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
;
103 #define __drbd_printk_device(level, device, fmt, args...) \
104 dev_printk(level, disk_to_dev((device)->vdisk), fmt, ## args)
105 #define __drbd_printk_peer_device(level, peer_device, fmt, args...) \
106 dev_printk(level, disk_to_dev((peer_device)->device->vdisk), fmt, ## args)
107 #define __drbd_printk_resource(level, resource, fmt, args...) \
108 printk(level "drbd %s: " fmt, (resource)->name, ## args)
109 #define __drbd_printk_connection(level, connection, fmt, args...) \
110 printk(level "drbd %s: " fmt, (connection)->resource->name, ## args)
112 void drbd_printk_with_wrong_object_type(void);
114 #define __drbd_printk_if_same_type(obj, type, func, level, fmt, args...) \
115 (__builtin_types_compatible_p(typeof(obj), type) || \
116 __builtin_types_compatible_p(typeof(obj), const type)), \
117 func(level, (const type)(obj), fmt, ## args)
119 #define drbd_printk(level, obj, fmt, args...) \
120 __builtin_choose_expr( \
121 __drbd_printk_if_same_type(obj, struct drbd_device *, \
122 __drbd_printk_device, level, fmt, ## args), \
123 __builtin_choose_expr( \
124 __drbd_printk_if_same_type(obj, struct drbd_resource *, \
125 __drbd_printk_resource, level, fmt, ## args), \
126 __builtin_choose_expr( \
127 __drbd_printk_if_same_type(obj, struct drbd_connection *, \
128 __drbd_printk_connection, level, fmt, ## args), \
129 __builtin_choose_expr( \
130 __drbd_printk_if_same_type(obj, struct drbd_peer_device *, \
131 __drbd_printk_peer_device, level, fmt, ## args), \
132 drbd_printk_with_wrong_object_type()))))
134 #define drbd_dbg(obj, fmt, args...) \
135 drbd_printk(KERN_DEBUG, obj, fmt, ## args)
136 #define drbd_alert(obj, fmt, args...) \
137 drbd_printk(KERN_ALERT, obj, fmt, ## args)
138 #define drbd_err(obj, fmt, args...) \
139 drbd_printk(KERN_ERR, obj, fmt, ## args)
140 #define drbd_warn(obj, fmt, args...) \
141 drbd_printk(KERN_WARNING, obj, fmt, ## args)
142 #define drbd_info(obj, fmt, args...) \
143 drbd_printk(KERN_INFO, obj, fmt, ## args)
144 #define drbd_emerg(obj, fmt, args...) \
145 drbd_printk(KERN_EMERG, obj, fmt, ## args)
147 #define dynamic_drbd_dbg(device, fmt, args...) \
148 dynamic_dev_dbg(disk_to_dev(device->vdisk), fmt, ## args)
150 #define conn_printk(LEVEL, TCONN, FMT, ARGS...) \
151 printk(LEVEL "d-con %s: " FMT, TCONN->resource->name , ## ARGS)
152 #define conn_alert(TCONN, FMT, ARGS...) conn_printk(KERN_ALERT, TCONN, FMT, ## ARGS)
153 #define conn_crit(TCONN, FMT, ARGS...) conn_printk(KERN_CRIT, TCONN, FMT, ## ARGS)
154 #define conn_err(TCONN, FMT, ARGS...) conn_printk(KERN_ERR, TCONN, FMT, ## ARGS)
155 #define conn_warn(TCONN, FMT, ARGS...) conn_printk(KERN_WARNING, TCONN, FMT, ## ARGS)
156 #define conn_notice(TCONN, FMT, ARGS...) conn_printk(KERN_NOTICE, TCONN, FMT, ## ARGS)
157 #define conn_info(TCONN, FMT, ARGS...) conn_printk(KERN_INFO, TCONN, FMT, ## ARGS)
158 #define conn_dbg(TCONN, FMT, ARGS...) conn_printk(KERN_DEBUG, TCONN, FMT, ## ARGS)
160 #define D_ASSERT(exp) if (!(exp)) \
161 drbd_err(device, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
164 * expect - Make an assertion
166 * Unlike the assert macro, this macro returns a boolean result.
168 #define expect(exp) ({ \
169 bool _bool = (exp); \
171 drbd_err(device, "ASSERTION %s FAILED in %s\n", \
176 /* Defines to control fault insertion */
178 DRBD_FAULT_MD_WR
= 0, /* meta data write */
179 DRBD_FAULT_MD_RD
= 1, /* read */
180 DRBD_FAULT_RS_WR
= 2, /* resync */
181 DRBD_FAULT_RS_RD
= 3,
182 DRBD_FAULT_DT_WR
= 4, /* data */
183 DRBD_FAULT_DT_RD
= 5,
184 DRBD_FAULT_DT_RA
= 6, /* data read ahead */
185 DRBD_FAULT_BM_ALLOC
= 7, /* bitmap allocation */
186 DRBD_FAULT_AL_EE
= 8, /* alloc ee */
187 DRBD_FAULT_RECEIVE
= 9, /* Changes some bytes upon receiving a [rs]data block */
193 _drbd_insert_fault(struct drbd_device
*device
, unsigned int type
);
196 drbd_insert_fault(struct drbd_device
*device
, unsigned int type
) {
197 #ifdef CONFIG_DRBD_FAULT_INJECTION
199 (enable_faults
& (1<<type
)) &&
200 _drbd_insert_fault(device
, type
);
206 /* integer division, round _UP_ to the next integer */
207 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
208 /* usual integer division */
209 #define div_floor(A, B) ((A)/(B))
211 extern struct ratelimit_state drbd_ratelimit_state
;
212 extern struct idr drbd_devices
; /* RCU, updates: genl_lock() */
213 extern struct list_head drbd_resources
; /* RCU, updates: genl_lock() */
215 extern const char *cmdname(enum drbd_packet cmd
);
217 /* for sending/receiving the bitmap,
218 * possibly in some encoding scheme */
221 * stores total bits and long words
222 * of the bitmap, so we don't need to
223 * call the accessor functions over and again. */
224 unsigned long bm_bits
;
225 unsigned long bm_words
;
226 /* during xfer, current position within the bitmap */
227 unsigned long bit_offset
;
228 unsigned long word_offset
;
230 /* statistics; index: (h->command == P_BITMAP) */
235 extern void INFO_bm_xfer_stats(struct drbd_device
*device
,
236 const char *direction
, struct bm_xfer_ctx
*c
);
238 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx
*c
)
240 /* word_offset counts "native long words" (32 or 64 bit),
242 * Encoded packet may end at an unaligned bit offset.
243 * In case a fallback clear text packet is transmitted in
244 * between, we adjust this offset back to the last 64bit
245 * aligned "native long word", which makes coding and decoding
246 * the plain text bitmap much more convenient. */
247 #if BITS_PER_LONG == 64
248 c
->word_offset
= c
->bit_offset
>> 6;
249 #elif BITS_PER_LONG == 32
250 c
->word_offset
= c
->bit_offset
>> 5;
251 c
->word_offset
&= ~(1UL);
253 # error "unsupported BITS_PER_LONG"
257 extern unsigned int drbd_header_size(struct drbd_connection
*connection
);
259 /**********************************************************************/
260 enum drbd_thread_state
{
269 struct task_struct
*task
;
270 struct completion stop
;
271 enum drbd_thread_state t_state
;
272 int (*function
) (struct drbd_thread
*);
273 struct drbd_connection
*connection
;
278 static inline enum drbd_thread_state
get_t_state(struct drbd_thread
*thi
)
280 /* THINK testing the t_state seems to be uncritical in all cases
281 * (but thread_{start,stop}), so we can read it *without* the lock.
289 struct list_head list
;
290 int (*cb
)(struct drbd_work
*, int cancel
);
292 struct drbd_device
*device
;
293 struct drbd_connection
*connection
;
297 #include "drbd_interval.h"
299 extern int drbd_wait_misc(struct drbd_device
*, struct drbd_interval
*);
301 struct drbd_request
{
304 /* if local IO is not allowed, will be NULL.
305 * if local IO _is_ allowed, holds the locally submitted bio clone,
306 * or, after local IO completion, the ERR_PTR(error).
307 * see drbd_request_endio(). */
308 struct bio
*private_bio
;
310 struct drbd_interval i
;
312 /* epoch: used to check on "completion" whether this req was in
313 * the current epoch, and we therefore have to close it,
314 * causing a p_barrier packet to be send, starting a new epoch.
316 * This corresponds to "barrier" in struct p_barrier[_ack],
317 * and to "barrier_nr" in struct drbd_epoch (and various
318 * comments/function parameters/local variable names).
