queue_head_t ping_wait
; /* Woken upon reception of a ping, and a state change */
868 struct res_opts res_opts
;
870 struct drbd_socket data
; /* data/barrier/cstate/parameter packets */
871 struct drbd_socket meta
; /* ping/ack (metadata) packets */
872 int agreed_pro_version
; /* actually used protocol version */
873 unsigned long last_received
; /* in jiffies, either socket */
874 unsigned int ko_count
;
877 struct drbd_tl_epoch
*unused_spare_tle
; /* for pre-allocation */
878 struct drbd_tl_epoch
*newest_tle
;
879 struct drbd_tl_epoch
*oldest_tle
;
880 struct list_head out_of_sequence_requests
;
882 struct crypto_hash
*cram_hmac_tfm
;
883 struct crypto_hash
*integrity_w_tfm
; /* to be used by the worker thread */
884 struct crypto_hash
*integrity_r_tfm
; /* to be used by the receiver thread */
885 struct crypto_hash
*csums_tfm
;
886 struct crypto_hash
*verify_tfm
;
890 struct drbd_thread receiver
;
891 struct drbd_thread worker
;
892 struct drbd_thread asender
;
893 cpumask_var_t cpu_mask
;
897 struct drbd_tconn
*tconn
;
898 int vnr
; /* volume number within the connection */
900 /* things that are stored as / read from meta data on disk */
903 /* configured by drbdsetup */
904 struct drbd_backing_dev
*ldev
__protected_by(local
);
906 sector_t p_size
; /* partner's disk size */
907 struct request_queue
*rq_queue
;
908 struct block_device
*this_bdev
;
909 struct gendisk
*vdisk
;
911 struct drbd_work resync_work
,
916 struct timer_list resync_timer
;
917 struct timer_list md_sync_timer
;
918 struct timer_list start_resync_timer
;
919 struct timer_list request_timer
;
920 #ifdef DRBD_DEBUG_MD_SYNC
924 } last_md_mark_dirty
;
927 /* Used after attach while negotiating new disk state. */
928 union drbd_state new_state_tmp
;
930 union drbd_dev_state state
;
931 wait_queue_head_t misc_wait
;
932 wait_queue_head_t state_wait
; /* upon each state change. */
933 unsigned int send_cnt
;
934 unsigned int recv_cnt
;
935 unsigned int read_cnt
;
936 unsigned int writ_cnt
;
937 unsigned int al_writ_cnt
;
938 unsigned int bm_writ_cnt
;
939 atomic_t ap_bio_cnt
; /* Requests we need to complete */
940 atomic_t ap_pending_cnt
; /* AP data packets on the wire, ack expected */
941 atomic_t rs_pending_cnt
; /* RS request/data packets on the wire */
942 atomic_t unacked_cnt
; /* Need to send replys for */
943 atomic_t local_cnt
; /* Waiting for local completion */
945 /* Interval tree of pending local requests */
946 struct rb_root read_requests
;
947 struct rb_root write_requests
;
949 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
950 unsigned long rs_total
;
951 /* number of resync blocks that failed in this run */
952 unsigned long rs_failed
;
953 /* Syncer's start time [unit jiffies] */
954 unsigned long rs_start
;
955 /* cumulated time in PausedSyncX state [unit jiffies] */
956 unsigned long rs_paused
;
957 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
958 unsigned long rs_same_csum
;
959 #define DRBD_SYNC_MARKS 8
960 #define DRBD_SYNC_MARK_STEP (3*HZ)
961 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
962 unsigned long rs_mark_left
[DRBD_SYNC_MARKS
];
963 /* marks's time [unit jiffies] */
964 unsigned long rs_mark_time
[DRBD_SYNC_MARKS
];
965 /* current index into rs_mark_{left,time} */
968 /* where does the admin want us to start? (sector) */
969 sector_t ov_start_sector
;
970 /* where are we now? (sector) */
971 sector_t ov_position
;
972 /* Start sector of out of sync range (to merge printk reporting). */
973 sector_t ov_last_oos_start
;
974 /* size of out-of-sync range in sectors. */
975 sector_t ov_last_oos_size
;
976 unsigned long ov_left
; /* in bits */
978 struct drbd_bitmap
*bitmap
;
979 unsigned long bm_resync_fo
; /* bit offset for drbd_bm_find_next */
981 /* Used to track operations of resync... */
982 struct lru_cache
*resync
;
983 /* Number of locked elements in resync LRU */
984 unsigned int resync_locked
;
985 /* resync extent number waiting for application requests */
986 unsigned int resync_wenr
;
990 struct drbd_epoch
*current_epoch
;
991 spinlock_t epoch_lock
;
993 enum write_ordering_e write_ordering
;
994 struct list_head active_ee
; /* IO in progress (P_DATA gets written to disk) */
995 struct list_head sync_ee
; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
996 struct list_head done_ee
; /* need to send P_WRITE_ACK */
997 struct list_head read_ee
; /* [RS]P_DATA_REQUEST being read */
998 struct list_head net_ee
; /* zero-copy network send in progress */
1000 int next_barrier_nr
;
1001 struct list_head resync_reads
;
1002 atomic_t pp_in_use
; /* allocated from page pool */
1003 atomic_t pp_in_use_by_net
; /* sendpage()d, still referenced by tcp */
1004 wait_queue_head_t ee_wait
;
1005 struct page
*md_io_page
; /* one page buffer for md_io */
1006 struct mutex md_io_mutex
; /* protects the md_io_buffer */
1008 wait_queue_head_t al_wait
;
1009 struct lru_cache
*act_log
; /* activity log */
1010 unsigned int al_tr_number
;
1012 int al_tr_pos
; /* position of the next transaction in the journal */
1013 wait_queue_head_t seq_wait
;
1014 atomic_t packet_seq
;
1015 unsigned int peer_seq
;
1016 spinlock_t peer_seq_lock
;
1018 unsigned long comm_bm_set
; /* communicated number of set bits. */
1019 struct bm_io_work bm_io_work
;
1020 u64 ed_uuid
; /* UUID of the exposed data */
1021 struct mutex own_state_mutex
;
1022 struct mutex
*state_mutex
; /* either own_state_mutex or mdev->tconn->cstate_mutex */
1023 char congestion_reason
; /* Why we where congested... */
1024 atomic_t rs_sect_in
; /* for incoming resync data rate, SyncTarget */
1025 atomic_t rs_sect_ev
; /* for submitted resync data rate, both */
1026 int rs_last_sect_ev
; /* counter to compare with */
1027 int rs_last_events
; /* counter of read or write "events" (unit sectors)
1028 * on the lower level device when we last looked. */
1029 int c_sync_rate
; /* current resync rate after syncer throttle magic */
1030 struct fifo_buffer rs_plan_s
; /* correction values of resync planer */
1031 int rs_in_flight
; /* resync sectors in flight (to proxy, in proxy and from proxy) */
1032 int rs_planed
; /* resync sectors already planned */
1033 atomic_t ap_in_flight
; /* App sectors in flight (waiting for ack) */
1034 int peer_max_bio_size
;
1035 int local_max_bio_size
;
1038 static inline struct drbd_conf
*minor_to_mdev(unsigned int minor
)
1040 return (struct drbd_conf
