4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
6 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
7 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
8 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
10 drbd is free software; you can redistribute it and/or modify
11 it under the terms of the GNU General Public License as published by
12 the Free Software Foundation; either version 2, or (at your option)
15 drbd is distributed in the hope that it will be useful,
16 but WITHOUT ANY WARRANTY; without even the implied warranty of
17 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 GNU General Public License for more details.
20 You should have received a copy of the GNU General Public License
21 along with drbd; see the file COPYING. If not, write to
22 the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
29 #include <linux/compiler.h>
30 #include <linux/types.h>
31 #include <linux/list.h>
32 #include <linux/sched.h>
33 #include <linux/bitops.h>
34 #include <linux/slab.h>
35 #include <linux/crypto.h>
36 #include <linux/ratelimit.h>
37 #include <linux/tcp.h>
38 #include <linux/mutex.h>
39 #include <linux/major.h>
40 #include <linux/blkdev.h>
41 #include <linux/genhd.h>
42 #include <linux/idr.h>
44 #include <linux/lru_cache.h>
45 #include <linux/prefetch.h>
46 #include <linux/drbd_genl_api.h>
47 #include <linux/drbd.h>
48 #include "drbd_state.h"
49 #include "drbd_protocol.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 bool disable_sendpage
;
68 extern bool allow_oos
;
69 void tl_abort_disk_io(struct drbd_device
*device
);
71 #ifdef CONFIG_DRBD_FAULT_INJECTION
72 extern int enable_faults
;
73 extern int fault_rate
;
74 extern int fault_devs
;
77 extern char usermode_helper
[];
80 /* I don't remember why XCPU ...
81 * This is used to wake the asender,
82 * and to interrupt sending the sending task
85 #define DRBD_SIG SIGXCPU
87 /* This is used to stop/restart our threads.
88 * Cannot use SIGTERM nor SIGKILL, since these
89 * are sent out by init on runlevel changes
90 * I choose SIGHUP for now.
92 #define DRBD_SIGKILL SIGHUP
94 #define ID_IN_SYNC (4711ULL)
95 #define ID_OUT_OF_SYNC (4712ULL)
96 #define ID_SYNCER (-1ULL)
98 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
101 struct drbd_connection
;
104 /* to shorten dev_warn(DEV, "msg"); and relatives statements */
105 #define DEV (disk_to_dev(device->vdisk))
107 #define conn_printk(LEVEL, TCONN, FMT, ARGS...) \
108 printk(LEVEL "d-con %s: " FMT, TCONN->resource->name , ## ARGS)
109 #define conn_alert(TCONN, FMT, ARGS...) conn_printk(KERN_ALERT, TCONN, FMT, ## ARGS)
110 #define conn_crit(TCONN, FMT, ARGS...) conn_printk(KERN_CRIT, TCONN, FMT, ## ARGS)
111 #define conn_err(TCONN, FMT, ARGS...) conn_printk(KERN_ERR, TCONN, FMT, ## ARGS)
112 #define conn_warn(TCONN, FMT, ARGS...) conn_printk(KERN_WARNING, TCONN, FMT, ## ARGS)
113 #define conn_notice(TCONN, FMT, ARGS...) conn_printk(KERN_NOTICE, TCONN, FMT, ## ARGS)
114 #define conn_info(TCONN, FMT, ARGS...) conn_printk(KERN_INFO, TCONN, FMT, ## ARGS)
115 #define conn_dbg(TCONN, FMT, ARGS...) conn_printk(KERN_DEBUG, TCONN, FMT, ## ARGS)
117 #define D_ASSERT(exp) if (!(exp)) \
118 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
121 * expect - Make an assertion
123 * Unlike the assert macro, this macro returns a boolean result.
125 #define expect(exp) ({ \
126 bool _bool = (exp); \
128 dev_err(DEV, "ASSERTION %s FAILED in %s\n", \
133 /* Defines to control fault insertion */
135 DRBD_FAULT_MD_WR
= 0, /* meta data write */
136 DRBD_FAULT_MD_RD
= 1, /* read */
137 DRBD_FAULT_RS_WR
= 2, /* resync */
138 DRBD_FAULT_RS_RD
= 3,
139 DRBD_FAULT_DT_WR
= 4, /* data */
140 DRBD_FAULT_DT_RD
= 5,
141 DRBD_FAULT_DT_RA
= 6, /* data read ahead */
142 DRBD_FAULT_BM_ALLOC
= 7, /* bitmap allocation */
143 DRBD_FAULT_AL_EE
= 8, /* alloc ee */
144 DRBD_FAULT_RECEIVE
= 9, /* Changes some bytes upon receiving a [rs]data block */
150 _drbd_insert_fault(struct drbd_device
*device
, unsigned int type
);
153 drbd_insert_fault(struct drbd_device
*device
, unsigned int type
) {
154 #ifdef CONFIG_DRBD_FAULT_INJECTION
156 (enable_faults
& (1<<type
)) &&
157 _drbd_insert_fault(device
, type
);
163 /* integer division, round _UP_ to the next integer */
164 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
165 /* usual integer division */
166 #define div_floor(A, B) ((A)/(B))
168 extern struct ratelimit_state drbd_ratelimit_state
;
169 extern struct idr drbd_devices
; /* RCU, updates: genl_lock() */
170 extern struct list_head drbd_resources
; /* RCU, updates: genl_lock() */
172 extern const char *cmdname(enum drbd_packet cmd
);
174 /* for sending/receiving the bitmap,
175 * possibly in some encoding scheme */
178 * stores total bits and long words
179 * of the bitmap, so we don't need to
180 * call the accessor functions over and again. */
181 unsigned long bm_bits
;
182 unsigned long bm_words
;
183 /* during xfer, current position within the bitmap */
184 unsigned long bit_offset
;
185 unsigned long word_offset
;
187 /* statistics; index: (h->command == P_BITMAP) */
192 extern void INFO_bm_xfer_stats(struct drbd_device
*device
,
193 const char *direction
, struct bm_xfer_ctx
*c
);
195 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx
*c
)
197 /* word_offset counts "native long words" (32 or 64 bit),
199 * Encoded packet may end at an unaligned bit offset.
200 * In case a fallback clear text packet is transmitted in
201 * between, we adjust this offset back to the last 64bit
202 * aligned "native long word", which makes coding and decoding
203 * the plain text bitmap much more convenient. */
204 #if BITS_PER_LONG == 64
205 c
->word_offset
= c
->bit_offset
>> 6;
206 #elif BITS_PER_LONG == 32
207 c
->word_offset
= c
->bit_offset
>> 5;
208 c
->word_offset
&= ~(1UL);
210 # error "unsupported BITS_PER_LONG"
214 extern unsigned int drbd_header_size(struct drbd_connection
*connection
);
216 /**********************************************************************/
217 enum drbd_thread_state
{
226 struct task_struct
*task
;
227 struct completion stop
;
228 enum drbd_thread_state t_state
;
229 int (*function
) (struct drbd_thread
*);
230 struct drbd_connection
*connection
;
235 static inline enum drbd_thread_state
get_t_state(struct drbd_thread
*thi
)
237 /* THINK testing the t_state seems to be uncritical in all cases
238 * (but thread_{start,stop}), so we can read it *without* the lock.
246 struct list_head list
;
247 int (*cb
)(struct drbd_work
*, int cancel
);
249 struct drbd_device
*device
;
250 struct drbd_connection
*connection
;
254 #include "drbd_interval.h"
256 extern int drbd_wait_misc(struct drbd_device
*, struct drbd_interval
*);
258 struct drbd_request
{
261 /* if local IO is not allowed, will be NULL.
262 * if local IO _is_ allowed, holds the locally submitted bio clone,
263 * or, after local IO completion, the ERR_PTR(error).
264 * see drbd_request_endio(). */
265 struct bio
*private_bio
;
267 struct drbd_interval i
;
269 /* epoch: used to check on "completion" whether this req was in
270 * the current epoch, and we therefore have to close it,
271 * causing a p_barrier packet to be send, starting a new epoch.
273 * This corresponds to "barrier" in struct p_barrier[_ack],
274 * and to "barrier_nr" in struct drbd_epoch (and various
275 * comments/function parameters/local variable names).
279 struct list_head tl_requests
; /* ring list in the transfer log */
280 struct bio
*master_bio
; /* master bio pointer */
281 unsigned long start_time
;
283 /* once it hits 0, we may complete the master_bio */
284 atomic_t completion_ref
;
285 /* once it hits 0, we may destroy this drbd_request object */
288 unsigned rq_state
; /* see comments above _req_mod() */
292 struct drbd_connection
*connection
;
293 struct list_head list
;
294 unsigned int barrier_nr
;
295 atomic_t epoch_size
; /* increased on every request added. */
296 atomic_t active
; /* increased on every req. added, and dec on every finished. */
300 /* Prototype declaration of function defined in drbd_receiver.c */
301 int drbdd_init(struct drbd_thread
*);
302 int drbd_asender(struct drbd_thread
*);
304 /* drbd_epoch flag bits */
306 DE_HAVE_BARRIER_NUMBER
,
313 EV_CLEANUP
= 32, /* used as flag */
316 struct drbd_wq_barrier
{
318 struct completion done
;
326 struct drbd_peer_request
{
328 struct drbd_epoch
*epoch
; /* for writes */
330 atomic_t pending_bios
;
331 struct drbd_interval i
;
332 /* see comments on ee flag bits below */
336 struct digest_info
*digest
;
341 * While corresponding bios are in flight, the only modification will be
342 * set_bit WAS_ERROR, which has to be atomic.
343 * If no bios are in flight yet, or all have been completed,
344 * non-atomic modification to ee->flags is ok.