322 struct list_head tl_requests
; /* ring list in the transfer log */
323 struct bio
*master_bio
; /* master bio pointer */
324 unsigned long start_time
;
326 /* once it hits 0, we may complete the master_bio */
327 atomic_t completion_ref
;
328 /* once it hits 0, we may destroy this drbd_request object */
331 unsigned rq_state
; /* see comments above _req_mod() */
335 struct drbd_connection
*connection
;
336 struct list_head list
;
337 unsigned int barrier_nr
;
338 atomic_t epoch_size
; /* increased on every request added. */
339 atomic_t active
; /* increased on every req. added, and dec on every finished. */
343 /* Prototype declaration of function defined in drbd_receiver.c */
344 int drbdd_init(struct drbd_thread
*);
345 int drbd_asender(struct drbd_thread
*);
347 /* drbd_epoch flag bits */
349 DE_HAVE_BARRIER_NUMBER
,
356 EV_CLEANUP
= 32, /* used as flag */
359 struct drbd_wq_barrier
{
361 struct completion done
;
369 struct drbd_peer_request
{
371 struct drbd_epoch
*epoch
; /* for writes */
373 atomic_t pending_bios
;
374 struct drbd_interval i
;
375 /* see comments on ee flag bits below */
379 struct digest_info
*digest
;
384 * While corresponding bios are in flight, the only modification will be
385 * set_bit WAS_ERROR, which has to be atomic.
386 * If no bios are in flight yet, or all have been completed,
387 * non-atomic modification to ee->flags is ok.
390 __EE_CALL_AL_COMPLETE_IO
,
391 __EE_MAY_SET_IN_SYNC
,
393 /* In case a barrier failed,
394 * we need to resubmit without the barrier flag. */
397 /* we may have several bios per peer request.
398 * if any of those fail, we set this flag atomically
399 * from the endio callback */
402 /* This ee has a pointer to a digest instead of a block id */
405 /* Conflicting local requests need to be restarted after this request */
406 __EE_RESTART_REQUESTS
,
408 /* The peer wants a write ACK for this (wire proto C) */
411 /* Is set when net_conf had two_primaries set while creating this peer_req */
412 __EE_IN_INTERVAL_TREE
,
414 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
415 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
416 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
417 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
418 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
419 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
420 #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK)
421 #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE)
423 /* flag bits per device */
425 UNPLUG_REMOTE
, /* sending a "UnplugRemote" could help */
426 MD_DIRTY
, /* current uuids and flags not yet on disk */
427 USE_DEGR_WFC_T
, /* degr-wfc-timeout instead of wfc-timeout. */
430 CRASHED_PRIMARY
, /* This node was a crashed primary.
431 * Gets cleared when the state.conn
432 * goes into C_CONNECTED state. */
435 MD_NO_FUA
, /* Users wants us to not use FUA/FLUSH on meta data dev */
436 SUSPEND_IO
, /* suspend application io */
437 BITMAP_IO
, /* suspend application io;
438 once no more io in flight, start bitmap io */
439 BITMAP_IO_QUEUED
, /* Started bitmap IO */
440 GO_DISKLESS
, /* Disk is being detached, on io-error or admin request. */
441 WAS_IO_ERROR
, /* Local disk failed, returned IO error */
442 WAS_READ_ERROR
, /* Local disk READ failed (set additionally to the above) */
443 FORCE_DETACH
, /* Force-detach from local disk, aborting any pending local IO */
444 RESYNC_AFTER_NEG
, /* Resync after online grow after the attach&negotiate finished. */
445 RESIZE_PENDING
, /* Size change detected locally, waiting for the response from
446 * the peer, if it changed there as well. */
447 NEW_CUR_UUID
, /* Create new current UUID when thawing IO */
448 AL_SUSPENDED
, /* Activity logging is currently suspended. */
449 AHEAD_TO_SYNC_SOURCE
, /* Ahead -> SyncSource queued */
450 B_RS_H_DONE
, /* Before resync handler done (already executed) */
451 DISCARD_MY_DATA
, /* discard_my_data flag per volume */
455 struct drbd_bitmap
; /* opaque for drbd_device */
457 /* definition of bits in bm_flags to be used in drbd_bm_lock
458 * and drbd_bitmap_io and friends. */
460 /* do we need to kfree, or vfree bm_pages? */
461 BM_P_VMALLOCED
= 0x10000, /* internal use only, will be masked out */
463 /* currently locked for bulk operation */
464 BM_LOCKED_MASK
= 0xf,
466 /* in detail, that is: */
471 /* so we can mark it locked for bulk operation,
472 * and still allow all non-bulk operations */
475 /* (test bit, count bit) allowed (common case) */
476 BM_LOCKED_TEST_ALLOWED
= BM_DONT_CLEAR
| BM_DONT_SET
| BM_IS_LOCKED
,
478 /* testing bits, as well as setting new bits allowed, but clearing bits
479 * would be unexpected. Used during bitmap receive. Setting new bits
480 * requires sending of "out-of-sync" information, though. */
481 BM_LOCKED_SET_ALLOWED
= BM_DONT_CLEAR
| BM_IS_LOCKED
,
483 /* for drbd_bm_write_copy_pages, everything is allowed,
484 * only concurrent bulk operations are locked out. */
485 BM_LOCKED_CHANGE_ALLOWED
= BM_IS_LOCKED
,
488 struct drbd_work_queue
{
490 spinlock_t q_lock
; /* to protect the list. */
491 wait_queue_head_t q_wait
;
496 struct socket
*socket
;
497 /* this way we get our
498 * send/receive buffers off the stack */
504 u64 md_offset
; /* sector offset to 'super' block */
506 u64 la_size_sect
; /* last agreed size, unit sectors */
507 spinlock_t uuid_lock
;
513 s32 al_offset
; /* signed relative sector offset to activity log */
514 s32 bm_offset
; /* signed relative sector offset to bitmap */
516 /* cached value of bdev->disk_conf->meta_dev_idx (see below) */
519 /* see al_tr_number_to_on_disk_sector() */
521 u32 al_stripe_size_4k
;
522 u32 al_size_4k
; /* cached product of the above */
525 struct drbd_backing_dev
{
526 struct block_device
*backing_bdev
;
527 struct block_device
*md_bdev
;
529 struct disk_conf
*disk_conf
; /* RCU, for updates: first_peer_device(device)->connection->conf_update */
530 sector_t known_size
; /* last known size of that backing device */
542 int (*io_fn
)(struct drbd_device
*device
);
543 void (*done
)(struct drbd_device
*device
, int rv
);
546 enum write_ordering_e
{
553 unsigned int head_index
;
555 int total
; /* sum of all values */
558 extern struct fifo_buffer
*fifo_alloc(int fifo_size
);
560 /* flag bits per connection */
562 NET_CONGESTED
, /* The data socket is congested */
563 RESOLVE_CONFLICTS
, /* Set on one node, cleared on the peer! */
564 SEND_PING
, /* whether asender should send a ping asap */
565 SIGNAL_ASENDER
, /* whether asender wants to be interrupted */
566 GOT_PING_ACK
, /* set when we receive a ping_ack packet, ping_wait gets woken */
567 CONN_WD_ST_CHG_REQ
, /* A cluster wide state change on the connection is active */
570 CONN_DRY_RUN
, /* Expect disconnect after resync handshake. */
571 CREATE_BARRIER
, /* next P_DATA is preceded by a P_BARRIER */
572 STATE_SENT
, /* Do not change state/UUIDs while this is set */
573 CALLBACK_PENDING
, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
574 * pending, from drbd worker context.
575 * If set, bdi_write_congested() returns true,
576 * so shrink_page_list() would not recurse into,
577 * and potentially deadlock on, this drbd worker.