*)idr_find(&minors
, minor
);
1043 static inline unsigned int mdev_to_minor(struct drbd_conf
*mdev
)
1048 static inline struct drbd_conf
*vnr_to_mdev(struct drbd_tconn
*tconn
, int vnr
)
1050 return (struct drbd_conf
*)idr_find(&tconn
->volumes
, vnr
);
1054 * function declarations
1055 *************************/
1061 DDSF_NO_RESYNC
= 2, /* Do not run a resync for the new space */
1064 extern void drbd_init_set_defaults(struct drbd_conf
*mdev
);
1065 extern int drbd_thread_start(struct drbd_thread
*thi
);
1066 extern void _drbd_thread_stop(struct drbd_thread
*thi
, int restart
, int wait
);
1067 extern char *drbd_task_to_thread_name(struct drbd_tconn
*tconn
, struct task_struct
*task
);
1069 extern void drbd_thread_current_set_cpu(struct drbd_thread
*thi
);
1070 extern void drbd_calc_cpu_mask(struct drbd_tconn
*tconn
);
1072 #define drbd_thread_current_set_cpu(A) ({})
1073 #define drbd_calc_cpu_mask(A) ({})
1075 extern void drbd_free_resources(struct drbd_conf
*mdev
);
1076 extern void tl_release(struct drbd_tconn
*, unsigned int barrier_nr
,
1077 unsigned int set_size
);
1078 extern void tl_clear(struct drbd_tconn
*);
1079 extern void _tl_add_barrier(struct drbd_tconn
*, struct drbd_tl_epoch
*);
1080 extern void drbd_free_sock(struct drbd_tconn
*tconn
);
1081 extern int drbd_send(struct drbd_tconn
*tconn
, struct socket
*sock
,
1082 void *buf
, size_t size
, unsigned msg_flags
);
1083 extern int drbd_send_all(struct drbd_tconn
*, struct socket
*, void *, size_t,
1086 extern int drbd_send_protocol(struct drbd_tconn
*tconn
);
1087 extern int drbd_send_uuids(struct drbd_conf
*mdev
);
1088 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf
*mdev
);
1089 extern void drbd_gen_and_send_sync_uuid(struct drbd_conf
*mdev
);
1090 extern int drbd_send_sizes(struct drbd_conf
*mdev
, int trigger_reply
, enum dds_flags flags
);
1091 extern int _conn_send_state_req(struct drbd_tconn
*, int vnr
, enum drbd_packet cmd
,
1092 union drbd_state
, union drbd_state
);
1093 extern int _drbd_send_state(struct drbd_conf
*mdev
);
1094 extern int drbd_send_state(struct drbd_conf
*mdev
);
1095 extern int drbd_send_sync_param(struct drbd_conf
*mdev
);
1096 extern void drbd_send_b_ack(struct drbd_conf
*mdev
, u32 barrier_nr
,
1098 extern int drbd_send_ack(struct drbd_conf
*, enum drbd_packet
,
1099 struct drbd_peer_request
*);
1100 extern void drbd_send_ack_rp(struct drbd_conf
*mdev
, enum drbd_packet cmd
,
1101 struct p_block_req
*rp
);
1102 extern void drbd_send_ack_dp(struct drbd_conf
*mdev
, enum drbd_packet cmd
,
1103 struct p_data
*dp
, int data_size
);
1104 extern int drbd_send_ack_ex(struct drbd_conf
*mdev
, enum drbd_packet cmd
,
1105 sector_t sector
, int blksize
, u64 block_id
);
1106 extern int drbd_send_out_of_sync(struct drbd_conf
*, struct drbd_request
*);
1107 extern int drbd_send_block(struct drbd_conf
*, enum drbd_packet
,
1108 struct drbd_peer_request
*);
1109 extern int drbd_send_dblock(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1110 extern int drbd_send_drequest(struct drbd_conf
*mdev
, int cmd
,
1111 sector_t sector
, int size
, u64 block_id
);
1112 extern int drbd_send_drequest_csum(struct drbd_conf
*mdev
, sector_t sector
,
1113 int size
, void *digest
, int digest_size
,
1114 enum drbd_packet cmd
);
1115 extern int drbd_send_ov_request(struct drbd_conf
*mdev
,sector_t sector
,int size
);
1117 extern int drbd_send_bitmap(struct drbd_conf
*mdev
);
1118 extern void drbd_send_sr_reply(struct drbd_conf
*mdev
, enum drbd_state_rv retcode
);
1119 extern void conn_send_sr_reply(struct drbd_tconn
*tconn
, enum drbd_state_rv retcode
);
1120 extern void drbd_free_bc(struct drbd_backing_dev
*ldev
);
1121 extern void drbd_mdev_cleanup(struct drbd_conf
*mdev
);
1122 void drbd_print_uuids(struct drbd_conf
*mdev
, const char *text
);
1124 extern void drbd_md_sync(struct drbd_conf
*mdev
);
1125 extern int drbd_md_read(struct drbd_conf
*mdev
, struct drbd_backing_dev
*bdev
);
1126 extern void drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1127 extern void _drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1128 extern void drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1129 extern void _drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1130 extern void drbd_uuid_set_bm(struct drbd_conf
*mdev
, u64 val
) __must_hold(local
);
1131 extern void drbd_md_set_flag(struct drbd_conf
*mdev
, int flags
) __must_hold(local
);
1132 extern void drbd_md_clear_flag(struct drbd_conf
*mdev
, int flags
)__must_hold(local
);
1133 extern int drbd_md_test_flag(struct drbd_backing_dev
*, int);
1134 #ifndef DRBD_DEBUG_MD_SYNC
1135 extern void drbd_md_mark_dirty(struct drbd_conf
*mdev
);
1137 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1138 extern void drbd_md_mark_dirty_(struct drbd_conf
*mdev
,
1139 unsigned int line
, const char *func
);
1141 extern void drbd_queue_bitmap_io(struct drbd_conf
*mdev
,
1142 int (*io_fn
)(struct drbd_conf
*),
1143 void (*done
)(struct drbd_conf
*, int),
1144 char *why
, enum bm_flag flags
);
1145 extern int drbd_bitmap_io(struct drbd_conf
*mdev
,
1146 int (*io_fn
)(struct drbd_conf
*),
1147 char *why
, enum bm_flag flags
);
1148 extern int drbd_bmio_set_n_write(struct drbd_conf
*mdev
);
1149 extern int drbd_bmio_clear_n_write(struct drbd_conf
*mdev
);
1150 extern void drbd_go_diskless(struct drbd_conf
*mdev
);
1151 extern void drbd_ldev_destroy(struct drbd_conf
*mdev
);
1154 We reserve a 128MB Block (4k aligned)
1155 * either at the end of the backing device
1156 * or on a separate meta data device. */
1158 /* The following numbers are sectors */
1159 /* Allows up to about 3.8TB, so if you want more,
1160 * you need to use the "flexible" meta data format. */
1161 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1162 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1163 #define MD_AL_SECTORS 64 /* = 32 kB on disk activity log ring buffer */
1164 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_SECTORS)
1166 /* we do all meta data IO in 4k blocks */
1167 #define MD_BLOCK_SHIFT 12
1168 #define MD_BLOCK_SIZE (1<<MD_BLOCK_SHIFT)
1170 /* One activity log extent represents 4M of storage */
1171 #define AL_EXTENT_SHIFT 22
1172 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1174 /* We could make these currently hardcoded constants configurable
1175 * variables at create-md time (or even re-configurable at runtime?).