347 __EE_CALL_AL_COMPLETE_IO
,
348 __EE_MAY_SET_IN_SYNC
,
350 /* In case a barrier failed,
351 * we need to resubmit without the barrier flag. */
354 /* we may have several bios per peer request.
355 * if any of those fail, we set this flag atomically
356 * from the endio callback */
359 /* This ee has a pointer to a digest instead of a block id */
362 /* Conflicting local requests need to be restarted after this request */
363 __EE_RESTART_REQUESTS
,
365 /* The peer wants a write ACK for this (wire proto C) */
368 /* Is set when net_conf had two_primaries set while creating this peer_req */
369 __EE_IN_INTERVAL_TREE
,
371 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
372 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
373 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
374 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
375 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
376 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
377 #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK)
378 #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE)
380 /* flag bits per device */
382 UNPLUG_REMOTE
, /* sending a "UnplugRemote" could help */
383 MD_DIRTY
, /* current uuids and flags not yet on disk */
384 USE_DEGR_WFC_T
, /* degr-wfc-timeout instead of wfc-timeout. */
387 CRASHED_PRIMARY
, /* This node was a crashed primary.
388 * Gets cleared when the state.conn
389 * goes into C_CONNECTED state. */
392 MD_NO_FUA
, /* Users wants us to not use FUA/FLUSH on meta data dev */
393 SUSPEND_IO
, /* suspend application io */
394 BITMAP_IO
, /* suspend application io;
395 once no more io in flight, start bitmap io */
396 BITMAP_IO_QUEUED
, /* Started bitmap IO */
397 GO_DISKLESS
, /* Disk is being detached, on io-error or admin request. */
398 WAS_IO_ERROR
, /* Local disk failed, returned IO error */
399 WAS_READ_ERROR
, /* Local disk READ failed (set additionally to the above) */
400 FORCE_DETACH
, /* Force-detach from local disk, aborting any pending local IO */
401 RESYNC_AFTER_NEG
, /* Resync after online grow after the attach&negotiate finished. */
402 RESIZE_PENDING
, /* Size change detected locally, waiting for the response from
403 * the peer, if it changed there as well. */
404 NEW_CUR_UUID
, /* Create new current UUID when thawing IO */
405 AL_SUSPENDED
, /* Activity logging is currently suspended. */
406 AHEAD_TO_SYNC_SOURCE
, /* Ahead -> SyncSource queued */
407 B_RS_H_DONE
, /* Before resync handler done (already executed) */
408 DISCARD_MY_DATA
, /* discard_my_data flag per volume */
412 struct drbd_bitmap
; /* opaque for drbd_device */
414 /* definition of bits in bm_flags to be used in drbd_bm_lock
415 * and drbd_bitmap_io and friends. */
417 /* do we need to kfree, or vfree bm_pages? */
418 BM_P_VMALLOCED
= 0x10000, /* internal use only, will be masked out */
420 /* currently locked for bulk operation */
421 BM_LOCKED_MASK
= 0xf,
423 /* in detail, that is: */
428 /* so we can mark it locked for bulk operation,
429 * and still allow all non-bulk operations */
432 /* (test bit, count bit) allowed (common case) */
433 BM_LOCKED_TEST_ALLOWED
= BM_DONT_CLEAR
| BM_DONT_SET
| BM_IS_LOCKED
,
435 /* testing bits, as well as setting new bits allowed, but clearing bits
436 * would be unexpected. Used during bitmap receive. Setting new bits
437 * requires sending of "out-of-sync" information, though. */
438 BM_LOCKED_SET_ALLOWED
= BM_DONT_CLEAR
| BM_IS_LOCKED
,
440 /* for drbd_bm_write_copy_pages, everything is allowed,
441 * only concurrent bulk operations are locked out. */
442 BM_LOCKED_CHANGE_ALLOWED
= BM_IS_LOCKED
,
445 struct drbd_work_queue
{
447 spinlock_t q_lock
; /* to protect the list. */
448 wait_queue_head_t q_wait
;
453 struct socket
*socket
;
454 /* this way we get our
455 * send/receive buffers off the stack */
461 u64 md_offset
; /* sector offset to 'super' block */
463 u64 la_size_sect
; /* last agreed size, unit sectors */
464 spinlock_t uuid_lock
;
470 s32 al_offset
; /* signed relative sector offset to activity log */
471 s32 bm_offset
; /* signed relative sector offset to bitmap */
473 /* cached value of bdev->disk_conf->meta_dev_idx (see below) */
476 /* see al_tr_number_to_on_disk_sector() */
478 u32 al_stripe_size_4k
;
479 u32 al_size_4k
; /* cached product of the above */
482 struct drbd_backing_dev
{
483 struct block_device
*backing_bdev
;
484 struct block_device
*md_bdev
;
486 struct disk_conf
*disk_conf
; /* RCU, for updates: first_peer_device(device)->connection->conf_update */
487 sector_t known_size
; /* last known size of that backing device */
499 int (*io_fn
)(struct drbd_device
*device
);
500 void (*done
)(struct drbd_device
*device
, int rv
);
503 enum write_ordering_e
{
510 unsigned int head_index
;
512 int total
; /* sum of all values */
515 extern struct fifo_buffer
*fifo_alloc(int fifo_size
);
517 /* flag bits per connection */
519 NET_CONGESTED
, /* The data socket is congested */
520 RESOLVE_CONFLICTS
, /* Set on one node, cleared on the peer! */
521 SEND_PING
, /* whether asender should send a ping asap */
522 SIGNAL_ASENDER
, /* whether asender wants to be interrupted */
523 GOT_PING_ACK
, /* set when we receive a ping_ack packet, ping_wait gets woken */
524 CONN_WD_ST_CHG_REQ
, /* A cluster wide state change on the connection is active */
527 CONN_DRY_RUN
, /* Expect disconnect after resync handshake. */
528 CREATE_BARRIER
, /* next P_DATA is preceded by a P_BARRIER */
529 STATE_SENT
, /* Do not change state/UUIDs while this is set */
530 CALLBACK_PENDING
, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
531 * pending, from drbd worker context.
532 * If set, bdi_write_congested() returns true,
533 * so shrink_page_list() would not recurse into,
534 * and potentially deadlock on, this drbd worker.
539 struct drbd_resource
{
542 struct idr devices
; /* volume number to device mapping */
543 struct list_head connections
;
544 struct list_head resources
;
545 struct res_opts res_opts
;
548 struct drbd_connection
{
549 struct list_head connections
;
550 struct drbd_resource
*resource
;
552 struct idr peer_devices
; /* volume number to peer device mapping */
553 enum drbd_conns cstate
; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
554 unsigned susp
:1; /* IO suspended by user */
555 unsigned susp_nod
:1; /* IO suspended because no data */
556 unsigned susp_fen
:1; /* IO suspended because fence peer handler runs */
557 struct mutex cstate_mutex
; /* Protects graceful disconnects */
558 unsigned int connect_cnt
; /* Inc each time a connection is established */
561 struct net_conf
*net_conf
; /* content protected by rcu */
562 struct mutex conf_update
; /* mutex for ready-copy-update of net_conf and disk_conf */
563 wait_queue_head_t ping_wait
; /* Woken upon reception of a ping, and a state change */
565 struct sockaddr_storage my_addr
;
567 struct sockaddr_storage peer_addr
;
570 struct drbd_socket data
; /* data/barrier/cstate/parameter packets */
571 struct drbd_socket meta
; /* ping/ack (metadata) packets */
572 int agreed_pro_version
; /* actually used protocol version */
573 unsigned long last_received
; /* in jiffies, either socket */
574 unsigned int ko_count
;
578 struct list_head transfer_log
; /* all requests not yet fully processed */
580 struct crypto_hash
*cram_hmac_tfm
;
581 struct crypto_hash
*integrity_tfm
; /* checksums we compute, updates protected by connection->data->mutex */
582 struct crypto_hash
*peer_integrity_tfm
; /* checksums we verify, only accessed from receiver thread */
583 struct crypto_hash
*csums_tfm
;
584 struct crypto_hash
*verify_tfm
;
589 struct drbd_epoch
*current_epoch
;
590 spinlock_t epoch_lock
;
592 enum write_ordering_e write_ordering
;
593 atomic_t current_tle_nr
; /* transfer log epoch number */
594 unsigned current_tle_writes
; /* writes seen within this tl epoch */
596 unsigned long last_reconnect_jif
;
597 struct drbd_thread receiver
;
598 struct drbd_thread worker
;
599 struct drbd_thread asender
;
600 cpumask_var_t cpu_mask
;
603 struct drbd_work_queue sender_work
;
606 /* whether this sender thread
607 * has processed a single write yet. */
608 bool seen_any_write_yet
;
610 /* Which barrier number to send with the next P_BARRIER */
611 int current_epoch_nr
;
613 /* how many write requests have been sent
614 * with req->epoch == current_epoch_nr.