582 struct drbd_resource
{
585 struct idr devices
; /* volume number to device mapping */
586 struct list_head connections
;
587 struct list_head resources
;
588 struct res_opts res_opts
;
591 struct drbd_connection
{
592 struct list_head connections
;
593 struct drbd_resource
*resource
;
595 struct idr peer_devices
; /* volume number to peer device mapping */
596 enum drbd_conns cstate
; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
597 unsigned susp
:1; /* IO suspended by user */
598 unsigned susp_nod
:1; /* IO suspended because no data */
599 unsigned susp_fen
:1; /* IO suspended because fence peer handler runs */
600 struct mutex cstate_mutex
; /* Protects graceful disconnects */
601 unsigned int connect_cnt
; /* Inc each time a connection is established */
604 struct net_conf
*net_conf
; /* content protected by rcu */
605 struct mutex conf_update
; /* mutex for ready-copy-update of net_conf and disk_conf */
606 wait_queue_head_t ping_wait
; /* Woken upon reception of a ping, and a state change */
608 struct sockaddr_storage my_addr
;
610 struct sockaddr_storage peer_addr
;
613 struct drbd_socket data
; /* data/barrier/cstate/parameter packets */
614 struct drbd_socket meta
; /* ping/ack (metadata) packets */
615 int agreed_pro_version
; /* actually used protocol version */
616 unsigned long last_received
; /* in jiffies, either socket */
617 unsigned int ko_count
;
621 struct list_head transfer_log
; /* all requests not yet fully processed */
623 struct crypto_hash
*cram_hmac_tfm
;
624 struct crypto_hash
*integrity_tfm
; /* checksums we compute, updates protected by connection->data->mutex */
625 struct crypto_hash
*peer_integrity_tfm
; /* checksums we verify, only accessed from receiver thread */
626 struct crypto_hash
*csums_tfm
;
627 struct crypto_hash
*verify_tfm
;
632 struct drbd_epoch
*current_epoch
;
633 spinlock_t epoch_lock
;
635 enum write_ordering_e write_ordering
;
636 atomic_t current_tle_nr
; /* transfer log epoch number */
637 unsigned current_tle_writes
; /* writes seen within this tl epoch */
639 unsigned long last_reconnect_jif
;
640 struct drbd_thread receiver
;
641 struct drbd_thread worker
;
642 struct drbd_thread asender
;
643 cpumask_var_t cpu_mask
;
646 struct drbd_work_queue sender_work
;
649 /* whether this sender thread
650 * has processed a single write yet. */
651 bool seen_any_write_yet
;
653 /* Which barrier number to send with the next P_BARRIER */
654 int current_epoch_nr
;
656 /* how many write requests have been sent
657 * with req->epoch == current_epoch_nr.
658 * If none, no P_BARRIER will be sent. */
659 unsigned current_epoch_writes
;
663 struct submit_worker
{
664 struct workqueue_struct
*wq
;
665 struct work_struct worker
;
668 struct list_head writes
;
671 struct drbd_peer_device
{
672 struct list_head peer_devices
;
673 struct drbd_device
*device
;
674 struct drbd_connection
*connection
;
678 struct drbd_resource
*resource
;
679 struct list_head peer_devices
;
680 int vnr
; /* volume number within the connection */
683 /* things that are stored as / read from meta data on disk */
686 /* configured by drbdsetup */
687 struct drbd_backing_dev
*ldev
__protected_by(local
);
689 sector_t p_size
; /* partner's disk size */
690 struct request_queue
*rq_queue
;
691 struct block_device
*this_bdev
;
692 struct gendisk
*vdisk
;
694 unsigned long last_reattach_jif
;
695 struct drbd_work resync_work
,
700 struct timer_list resync_timer
;
701 struct timer_list md_sync_timer
;
702 struct timer_list start_resync_timer
;
703 struct timer_list request_timer
;
704 #ifdef DRBD_DEBUG_MD_SYNC
708 } last_md_mark_dirty
;
711 /* Used after attach while negotiating new disk state. */
712 union drbd_state new_state_tmp
;
714 union drbd_dev_state state
;
715 wait_queue_head_t misc_wait
;
716 wait_queue_head_t state_wait
; /* upon each state change. */
717 unsigned int send_cnt
;
718 unsigned int recv_cnt
;
719 unsigned int read_cnt
;
720 unsigned int writ_cnt
;
721 unsigned int al_writ_cnt
;
722 unsigned int bm_writ_cnt
;
723 atomic_t ap_bio_cnt
; /* Requests we need to complete */
724 atomic_t ap_pending_cnt
; /* AP data packets on the wire, ack expected */
725 atomic_t rs_pending_cnt
; /* RS request/data packets on the wire */
726 atomic_t unacked_cnt
; /* Need to send replies for */
727 atomic_t local_cnt
; /* Waiting for local completion */
729 /* Interval tree of pending local requests */
730 struct rb_root read_requests
;
731 struct rb_root write_requests
;
733 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
734 unsigned long rs_total
;
735 /* number of resync blocks that failed in this run */
736 unsigned long rs_failed
;
737 /* Syncer's start time [unit jiffies] */
738 unsigned long rs_start
;
739 /* cumulated time in PausedSyncX state [unit jiffies] */
740 unsigned long rs_paused
;
741 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
742 unsigned long rs_same_csum
;
743 #define DRBD_SYNC_MARKS 8
744 #define DRBD_SYNC_MARK_STEP (3*HZ)
745 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
746 unsigned long rs_mark_left
[DRBD_SYNC_MARKS
];
747 /* marks's time [unit jiffies] */
748 unsigned long rs_mark_time
[DRBD_SYNC_MARKS
];
749 /* current index into rs_mark_{left,time} */
751 unsigned long rs_last_bcast
; /* [unit jiffies] */
753 /* where does the admin want us to start? (sector) */
754 sector_t ov_start_sector
;
755 sector_t ov_stop_sector
;
756 /* where are we now? (sector) */
757 sector_t ov_position
;
758 /* Start sector of out of sync range (to merge printk reporting). */
759 sector_t ov_last_oos_start
;
760 /* size of out-of-sync range in sectors. */
761 sector_t ov_last_oos_size
;
762 unsigned long ov_left
; /* in bits */
764 struct drbd_bitmap
*bitmap
;
765 unsigned long bm_resync_fo
; /* bit offset for drbd_bm_find_next */
767 /* Used to track operations of resync... */
768 struct lru_cache
*resync
;
769 /* Number of locked elements in resync LRU */
770 unsigned int resync_locked
;
771 /* resync extent number waiting for application requests */
772 unsigned int resync_wenr
;
777 struct list_head active_ee
; /* IO in progress (P_DATA gets written to disk) */
778 struct list_head sync_ee
; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
779 struct list_head done_ee
; /* need to send P_WRITE_ACK */
780 struct list_head read_ee
; /* [RS]P_DATA_REQUEST being read */
781 struct list_head net_ee
; /* zero-copy network send in progress */
784 struct list_head resync_reads
;
785 atomic_t pp_in_use
; /* allocated from page pool */
786 atomic_t pp_in_use_by_net
; /* sendpage()d, still referenced by tcp */
787 wait_queue_head_t ee_wait
;
788 struct page
*md_io_page
; /* one page buffer for md_io */
789 struct drbd_md_io md_io
;
790 atomic_t md_io_in_use
; /* protects the md_io, md_io_page and md_io_tmpp */
792 wait_queue_head_t al_wait
;
793 struct lru_cache
*act_log
; /* activity log */
794 unsigned int al_tr_number
;
796 wait_queue_head_t seq_wait
;
798 unsigned int peer_seq
;
799 spinlock_t peer_seq_lock
;
801 unsigned long comm_bm_set
; /* communicated number of set bits. */
802 struct bm_io_work bm_io_work
;
803 u64 ed_uuid
; /* UUID of the exposed data */
804 struct mutex own_state_mutex
;
805 struct mutex
*state_mutex
; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */
806 char congestion_reason
; /* Why we where congested... */
807 atomic_t rs_sect_in
; /* for incoming resync data rate, SyncTarget */
808 atomic_t rs_sect_ev
; /* for submitted resync data rate, both */
809 int rs_last_sect_ev
; /* counter to compare with */
810 int rs_last_events
; /* counter of read or write "events" (unit sectors)
811 * on the lower level device when we last looked. */
812 int c_sync_rate
; /* current resync rate after syncer throttle magic */
813 struct fifo_buffer
*rs_plan_s
; /* correction values of resync planer (RCU, connection->conn_update) */
814 int rs_in_flight
; /* resync sectors in flight (to proxy, in proxy and from proxy) */
815 atomic_t ap_in_flight
; /* App sectors in flight (waiting for ack) */
816 unsigned int peer_max_bio_size
;
817 unsigned int local_max_bio_size
;
819 /* any requests that would block in drbd_make_request()
820 * are deferred to this single-threaded work queue */
821 struct submit_worker submit
;
824 static inline struct drbd_device
*minor_to_device(unsigned int minor
)
826 return (struct drbd_device
*)idr_find(&drbd_devices
, minor
);
829 static inline struct drbd_peer_device
*first_peer_device(struct drbd_device
*device
)
831 return list_first_entry(&device
->peer_devices
, struct drbd_peer_device
, peer_devices
);
834 #define for_each_resource(resource, _resources) \
835 list_for_each_entry(resource, _resources, resources)
837 #define for_each_resource_rcu(resource, _resources) \
838 list_for_each_entry_rcu(resource, _resources, resources)