1176 * Which will require some more changes to the DRBD "super block"
1179 * updates per transaction:
1180 * This many changes to the active set can be logged with one transaction.
1181 * This number is arbitrary.
1182 * context per transaction:
1183 * This many context extent numbers are logged with each transaction.
1184 * This number is resulting from the transaction block size (4k), the layout
1185 * of the transaction header, and the number of updates per transaction.
1186 * See drbd_actlog.c:struct al_transaction_on_disk
1188 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1189 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1191 #if BITS_PER_LONG == 32
1193 #define cpu_to_lel(A) cpu_to_le32(A)
1194 #define lel_to_cpu(A) le32_to_cpu(A)
1195 #elif BITS_PER_LONG == 64
1197 #define cpu_to_lel(A) cpu_to_le64(A)
1198 #define lel_to_cpu(A) le64_to_cpu(A)
1200 #error "LN2 of BITS_PER_LONG unknown!"
1204 /* 16MB sized 'bitmap extent' to track syncer usage */
1206 int rs_left
; /* number of bits set (out of sync) in this extent. */
1207 int rs_failed
; /* number of failed resync requests in this extent. */
1208 unsigned long flags
;
1209 struct lc_element lce
;
1212 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1213 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1214 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1218 * We need to store one bit for a block.
1219 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1220 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1221 * Bit 1 ==> local node thinks this block needs to be synced.
1224 #define SLEEP_TIME (HZ/10)
1226 /* We do bitmap IO in units of 4k blocks.
1227 * We also still have a hardcoded 4k per bit relation. */
1228 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1229 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1230 /* mostly arbitrarily set the represented size of one bitmap extent,
1231 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1232 * at 4k per bit resolution) */
1233 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
1234 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1236 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1237 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1240 /* thus many _storage_ sectors are described by one bit */
1241 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1242 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1243 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1245 /* bit to represented kilo byte conversion */
1246 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1248 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1249 * _storage_ sector is located in */
1250 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1252 /* how much _storage_ sectors we have per bitmap sector */
1253 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1254 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1256 /* in one sector of the bitmap, we have this many activity_log extents. */
1257 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1258 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1260 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1261 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1263 /* the extent in "PER_EXTENT" below is an activity log extent
1264 * we need that many (long words/bytes) to store the bitmap
1265 * of one AL_EXTENT_SIZE chunk of storage.
1266 * we can store the bitmap for that many AL_EXTENTS within
1267 * one sector of the _on_disk_ bitmap:
1268 * bit 0 bit 37 bit 38 bit (512*8)-1
1269 * ...|........|........|.. // ..|........|
1270 * sect. 0 `296 `304 ^(512*8*8)-1
1272 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1273 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1274 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1277 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1278 #define DRBD_MAX_SECTORS_BM \
1279 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1280 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1281 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1282 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1283 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1284 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1285 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1287 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1288 /* 16 TB in units of sectors */
1289 #if BITS_PER_LONG == 32
1290 /* adjust by one page worth of bitmap,
1291 * so we won't wrap around in drbd_bm_find_next_bit.
1292 * you should use 64bit OS for that much storage, anyways. */
1293 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1295 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1296 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1297 /* corresponds to (1UL << 38) bits right now. */
1301 /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE,
1302 * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte.
1303 * Since we may live in a mixed-platform cluster,
1304 * we limit us to a platform agnostic constant here for now.
1305 * A followup commit may allow even bigger BIO sizes,
1306 * once we thought that through. */
1307 #define DRBD_MAX_BIO_SIZE (1 << 20)
1308 #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1309 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1311 #define DRBD_MAX_BIO_SIZE_SAFE (1 << 12) /* Works always = 4k */
1313 #define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */
1315 extern int drbd_bm_init(struct drbd_conf
*mdev
);
1316 extern int drbd_bm_resize(struct drbd_conf
*mdev
, sector_t sectors
, int set_new_bits
);
1317 extern void drbd_bm_cleanup(struct drbd_conf
*mdev
);
1318 extern void drbd_bm_set_all(struct drbd_conf
*mdev
);
1319 extern void drbd_bm_clear_all(struct drbd_conf
*mdev
);
1320 /* set/clear/test only a few bits at a time */
1321 extern int drbd_bm_set_bits(
1322 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1323 extern int drbd_bm_clear_bits(
1324 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1325 extern int drbd_bm_count_bits(
1326 struct drbd_conf
*mdev
, const unsigned long s
, const unsigned long e
);
1327 /* bm_set_bits variant for use while holding drbd_bm_lock,
1328 * may process the whole bitmap in one go */
1329 extern void _drbd_bm_set_bits(struct drbd_conf
*mdev
,
1330 const unsigned long s
, const unsigned long e
);
1331 extern int drbd_bm_test_bit(struct drbd_conf
*mdev
, unsigned long bitnr
);
1332 extern int drbd_bm_e_weight(struct drbd_conf
*mdev
, unsigned long enr
);
1333 extern int drbd_bm_write_page(struct drbd_conf
*mdev
, unsigned int idx
) __must_hold(local
);
1334 extern int drbd_bm_read(struct drbd_conf
*mdev
) __must_hold(local
);
1335 extern void drbd_bm_mark_for_writeout(struct drbd_conf
*mdev
, int page_nr
);
1336 extern int drbd_bm_write(struct drbd_conf
*mdev
) __must_hold(local
);
1337 extern int drbd_bm_write_hinted(struct drbd_conf
*mdev
) __must_hold(local
);
1338 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf
*mdev
,
1339 unsigned long al_enr
);
1340 extern size_t drbd_bm_words(struct drbd_conf
*mdev
);
1341 extern unsigned long drbd_bm_bits(struct drbd_conf
*mdev
);
1342 extern sector_t
drbd_bm_capacity(struct drbd_conf
*mdev
);
1344 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1345 extern unsigned long drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1346 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1347 extern unsigned long _drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1348 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1349 extern unsigned long _drbd_bm_total_weight(struct drbd_conf
*mdev
);
1350 extern unsigned long drbd_bm_total_weight(struct drbd_conf
*mdev
);
1351 extern int drbd_bm_rs_done(struct drbd_conf
*mdev
);
1352 /* for receive_bitmap */
1353 extern void drbd_bm_merge_lel(struct drbd_conf
*mdev
, size_t offset
,
1354 size_t number
, unsigned long *buffer
);
1355 /* for _drbd_send_bitmap */
1356 extern void drbd_bm_get_lel(struct drbd_conf
*mdev
, size_t offset
,
1357 size_t number
, unsigned long *buffer
);
1359 extern void drbd_bm_lock(struct drbd_conf
*mdev
, char *why
, enum bm_flag flags
);
1360 extern void drbd_bm_unlock(struct drbd_conf
*mdev
);
1363 extern struct kmem_cache
*drbd_request_cache
;
1364 extern struct kmem_cache
*drbd_ee_cache
; /* peer requests */
1365 extern struct kmem_cache
*drbd_bm_ext_cache
; /* bitmap extents */
1366 extern struct kmem_cache
*drbd_al_ext_cache
; /* activity log extents */
1367 extern mempool_t
*drbd_request_mempool
;
1368 extern mempool_t
*drbd_ee_mempool
;
1370 /* drbd's page pool, used to buffer data received from the peer,
1371 * or data requested by the peer.