615 * If none, no P_BARRIER will be sent. */
616 unsigned current_epoch_writes
;
620 struct submit_worker
{
621 struct workqueue_struct
*wq
;
622 struct work_struct worker
;
625 struct list_head writes
;
628 struct drbd_peer_device
{
629 struct list_head peer_devices
;
630 struct drbd_device
*device
;
631 struct drbd_connection
*connection
;
635 struct drbd_resource
*resource
;
636 struct list_head peer_devices
;
637 int vnr
; /* volume number within the connection */
640 /* things that are stored as / read from meta data on disk */
643 /* configured by drbdsetup */
644 struct drbd_backing_dev
*ldev
__protected_by(local
);
646 sector_t p_size
; /* partner's disk size */
647 struct request_queue
*rq_queue
;
648 struct block_device
*this_bdev
;
649 struct gendisk
*vdisk
;
651 unsigned long last_reattach_jif
;
652 struct drbd_work resync_work
,
657 struct timer_list resync_timer
;
658 struct timer_list md_sync_timer
;
659 struct timer_list start_resync_timer
;
660 struct timer_list request_timer
;
661 #ifdef DRBD_DEBUG_MD_SYNC
665 } last_md_mark_dirty
;
668 /* Used after attach while negotiating new disk state. */
669 union drbd_state new_state_tmp
;
671 union drbd_dev_state state
;
672 wait_queue_head_t misc_wait
;
673 wait_queue_head_t state_wait
; /* upon each state change. */
674 unsigned int send_cnt
;
675 unsigned int recv_cnt
;
676 unsigned int read_cnt
;
677 unsigned int writ_cnt
;
678 unsigned int al_writ_cnt
;
679 unsigned int bm_writ_cnt
;
680 atomic_t ap_bio_cnt
; /* Requests we need to complete */
681 atomic_t ap_pending_cnt
; /* AP data packets on the wire, ack expected */
682 atomic_t rs_pending_cnt
; /* RS request/data packets on the wire */
683 atomic_t unacked_cnt
; /* Need to send replies for */
684 atomic_t local_cnt
; /* Waiting for local completion */
686 /* Interval tree of pending local requests */
687 struct rb_root read_requests
;
688 struct rb_root write_requests
;
690 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
691 unsigned long rs_total
;
692 /* number of resync blocks that failed in this run */
693 unsigned long rs_failed
;
694 /* Syncer's start time [unit jiffies] */
695 unsigned long rs_start
;
696 /* cumulated time in PausedSyncX state [unit jiffies] */
697 unsigned long rs_paused
;
698 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
699 unsigned long rs_same_csum
;
700 #define DRBD_SYNC_MARKS 8
701 #define DRBD_SYNC_MARK_STEP (3*HZ)
702 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
703 unsigned long rs_mark_left
[DRBD_SYNC_MARKS
];
704 /* marks's time [unit jiffies] */
705 unsigned long rs_mark_time
[DRBD_SYNC_MARKS
];
706 /* current index into rs_mark_{left,time} */
708 unsigned long rs_last_bcast
; /* [unit jiffies] */
710 /* where does the admin want us to start? (sector) */
711 sector_t ov_start_sector
;
712 sector_t ov_stop_sector
;
713 /* where are we now? (sector) */
714 sector_t ov_position
;
715 /* Start sector of out of sync range (to merge printk reporting). */
716 sector_t ov_last_oos_start
;
717 /* size of out-of-sync range in sectors. */
718 sector_t ov_last_oos_size
;
719 unsigned long ov_left
; /* in bits */
721 struct drbd_bitmap
*bitmap
;
722 unsigned long bm_resync_fo
; /* bit offset for drbd_bm_find_next */
724 /* Used to track operations of resync... */
725 struct lru_cache
*resync
;
726 /* Number of locked elements in resync LRU */
727 unsigned int resync_locked
;
728 /* resync extent number waiting for application requests */
729 unsigned int resync_wenr
;
734 struct list_head active_ee
; /* IO in progress (P_DATA gets written to disk) */
735 struct list_head sync_ee
; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
736 struct list_head done_ee
; /* need to send P_WRITE_ACK */
737 struct list_head read_ee
; /* [RS]P_DATA_REQUEST being read */
738 struct list_head net_ee
; /* zero-copy network send in progress */
741 struct list_head resync_reads
;
742 atomic_t pp_in_use
; /* allocated from page pool */
743 atomic_t pp_in_use_by_net
; /* sendpage()d, still referenced by tcp */
744 wait_queue_head_t ee_wait
;
745 struct page
*md_io_page
; /* one page buffer for md_io */
746 struct drbd_md_io md_io
;
747 atomic_t md_io_in_use
; /* protects the md_io, md_io_page and md_io_tmpp */
749 wait_queue_head_t al_wait
;
750 struct lru_cache
*act_log
; /* activity log */
751 unsigned int al_tr_number
;
753 wait_queue_head_t seq_wait
;
755 unsigned int peer_seq
;
756 spinlock_t peer_seq_lock
;
758 unsigned long comm_bm_set
; /* communicated number of set bits. */
759 struct bm_io_work bm_io_work
;
760 u64 ed_uuid
; /* UUID of the exposed data */
761 struct mutex own_state_mutex
;
762 struct mutex
*state_mutex
; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */
763 char congestion_reason
; /* Why we where congested... */
764 atomic_t rs_sect_in
; /* for incoming resync data rate, SyncTarget */
765 atomic_t rs_sect_ev
; /* for submitted resync data rate, both */
766 int rs_last_sect_ev
; /* counter to compare with */
767 int rs_last_events
; /* counter of read or write "events" (unit sectors)
768 * on the lower level device when we last looked. */
769 int c_sync_rate
; /* current resync rate after syncer throttle magic */
770 struct fifo_buffer
*rs_plan_s
; /* correction values of resync planer (RCU, connection->conn_update) */
771 int rs_in_flight
; /* resync sectors in flight (to proxy, in proxy and from proxy) */
772 atomic_t ap_in_flight
; /* App sectors in flight (waiting for ack) */
773 unsigned int peer_max_bio_size
;
774 unsigned int local_max_bio_size
;
776 /* any requests that would block in drbd_make_request()
777 * are deferred to this single-threaded work queue */
778 struct submit_worker submit
;
781 static inline struct drbd_device
*minor_to_device(unsigned int minor
)
783 return (struct drbd_device
*)idr_find(&drbd_devices
, minor
);
786 static inline struct drbd_peer_device
*first_peer_device(struct drbd_device
*device
)
788 return list_first_entry(&device
->peer_devices
, struct drbd_peer_device
, peer_devices
);
791 #define for_each_resource(resource, _resources) \
792 list_for_each_entry(resource, _resources, resources)
794 #define for_each_resource_rcu(resource, _resources) \
795 list_for_each_entry_rcu(resource, _resources, resources)
797 #define for_each_resource_safe(resource, tmp, _resources) \
798 list_for_each_entry_safe(resource, tmp, _resources, resources)
800 #define for_each_connection(connection, resource) \
801 list_for_each_entry(connection, &resource->connections, connections)
803 #define for_each_connection_rcu(connection, resource) \
804 list_for_each_entry_rcu(connection, &resource->connections, connections)
806 #define for_each_connection_safe(connection, tmp, resource) \
807 list_for_each_entry_safe(connection, tmp, &resource->connections, connections)
809 #define for_each_peer_device(peer_device, device) \
810 list_for_each_entry(peer_device, &device->peer_devices, peer_devices)
812 #define for_each_peer_device_rcu(peer_device, device) \
813 list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices)
815 #define for_each_peer_device_safe(peer_device, tmp, device) \
816 list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices)
818 static inline unsigned int device_to_minor(struct drbd_device
*device
)
820 return device
->minor
;
823 static inline struct drbd_device
*vnr_to_device(struct drbd_connection
*connection
, int vnr
)
825 struct drbd_peer_device
*peer_device
;
827 peer_device
= idr_find(&connection
->peer_devices
, vnr
);
828 return peer_device
? peer_device
->device
: NULL
;
832 * function declarations
833 *************************/
839 DDSF_NO_RESYNC
= 2, /* Do not run a resync for the new space */
842 extern void drbd_init_set_defaults(struct drbd_device
*device
);
843 extern int drbd_thread_start(struct drbd_thread
*thi
);
844 extern void _drbd_thread_stop(struct drbd_thread
*thi
, int restart
, int wait
);
845 extern char *drbd_task_to_thread_name(struct drbd_connection
*connection
, struct task_struct
*task
);
847 extern void drbd_thread_current_set_cpu(struct drbd_thread
*thi
);
848 extern void drbd_calc_cpu_mask(struct drbd_connection
*connection
);
850 #define drbd_thread_current_set_cpu(A) ({})
851 #define drbd_calc_cpu_mask(A) ({})
853 extern void tl_release(struct drbd_connection
*, unsigned int barrier_nr
,
854 unsigned int set_size
);
855 extern void tl_clear(struct drbd_connection
*);
856 extern void drbd_free_sock(struct drbd_connection
*connection
);
857 extern int drbd_send(struct drbd_connection
*connection
, struct socket
*sock
,
858 void *buf
, size_t size
, unsigned msg_flags
);
859 extern int drbd_send_all(struct drbd_connection
*, struct socket
*, void *, size_t,
862 extern int __drbd_send_protocol(struct drbd_connection
*connection
, enum drbd_packet cmd
);
863 extern int drbd_send_protocol(struct drbd_connection
*connection
);
864 extern int drbd_send_uuids(struct drbd_device
*device
);
865 extern int drbd_send_uuids_skip_initial_sync(struct drbd_device
*device
);
866 extern void drbd_gen_and_send_sync_uuid(struct drbd_device
*device
);
867 extern int drbd_send_sizes(struct drbd_device
*device
, int trigger_reply
, enum dds_flags flags
);
868 extern int drbd_send_state(struct drbd_device
*device
, union drbd_state s
);
869 extern int drbd_send_current_state(struct drbd_device
*device
);
870 extern int drbd_send_sync_param(struct drbd_device
*device
);
871 extern void drbd_send_b_ack(struct drbd_connection
*connection
, u32 barrier_nr
,
873 extern int drbd_send_ack(struct drbd_device
*, enum drbd_packet
,
874 struct drbd_peer_request
*);
875 extern void drbd_send_ack_rp(struct drbd_device
*device
, enum drbd_packet cmd
,
876 struct p_block_req
*rp
);
877 extern void drbd_send_ack_dp(struct drbd_device
*device
, enum drbd_packet cmd
,
878 struct p_data
*dp
, int data_size
);