840 #define for_each_resource_safe(resource, tmp, _resources) \
841 list_for_each_entry_safe(resource, tmp, _resources, resources)
843 #define for_each_connection(connection, resource) \
844 list_for_each_entry(connection, &resource->connections, connections)
846 #define for_each_connection_rcu(connection, resource) \
847 list_for_each_entry_rcu(connection, &resource->connections, connections)
849 #define for_each_connection_safe(connection, tmp, resource) \
850 list_for_each_entry_safe(connection, tmp, &resource->connections, connections)
852 #define for_each_peer_device(peer_device, device) \
853 list_for_each_entry(peer_device, &device->peer_devices, peer_devices)
855 #define for_each_peer_device_rcu(peer_device, device) \
856 list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices)
858 #define for_each_peer_device_safe(peer_device, tmp, device) \
859 list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices)
861 static inline unsigned int device_to_minor(struct drbd_device
*device
)
863 return device
->minor
;
866 static inline struct drbd_device
*vnr_to_device(struct drbd_connection
*connection
, int vnr
)
868 struct drbd_peer_device
*peer_device
;
870 peer_device
= idr_find(&connection
->peer_devices
, vnr
);
871 return peer_device
? peer_device
->device
: NULL
;
875 * function declarations
876 *************************/
882 DDSF_NO_RESYNC
= 2, /* Do not run a resync for the new space */
885 extern void drbd_init_set_defaults(struct drbd_device
*device
);
886 extern int drbd_thread_start(struct drbd_thread
*thi
);
887 extern void _drbd_thread_stop(struct drbd_thread
*thi
, int restart
, int wait
);
888 extern char *drbd_task_to_thread_name(struct drbd_connection
*connection
, struct task_struct
*task
);
890 extern void drbd_thread_current_set_cpu(struct drbd_thread
*thi
);
891 extern void drbd_calc_cpu_mask(struct drbd_connection
*connection
);
893 #define drbd_thread_current_set_cpu(A) ({})
894 #define drbd_calc_cpu_mask(A) ({})
896 extern void tl_release(struct drbd_connection
*, unsigned int barrier_nr
,
897 unsigned int set_size
);
898 extern void tl_clear(struct drbd_connection
*);
899 extern void drbd_free_sock(struct drbd_connection
*connection
);
900 extern int drbd_send(struct drbd_connection
*connection
, struct socket
*sock
,
901 void *buf
, size_t size
, unsigned msg_flags
);
902 extern int drbd_send_all(struct drbd_connection
*, struct socket
*, void *, size_t,
905 extern int __drbd_send_protocol(struct drbd_connection
*connection
, enum drbd_packet cmd
);
906 extern int drbd_send_protocol(struct drbd_connection
*connection
);
907 extern int drbd_send_uuids(struct drbd_device
*device
);
908 extern int drbd_send_uuids_skip_initial_sync(struct drbd_device
*device
);
909 extern void drbd_gen_and_send_sync_uuid(struct drbd_device
*device
);
910 extern int drbd_send_sizes(struct drbd_device
*device
, int trigger_reply
, enum dds_flags flags
);
911 extern int drbd_send_state(struct drbd_device
*device
, union drbd_state s
);
912 extern int drbd_send_current_state(struct drbd_device
*device
);
913 extern int drbd_send_sync_param(struct drbd_device
*device
);
914 extern void drbd_send_b_ack(struct drbd_connection
*connection
, u32 barrier_nr
,
916 extern int drbd_send_ack(struct drbd_device
*, enum drbd_packet
,
917 struct drbd_peer_request
*);
918 extern void drbd_send_ack_rp(struct drbd_device
*device
, enum drbd_packet cmd
,
919 struct p_block_req
*rp
);
920 extern void drbd_send_ack_dp(struct drbd_device
*device
, enum drbd_packet cmd
,
921 struct p_data
*dp
, int data_size
);
922 extern int drbd_send_ack_ex(struct drbd_device
*device
, enum drbd_packet cmd
,
923 sector_t sector
, int blksize
, u64 block_id
);
924 extern int drbd_send_out_of_sync(struct drbd_device
*, struct drbd_request
*);
925 extern int drbd_send_block(struct drbd_device
*, enum drbd_packet
,
926 struct drbd_peer_request
*);
927 extern int drbd_send_dblock(struct drbd_device
*device
, struct drbd_request
*req
);
928 extern int drbd_send_drequest(struct drbd_device
*device
, int cmd
,
929 sector_t sector
, int size
, u64 block_id
);
930 extern int drbd_send_drequest_csum(struct drbd_device
*device
, sector_t sector
,
931 int size
, void *digest
, int digest_size
,
932 enum drbd_packet cmd
);
933 extern int drbd_send_ov_request(struct drbd_device
*device
, sector_t sector
, int size
);
935 extern int drbd_send_bitmap(struct drbd_device
*device
);
936 extern void drbd_send_sr_reply(struct drbd_device
*device
, enum drbd_state_rv retcode
);
937 extern void conn_send_sr_reply(struct drbd_connection
*connection
, enum drbd_state_rv retcode
);
938 extern void drbd_free_bc(struct drbd_backing_dev
*ldev
);
939 extern void drbd_device_cleanup(struct drbd_device
*device
);
940 void drbd_print_uuids(struct drbd_device
*device
, const char *text
);
942 extern void conn_md_sync(struct drbd_connection
*connection
);
943 extern void drbd_md_write(struct drbd_device
*device
, void *buffer
);
944 extern void drbd_md_sync(struct drbd_device
*device
);
945 extern int drbd_md_read(struct drbd_device
*device
, struct drbd_backing_dev
*bdev
);
946 extern void drbd_uuid_set(struct drbd_device
*device
, int idx
, u64 val
) __must_hold(local
);
947 extern void _drbd_uuid_set(struct drbd_device
*device
, int idx
, u64 val
) __must_hold(local
);
948 extern void drbd_uuid_new_current(struct drbd_device
*device
) __must_hold(local
);
949 extern void drbd_uuid_set_bm(struct drbd_device
*device
, u64 val
) __must_hold(local
);
950 extern void drbd_uuid_move_history(struct drbd_device
*device
) __must_hold(local
);
951 extern void __drbd_uuid_set(struct drbd_device
*device
, int idx
, u64 val
) __must_hold(local
);
952 extern void drbd_md_set_flag(struct drbd_device
*device
, int flags
) __must_hold(local
);
953 extern void drbd_md_clear_flag(struct drbd_device
*device
, int flags
)__must_hold(local
);
954 extern int drbd_md_test_flag(struct drbd_backing_dev
*, int);
955 #ifndef DRBD_DEBUG_MD_SYNC
956 extern void drbd_md_mark_dirty(struct drbd_device
*device
);
958 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
959 extern void drbd_md_mark_dirty_(struct drbd_device
*device
,
960 unsigned int line
, const char *func
);
962 extern void drbd_queue_bitmap_io(struct drbd_device
*device
,
963 int (*io_fn
)(struct drbd_device
*),
964 void (*done
)(struct drbd_device
*, int),
965 char *why
, enum bm_flag flags
);
966 extern int drbd_bitmap_io(struct drbd_device
*device
,
967 int (*io_fn
)(struct drbd_device
*),
968 char *why
, enum bm_flag flags
);
969 extern int drbd_bitmap_io_from_worker(struct drbd_device
*device
,
970 int (*io_fn
)(struct drbd_device
*),
971 char *why
, enum bm_flag flags
);
972 extern int drbd_bmio_set_n_write(struct drbd_device
*device
);
973 extern int drbd_bmio_clear_n_write(struct drbd_device
*device
);
974 extern void drbd_ldev_destroy(struct drbd_device
*device
);
978 * We currently have two possible layouts.
979 * Offsets in (512 byte) sectors.
981 * |----------- md_size_sect ------------------|
982 * [ 4k superblock ][ activity log ][ Bitmap ]
984 * | bm_offset = al_offset + X |
985 * ==> bitmap sectors = md_size_sect - bm_offset
988 * old, indexed fixed size meta data:
991 * |----------- md_size_sect ------------------|
992 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*]
994 * | bm_offset = al_offset - Y |
995 * ==> bitmap sectors = Y = al_offset - bm_offset
997 * [padding*] are zero or up to 7 unused 512 Byte sectors to the
998 * end of the device, so that the [4k superblock] will be 4k aligned.
1000 * The activity log consists of 4k transaction blocks,
1001 * which are written in a ring-buffer, or striped ring-buffer like fashion,
1002 * which are writtensize used to be fixed 32kB,
1003 * but is about to become configurable.
1006 /* Our old fixed size meta data layout
1007 * allows up to about 3.8TB, so if you want more,
1008 * you need to use the "flexible" meta data format. */
1009 #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */
1010 #define MD_4kB_SECT 8
1011 #define MD_32kB_SECT 64
1013 /* One activity log extent represents 4M of storage */
1014 #define AL_EXTENT_SHIFT 22
1015 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1017 /* We could make these currently hardcoded constants configurable
1018 * variables at create-md time (or even re-configurable at runtime?).
1019 * Which will require some more changes to the DRBD "super block"
1022 * updates per transaction:
1023 * This many changes to the active set can be logged with one transaction.
1024 * This number is arbitrary.
1025 * context per transaction:
1026 * This many context extent numbers are logged with each transaction.
1027 * This number is resulting from the transaction block size (4k), the layout
1028 * of the transaction header, and the number of updates per transaction.