1373 * This does not have an emergency reserve.
1375 * When allocating from this pool, it first takes pages from the pool.
1376 * Only if the pool is depleted will try to allocate from the system.
1378 * The assumption is that pages taken from this pool will be processed,
1379 * and given back, "quickly", and then can be recycled, so we can avoid
1380 * frequent calls to alloc_page(), and still will be able to make progress even
1381 * under memory pressure.
1383 extern struct page
*drbd_pp_pool
;
1384 extern spinlock_t drbd_pp_lock
;
1385 extern int drbd_pp_vacant
;
1386 extern wait_queue_head_t drbd_pp_wait
;
1388 /* We also need a standard (emergency-reserve backed) page pool
1389 * for meta data IO (activity log, bitmap).
1390 * We can keep it global, as long as it is used as "N pages at a time".
1391 * 128 should be plenty, currently we probably can get away with as few as 1.
1393 #define DRBD_MIN_POOL_PAGES 128
1394 extern mempool_t
*drbd_md_io_page_pool
;
1396 /* We also need to make sure we get a bio
1397 * when we need it for housekeeping purposes */
1398 extern struct bio_set
*drbd_md_io_bio_set
;
1399 /* to allocate from that set */
1400 extern struct bio
*bio_alloc_drbd(gfp_t gfp_mask
);
1402 extern rwlock_t global_state_lock
;
1404 extern int conn_lowest_minor(struct drbd_tconn
*tconn
);
1405 enum drbd_ret_code
conn_new_minor(struct drbd_tconn
*tconn
, unsigned int minor
, int vnr
);
1406 extern void drbd_free_mdev(struct drbd_conf
*mdev
);
1407 extern void drbd_delete_device(unsigned int minor
);
1409 struct drbd_tconn
*drbd_new_tconn(const char *name
);
1410 extern void drbd_free_tconn(struct drbd_tconn
*tconn
);
1411 struct drbd_tconn
*conn_by_name(const char *name
);
1413 extern int proc_details
;
1416 extern int __drbd_make_request(struct drbd_conf
*, struct bio
*, unsigned long);
1417 extern int drbd_make_request(struct request_queue
*q
, struct bio
*bio
);
1418 extern int drbd_read_remote(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1419 extern int drbd_merge_bvec(struct request_queue
*q
, struct bvec_merge_data
*bvm
, struct bio_vec
*bvec
);
1420 extern int is_valid_ar_handle(struct drbd_request
*, sector_t
);
1424 extern int drbd_msg_put_info(const char *info
);
1425 extern void drbd_suspend_io(struct drbd_conf
*mdev
);
1426 extern void drbd_resume_io(struct drbd_conf
*mdev
);
1427 extern char *ppsize(char *buf
, unsigned long long size
);
1428 extern sector_t
drbd_new_dev_size(struct drbd_conf
*, struct drbd_backing_dev
*, int);
1429 enum determine_dev_size
{ dev_size_error
= -1, unchanged
= 0, shrunk
= 1, grew
= 2 };
1430 extern enum determine_dev_size
drbd_determine_dev_size(struct drbd_conf
*, enum dds_flags
) __must_hold(local
);
1431 extern void resync_after_online_grow(struct drbd_conf
*);
1432 extern void drbd_reconsider_max_bio_size(struct drbd_conf
*mdev
);
1433 extern enum drbd_state_rv
drbd_set_role(struct drbd_conf
*mdev
,
1434 enum drbd_role new_role
,
1436 extern bool conn_try_outdate_peer(struct drbd_tconn
*tconn
);
1437 extern void conn_try_outdate_peer_async(struct drbd_tconn
*tconn
);
1438 extern int drbd_khelper(struct drbd_conf
*mdev
, char *cmd
);
1441 extern int drbd_worker(struct drbd_thread
*thi
);
1442 extern int drbd_alter_sa(struct drbd_conf
*mdev
, int na
);
1443 extern void drbd_start_resync(struct drbd_conf
*mdev
, enum drbd_conns side
);
1444 extern void resume_next_sg(struct drbd_conf
*mdev
);
1445 extern void suspend_other_sg(struct drbd_conf
*mdev
);
1446 extern int drbd_resync_finished(struct drbd_conf
*mdev
);
1447 /* maybe rather drbd_main.c ? */
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 drbd_rs_controller_reset(struct drbd_conf
*mdev
);
1453 static inline void ov_out_of_sync_print(struct drbd_conf
*mdev
)
1455 if (mdev
->ov_last_oos_size
) {
1456 dev_err(DEV
, "Out of sync: start=%llu, size=%lu (sectors)\n",
1457 (unsigned long long)mdev
->ov_last_oos_start
,
1458 (unsigned long)mdev
->ov_last_oos_size
);
1460 mdev
->ov_last_oos_size
=0;
1464 extern void drbd_csum_bio(struct drbd_conf
*, struct crypto_hash
*, struct bio
*, void *);
1465 extern void drbd_csum_ee(struct drbd_conf
*, struct crypto_hash
*,
1466 struct drbd_peer_request
*, void *);
1467 /* worker callbacks */
1468 extern int w_read_retry_remote(struct drbd_work
*, int);
1469 extern int w_e_end_data_req(struct drbd_work
*, int);
1470 extern int w_e_end_rsdata_req(struct drbd_work
*, int);
1471 extern int w_e_end_csum_rs_req(struct drbd_work
*, int);
1472 extern int w_e_end_ov_reply(struct drbd_work
*, int);
1473 extern int w_e_end_ov_req(struct drbd_work
*, int);
1474 extern int w_ov_finished(struct drbd_work
*, int);
1475 extern int w_resync_timer(struct drbd_work
*, int);
1476 extern int w_send_write_hint(struct drbd_work
*, int);
1477 extern int w_make_resync_request(struct drbd_work
*, int);
1478 extern int w_send_dblock(struct drbd_work
*, int);
1479 extern int w_send_barrier(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_release_ee(struct drbd_conf
*mdev
, struct list_head
*list
);
1496 extern struct drbd_peer_request
*drbd_alloc_ee(struct drbd_conf
*,
1497 u64
, sector_t
, unsigned int,
1498 gfp_t
) __must_hold(local
);
1499 extern void drbd_free_some_ee(struct drbd_conf
*, struct drbd_peer_request
*,
1501 #define drbd_free_ee(m,e) drbd_free_some_ee(m, e, 0)
1502 #define drbd_free_net_ee(m,e) drbd_free_some_ee(m, e, 1)
1503 extern void drbd_wait_ee_list_empty(struct drbd_conf
*mdev
,
1504 struct list_head
*head
);
1505 extern void _drbd_wait_ee_list_empty(struct drbd_conf
*mdev
,
1506 struct list_head
*head
);
1507 extern void drbd_set_recv_tcq(struct drbd_conf
*mdev
, int tcq_enabled
);