879 extern int drbd_send_ack_ex(struct drbd_device
*device
, enum drbd_packet cmd
,
880 sector_t sector
, int blksize
, u64 block_id
);
881 extern int drbd_send_out_of_sync(struct drbd_device
*, struct drbd_request
*);
882 extern int drbd_send_block(struct drbd_device
*, enum drbd_packet
,
883 struct drbd_peer_request
*);
884 extern int drbd_send_dblock(struct drbd_device
*device
, struct drbd_request
*req
);
885 extern int drbd_send_drequest(struct drbd_device
*device
, int cmd
,
886 sector_t sector
, int size
, u64 block_id
);
887 extern int drbd_send_drequest_csum(struct drbd_device
*device
, sector_t sector
,
888 int size
, void *digest
, int digest_size
,
889 enum drbd_packet cmd
);
890 extern int drbd_send_ov_request(struct drbd_device
*device
, sector_t sector
, int size
);
892 extern int drbd_send_bitmap(struct drbd_device
*device
);
893 extern void drbd_send_sr_reply(struct drbd_device
*device
, enum drbd_state_rv retcode
);
894 extern void conn_send_sr_reply(struct drbd_connection
*connection
, enum drbd_state_rv retcode
);
895 extern void drbd_free_bc(struct drbd_backing_dev
*ldev
);
896 extern void drbd_device_cleanup(struct drbd_device
*device
);
897 void drbd_print_uuids(struct drbd_device
*device
, const char *text
);
899 extern void conn_md_sync(struct drbd_connection
*connection
);
900 extern void drbd_md_write(struct drbd_device
*device
, void *buffer
);
901 extern void drbd_md_sync(struct drbd_device
*device
);
902 extern int drbd_md_read(struct drbd_device
*device
, struct drbd_backing_dev
*bdev
);
903 extern void drbd_uuid_set(struct drbd_device
*device
, int idx
, u64 val
) __must_hold(local
);
904 extern void _drbd_uuid_set(struct drbd_device
*device
, int idx
, u64 val
) __must_hold(local
);
905 extern void drbd_uuid_new_current(struct drbd_device
*device
) __must_hold(local
);
906 extern void drbd_uuid_set_bm(struct drbd_device
*device
, u64 val
) __must_hold(local
);
907 extern void drbd_uuid_move_history(struct drbd_device
*device
) __must_hold(local
);
908 extern void __drbd_uuid_set(struct drbd_device
*device
, int idx
, u64 val
) __must_hold(local
);
909 extern void drbd_md_set_flag(struct drbd_device
*device
, int flags
) __must_hold(local
);
910 extern void drbd_md_clear_flag(struct drbd_device
*device
, int flags
)__must_hold(local
);
911 extern int drbd_md_test_flag(struct drbd_backing_dev
*, int);
912 #ifndef DRBD_DEBUG_MD_SYNC
913 extern void drbd_md_mark_dirty(struct drbd_device
*device
);
915 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
916 extern void drbd_md_mark_dirty_(struct drbd_device
*device
,
917 unsigned int line
, const char *func
);
919 extern void drbd_queue_bitmap_io(struct drbd_device
*device
,
920 int (*io_fn
)(struct drbd_device
*),
921 void (*done
)(struct drbd_device
*, int),
922 char *why
, enum bm_flag flags
);
923 extern int drbd_bitmap_io(struct drbd_device
*device
,
924 int (*io_fn
)(struct drbd_device
*),
925 char *why
, enum bm_flag flags
);
926 extern int drbd_bitmap_io_from_worker(struct drbd_device
*device
,
927 int (*io_fn
)(struct drbd_device
*),
928 char *why
, enum bm_flag flags
);
929 extern int drbd_bmio_set_n_write(struct drbd_device
*device
);
930 extern int drbd_bmio_clear_n_write(struct drbd_device
*device
);
931 extern void drbd_ldev_destroy(struct drbd_device
*device
);
935 * We currently have two possible layouts.
936 * Offsets in (512 byte) sectors.
938 * |----------- md_size_sect ------------------|
939 * [ 4k superblock ][ activity log ][ Bitmap ]
941 * | bm_offset = al_offset + X |
942 * ==> bitmap sectors = md_size_sect - bm_offset
945 * old, indexed fixed size meta data:
948 * |----------- md_size_sect ------------------|
949 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*]
951 * | bm_offset = al_offset - Y |
952 * ==> bitmap sectors = Y = al_offset - bm_offset
954 * [padding*] are zero or up to 7 unused 512 Byte sectors to the
955 * end of the device, so that the [4k superblock] will be 4k aligned.
957 * The activity log consists of 4k transaction blocks,
958 * which are written in a ring-buffer, or striped ring-buffer like fashion,
959 * which are writtensize used to be fixed 32kB,
960 * but is about to become configurable.
963 /* Our old fixed size meta data layout
964 * allows up to about 3.8TB, so if you want more,
965 * you need to use the "flexible" meta data format. */
966 #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */
967 #define MD_4kB_SECT 8
968 #define MD_32kB_SECT 64
970 /* One activity log extent represents 4M of storage */
971 #define AL_EXTENT_SHIFT 22
972 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
974 /* We could make these currently hardcoded constants configurable
975 * variables at create-md time (or even re-configurable at runtime?).
976 * Which will require some more changes to the DRBD "super block"
979 * updates per transaction:
980 * This many changes to the active set can be logged with one transaction.
981 * This number is arbitrary.
982 * context per transaction:
983 * This many context extent numbers are logged with each transaction.
984 * This number is resulting from the transaction block size (4k), the layout
985 * of the transaction header, and the number of updates per transaction.
986 * See drbd_actlog.c:struct al_transaction_on_disk
988 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
989 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
991 #if BITS_PER_LONG == 32
993 #define cpu_to_lel(A) cpu_to_le32(A)
994 #define lel_to_cpu(A) le32_to_cpu(A)
995 #elif BITS_PER_LONG == 64
997 #define cpu_to_lel(A) cpu_to_le64(A)
998 #define lel_to_cpu(A) le64_to_cpu(A)
1000 #error "LN2 of BITS_PER_LONG unknown!"
1004 /* 16MB sized 'bitmap extent' to track syncer usage */
1006 int rs_left
; /* number of bits set (out of sync) in this extent. */
1007 int rs_failed
; /* number of failed resync requests in this extent. */
1008 unsigned long flags
;
1009 struct lc_element lce
;
1012 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1013 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1014 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1018 * We need to store one bit for a block.
1019 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1020 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1021 * Bit 1 ==> local node thinks this block needs to be synced.
1024 #define SLEEP_TIME (HZ/10)
1026 /* We do bitmap IO in units of 4k blocks.
1027 * We also still have a hardcoded 4k per bit relation. */
1028 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1029 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1030 /* mostly arbitrarily set the represented size of one bitmap extent,
1031 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1032 * at 4k per bit resolution) */
1033 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
1034 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1036 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1037 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1040 /* thus many _storage_ sectors are described by one bit */
1041 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1042 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1043 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1045 /* bit to represented kilo byte conversion */
1046 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1048 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1049 * _storage_ sector is located in */
1050 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1052 /* how much _storage_ sectors we have per bitmap sector */
1053 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1054 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1056 /* in one sector of the bitmap, we have this many activity_log extents. */
1057 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1059 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1060 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1062 /* the extent in "PER_EXTENT" below is an activity log extent
1063 * we need that many (long words/bytes) to store the bitmap
1064 * of one AL_EXTENT_SIZE chunk of storage.
1065 * we can store the bitmap for that many AL_EXTENTS within
1066 * one sector of the _on_disk_ bitmap:
1067 * bit 0 bit 37 bit 38 bit (512*8)-1
1068 * ...|........|........|.. // ..|........|
1069 * sect. 0 `296 `304 ^(512*8*8)-1
1071 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1072 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1073 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1076 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1077 /* we have a certain meta data variant that has a fixed on-disk size of 128
1078 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity
1079 * log, leaving this many sectors for the bitmap.
1082 #define DRBD_MAX_SECTORS_FIXED_BM \
1083 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9)))
1084 #if !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1085 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1086 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1088 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM
1089 /* 16 TB in units of sectors */
1090 #if BITS_PER_LONG == 32
1091 /* adjust by one page worth of bitmap,
1092 * so we won't wrap around in drbd_bm_find_next_bit.
1093 * you should use 64bit OS for that much storage, anyways. */
1094 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1096 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1097 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1098 /* corresponds to (1UL << 38) bits right now. */
1102 /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE,
1103 * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte.
1104 * Since we may live in a mixed-platform cluster,
1105 * we limit us to a platform agnostic constant here for now.