1029 * See drbd_actlog.c:struct al_transaction_on_disk
1031 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1032 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1034 #if BITS_PER_LONG == 32
1036 #define cpu_to_lel(A) cpu_to_le32(A)
1037 #define lel_to_cpu(A) le32_to_cpu(A)
1038 #elif BITS_PER_LONG == 64
1040 #define cpu_to_lel(A) cpu_to_le64(A)
1041 #define lel_to_cpu(A) le64_to_cpu(A)
1043 #error "LN2 of BITS_PER_LONG unknown!"
1047 /* 16MB sized 'bitmap extent' to track syncer usage */
1049 int rs_left
; /* number of bits set (out of sync) in this extent. */
1050 int rs_failed
; /* number of failed resync requests in this extent. */
1051 unsigned long flags
;
1052 struct lc_element lce
;
1055 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1056 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1057 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1061 * We need to store one bit for a block.
1062 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1063 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1064 * Bit 1 ==> local node thinks this block needs to be synced.
1067 #define SLEEP_TIME (HZ/10)
1069 /* We do bitmap IO in units of 4k blocks.
1070 * We also still have a hardcoded 4k per bit relation. */
1071 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1072 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1073 /* mostly arbitrarily set the represented size of one bitmap extent,
1074 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1075 * at 4k per bit resolution) */
1076 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
1077 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1079 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1080 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1083 /* thus many _storage_ sectors are described by one bit */
1084 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1085 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1086 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1088 /* bit to represented kilo byte conversion */
1089 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1091 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1092 * _storage_ sector is located in */
1093 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1095 /* how much _storage_ sectors we have per bitmap sector */
1096 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1097 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1099 /* in one sector of the bitmap, we have this many activity_log extents. */
1100 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1102 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1103 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1105 /* the extent in "PER_EXTENT" below is an activity log extent
1106 * we need that many (long words/bytes) to store the bitmap
1107 * of one AL_EXTENT_SIZE chunk of storage.
1108 * we can store the bitmap for that many AL_EXTENTS within
1109 * one sector of the _on_disk_ bitmap:
1110 * bit 0 bit 37 bit 38 bit (512*8)-1
1111 * ...|........|........|.. // ..|........|
1112 * sect. 0 `296 `304 ^(512*8*8)-1
1114 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1115 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1116 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1119 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1120 /* we have a certain meta data variant that has a fixed on-disk size of 128
1121 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity
1122 * log, leaving this many sectors for the bitmap.
1125 #define DRBD_MAX_SECTORS_FIXED_BM \
1126 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9)))
1127 #if !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1128 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1129 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1131 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM
1132 /* 16 TB in units of sectors */
1133 #if BITS_PER_LONG == 32
1134 /* adjust by one page worth of bitmap,
1135 * so we won't wrap around in drbd_bm_find_next_bit.
1136 * you should use 64bit OS for that much storage, anyways. */
1137 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1139 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1140 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1141 /* corresponds to (1UL << 38) bits right now. */
1145 /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE,
1146 * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte.
1147 * Since we may live in a mixed-platform cluster,
1148 * we limit us to a platform agnostic constant here for now.
1149 * A followup commit may allow even bigger BIO sizes,
1150 * once we thought that through. */
1151 #define DRBD_MAX_BIO_SIZE (1U << 20)
1152 #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1153 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1155 #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */
1157 #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */
1158 #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */
1160 extern int drbd_bm_init(struct drbd_device
*device
);
1161 extern int drbd_bm_resize(struct drbd_device
*device
, sector_t sectors
, int set_new_bits
);
1162 extern void drbd_bm_cleanup(struct drbd_device
*device
);
1163 extern void drbd_bm_set_all(struct drbd_device
*device
);
1164 extern void drbd_bm_clear_all(struct drbd_device
*device
);
1165 /* set/clear/test only a few bits at a time */
1166 extern int drbd_bm_set_bits(
1167 struct drbd_device
*device
, unsigned long s
, unsigned long e
);
1168 extern int drbd_bm_clear_bits(
1169 struct drbd_device
*device
, unsigned long s
, unsigned long e
);
1170 extern int drbd_bm_count_bits(
1171 struct drbd_device
*device
, const unsigned long s
, const unsigned long e
);
1172 /* bm_set_bits variant for use while holding drbd_bm_lock,
1173 * may process the whole bitmap in one go */
1174 extern void _drbd_bm_set_bits(struct drbd_device
*device
,
1175 const unsigned long s
, const unsigned long e
);
1176 extern int drbd_bm_test_bit(struct drbd_device
*device
, unsigned long bitnr
);
1177 extern int drbd_bm_e_weight(struct drbd_device
*device
, unsigned long enr
);
1178 extern int drbd_bm_write_page(struct drbd_device
*device
, unsigned int idx
) __must_hold(local
);
1179 extern int drbd_bm_read(struct drbd_device
*device
) __must_hold(local
);
1180 extern void drbd_bm_mark_for_writeout(struct drbd_device
*device
, int page_nr
);
1181 extern int drbd_bm_write(struct drbd_device
*device
) __must_hold(local
);
1182 extern int drbd_bm_write_hinted(struct drbd_device
*device
) __must_hold(local
);
1183 extern int drbd_bm_write_all(struct drbd_device
*device
) __must_hold(local
);
1184 extern int drbd_bm_write_copy_pages(struct drbd_device
*device
) __must_hold(local
);
1185 extern size_t drbd_bm_words(struct drbd_device
*device
);
1186 extern unsigned long drbd_bm_bits(struct drbd_device
*device
);
1187 extern sector_t
drbd_bm_capacity(struct drbd_device
*device
);
1189 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1190 extern unsigned long drbd_bm_find_next(struct drbd_device
*device
, unsigned long bm_fo
);
1191 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1192 extern unsigned long _drbd_bm_find_next(struct drbd_device
*device
, unsigned long bm_fo
);
1193 extern unsigned long _drbd_bm_find_next_zero(struct drbd_device
*device
, unsigned long bm_fo
);
1194 extern unsigned long _drbd_bm_total_weight(struct drbd_device
*device
);
1195 extern unsigned long drbd_bm_total_weight(struct drbd_device
*device
);
1196 extern int drbd_bm_rs_done(struct drbd_device
*device
);
1197 /* for receive_bitmap */
1198 extern void drbd_bm_merge_lel(struct drbd_device
*device
, size_t offset
,
1199 size_t number
, unsigned long *buffer
);
1200 /* for _drbd_send_bitmap */
1201 extern void drbd_bm_get_lel(struct drbd_device
*device
, size_t offset
,
1202 size_t number
, unsigned long *buffer
);
1204 extern void drbd_bm_lock(struct drbd_device
*device
, char *why
, enum bm_flag flags
);
1205 extern void drbd_bm_unlock(struct drbd_device
*device
);
1208 extern struct kmem_cache
*drbd_request_cache
;
1209 extern struct kmem_cache
*drbd_ee_cache
; /* peer requests */
1210 extern struct kmem_cache
*drbd_bm_ext_cache
; /* bitmap extents */
1211 extern struct kmem_cache
*drbd_al_ext_cache
; /* activity log extents */
1212 extern mempool_t
*drbd_request_mempool
;
1213 extern mempool_t
*drbd_ee_mempool
;
1215 /* drbd's page pool, used to buffer data received from the peer,
1216 * or data requested by the peer.
1218 * This does not have an emergency reserve.
1220 * When allocating from this pool, it first takes pages from the pool.
1221 * Only if the pool is depleted will try to allocate from the system.
1223 * The assumption is that pages taken from this pool will be processed,
1224 * and given back, "quickly", and then can be recycled, so we can avoid
1225 * frequent calls to alloc_page(), and still will be able to make progress even
1226 * under memory pressure.
1228 extern struct page
*drbd_pp_pool
;
1229 extern spinlock_t drbd_pp_lock
;
1230 extern int drbd_pp_vacant
;
1231 extern wait_queue_head_t drbd_pp_wait
;
1233 /* We also need a standard (emergency-reserve backed) page pool
1234 * for meta data IO (activity log, bitmap).
1235 * We can keep it global, as long as it is used as "N pages at a time".
1236 * 128 should be plenty, currently we probably can get away with as few as 1.