1508 extern void _drbd_clear_done_ee(struct drbd_conf
*mdev
, struct list_head
*to_be_freed
);
1509 extern void conn_flush_workqueue(struct drbd_tconn
*tconn
);
1510 extern int drbd_connected(int vnr
, void *p
, void *data
);
1511 static inline void drbd_flush_workqueue(struct drbd_conf
*mdev
)
1513 conn_flush_workqueue(mdev
->tconn
);
1516 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1517 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1518 static inline int drbd_setsockopt(struct socket
*sock
, int level
, int optname
,
1519 char __user
*optval
, int optlen
)
1522 if (level
== SOL_SOCKET
)
1523 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
1525 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1530 static inline void drbd_tcp_cork(struct socket
*sock
)
1533 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1534 (char __user
*)&val
, sizeof(val
));
1537 static inline void drbd_tcp_uncork(struct socket
*sock
)
1540 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1541 (char __user
*)&val
, sizeof(val
));
1544 static inline void drbd_tcp_nodelay(struct socket
*sock
)
1547 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_NODELAY
,
1548 (char __user
*)&val
, sizeof(val
));
1551 static inline void drbd_tcp_quickack(struct socket
*sock
)
1554 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_QUICKACK
,
1555 (char __user
*)&val
, sizeof(val
));
1558 void drbd_bump_write_ordering(struct drbd_conf
*mdev
, enum write_ordering_e wo
);
1561 extern struct proc_dir_entry
*drbd_proc
;
1562 extern const struct file_operations drbd_proc_fops
;
1563 extern const char *drbd_conn_str(enum drbd_conns s
);
1564 extern const char *drbd_role_str(enum drbd_role s
);
1567 extern void drbd_al_begin_io(struct drbd_conf
*mdev
, struct drbd_interval
*i
);
1568 extern void drbd_al_complete_io(struct drbd_conf
*mdev
, struct drbd_interval
*i
);
1569 extern void drbd_rs_complete_io(struct drbd_conf
*mdev
, sector_t sector
);
1570 extern int drbd_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1571 extern int drbd_try_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1572 extern void drbd_rs_cancel_all(struct drbd_conf
*mdev
);
1573 extern int drbd_rs_del_all(struct drbd_conf
*mdev
);
1574 extern void drbd_rs_failed_io(struct drbd_conf
*mdev
,
1575 sector_t sector
, int size
);
1576 extern int drbd_al_read_log(struct drbd_conf
*mdev
, struct drbd_backing_dev
*);
1577 extern void drbd_advance_rs_marks(struct drbd_conf
*mdev
, unsigned long still_to_go
);
1578 extern void __drbd_set_in_sync(struct drbd_conf
*mdev
, sector_t sector
,
1579 int size
, const char *file
, const unsigned int line
);
1580 #define drbd_set_in_sync(mdev, sector, size) \
1581 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1582 extern int __drbd_set_out_of_sync(struct drbd_conf
*mdev
, sector_t sector
,
1583 int size
, const char *file
, const unsigned int line
);
1584 #define drbd_set_out_of_sync(mdev, sector, size) \
1585 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1586 extern void drbd_al_apply_to_bm(struct drbd_conf
*mdev
);
1587 extern void drbd_al_shrink(struct drbd_conf
*mdev
);
1590 /* state info broadcast */
1592 enum drbd_state_info_bcast_reason sib_reason
;
1596 unsigned helper_exit_code
;
1599 union drbd_state os
;
1600 union drbd_state ns
;
1604 void drbd_bcast_event(struct drbd_conf
*mdev
, const struct sib_info
*sib
);
1607 * inline helper functions
1608 *************************/
1610 /* see also page_chain_add and friends in drbd_receiver.c */
1611 static inline struct page
*page_chain_next(struct page
*page
)
1613 return (struct page
*)page_private(page
);
1615 #define page_chain_for_each(page) \
1616 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1617 page = page_chain_next(page))
1618 #define page_chain_for_each_safe(page, n) \
1619 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1621 static inline int drbd_bio_has_active_page(struct bio
*bio
)
1623 struct bio_vec
*bvec
;
1626 __bio_for_each_segment(bvec
, bio
, i
, 0) {
1627 if (page_count(bvec
->bv_page
) > 1)
1634 static inline int drbd_ee_has_active_page(struct drbd_peer_request
*peer_req
)
1636 struct page
*page
= peer_req
->pages
;
1637 page_chain_for_each(page
) {
1638 if (page_count(page
) > 1)
1644 static inline enum drbd_state_rv
1645 _drbd_set_state(struct drbd_conf
*mdev
, union drbd_state ns
,
1646 enum chg_state_flags flags
, struct completion
*done
)
1648 enum drbd_state_rv rv
;
1650 read_lock(&global_state_lock
);
1651 rv
= __drbd_set_state(mdev
, ns
, flags
, done
);
1652 read_unlock(&global_state_lock
);
1657 static inline union drbd_state
drbd_read_state(struct drbd_conf
*mdev
)
1659 union drbd_state rv
;
1661 rv
.i
= mdev
->state
.i
;
1662 rv
.susp
= mdev
->tconn
->susp
;
1663 rv
.susp_nod
= mdev
->tconn
->susp_nod
;
1664 rv
.susp_fen
= mdev
->tconn
->susp_fen
;
1669 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1670 static inline void __drbd_chk_io_error_(struct drbd_conf
*mdev
, int forcedetach
, const char *where
)
1672 switch (mdev
->ldev
->dc
.on_io_error
) {
1675 if (__ratelimit(&drbd_ratelimit_state
))
1676 dev_err(DEV
, "Local IO failed in %s.\n", where
);
1677 if (mdev
->state
.disk
> D_INCONSISTENT
)
1678 _drbd_set_state(_NS(mdev
, disk
, D_INCONSISTENT
), CS_HARD
, NULL
);
1681 /* NOTE fall through to detach case if forcedetach set */
1683 case EP_CALL_HELPER
:
1684 set_bit(WAS_IO_ERROR
, &mdev
->flags
);
1685 if (mdev
->state
.disk
> D_FAILED
) {
1686 _drbd_set_state(_NS(mdev
, disk
, D_FAILED
), CS_HARD
, NULL
);
1688 "Local IO failed in %s. Detaching...\n", where
);
1695 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1696 * @mdev: DRBD device.