1106 * A followup commit may allow even bigger BIO sizes,
1107 * once we thought that through. */
1108 #define DRBD_MAX_BIO_SIZE (1U << 20)
1109 #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1110 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1112 #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */
1114 #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */
1115 #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */
1117 extern int drbd_bm_init(struct drbd_device
*device
);
1118 extern int drbd_bm_resize(struct drbd_device
*device
, sector_t sectors
, int set_new_bits
);
1119 extern void drbd_bm_cleanup(struct drbd_device
*device
);
1120 extern void drbd_bm_set_all(struct drbd_device
*device
);
1121 extern void drbd_bm_clear_all(struct drbd_device
*device
);
1122 /* set/clear/test only a few bits at a time */
1123 extern int drbd_bm_set_bits(
1124 struct drbd_device
*device
, unsigned long s
, unsigned long e
);
1125 extern int drbd_bm_clear_bits(
1126 struct drbd_device
*device
, unsigned long s
, unsigned long e
);
1127 extern int drbd_bm_count_bits(
1128 struct drbd_device
*device
, const unsigned long s
, const unsigned long e
);
1129 /* bm_set_bits variant for use while holding drbd_bm_lock,
1130 * may process the whole bitmap in one go */
1131 extern void _drbd_bm_set_bits(struct drbd_device
*device
,
1132 const unsigned long s
, const unsigned long e
);
1133 extern int drbd_bm_test_bit(struct drbd_device
*device
, unsigned long bitnr
);
1134 extern int drbd_bm_e_weight(struct drbd_device
*device
, unsigned long enr
);
1135 extern int drbd_bm_write_page(struct drbd_device
*device
, unsigned int idx
) __must_hold(local
);
1136 extern int drbd_bm_read(struct drbd_device
*device
) __must_hold(local
);
1137 extern void drbd_bm_mark_for_writeout(struct drbd_device
*device
, int page_nr
);
1138 extern int drbd_bm_write(struct drbd_device
*device
) __must_hold(local
);
1139 extern int drbd_bm_write_hinted(struct drbd_device
*device
) __must_hold(local
);
1140 extern int drbd_bm_write_all(struct drbd_device
*device
) __must_hold(local
);
1141 extern int drbd_bm_write_copy_pages(struct drbd_device
*device
) __must_hold(local
);
1142 extern size_t drbd_bm_words(struct drbd_device
*device
);
1143 extern unsigned long drbd_bm_bits(struct drbd_device
*device
);
1144 extern sector_t
drbd_bm_capacity(struct drbd_device
*device
);
1146 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1147 extern unsigned long drbd_bm_find_next(struct drbd_device
*device
, unsigned long bm_fo
);
1148 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1149 extern unsigned long _drbd_bm_find_next(struct drbd_device
*device
, unsigned long bm_fo
);
1150 extern unsigned long _drbd_bm_find_next_zero(struct drbd_device
*device
, unsigned long bm_fo
);
1151 extern unsigned long _drbd_bm_total_weight(struct drbd_device
*device
);
1152 extern unsigned long drbd_bm_total_weight(struct drbd_device
*device
);
1153 extern int drbd_bm_rs_done(struct drbd_device
*device
);
1154 /* for receive_bitmap */
1155 extern void drbd_bm_merge_lel(struct drbd_device
*device
, size_t offset
,
1156 size_t number
, unsigned long *buffer
);
1157 /* for _drbd_send_bitmap */
1158 extern void drbd_bm_get_lel(struct drbd_device
*device
, size_t offset
,
1159 size_t number
, unsigned long *buffer
);
1161 extern void drbd_bm_lock(struct drbd_device
*device
, char *why
, enum bm_flag flags
);
1162 extern void drbd_bm_unlock(struct drbd_device
*device
);
1165 extern struct kmem_cache
*drbd_request_cache
;
1166 extern struct kmem_cache
*drbd_ee_cache
; /* peer requests */
1167 extern struct kmem_cache
*drbd_bm_ext_cache
; /* bitmap extents */
1168 extern struct kmem_cache
*drbd_al_ext_cache
; /* activity log extents */
1169 extern mempool_t
*drbd_request_mempool
;
1170 extern mempool_t
*drbd_ee_mempool
;
1172 /* drbd's page pool, used to buffer data received from the peer,
1173 * or data requested by the peer.
1175 * This does not have an emergency reserve.
1177 * When allocating from this pool, it first takes pages from the pool.
1178 * Only if the pool is depleted will try to allocate from the system.
1180 * The assumption is that pages taken from this pool will be processed,
1181 * and given back, "quickly", and then can be recycled, so we can avoid
1182 * frequent calls to alloc_page(), and still will be able to make progress even
1183 * under memory pressure.
1185 extern struct page
*drbd_pp_pool
;
1186 extern spinlock_t drbd_pp_lock
;
1187 extern int drbd_pp_vacant
;
1188 extern wait_queue_head_t drbd_pp_wait
;
1190 /* We also need a standard (emergency-reserve backed) page pool
1191 * for meta data IO (activity log, bitmap).
1192 * We can keep it global, as long as it is used as "N pages at a time".
1193 * 128 should be plenty, currently we probably can get away with as few as 1.
1195 #define DRBD_MIN_POOL_PAGES 128
1196 extern mempool_t
*drbd_md_io_page_pool
;
1198 /* We also need to make sure we get a bio
1199 * when we need it for housekeeping purposes */
1200 extern struct bio_set
*drbd_md_io_bio_set
;
1201 /* to allocate from that set */
1202 extern struct bio
*bio_alloc_drbd(gfp_t gfp_mask
);
1204 extern rwlock_t global_state_lock
;
1206 extern int conn_lowest_minor(struct drbd_connection
*connection
);
1207 enum drbd_ret_code
drbd_create_minor(struct drbd_connection
*connection
, unsigned int minor
, int vnr
);
1208 extern void drbd_destroy_device(struct kref
*kref
);
1209 extern void drbd_delete_minor(struct drbd_device
*mdev
);
1211 extern struct drbd_resource
*drbd_create_resource(const char *name
);
1212 extern void drbd_free_resource(struct drbd_resource
*resource
);
1214 extern int set_resource_options(struct drbd_resource
*resource
, struct res_opts
*res_opts
);
1215 extern struct drbd_connection
*conn_create(const char *name
, struct res_opts
*res_opts
);
1216 extern void drbd_destroy_connection(struct kref
*kref
);
1217 extern struct drbd_connection
*conn_get_by_addrs(void *my_addr
, int my_addr_len
,
1218 void *peer_addr
, int peer_addr_len
);
1219 extern struct drbd_resource
*drbd_find_resource(const char *name
);
1220 extern void drbd_destroy_resource(struct kref
*kref
);
1221 extern void conn_free_crypto(struct drbd_connection
*connection
);
1223 extern int proc_details
;
1226 extern void do_submit(struct work_struct
*ws
);
1227 extern void __drbd_make_request(struct drbd_device
*, struct bio
*, unsigned long);
1228 extern void drbd_make_request(struct request_queue
*q
, struct bio
*bio
);
1229 extern int drbd_read_remote(struct drbd_device
*device
, struct drbd_request
*req
);
1230 extern int drbd_merge_bvec(struct request_queue
*q
, struct bvec_merge_data
*bvm
, struct bio_vec
*bvec
);
1231 extern int is_valid_ar_handle(struct drbd_request
*, sector_t
);
1235 extern int drbd_msg_put_info(const char *info
);
1236 extern void drbd_suspend_io(struct drbd_device
*device
);
1237 extern void drbd_resume_io(struct drbd_device
*device
);
1238 extern char *ppsize(char *buf
, unsigned long long size
);
1239 extern sector_t
drbd_new_dev_size(struct drbd_device
*, struct drbd_backing_dev
*, sector_t
, int);
1240 enum determine_dev_size
{
1241 DS_ERROR_SHRINK
= -3,
1242 DS_ERROR_SPACE_MD
= -2,
1247 DS_GREW_FROM_ZERO
= 3,
1249 extern enum determine_dev_size
1250 drbd_determine_dev_size(struct drbd_device
*, enum dds_flags
, struct resize_parms
*) __must_hold(local
);
1251 extern void resync_after_online_grow(struct drbd_device
*);
1252 extern void drbd_reconsider_max_bio_size(struct drbd_device
*device
);
1253 extern enum drbd_state_rv
drbd_set_role(struct drbd_device
*device
,
1254 enum drbd_role new_role
,
1256 extern bool conn_try_outdate_peer(struct drbd_connection
*connection
);
1257 extern void conn_try_outdate_peer_async(struct drbd_connection
*connection
);
1258 extern int drbd_khelper(struct drbd_device
*device
, char *cmd
);
1261 extern int drbd_worker(struct drbd_thread
*thi
);
1262 enum drbd_ret_code
drbd_resync_after_valid(struct drbd_device
*device
, int o_minor
);
1263 void drbd_resync_after_changed(struct drbd_device
*device
);
1264 extern void drbd_start_resync(struct drbd_device
*device
, enum drbd_conns side
);
1265 extern void resume_next_sg(struct drbd_device
*device
);
1266 extern void suspend_other_sg(struct drbd_device
*device
);
1267 extern int drbd_resync_finished(struct drbd_device
*device
);
1268 /* maybe rather drbd_main.c ? */
1269 extern void *drbd_md_get_buffer(struct drbd_device
*device
);
1270 extern void drbd_md_put_buffer(struct drbd_device
*device
);
1271 extern int drbd_md_sync_page_io(struct drbd_device
*device
,
1272 struct drbd_backing_dev
*bdev
, sector_t sector
, int rw
);
1273 extern void drbd_ov_out_of_sync_found(struct drbd_device
*, sector_t
, int);
1274 extern void wait_until_done_or_force_detached(struct drbd_device
*device
,
1275 struct drbd_backing_dev
*bdev
, unsigned int *done
);
1276 extern void drbd_rs_controller_reset(struct drbd_device
*device
);
1278 static inline void ov_out_of_sync_print(struct drbd_device
*device
)
1280 if (device
->ov_last_oos_size
) {
1281 dev_err(DEV
, "Out of sync: start=%llu, size=%lu (sectors)\n",
1282 (unsigned long long)device
->ov_last_oos_start
,
1283 (unsigned long)device
->ov_last_oos_size
);
1285 device
->ov_last_oos_size
= 0;
1289 extern void drbd_csum_bio(struct drbd_device
*, struct crypto_hash
*, struct bio
*, void *);
1290 extern void drbd_csum_ee(struct drbd_device
*, struct crypto_hash
*,
1291 struct drbd_peer_request
*, void *);
1292 /* worker callbacks */
1293 extern int w_e_end_data_req(struct drbd_work
*, int);
1294 extern int w_e_end_rsdata_req(struct drbd_work
*, int);
1295 extern int w_e_end_csum_rs_req(struct drbd_work
*, int);
1296 extern int w_e_end_ov_reply(struct drbd_work
*, int);
1297 extern int w_e_end_ov_req(struct drbd_work
*, int);
1298 extern int w_ov_finished(struct drbd_work
*, int);
1299 extern int w_resync_timer(struct drbd_work
*, int);
1300 extern int w_send_write_hint(struct drbd_work
*, int);
1301 extern int w_make_resync_request(struct drbd_work
*, int);
1302 extern int w_send_dblock(struct drbd_work
*, int);
1303 extern int w_send_read_req(struct drbd_work
*, int);
1304 extern int w_prev_work_done(struct drbd_work
*, int);
1305 extern int w_e_reissue(struct drbd_work
*, int);
1306 extern int w_restart_disk_io(struct drbd_work
*, int);
1307 extern int w_send_out_of_sync(struct drbd_work
*, int);
1308 extern int w_start_resync(struct drbd_work
*, int);
1310 extern void resync_timer_fn(unsigned long data
);
1311 extern void start_resync_timer_fn(unsigned long data
);
1313 /* drbd_receiver.c */
1314 extern int drbd_rs_should_slow_down(struct drbd_device
*device
, sector_t sector
);
1315 extern int drbd_submit_peer_request(struct drbd_device
*,
1316 struct drbd_peer_request
*, const unsigned,
1318 extern int drbd_free_peer_reqs(struct drbd_device
*, struct list_head
*);
1319 extern struct drbd_peer_request
*drbd_alloc_peer_req(struct drbd_device
*, u64
,
1320 sector_t
, unsigned int,
1321 gfp_t
) __must_hold(local
);
1322 extern void __drbd_free_peer_req(struct drbd_device
*, struct drbd_peer_request
*,
1324 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1325 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
1326 extern struct page
*drbd_alloc_pages(struct drbd_device
*, unsigned int, bool);
1327 extern void drbd_set_recv_tcq(struct drbd_device
*device
, int tcq_enabled
);
1328 extern void _drbd_clear_done_ee(struct drbd_device
*device
, struct list_head
*to_be_freed
);
1329 extern void conn_flush_workqueue(struct drbd_connection
*connection
);
1330 extern int drbd_connected(struct drbd_device
*device
);
1331 static inline void drbd_flush_workqueue(struct drbd_device
*device
)
1333 conn_flush_workqueue(first_peer_device(device
)->connection
);
1336 /* Yes, there is kernel_setsockopt, but only since 2.6.18.