1238 #define DRBD_MIN_POOL_PAGES 128
1239 extern mempool_t
*drbd_md_io_page_pool
;
1241 /* We also need to make sure we get a bio
1242 * when we need it for housekeeping purposes */
1243 extern struct bio_set
*drbd_md_io_bio_set
;
1244 /* to allocate from that set */
1245 extern struct bio
*bio_alloc_drbd(gfp_t gfp_mask
);
1247 extern rwlock_t global_state_lock
;
1249 extern int conn_lowest_minor(struct drbd_connection
*connection
);
1250 enum drbd_ret_code
drbd_create_minor(struct drbd_connection
*connection
, unsigned int minor
, int vnr
);
1251 extern void drbd_destroy_device(struct kref
*kref
);
1252 extern void drbd_delete_minor(struct drbd_device
*mdev
);
1254 extern struct drbd_resource
*drbd_create_resource(const char *name
);
1255 extern void drbd_free_resource(struct drbd_resource
*resource
);
1257 extern int set_resource_options(struct drbd_resource
*resource
, struct res_opts
*res_opts
);
1258 extern struct drbd_connection
*conn_create(const char *name
, struct res_opts
*res_opts
);
1259 extern void drbd_destroy_connection(struct kref
*kref
);
1260 extern struct drbd_connection
*conn_get_by_addrs(void *my_addr
, int my_addr_len
,
1261 void *peer_addr
, int peer_addr_len
);
1262 extern struct drbd_resource
*drbd_find_resource(const char *name
);
1263 extern void drbd_destroy_resource(struct kref
*kref
);
1264 extern void conn_free_crypto(struct drbd_connection
*connection
);
1266 extern int proc_details
;
1269 extern void do_submit(struct work_struct
*ws
);
1270 extern void __drbd_make_request(struct drbd_device
*, struct bio
*, unsigned long);
1271 extern void drbd_make_request(struct request_queue
*q
, struct bio
*bio
);
1272 extern int drbd_read_remote(struct drbd_device
*device
, struct drbd_request
*req
);
1273 extern int drbd_merge_bvec(struct request_queue
*q
, struct bvec_merge_data
*bvm
, struct bio_vec
*bvec
);
1274 extern int is_valid_ar_handle(struct drbd_request
*, sector_t
);
1278 extern int drbd_msg_put_info(const char *info
);
1279 extern void drbd_suspend_io(struct drbd_device
*device
);
1280 extern void drbd_resume_io(struct drbd_device
*device
);
1281 extern char *ppsize(char *buf
, unsigned long long size
);
1282 extern sector_t
drbd_new_dev_size(struct drbd_device
*, struct drbd_backing_dev
*, sector_t
, int);
1283 enum determine_dev_size
{
1284 DS_ERROR_SHRINK
= -3,
1285 DS_ERROR_SPACE_MD
= -2,
1290 DS_GREW_FROM_ZERO
= 3,
1292 extern enum determine_dev_size
1293 drbd_determine_dev_size(struct drbd_device
*, enum dds_flags
, struct resize_parms
*) __must_hold(local
);
1294 extern void resync_after_online_grow(struct drbd_device
*);
1295 extern void drbd_reconsider_max_bio_size(struct drbd_device
*device
);
1296 extern enum drbd_state_rv
drbd_set_role(struct drbd_device
*device
,
1297 enum drbd_role new_role
,
1299 extern bool conn_try_outdate_peer(struct drbd_connection
*connection
);
1300 extern void conn_try_outdate_peer_async(struct drbd_connection
*connection
);
1301 extern int drbd_khelper(struct drbd_device
*device
, char *cmd
);
1304 extern int drbd_worker(struct drbd_thread
*thi
);
1305 enum drbd_ret_code
drbd_resync_after_valid(struct drbd_device
*device
, int o_minor
);
1306 void drbd_resync_after_changed(struct drbd_device
*device
);
1307 extern void drbd_start_resync(struct drbd_device
*device
, enum drbd_conns side
);
1308 extern void resume_next_sg(struct drbd_device
*device
);
1309 extern void suspend_other_sg(struct drbd_device
*device
);
1310 extern int drbd_resync_finished(struct drbd_device
*device
);
1311 /* maybe rather drbd_main.c ? */
1312 extern void *drbd_md_get_buffer(struct drbd_device
*device
);
1313 extern void drbd_md_put_buffer(struct drbd_device
*device
);
1314 extern int drbd_md_sync_page_io(struct drbd_device
*device
,
1315 struct drbd_backing_dev
*bdev
, sector_t sector
, int rw
);
1316 extern void drbd_ov_out_of_sync_found(struct drbd_device
*, sector_t
, int);
1317 extern void wait_until_done_or_force_detached(struct drbd_device
*device
,
1318 struct drbd_backing_dev
*bdev
, unsigned int *done
);
1319 extern void drbd_rs_controller_reset(struct drbd_device
*device
);
1321 static inline void ov_out_of_sync_print(struct drbd_device
*device
)
1323 if (device
->ov_last_oos_size
) {
1324 drbd_err(device
, "Out of sync: start=%llu, size=%lu (sectors)\n",
1325 (unsigned long long)device
->ov_last_oos_start
,
1326 (unsigned long)device
->ov_last_oos_size
);
1328 device
->ov_last_oos_size
= 0;
1332 extern void drbd_csum_bio(struct drbd_device
*, struct crypto_hash
*, struct bio
*, void *);
1333 extern void drbd_csum_ee(struct drbd_device
*, struct crypto_hash
*,
1334 struct drbd_peer_request
*, void *);
1335 /* worker callbacks */
1336 extern int w_e_end_data_req(struct drbd_work
*, int);
1337 extern int w_e_end_rsdata_req(struct drbd_work
*, int);
1338 extern int w_e_end_csum_rs_req(struct drbd_work
*, int);
1339 extern int w_e_end_ov_reply(struct drbd_work
*, int);
1340 extern int w_e_end_ov_req(struct drbd_work
*, int);
1341 extern int w_ov_finished(struct drbd_work
*, int);
1342 extern int w_resync_timer(struct drbd_work
*, int);
1343 extern int w_send_write_hint(struct drbd_work
*, int);
1344 extern int w_make_resync_request(struct drbd_work
*, int);
1345 extern int w_send_dblock(struct drbd_work
*, int);
1346 extern int w_send_read_req(struct drbd_work
*, int);
1347 extern int w_prev_work_done(struct drbd_work
*, int);
1348 extern int w_e_reissue(struct drbd_work
*, int);
1349 extern int w_restart_disk_io(struct drbd_work
*, int);
1350 extern int w_send_out_of_sync(struct drbd_work
*, int);
1351 extern int w_start_resync(struct drbd_work
*, int);
1353 extern void resync_timer_fn(unsigned long data
);
1354 extern void start_resync_timer_fn(unsigned long data
);
1356 /* drbd_receiver.c */
1357 extern int drbd_rs_should_slow_down(struct drbd_device
*device
, sector_t sector
);
1358 extern int drbd_submit_peer_request(struct drbd_device
*,
1359 struct drbd_peer_request
*, const unsigned,
1361 extern int drbd_free_peer_reqs(struct drbd_device
*, struct list_head
*);
1362 extern struct drbd_peer_request
*drbd_alloc_peer_req(struct drbd_device
*, u64
,
1363 sector_t
, unsigned int,
1364 gfp_t
) __must_hold(local
);
1365 extern void __drbd_free_peer_req(struct drbd_device
*, struct drbd_peer_request
*,
1367 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1368 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
1369 extern struct page
*drbd_alloc_pages(struct drbd_device
*, unsigned int, bool);
1370 extern void drbd_set_recv_tcq(struct drbd_device
*device
, int tcq_enabled
);
1371 extern void _drbd_clear_done_ee(struct drbd_device
*device
, struct list_head
*to_be_freed
);
1372 extern void conn_flush_workqueue(struct drbd_connection
*connection
);
1373 extern int drbd_connected(struct drbd_device
*device
);
1374 static inline void drbd_flush_workqueue(struct drbd_device
*device
)
1376 conn_flush_workqueue(first_peer_device(device
)->connection
);
1379 /* Yes, there is kernel_setsockopt, but only since 2.6.18.