1697 * @error: Error code passed to the IO completion callback
1698 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1700 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1702 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1703 static inline void drbd_chk_io_error_(struct drbd_conf
*mdev
,
1704 int error
, int forcedetach
, const char *where
)
1707 unsigned long flags
;
1708 spin_lock_irqsave(&mdev
->tconn
->req_lock
, flags
);
1709 __drbd_chk_io_error_(mdev
, forcedetach
, where
);
1710 spin_unlock_irqrestore(&mdev
->tconn
->req_lock
, flags
);
1716 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1717 * @bdev: Meta data block device.
1719 * BTW, for internal meta data, this happens to be the maximum capacity
1720 * we could agree upon with our peer node.
1722 static inline sector_t
drbd_md_first_sector(struct drbd_backing_dev
*bdev
)
1724 switch (bdev
->dc
.meta_dev_idx
) {
1725 case DRBD_MD_INDEX_INTERNAL
:
1726 case DRBD_MD_INDEX_FLEX_INT
:
1727 return bdev
->md
.md_offset
+ bdev
->md
.bm_offset
;
1728 case DRBD_MD_INDEX_FLEX_EXT
:
1730 return bdev
->md
.md_offset
;
1735 * drbd_md_last_sector() - Return the last sector number of the meta data area
1736 * @bdev: Meta data block device.
1738 static inline sector_t
drbd_md_last_sector(struct drbd_backing_dev
*bdev
)
1740 switch (bdev
->dc
.meta_dev_idx
) {
1741 case DRBD_MD_INDEX_INTERNAL
:
1742 case DRBD_MD_INDEX_FLEX_INT
:
1743 return bdev
->md
.md_offset
+ MD_AL_OFFSET
- 1;
1744 case DRBD_MD_INDEX_FLEX_EXT
:
1746 return bdev
->md
.md_offset
+ bdev
->md
.md_size_sect
;
1750 /* Returns the number of 512 byte sectors of the device */
1751 static inline sector_t
drbd_get_capacity(struct block_device
*bdev
)
1753 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1754 return bdev
? i_size_read(bdev
->bd_inode
) >> 9 : 0;
1758 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1759 * @bdev: Meta data block device.
1761 * returns the capacity we announce to out peer. we clip ourselves at the
1762 * various MAX_SECTORS, because if we don't, current implementation will
1763 * oops sooner or later
1765 static inline sector_t
drbd_get_max_capacity(struct drbd_backing_dev
*bdev
)
1768 switch (bdev
->dc
.meta_dev_idx
) {
1769 case DRBD_MD_INDEX_INTERNAL
:
1770 case DRBD_MD_INDEX_FLEX_INT
:
1771 s
= drbd_get_capacity(bdev
->backing_bdev
)
1772 ? min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1773 drbd_md_first_sector(bdev
))
1776 case DRBD_MD_INDEX_FLEX_EXT
:
1777 s
= min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1778 drbd_get_capacity(bdev
->backing_bdev
));
1779 /* clip at maximum size the meta device can support */
1780 s
= min_t(sector_t
, s
,
1781 BM_EXT_TO_SECT(bdev
->md
.md_size_sect
1782 - bdev
->md
.bm_offset
));
1785 s
= min_t(sector_t
, DRBD_MAX_SECTORS
,
1786 drbd_get_capacity(bdev
->backing_bdev
));
1792 * drbd_md_ss__() - Return the sector number of our meta data super block
1793 * @mdev: DRBD device.
1794 * @bdev: Meta data block device.
1796 static inline sector_t
drbd_md_ss__(struct drbd_conf
*mdev
,
1797 struct drbd_backing_dev
*bdev
)
1799 switch (bdev
->dc
.meta_dev_idx
) {
1800 default: /* external, some index */
1801 return MD_RESERVED_SECT
* bdev
->dc
.meta_dev_idx
;
1802 case DRBD_MD_INDEX_INTERNAL
:
1803 /* with drbd08, internal meta data is always "flexible" */
1804 case DRBD_MD_INDEX_FLEX_INT
:
1805 /* sizeof(struct md_on_disk_07) == 4k
1806 * position: last 4k aligned block of 4k size */
1807 if (!bdev
->backing_bdev
) {
1808 if (__ratelimit(&drbd_ratelimit_state
)) {
1809 dev_err(DEV
, "bdev->backing_bdev==NULL\n");
1814 return (drbd_get_capacity(bdev
->backing_bdev
) & ~7ULL)
1816 case DRBD_MD_INDEX_FLEX_EXT
:
1822 drbd_queue_work_front(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1824 unsigned long flags
;
1825 spin_lock_irqsave(&q
->q_lock
, flags
);
1826 list_add(&w
->list
, &q
->q
);
1827 up(&q
->s
); /* within the spinlock,
1828 see comment near end of drbd_worker() */
1829 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1833 drbd_queue_work(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1835 unsigned long flags
;
1836 spin_lock_irqsave(&q
->q_lock
, flags
);
1837 list_add_tail(&w
->list
, &q
->q
);
1838 up(&q
->s
); /* within the spinlock,
1839 see comment near end of drbd_worker() */
1840 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1843 static inline void wake_asender(struct drbd_tconn
*tconn
)
1845 if (test_bit(SIGNAL_ASENDER
, &tconn
->flags
))
1846 force_sig(DRBD_SIG
, tconn
->asender
.task
);
1849 static inline void request_ping(struct drbd_tconn
*tconn
)
1851 set_bit(SEND_PING
, &tconn
->flags
);
1852 wake_asender(tconn
);
1855 extern void *conn_prepare_command(struct drbd_tconn
*, struct drbd_socket
*);
1856 extern void *drbd_prepare_command(struct drbd_conf
*, struct drbd_socket
*);
1857 extern int conn_send_command(struct drbd_tconn
*, struct drbd_socket
*,
1858 enum drbd_packet
, unsigned int, void *,
1860 extern int drbd_send_command(struct drbd_conf
*, struct drbd_socket
*,
1861 enum drbd_packet
, unsigned int, void *,
1864 extern int drbd_send_ping(struct drbd_tconn
*tconn
);
1865 extern int drbd_send_ping_ack(struct drbd_tconn
*tconn
);
1866 extern int drbd_send_state_req(struct drbd_conf
*, union drbd_state
, union drbd_state
);
1867 extern int conn_send_state_req(struct drbd_tconn
*, union drbd_state
, union drbd_state
);
1869 static inline void drbd_thread_stop(struct drbd_thread
*thi
)
1871 _drbd_thread_stop(thi
, false, true);
1874 static inline void drbd_thread_stop_nowait(struct drbd_thread
*thi
)
1876 _drbd_thread_stop(thi
, false, false);
1879 static inline void drbd_thread_restart_nowait(struct drbd_thread
*thi
)
1881 _drbd_thread_stop(thi
, true, false);
1884 /* counts how many answer packets packets we expect from our peer,
1885 * for either explicit application requests,
1886 * or implicit barrier packets as necessary.