1337 * So we have our own copy of it here. */
1338 static inline int drbd_setsockopt(struct socket
*sock
, int level
, int optname
,
1339 char *optval
, int optlen
)
1341 mm_segment_t oldfs
= get_fs();
1342 char __user
*uoptval
;
1345 uoptval
= (char __user __force
*)optval
;
1348 if (level
== SOL_SOCKET
)
1349 err
= sock_setsockopt(sock
, level
, optname
, uoptval
, optlen
);
1351 err
= sock
->ops
->setsockopt(sock
, level
, optname
, uoptval
,
1357 static inline void drbd_tcp_cork(struct socket
*sock
)
1360 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1361 (char*)&val
, sizeof(val
));
1364 static inline void drbd_tcp_uncork(struct socket
*sock
)
1367 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1368 (char*)&val
, sizeof(val
));
1371 static inline void drbd_tcp_nodelay(struct socket
*sock
)
1374 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_NODELAY
,
1375 (char*)&val
, sizeof(val
));
1378 static inline void drbd_tcp_quickack(struct socket
*sock
)
1381 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_QUICKACK
,
1382 (char*)&val
, sizeof(val
));
1385 void drbd_bump_write_ordering(struct drbd_connection
*connection
, enum write_ordering_e wo
);
1388 extern struct proc_dir_entry
*drbd_proc
;
1389 extern const struct file_operations drbd_proc_fops
;
1390 extern const char *drbd_conn_str(enum drbd_conns s
);
1391 extern const char *drbd_role_str(enum drbd_role s
);
1394 extern int drbd_al_begin_io_nonblock(struct drbd_device
*device
, struct drbd_interval
*i
);
1395 extern void drbd_al_begin_io_commit(struct drbd_device
*device
, bool delegate
);
1396 extern bool drbd_al_begin_io_fastpath(struct drbd_device
*device
, struct drbd_interval
*i
);
1397 extern void drbd_al_begin_io(struct drbd_device
*device
, struct drbd_interval
*i
, bool delegate
);
1398 extern void drbd_al_complete_io(struct drbd_device
*device
, struct drbd_interval
*i
);
1399 extern void drbd_rs_complete_io(struct drbd_device
*device
, sector_t sector
);
1400 extern int drbd_rs_begin_io(struct drbd_device
*device
, sector_t sector
);
1401 extern int drbd_try_rs_begin_io(struct drbd_device
*device
, sector_t sector
);
1402 extern void drbd_rs_cancel_all(struct drbd_device
*device
);
1403 extern int drbd_rs_del_all(struct drbd_device
*device
);
1404 extern void drbd_rs_failed_io(struct drbd_device
*device
,
1405 sector_t sector
, int size
);
1406 extern void drbd_advance_rs_marks(struct drbd_device
*device
, unsigned long still_to_go
);
1407 extern void __drbd_set_in_sync(struct drbd_device
*device
, sector_t sector
,
1408 int size
, const char *file
, const unsigned int line
);
1409 #define drbd_set_in_sync(device, sector, size) \
1410 __drbd_set_in_sync(device, sector, size, __FILE__, __LINE__)
1411 extern int __drbd_set_out_of_sync(struct drbd_device
*device
, sector_t sector
,
1412 int size
, const char *file
, const unsigned int line
);
1413 #define drbd_set_out_of_sync(device, sector, size) \
1414 __drbd_set_out_of_sync(device, sector, size, __FILE__, __LINE__)
1415 extern void drbd_al_shrink(struct drbd_device
*device
);
1416 extern int drbd_initialize_al(struct drbd_device
*, void *);
1419 /* state info broadcast */
1421 enum drbd_state_info_bcast_reason sib_reason
;
1425 unsigned helper_exit_code
;
1428 union drbd_state os
;
1429 union drbd_state ns
;
1433 void drbd_bcast_event(struct drbd_device
*device
, const struct sib_info
*sib
);
1436 * inline helper functions
1437 *************************/
1439 /* see also page_chain_add and friends in drbd_receiver.c */
1440 static inline struct page
*page_chain_next(struct page
*page
)
1442 return (struct page
*)page_private(page
);
1444 #define page_chain_for_each(page) \
1445 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1446 page = page_chain_next(page))
1447 #define page_chain_for_each_safe(page, n) \
1448 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1451 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request
*peer_req
)
1453 struct page
*page
= peer_req
->pages
;
1454 page_chain_for_each(page
) {
1455 if (page_count(page
) > 1)
1461 static inline enum drbd_state_rv
1462 _drbd_set_state(struct drbd_device
*device
, union drbd_state ns
,
1463 enum chg_state_flags flags
, struct completion
*done
)
1465 enum drbd_state_rv rv
;
1467 read_lock(&global_state_lock
);
1468 rv
= __drbd_set_state(device
, ns
, flags
, done
);
1469 read_unlock(&global_state_lock
);
1474 static inline union drbd_state
drbd_read_state(struct drbd_device
*device
)
1476 union drbd_state rv
;
1478 rv
.i
= device
->state
.i
;
1479 rv
.susp
= first_peer_device(device
)->connection
->susp
;
1480 rv
.susp_nod
= first_peer_device(device
)->connection
->susp_nod
;
1481 rv
.susp_fen
= first_peer_device(device
)->connection
->susp_fen
;
1486 enum drbd_force_detach_flags
{
1493 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1494 static inline void __drbd_chk_io_error_(struct drbd_device
*device
,
1495 enum drbd_force_detach_flags df
,
1498 enum drbd_io_error_p ep
;
1501 ep
= rcu_dereference(device
->ldev
->disk_conf
)->on_io_error
;
1504 case EP_PASS_ON
: /* FIXME would this be better named "Ignore"? */
1505 if (df
== DRBD_READ_ERROR
|| df
== DRBD_WRITE_ERROR
) {
1506 if (__ratelimit(&drbd_ratelimit_state
))
1507 dev_err(DEV
, "Local IO failed in %s.\n", where
);
1508 if (device
->state
.disk
> D_INCONSISTENT
)
1509 _drbd_set_state(_NS(device
, disk
, D_INCONSISTENT
), CS_HARD
, NULL
);
1512 /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */
1514 case EP_CALL_HELPER
:
1515 /* Remember whether we saw a READ or WRITE error.
1517 * Recovery of the affected area for WRITE failure is covered
1518 * by the activity log.
1519 * READ errors may fall outside that area though. Certain READ
1520 * errors can be "healed" by writing good data to the affected
1521 * blocks, which triggers block re-allocation in lower layers.
1523 * If we can not write the bitmap after a READ error,
1524 * we may need to trigger a full sync (see w_go_diskless()).
1526 * Force-detach is not really an IO error, but rather a
1527 * desperate measure to try to deal with a completely
1528 * unresponsive lower level IO stack.
1529 * Still it should be treated as a WRITE error.
1531 * Meta IO error is always WRITE error:
1532 * we read meta data only once during attach,
1533 * which will fail in case of errors.