1380 * So we have our own copy of it here. */
1381 static inline int drbd_setsockopt(struct socket
*sock
, int level
, int optname
,
1382 char *optval
, int optlen
)
1384 mm_segment_t oldfs
= get_fs();
1385 char __user
*uoptval
;
1388 uoptval
= (char __user __force
*)optval
;
1391 if (level
== SOL_SOCKET
)
1392 err
= sock_setsockopt(sock
, level
, optname
, uoptval
, optlen
);
1394 err
= sock
->ops
->setsockopt(sock
, level
, optname
, uoptval
,
1400 static inline void drbd_tcp_cork(struct socket
*sock
)
1403 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1404 (char*)&val
, sizeof(val
));
1407 static inline void drbd_tcp_uncork(struct socket
*sock
)
1410 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1411 (char*)&val
, sizeof(val
));
1414 static inline void drbd_tcp_nodelay(struct socket
*sock
)
1417 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_NODELAY
,
1418 (char*)&val
, sizeof(val
));
1421 static inline void drbd_tcp_quickack(struct socket
*sock
)
1424 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_QUICKACK
,
1425 (char*)&val
, sizeof(val
));
1428 void drbd_bump_write_ordering(struct drbd_connection
*connection
, enum write_ordering_e wo
);
1431 extern struct proc_dir_entry
*drbd_proc
;
1432 extern const struct file_operations drbd_proc_fops
;
1433 extern const char *drbd_conn_str(enum drbd_conns s
);
1434 extern const char *drbd_role_str(enum drbd_role s
);
1437 extern int drbd_al_begin_io_nonblock(struct drbd_device
*device
, struct drbd_interval
*i
);
1438 extern void drbd_al_begin_io_commit(struct drbd_device
*device
, bool delegate
);
1439 extern bool drbd_al_begin_io_fastpath(struct drbd_device
*device
, struct drbd_interval
*i
);
1440 extern void drbd_al_begin_io(struct drbd_device
*device
, struct drbd_interval
*i
, bool delegate
);
1441 extern void drbd_al_complete_io(struct drbd_device
*device
, struct drbd_interval
*i
);
1442 extern void drbd_rs_complete_io(struct drbd_device
*device
, sector_t sector
);
1443 extern int drbd_rs_begin_io(struct drbd_device
*device
, sector_t sector
);
1444 extern int drbd_try_rs_begin_io(struct drbd_device
*device
, sector_t sector
);
1445 extern void drbd_rs_cancel_all(struct drbd_device
*device
);
1446 extern int drbd_rs_del_all(struct drbd_device
*device
);
1447 extern void drbd_rs_failed_io(struct drbd_device
*device
,
1448 sector_t sector
, int size
);
1449 extern void drbd_advance_rs_marks(struct drbd_device
*device
, unsigned long still_to_go
);
1450 extern void __drbd_set_in_sync(struct drbd_device
*device
, sector_t sector
,
1451 int size
, const char *file
, const unsigned int line
);
1452 #define drbd_set_in_sync(device, sector, size) \
1453 __drbd_set_in_sync(device, sector, size, __FILE__, __LINE__)
1454 extern int __drbd_set_out_of_sync(struct drbd_device
*device
, sector_t sector
,
1455 int size
, const char *file
, const unsigned int line
);
1456 #define drbd_set_out_of_sync(device, sector, size) \
1457 __drbd_set_out_of_sync(device, sector, size, __FILE__, __LINE__)
1458 extern void drbd_al_shrink(struct drbd_device
*device
);
1459 extern int drbd_initialize_al(struct drbd_device
*, void *);
1462 /* state info broadcast */
1464 enum drbd_state_info_bcast_reason sib_reason
;
1468 unsigned helper_exit_code
;
1471 union drbd_state os
;
1472 union drbd_state ns
;
1476 void drbd_bcast_event(struct drbd_device
*device
, const struct sib_info
*sib
);
1479 * inline helper functions
1480 *************************/
1482 /* see also page_chain_add and friends in drbd_receiver.c */
1483 static inline struct page
*page_chain_next(struct page
*page
)
1485 return (struct page
*)page_private(page
);
1487 #define page_chain_for_each(page) \
1488 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1489 page = page_chain_next(page))
1490 #define page_chain_for_each_safe(page, n) \
1491 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1494 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request
*peer_req
)
1496 struct page
*page
= peer_req
->pages
;
1497 page_chain_for_each(page
) {
1498 if (page_count(page
) > 1)
1504 static inline enum drbd_state_rv
1505 _drbd_set_state(struct drbd_device
*device
, union drbd_state ns
,
1506 enum chg_state_flags flags
, struct completion
*done
)
1508 enum drbd_state_rv rv
;
1510 read_lock(&global_state_lock
);
1511 rv
= __drbd_set_state(device
, ns
, flags
, done
);
1512 read_unlock(&global_state_lock
);
1517 static inline union drbd_state
drbd_read_state(struct drbd_device
*device
)
1519 union drbd_state rv
;
1521 rv
.i
= device
->state
.i
;
1522 rv
.susp
= first_peer_device(device
)->connection
->susp
;
1523 rv
.susp_nod
= first_peer_device(device
)->connection
->susp_nod
;
1524 rv
.susp_fen
= first_peer_device(device
)->connection
->susp_fen
;
1529 enum drbd_force_detach_flags
{
1536 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1537 static inline void __drbd_chk_io_error_(struct drbd_device
*device
,
1538 enum drbd_force_detach_flags df
,
1541 enum drbd_io_error_p ep
;
1544 ep
= rcu_dereference(device
->ldev
->disk_conf
)->on_io_error
;
1547 case EP_PASS_ON
: /* FIXME would this be better named "Ignore"? */
1548 if (df
== DRBD_READ_ERROR
|| df
== DRBD_WRITE_ERROR
) {
1549 if (__ratelimit(&drbd_ratelimit_state
))
1550 drbd_err(device
, "Local IO failed in %s.\n", where
);
1551 if (device
->state
.disk
> D_INCONSISTENT
)
1552 _drbd_set_state(_NS(device
, disk
, D_INCONSISTENT
), CS_HARD
, NULL
);
1555 /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */
1557 case EP_CALL_HELPER
:
1558 /* Remember whether we saw a READ or WRITE error.
1560 * Recovery of the affected area for WRITE failure is covered
1561 * by the activity log.
1562 * READ errors may fall outside that area though. Certain READ
1563 * errors can be "healed" by writing good data to the affected
1564 * blocks, which triggers block re-allocation in lower layers.
1566 * If we can not write the bitmap after a READ error,
1567 * we may need to trigger a full sync (see w_go_diskless()).
1569 * Force-detach is not really an IO error, but rather a
1570 * desperate measure to try to deal with a completely
1571 * unresponsive lower level IO stack.
1572 * Still it should be treated as a WRITE error.
1574 * Meta IO error is always WRITE error:
1575 * we read meta data only once during attach,
1576 * which will fail in case of errors.
1578 set_bit(WAS_IO_ERROR
, &device
->flags
);
1579 if (df
== DRBD_READ_ERROR
)
1580 set_bit(WAS_READ_ERROR
, &device
->flags
);
1581 if (df
== DRBD_FORCE_DETACH
)
1582 set_bit(FORCE_DETACH
, &device
->flags
);
1583 if (device
->state
.disk
> D_FAILED
) {
1584 _drbd_set_state(_NS(device
, disk
, D_FAILED
), CS_HARD
, NULL
);
1586 "Local IO failed in %s. Detaching...\n", where
);
1593 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1594 * @device: DRBD device.
1595 * @error: Error code passed to the IO completion callback
1596 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1598 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1600 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1601 static inline void drbd_chk_io_error_(struct drbd_device
*device
,
1602 int error
, enum drbd_force_detach_flags forcedetach
, const char *where
)
1605 unsigned long flags
;
1606 spin_lock_irqsave(&first_peer_device(device
)->connection
->req_lock
, flags
);
1607 __drbd_chk_io_error_(device
, forcedetach
, where
);
1608 spin_unlock_irqrestore(&first_peer_device(device
)->connection
->req_lock
, flags
);
1614 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1615 * @bdev: Meta data block device.
1617 * BTW, for internal meta data, this happens to be the maximum capacity
1618 * we could agree upon with our peer node.
1620 static inline sector_t
drbd_md_first_sector(struct drbd_backing_dev
*bdev
)
1622 switch (bdev
->md
.meta_dev_idx
) {
1623 case DRBD_MD_INDEX_INTERNAL
:
1624 case DRBD_MD_INDEX_FLEX_INT
:
1625 return bdev
->md
.md_offset
+ bdev
->md
.bm_offset
;
1626 case DRBD_MD_INDEX_FLEX_EXT
:
1628 return bdev
->md
.md_offset
;
1633 * drbd_md_last_sector() - Return the last sector number of the meta data area
1634 * @bdev: Meta data block device.
1636 static inline sector_t
drbd_md_last_sector(struct drbd_backing_dev
*bdev
)
1638 switch (bdev
->md
.meta_dev_idx
) {
1639 case DRBD_MD_INDEX_INTERNAL
:
1640 case DRBD_MD_INDEX_FLEX_INT
:
1641 return bdev
->md
.md_offset
+ MD_4kB_SECT
-1;
1642 case DRBD_MD_INDEX_FLEX_EXT
:
1644 return bdev
->md
.md_offset
+ bdev
->md
.md_size_sect
-1;
1648 /* Returns the number of 512 byte sectors of the device */
1649 static inline sector_t
drbd_get_capacity(struct block_device
*bdev
)
1651 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1652 return bdev
? i_size_read(bdev
->bd_inode
) >> 9 : 0;
1656 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1657 * @bdev: Meta data block device.