1889 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1890 * it is much easier and equally valid to count what we queue for the
1891 * worker, even before it actually was queued or send.
1892 * (drbd_make_request_common; recovery path on read io-error)
1894 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1895 * _req_mod(req, DATA_RECEIVED)
1896 * [from receive_DataReply]
1897 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1898 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1899 * for some reason it is NOT decreased in got_NegAck,
1900 * but in the resulting cleanup code from report_params.
1901 * we should try to remember the reason for that...
1902 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
1903 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
1904 * [from tl_clear_barrier]
1906 static inline void inc_ap_pending(struct drbd_conf
*mdev
)
1908 atomic_inc(&mdev
->ap_pending_cnt
);
1911 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \
1912 if (atomic_read(&mdev->which) < 0) \
1913 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
1915 atomic_read(&mdev->which))
1917 #define dec_ap_pending(mdev) _dec_ap_pending(mdev, __FUNCTION__, __LINE__)
1918 static inline void _dec_ap_pending(struct drbd_conf
*mdev
, const char *func
, int line
)
1920 if (atomic_dec_and_test(&mdev
->ap_pending_cnt
))
1921 wake_up(&mdev
->misc_wait
);
1922 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt
, func
, line
);
1925 /* counts how many resync-related answers we still expect from the peer
1927 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1928 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
1929 * (or P_NEG_ACK with ID_SYNCER)
1931 static inline void inc_rs_pending(struct drbd_conf
*mdev
)
1933 atomic_inc(&mdev
->rs_pending_cnt
);
1936 #define dec_rs_pending(mdev) _dec_rs_pending(mdev, __FUNCTION__, __LINE__)
1937 static inline void _dec_rs_pending(struct drbd_conf
*mdev
, const char *func
, int line
)
1939 atomic_dec(&mdev
->rs_pending_cnt
);
1940 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt
, func
, line
);
1943 /* counts how many answers we still need to send to the peer.
1945 * receive_Data unless protocol A;
1946 * we need to send a P_RECV_ACK (proto B)
1947 * or P_WRITE_ACK (proto C)
1948 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1949 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1950 * receive_Barrier_* we need to send a P_BARRIER_ACK
1952 static inline void inc_unacked(struct drbd_conf
*mdev
)
1954 atomic_inc(&mdev
->unacked_cnt
);
1957 #define dec_unacked(mdev) _dec_unacked(mdev, __FUNCTION__, __LINE__)
1958 static inline void _dec_unacked(struct drbd_conf
*mdev
, const char *func
, int line
)
1960 atomic_dec(&mdev
->unacked_cnt
);
1961 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt
, func
, line
);
1964 #define sub_unacked(mdev, n) _sub_unacked(mdev, n, __FUNCTION__, __LINE__)
1965 static inline void _sub_unacked(struct drbd_conf
*mdev
, int n
, const char *func
, int line
)
1967 atomic_sub(n
, &mdev
->unacked_cnt
);
1968 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt
, func
, line
);
1971 static inline void put_net_conf(struct drbd_tconn
*tconn
)
1973 if (atomic_dec_and_test(&tconn
->net_cnt
))
1974 wake_up(&tconn
->net_cnt_wait
);
1978 * get_net_conf() - Increase ref count on mdev->tconn->net_conf; Returns 0 if nothing there
1979 * @mdev: DRBD device.
1981 * You have to call put_net_conf() when finished working with mdev->tconn->net_conf.
1983 static inline int get_net_conf(struct drbd_tconn
*tconn
)
1987 atomic_inc(&tconn
->net_cnt
);
1988 have_net_conf
= tconn
->cstate
>= C_UNCONNECTED
;
1990 put_net_conf(tconn
);
1991 return have_net_conf
;
1995 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
1998 * You have to call put_ldev() when finished working with mdev->ldev.
2000 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2001 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2003 static inline void put_ldev(struct drbd_conf
*mdev
)
2005 int i
= atomic_dec_return(&mdev
->local_cnt
);
2007 /* This may be called from some endio handler,
2008 * so we must not sleep here. */
2013 if (mdev
->state
.disk
== D_DISKLESS
)
2014 /* even internal references gone, safe to destroy */
2015 drbd_ldev_destroy(mdev
);
2016 if (mdev
->state
.disk
== D_FAILED
)
2017 /* all application IO references gone. */
2018 drbd_go_diskless(mdev
);
2019 wake_up(&mdev
->misc_wait
);
2024 static inline int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
)
2028 /* never get a reference while D_DISKLESS */
2029 if (mdev
->state
.disk
== D_DISKLESS
)
2032 atomic_inc(&mdev
->local_cnt
);
2033 io_allowed
= (mdev
->state
.disk
>= mins
);
2039 extern int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
);
2042 /* you must have an "get_ldev" reference */
2043 static inline void drbd_get_syncer_progress(struct drbd_conf
*mdev
,
2044 unsigned long *bits_left
, unsigned int *per_mil_done
)
2046 /* this is to break it at compile time when we change that, in case we
2047 * want to support more than (1<<32) bits on a 32bit arch. */
2048 typecheck(unsigned long, mdev
->rs_total
);
2050 /* note: both rs_total and rs_left are in bits, i.e. in
2051 * units of BM_BLOCK_SIZE.
2052 * for the percentage, we don't care. */
2054 if (mdev
->state
.conn
== C_VERIFY_S
|| mdev
->state
.conn
== C_VERIFY_T
)
2055 *bits_left
= mdev
->ov_left
;
2057 *bits_left
= drbd_bm_total_weight(mdev
) - mdev
->rs_failed
;
2058 /* >> 10 to prevent overflow,
2059 * +1 to prevent division by zero */
2060 if (*bits_left
> mdev
->rs_total
) {
2061 /* doh. maybe a logic bug somewhere.