1535 set_bit(WAS_IO_ERROR
, &device
->flags
);
1536 if (df
== DRBD_READ_ERROR
)
1537 set_bit(WAS_READ_ERROR
, &device
->flags
);
1538 if (df
== DRBD_FORCE_DETACH
)
1539 set_bit(FORCE_DETACH
, &device
->flags
);
1540 if (device
->state
.disk
> D_FAILED
) {
1541 _drbd_set_state(_NS(device
, disk
, D_FAILED
), CS_HARD
, NULL
);
1543 "Local IO failed in %s. Detaching...\n", where
);
1550 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1551 * @device: DRBD device.
1552 * @error: Error code passed to the IO completion callback
1553 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1555 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1557 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1558 static inline void drbd_chk_io_error_(struct drbd_device
*device
,
1559 int error
, enum drbd_force_detach_flags forcedetach
, const char *where
)
1562 unsigned long flags
;
1563 spin_lock_irqsave(&first_peer_device(device
)->connection
->req_lock
, flags
);
1564 __drbd_chk_io_error_(device
, forcedetach
, where
);
1565 spin_unlock_irqrestore(&first_peer_device(device
)->connection
->req_lock
, flags
);
1571 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1572 * @bdev: Meta data block device.
1574 * BTW, for internal meta data, this happens to be the maximum capacity
1575 * we could agree upon with our peer node.
1577 static inline sector_t
drbd_md_first_sector(struct drbd_backing_dev
*bdev
)
1579 switch (bdev
->md
.meta_dev_idx
) {
1580 case DRBD_MD_INDEX_INTERNAL
:
1581 case DRBD_MD_INDEX_FLEX_INT
:
1582 return bdev
->md
.md_offset
+ bdev
->md
.bm_offset
;
1583 case DRBD_MD_INDEX_FLEX_EXT
:
1585 return bdev
->md
.md_offset
;
1590 * drbd_md_last_sector() - Return the last sector number of the meta data area
1591 * @bdev: Meta data block device.
1593 static inline sector_t
drbd_md_last_sector(struct drbd_backing_dev
*bdev
)
1595 switch (bdev
->md
.meta_dev_idx
) {
1596 case DRBD_MD_INDEX_INTERNAL
:
1597 case DRBD_MD_INDEX_FLEX_INT
:
1598 return bdev
->md
.md_offset
+ MD_4kB_SECT
-1;
1599 case DRBD_MD_INDEX_FLEX_EXT
:
1601 return bdev
->md
.md_offset
+ bdev
->md
.md_size_sect
-1;
1605 /* Returns the number of 512 byte sectors of the device */
1606 static inline sector_t
drbd_get_capacity(struct block_device
*bdev
)
1608 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1609 return bdev
? i_size_read(bdev
->bd_inode
) >> 9 : 0;
1613 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1614 * @bdev: Meta data block device.
1616 * returns the capacity we announce to out peer. we clip ourselves at the
1617 * various MAX_SECTORS, because if we don't, current implementation will
1618 * oops sooner or later
1620 static inline sector_t
drbd_get_max_capacity(struct drbd_backing_dev
*bdev
)
1624 switch (bdev
->md
.meta_dev_idx
) {
1625 case DRBD_MD_INDEX_INTERNAL
:
1626 case DRBD_MD_INDEX_FLEX_INT
:
1627 s
= drbd_get_capacity(bdev
->backing_bdev
)
1628 ? min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1629 drbd_md_first_sector(bdev
))
1632 case DRBD_MD_INDEX_FLEX_EXT
:
1633 s
= min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1634 drbd_get_capacity(bdev
->backing_bdev
));
1635 /* clip at maximum size the meta device can support */
1636 s
= min_t(sector_t
, s
,
1637 BM_EXT_TO_SECT(bdev
->md
.md_size_sect
1638 - bdev
->md
.bm_offset
));
1641 s
= min_t(sector_t
, DRBD_MAX_SECTORS
,
1642 drbd_get_capacity(bdev
->backing_bdev
));
1648 * drbd_md_ss() - Return the sector number of our meta data super block
1649 * @bdev: Meta data block device.
1651 static inline sector_t
drbd_md_ss(struct drbd_backing_dev
*bdev
)
1653 const int meta_dev_idx
= bdev
->md
.meta_dev_idx
;
1655 if (meta_dev_idx
== DRBD_MD_INDEX_FLEX_EXT
)
1658 /* Since drbd08, internal meta data is always "flexible".
1659 * position: last 4k aligned block of 4k size */
1660 if (meta_dev_idx
== DRBD_MD_INDEX_INTERNAL
||
1661 meta_dev_idx
== DRBD_MD_INDEX_FLEX_INT
)
1662 return (drbd_get_capacity(bdev
->backing_bdev
) & ~7ULL) - 8;
1664 /* external, some index; this is the old fixed size layout */
1665 return MD_128MB_SECT
* bdev
->md
.meta_dev_idx
;
1669 drbd_queue_work_front(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1671 unsigned long flags
;
1672 spin_lock_irqsave(&q
->q_lock
, flags
);
1673 list_add(&w
->list
, &q
->q
);
1674 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1675 wake_up(&q
->q_wait
);
1679 drbd_queue_work(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1681 unsigned long flags
;
1682 spin_lock_irqsave(&q
->q_lock
, flags
);
1683 list_add_tail(&w
->list
, &q
->q
);
1684 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1685 wake_up(&q
->q_wait
);
1688 static inline void wake_asender(struct drbd_connection
*connection
)
1690 if (test_bit(SIGNAL_ASENDER
, &connection
->flags
))
1691 force_sig(DRBD_SIG
, connection
->asender
.task
);
1694 static inline void request_ping(struct drbd_connection
*connection
)
1696 set_bit(SEND_PING
, &connection
->flags
);
1697 wake_asender(connection
);
1700 extern void *conn_prepare_command(struct drbd_connection
*, struct drbd_socket
*);
1701 extern void *drbd_prepare_command(struct drbd_device
*, struct drbd_socket
*);
1702 extern int conn_send_command(struct drbd_connection
*, struct drbd_socket
*,
1703 enum drbd_packet
, unsigned int, void *,
1705 extern int drbd_send_command(struct drbd_device
*, struct drbd_socket
*,
1706 enum drbd_packet
, unsigned int, void *,
1709 extern int drbd_send_ping(struct drbd_connection
*connection
);
1710 extern int drbd_send_ping_ack(struct drbd_connection
*connection
);
1711 extern int drbd_send_state_req(struct drbd_device
*, union drbd_state
, union drbd_state
);
1712 extern int conn_send_state_req(struct drbd_connection
*, union drbd_state
, union drbd_state
);
1714 static inline void drbd_thread_stop(struct drbd_thread
*thi
)
1716 _drbd_thread_stop(thi
, false, true);
1719 static inline void drbd_thread_stop_nowait(struct drbd_thread
*thi
)
1721 _drbd_thread_stop(thi
, false, false);
1724 static inline void drbd_thread_restart_nowait(struct drbd_thread
*thi
)
1726 _drbd_thread_stop(thi
, true, false);
1729 /* counts how many answer packets packets we expect from our peer,
1730 * for either explicit application requests,
1731 * or implicit barrier packets as necessary.
1734 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1735 * it is much easier and equally valid to count what we queue for the
1736 * worker, even before it actually was queued or send.
1737 * (drbd_make_request_common; recovery path on read io-error)
1739 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1740 * _req_mod(req, DATA_RECEIVED)
1741 * [from receive_DataReply]
1742 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1743 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1744 * for some reason it is NOT decreased in got_NegAck,
1745 * but in the resulting cleanup code from report_params.
1746 * we should try to remember the reason for that...
1747 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
1748 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
1749 * [from tl_clear_barrier]
1751 static inline void inc_ap_pending(struct drbd_device
*device
)
1753 atomic_inc(&device
->ap_pending_cnt
);
1756 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \
1757 if (atomic_read(&device->which) < 0) \
1758 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
1760 atomic_read(&device->which))
1762 #define dec_ap_pending(device) _dec_ap_pending(device, __FUNCTION__, __LINE__)
1763 static inline void _dec_ap_pending(struct drbd_device
*device
, const char *func
, int line
)
1765 if (atomic_dec_and_test(&device
->ap_pending_cnt
))
1766 wake_up(&device
->misc_wait
);
1767 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt
, func
, line
);
1770 /* counts how many resync-related answers we still expect from the peer
1772 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1773 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
1774 * (or P_NEG_ACK with ID_SYNCER)
1776 static inline void inc_rs_pending(struct drbd_device
*device
)
1778 atomic_inc(&device
->rs_pending_cnt
);
1781 #define dec_rs_pending(device) _dec_rs_pending(device, __FUNCTION__, __LINE__)
1782 static inline void _dec_rs_pending(struct drbd_device
*device
, const char *func
, int line
)
1784 atomic_dec(&device
->rs_pending_cnt
);
1785 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt
, func
, line
);
1788 /* counts how many answers we still need to send to the peer.
1790 * receive_Data unless protocol A;
1791 * we need to send a P_RECV_ACK (proto B)
1792 * or P_WRITE_ACK (proto C)
1793 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1794 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1795 * receive_Barrier_* we need to send a P_BARRIER_ACK
1797 static inline void inc_unacked(struct drbd_device
*device
)
1799 atomic_inc(&device
->unacked_cnt
);
1802 #define dec_unacked(device) _dec_unacked(device, __FUNCTION__, __LINE__)
1803 static inline void _dec_unacked(struct drbd_device
*device
, const char *func
, int line
)
1805 atomic_dec(&device
->unacked_cnt
);
1806 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt
, func
, line
);
1809 #define sub_unacked(device, n) _sub_unacked(device, n, __FUNCTION__, __LINE__)
1810 static inline void _sub_unacked(struct drbd_device
*device
, int n
, const char *func
, int line
)
1812 atomic_sub(n
, &device
->unacked_cnt
);
1813 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt
, func
, line
);
1817 * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev
1820 * You have to call put_ldev() when finished working with device->ldev.