2062 * may also be just a race condition
2063 * between this and a disconnect during sync.
2064 * for now, just prevent in-kernel buffer overflow.
2067 dev_warn(DEV
, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2068 drbd_conn_str(mdev
->state
.conn
),
2069 *bits_left
, mdev
->rs_total
, mdev
->rs_failed
);
2072 /* Make sure the division happens in long context.
2073 * We allow up to one petabyte storage right now,
2074 * at a granularity of 4k per bit that is 2**38 bits.
2075 * After shift right and multiplication by 1000,
2076 * this should still fit easily into a 32bit long,
2077 * so we don't need a 64bit division on 32bit arch.
2078 * Note: currently we don't support such large bitmaps on 32bit
2079 * arch anyways, but no harm done to be prepared for it here.
2081 unsigned int shift
= mdev
->rs_total
>= (1ULL << 32) ? 16 : 10;
2082 unsigned long left
= *bits_left
>> shift
;
2083 unsigned long total
= 1UL + (mdev
->rs_total
>> shift
);
2084 unsigned long tmp
= 1000UL - left
* 1000UL/total
;
2085 *per_mil_done
= tmp
;
2090 /* this throttles on-the-fly application requests
2091 * according to max_buffers settings;
2092 * maybe re-implement using semaphores? */
2093 static inline int drbd_get_max_buffers(struct drbd_conf
*mdev
)
2095 int mxb
= 1000000; /* arbitrary limit on open requests */
2096 if (get_net_conf(mdev
->tconn
)) {
2097 mxb
= mdev
->tconn
->net_conf
->max_buffers
;
2098 put_net_conf(mdev
->tconn
);
2103 static inline int drbd_state_is_stable(struct drbd_conf
*mdev
)
2105 union drbd_dev_state s
= mdev
->state
;
2107 /* DO NOT add a default clause, we want the compiler to warn us
2108 * for any newly introduced state we may have forgotten to add here */
2110 switch ((enum drbd_conns
)s
.conn
) {
2111 /* new io only accepted when there is no connection, ... */
2113 case C_WF_CONNECTION
:
2114 /* ... or there is a well established connection. */
2120 case C_PAUSED_SYNC_S
:
2121 case C_PAUSED_SYNC_T
:
2124 /* transitional states, IO allowed */
2125 case C_DISCONNECTING
:
2129 case C_NETWORK_FAILURE
:
2130 case C_PROTOCOL_ERROR
:
2132 case C_WF_REPORT_PARAMS
:
2133 case C_STARTING_SYNC_S
:
2134 case C_STARTING_SYNC_T
:
2137 /* Allow IO in BM exchange states with new protocols */
2139 if (mdev
->tconn
->agreed_pro_version
< 96)
2143 /* no new io accepted in these states */
2145 case C_WF_SYNC_UUID
:
2151 switch ((enum drbd_disk_state
)s
.disk
) {
2153 case D_INCONSISTENT
:
2157 /* disk state is stable as well. */
2160 /* no new io accepted during tansitional states */
2173 static inline int drbd_suspended(struct drbd_conf
*mdev
)
2175 struct drbd_tconn
*tconn
= mdev
->tconn
;
2177 return tconn
->susp
|| tconn
->susp_fen
|| tconn
->susp_nod
;
2180 static inline bool may_inc_ap_bio(struct drbd_conf
*mdev
)
2182 int mxb
= drbd_get_max_buffers(mdev
);
2184 if (drbd_suspended(mdev
))
2186 if (test_bit(SUSPEND_IO
, &mdev
->flags
))
2189 /* to avoid potential deadlock or bitmap corruption,
2190 * in various places, we only allow new application io
2191 * to start during "stable" states. */
2193 /* no new io accepted when attaching or detaching the disk */
2194 if (!drbd_state_is_stable(mdev
))
2197 /* since some older kernels don't have atomic_add_unless,
2198 * and we are within the spinlock anyways, we have this workaround. */
2199 if (atomic_read(&mdev
->ap_bio_cnt
) > mxb
)
2201 if (test_bit(BITMAP_IO
, &mdev
->flags
))
2206 static inline bool inc_ap_bio_cond(struct drbd_conf
*mdev
)
2210 spin_lock_irq(&mdev
->tconn
->req_lock
);
2211 rv
= may_inc_ap_bio(mdev
);
2213 atomic_inc(&mdev
->ap_bio_cnt
);
2214 spin_unlock_irq(&mdev
->tconn
->req_lock
);
2219 static inline void inc_ap_bio(struct drbd_conf
*mdev
)
2222 * as long as the device is suspended
2223 * until the bitmap is no longer on the fly during connection
2224 * handshake as long as we would exeed the max_buffer limit.
2226 * to avoid races with the reconnect code,
2227 * we need to atomic_inc within the spinlock. */
2229 wait_event(mdev
->misc_wait
, inc_ap_bio_cond(mdev
));
2232 static inline void dec_ap_bio(struct drbd_conf
*mdev
)
2234 int mxb
= drbd_get_max_buffers(mdev
);
2235 int ap_bio
= atomic_dec_return(&mdev
->ap_bio_cnt
);
2237 D_ASSERT(ap_bio
>= 0);
2238 /* this currently does wake_up for every dec_ap_bio!
2239 * maybe rather introduce some type of hysteresis?
2240 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2242 wake_up(&mdev
->misc_wait
);
2243 if (ap_bio
== 0 && test_bit(BITMAP_IO
, &mdev
->flags
)) {
2244 if (!test_and_set_bit(BITMAP_IO_QUEUED
, &mdev
->flags
))
2245 drbd_queue_work(&mdev
->tconn
->data
.work
, &mdev
->bm_io_work
.w
);
2249 static inline int drbd_set_ed_uuid(struct drbd_conf
*mdev
, u64 val
)
2251 int changed
= mdev
->ed_uuid
!= val
;
2252 mdev
->ed_uuid
= val
;
2256 static inline int drbd_queue_order_type(struct drbd_conf
*mdev
)
2258 /* sorry, we currently have no working implementation
2259 * of distributed TCQ stuff */
2260 #ifndef QUEUE_ORDERED_NONE
2261 #define QUEUE_ORDERED_NONE 0
2263 return QUEUE_ORDERED_NONE
;
2266 static inline void drbd_md_flush(struct drbd_conf
*mdev
)
2270 if (test_bit(MD_NO_FUA
, &mdev
->flags
))
2273 r
= blkdev_issue_flush(mdev
->ldev
->md_bdev
, GFP_KERNEL
, NULL
);
2275 set_bit(MD_NO_FUA
, &mdev
->flags
);
2276 dev_err(DEV
, "meta data flush failed with status %d, disabling md-flushes\n", r
);