1822 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
1823 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
1825 static inline void put_ldev(struct drbd_device
*device
)
1827 int i
= atomic_dec_return(&device
->local_cnt
);
1829 /* This may be called from some endio handler,
1830 * so we must not sleep here. */
1835 if (device
->state
.disk
== D_DISKLESS
)
1836 /* even internal references gone, safe to destroy */
1837 drbd_ldev_destroy(device
);
1838 if (device
->state
.disk
== D_FAILED
) {
1839 /* all application IO references gone. */
1840 if (!test_and_set_bit(GO_DISKLESS
, &device
->flags
))
1841 drbd_queue_work(&first_peer_device(device
)->connection
->sender_work
, &device
->go_diskless
);
1843 wake_up(&device
->misc_wait
);
1848 static inline int _get_ldev_if_state(struct drbd_device
*device
, enum drbd_disk_state mins
)
1852 /* never get a reference while D_DISKLESS */
1853 if (device
->state
.disk
== D_DISKLESS
)
1856 atomic_inc(&device
->local_cnt
);
1857 io_allowed
= (device
->state
.disk
>= mins
);
1863 extern int _get_ldev_if_state(struct drbd_device
*device
, enum drbd_disk_state mins
);
1866 /* you must have an "get_ldev" reference */
1867 static inline void drbd_get_syncer_progress(struct drbd_device
*device
,
1868 unsigned long *bits_left
, unsigned int *per_mil_done
)
1870 /* this is to break it at compile time when we change that, in case we
1871 * want to support more than (1<<32) bits on a 32bit arch. */
1872 typecheck(unsigned long, device
->rs_total
);
1874 /* note: both rs_total and rs_left are in bits, i.e. in
1875 * units of BM_BLOCK_SIZE.
1876 * for the percentage, we don't care. */
1878 if (device
->state
.conn
== C_VERIFY_S
|| device
->state
.conn
== C_VERIFY_T
)
1879 *bits_left
= device
->ov_left
;
1881 *bits_left
= drbd_bm_total_weight(device
) - device
->rs_failed
;
1882 /* >> 10 to prevent overflow,
1883 * +1 to prevent division by zero */
1884 if (*bits_left
> device
->rs_total
) {
1885 /* doh. maybe a logic bug somewhere.
1886 * may also be just a race condition
1887 * between this and a disconnect during sync.
1888 * for now, just prevent in-kernel buffer overflow.
1891 dev_warn(DEV
, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
1892 drbd_conn_str(device
->state
.conn
),
1893 *bits_left
, device
->rs_total
, device
->rs_failed
);
1896 /* Make sure the division happens in long context.
1897 * We allow up to one petabyte storage right now,
1898 * at a granularity of 4k per bit that is 2**38 bits.
1899 * After shift right and multiplication by 1000,
1900 * this should still fit easily into a 32bit long,
1901 * so we don't need a 64bit division on 32bit arch.
1902 * Note: currently we don't support such large bitmaps on 32bit
1903 * arch anyways, but no harm done to be prepared for it here.
1905 unsigned int shift
= device
->rs_total
> UINT_MAX
? 16 : 10;
1906 unsigned long left
= *bits_left
>> shift
;
1907 unsigned long total
= 1UL + (device
->rs_total
>> shift
);
1908 unsigned long tmp
= 1000UL - left
* 1000UL/total
;
1909 *per_mil_done
= tmp
;
1914 /* this throttles on-the-fly application requests
1915 * according to max_buffers settings;
1916 * maybe re-implement using semaphores? */
1917 static inline int drbd_get_max_buffers(struct drbd_device
*device
)
1919 struct net_conf
*nc
;
1923 nc
= rcu_dereference(first_peer_device(device
)->connection
->net_conf
);
1924 mxb
= nc
? nc
->max_buffers
: 1000000; /* arbitrary limit on open requests */
1930 static inline int drbd_state_is_stable(struct drbd_device
*device
)
1932 union drbd_dev_state s
= device
->state
;
1934 /* DO NOT add a default clause, we want the compiler to warn us
1935 * for any newly introduced state we may have forgotten to add here */
1937 switch ((enum drbd_conns
)s
.conn
) {
1938 /* new io only accepted when there is no connection, ... */
1940 case C_WF_CONNECTION
:
1941 /* ... or there is a well established connection. */
1947 case C_PAUSED_SYNC_S
:
1948 case C_PAUSED_SYNC_T
:
1951 /* transitional states, IO allowed */
1952 case C_DISCONNECTING
:
1956 case C_NETWORK_FAILURE
:
1957 case C_PROTOCOL_ERROR
:
1959 case C_WF_REPORT_PARAMS
:
1960 case C_STARTING_SYNC_S
:
1961 case C_STARTING_SYNC_T
:
1964 /* Allow IO in BM exchange states with new protocols */
1966 if (first_peer_device(device
)->connection
->agreed_pro_version
< 96)
1970 /* no new io accepted in these states */
1972 case C_WF_SYNC_UUID
:
1978 switch ((enum drbd_disk_state
)s
.disk
) {
1980 case D_INCONSISTENT
:
1985 /* disk state is stable as well. */
1988 /* no new io accepted during transitional states */
2000 static inline int drbd_suspended(struct drbd_device
*device
)
2002 struct drbd_connection
*connection
= first_peer_device(device
)->connection
;
2004 return connection
->susp
|| connection
->susp_fen
|| connection
->susp_nod
;
2007 static inline bool may_inc_ap_bio(struct drbd_device
*device
)
2009 int mxb
= drbd_get_max_buffers(device
);
2011 if (drbd_suspended(device
))
2013 if (test_bit(SUSPEND_IO
, &device
->flags
))
2016 /* to avoid potential deadlock or bitmap corruption,
2017 * in various places, we only allow new application io
2018 * to start during "stable" states. */
2020 /* no new io accepted when attaching or detaching the disk */
2021 if (!drbd_state_is_stable(device
))
2024 /* since some older kernels don't have atomic_add_unless,
2025 * and we are within the spinlock anyways, we have this workaround. */
2026 if (atomic_read(&device
->ap_bio_cnt
) > mxb
)
2028 if (test_bit(BITMAP_IO
, &device
->flags
))
2033 static inline bool inc_ap_bio_cond(struct drbd_device
*device
)
2037 spin_lock_irq(&first_peer_device(device
)->connection
->req_lock
);
2038 rv
= may_inc_ap_bio(device
);
2040 atomic_inc(&device
->ap_bio_cnt
);
2041 spin_unlock_irq(&first_peer_device(device
)->connection
->req_lock
);
2046 static inline void inc_ap_bio(struct drbd_device
*device
)
2049 * as long as the device is suspended
2050 * until the bitmap is no longer on the fly during connection
2051 * handshake as long as we would exceed the max_buffer limit.
2053 * to avoid races with the reconnect code,
2054 * we need to atomic_inc within the spinlock. */
2056 wait_event(device
->misc_wait
, inc_ap_bio_cond(device
));
2059 static inline void dec_ap_bio(struct drbd_device
*device
)
2061 int mxb
= drbd_get_max_buffers(device
);
2062 int ap_bio
= atomic_dec_return(&device
->ap_bio_cnt
);
2064 D_ASSERT(ap_bio
>= 0);
2066 if (ap_bio
== 0 && test_bit(BITMAP_IO
, &device
->flags
)) {
2067 if (!test_and_set_bit(BITMAP_IO_QUEUED
, &device
->flags
))
2068 drbd_queue_work(&first_peer_device(device
)->connection
->sender_work
, &device
->bm_io_work
.w
);
2071 /* this currently does wake_up for every dec_ap_bio!
2072 * maybe rather introduce some type of hysteresis?
2073 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2075 wake_up(&device
->misc_wait
);
2078 static inline bool verify_can_do_stop_sector(struct drbd_device
*device
)
2080 return first_peer_device(device
)->connection
->agreed_pro_version
>= 97 &&
2081 first_peer_device(device
)->connection
->agreed_pro_version
!= 100;
2084 static inline int drbd_set_ed_uuid(struct drbd_device
*device
, u64 val
)
2086 int changed
= device
->ed_uuid
!= val
;
2087 device
->ed_uuid
= val
;
2091 static inline int drbd_queue_order_type(struct drbd_device
*device
)
2093 /* sorry, we currently have no working implementation
2094 * of distributed TCQ stuff */
2095 #ifndef QUEUE_ORDERED_NONE
2096 #define QUEUE_ORDERED_NONE 0
2098 return QUEUE_ORDERED_NONE
;
2101 static inline void drbd_md_flush(struct drbd_device
*device
)
2105 if (device
->ldev
== NULL
) {
2106 dev_warn(DEV
, "device->ldev == NULL in drbd_md_flush\n");
2110 if (test_bit(MD_NO_FUA
, &device
->flags
))
2113 r
= blkdev_issue_flush(device
->ldev
->md_bdev
, GFP_NOIO
, NULL
);
2115 set_bit(MD_NO_FUA
, &device
->flags
);
2116 dev_err(DEV
, "meta data flush failed with status %d, disabling md-flushes\n", r
);
2120 static inline struct drbd_connection
*first_connection(struct drbd_resource
*resource
)
2122 return list_first_entry(&resource
->connections
,
2123 struct drbd_connection
, connections
);