1659 * returns the capacity we announce to out peer. we clip ourselves at the
1660 * various MAX_SECTORS, because if we don't, current implementation will
1661 * oops sooner or later
1663 static inline sector_t
drbd_get_max_capacity(struct drbd_backing_dev
*bdev
)
1667 switch (bdev
->md
.meta_dev_idx
) {
1668 case DRBD_MD_INDEX_INTERNAL
:
1669 case DRBD_MD_INDEX_FLEX_INT
:
1670 s
= drbd_get_capacity(bdev
->backing_bdev
)
1671 ? min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1672 drbd_md_first_sector(bdev
))
1675 case DRBD_MD_INDEX_FLEX_EXT
:
1676 s
= min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1677 drbd_get_capacity(bdev
->backing_bdev
));
1678 /* clip at maximum size the meta device can support */
1679 s
= min_t(sector_t
, s
,
1680 BM_EXT_TO_SECT(bdev
->md
.md_size_sect
1681 - bdev
->md
.bm_offset
));
1684 s
= min_t(sector_t
, DRBD_MAX_SECTORS
,
1685 drbd_get_capacity(bdev
->backing_bdev
));
1691 * drbd_md_ss() - Return the sector number of our meta data super block
1692 * @bdev: Meta data block device.
1694 static inline sector_t
drbd_md_ss(struct drbd_backing_dev
*bdev
)
1696 const int meta_dev_idx
= bdev
->md
.meta_dev_idx
;
1698 if (meta_dev_idx
== DRBD_MD_INDEX_FLEX_EXT
)
1701 /* Since drbd08, internal meta data is always "flexible".
1702 * position: last 4k aligned block of 4k size */
1703 if (meta_dev_idx
== DRBD_MD_INDEX_INTERNAL
||
1704 meta_dev_idx
== DRBD_MD_INDEX_FLEX_INT
)
1705 return (drbd_get_capacity(bdev
->backing_bdev
) & ~7ULL) - 8;
1707 /* external, some index; this is the old fixed size layout */
1708 return MD_128MB_SECT
* bdev
->md
.meta_dev_idx
;
1712 drbd_queue_work_front(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1714 unsigned long flags
;
1715 spin_lock_irqsave(&q
->q_lock
, flags
);
1716 list_add(&w
->list
, &q
->q
);
1717 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1718 wake_up(&q
->q_wait
);
1722 drbd_queue_work(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1724 unsigned long flags
;
1725 spin_lock_irqsave(&q
->q_lock
, flags
);
1726 list_add_tail(&w
->list
, &q
->q
);
1727 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1728 wake_up(&q
->q_wait
);
1731 static inline void wake_asender(struct drbd_connection
*connection
)
1733 if (test_bit(SIGNAL_ASENDER
, &connection
->flags
))
1734 force_sig(DRBD_SIG
, connection
->asender
.task
);
1737 static inline void request_ping(struct drbd_connection
*connection
)
1739 set_bit(SEND_PING
, &connection
->flags
);
1740 wake_asender(connection
);
1743 extern void *conn_prepare_command(struct drbd_connection
*, struct drbd_socket
*);
1744 extern void *drbd_prepare_command(struct drbd_device
*, struct drbd_socket
*);
1745 extern int conn_send_command(struct drbd_connection
*, struct drbd_socket
*,
1746 enum drbd_packet
, unsigned int, void *,
1748 extern int drbd_send_command(struct drbd_device
*, struct drbd_socket
*,
1749 enum drbd_packet
, unsigned int, void *,
1752 extern int drbd_send_ping(struct drbd_connection
*connection
);
1753 extern int drbd_send_ping_ack(struct drbd_connection
*connection
);
1754 extern int drbd_send_state_req(struct drbd_device
*, union drbd_state
, union drbd_state
);
1755 extern int conn_send_state_req(struct drbd_connection
*, union drbd_state
, union drbd_state
);
1757 static inline void drbd_thread_stop(struct drbd_thread
*thi
)
1759 _drbd_thread_stop(thi
, false, true);
1762 static inline void drbd_thread_stop_nowait(struct drbd_thread
*thi
)
1764 _drbd_thread_stop(thi
, false, false);
1767 static inline void drbd_thread_restart_nowait(struct drbd_thread
*thi
)
1769 _drbd_thread_stop(thi
, true, false);
1772 /* counts how many answer packets packets we expect from our peer,
1773 * for either explicit application requests,
1774 * or implicit barrier packets as necessary.
1777 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1778 * it is much easier and equally valid to count what we queue for the
1779 * worker, even before it actually was queued or send.
1780 * (drbd_make_request_common; recovery path on read io-error)
1782 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1783 * _req_mod(req, DATA_RECEIVED)
1784 * [from receive_DataReply]
1785 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1786 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1787 * for some reason it is NOT decreased in got_NegAck,
1788 * but in the resulting cleanup code from report_params.
1789 * we should try to remember the reason for that...
1790 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
1791 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
1792 * [from tl_clear_barrier]
1794 static inline void inc_ap_pending(struct drbd_device
*device
)
1796 atomic_inc(&device
->ap_pending_cnt
);
1799 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \
1800 if (atomic_read(&device->which) < 0) \
1801 drbd_err(device, "in %s:%d: " #which " = %d < 0 !\n", \
1803 atomic_read(&device->which))
1805 #define dec_ap_pending(device) _dec_ap_pending(device, __FUNCTION__, __LINE__)
1806 static inline void _dec_ap_pending(struct drbd_device
*device
, const char *func
, int line
)
1808 if (atomic_dec_and_test(&device
->ap_pending_cnt
))
1809 wake_up(&device
->misc_wait
);
1810 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt
, func
, line
);
1813 /* counts how many resync-related answers we still expect from the peer
1815 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1816 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
1817 * (or P_NEG_ACK with ID_SYNCER)
1819 static inline void inc_rs_pending(struct drbd_device
*device
)
1821 atomic_inc(&device
->rs_pending_cnt
);
1824 #define dec_rs_pending(device) _dec_rs_pending(device, __FUNCTION__, __LINE__)
1825 static inline void _dec_rs_pending(struct drbd_device
*device
, const char *func
, int line
)
1827 atomic_dec(&device
->rs_pending_cnt
);
1828 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt
, func
, line
);
1831 /* counts how many answers we still need to send to the peer.
1833 * receive_Data unless protocol A;
1834 * we need to send a P_RECV_ACK (proto B)
1835 * or P_WRITE_ACK (proto C)
1836 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1837 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1838 * receive_Barrier_* we need to send a P_BARRIER_ACK
1840 static inline void inc_unacked(struct drbd_device
*device
)
1842 atomic_inc(&device
->unacked_cnt
);
1845 #define dec_unacked(device) _dec_unacked(device, __FUNCTION__, __LINE__)
1846 static inline void _dec_unacked(struct drbd_device
*device
, const char *func
, int line
)
1848 atomic_dec(&device
->unacked_cnt
);
1849 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt
, func
, line
);
1852 #define sub_unacked(device, n) _sub_unacked(device, n, __FUNCTION__, __LINE__)
1853 static inline void _sub_unacked(struct drbd_device
*device
, int n
, const char *func
, int line
)
1855 atomic_sub(n
, &device
->unacked_cnt
);
1856 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt
, func
, line
);
1860 * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev
1863 * You have to call put_ldev() when finished working with device->ldev.
1865 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
1866 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
1868 static inline void put_ldev(struct drbd_device
*device
)
1870 int i
= atomic_dec_return(&device
->local_cnt
);
1872 /* This may be called from some endio handler,
1873 * so we must not sleep here. */
1878 if (device
->state
.disk
== D_DISKLESS
)
1879 /* even internal references gone, safe to destroy */
1880 drbd_ldev_destroy(device
);
1881 if (device
->state
.disk
== D_FAILED
) {
1882 /* all application IO references gone. */
1883 if (!test_and_set_bit(GO_DISKLESS
, &device
->flags
))
1884 drbd_queue_work(&first_peer_device(device
)->connection
->sender_work
, &device
->go_diskless
);
1886 wake_up(&device
->misc_wait
);
1891 static inline int _get_ldev_if_state(struct drbd_device
*device
, enum drbd_disk_state mins
)
1895 /* never get a reference while D_DISKLESS */
1896 if (device
->state
.disk
== D_DISKLESS
)
1899 atomic_inc(&device
->local_cnt
);
1900 io_allowed
= (device
->state
.disk
>= mins
);
1906 extern int _get_ldev_if_state(struct drbd_device
*device
, enum drbd_disk_state mins
);
1909 /* you must have an "get_ldev" reference */
1910 static inline void drbd_get_syncer_progress(struct drbd_device
*device
,
1911 unsigned long *bits_left
, unsigned int *per_mil_done
)
1913 /* this is to break it at compile time when we change that, in case we
1914 * want to support more than (1<<32) bits on a 32bit arch. */
1915 typecheck(unsigned long, device
->rs_total
);
1917 /* note: both rs_total and rs_left are in bits, i.e. in
1918 * units of BM_BLOCK_SIZE.
1919 * for the percentage, we don't care. */
1921 if (device
->state
.conn
== C_VERIFY_S
|| device
->state
.conn
== C_VERIFY_T
)
1922 *bits_left
= device
->ov_left
;
1924 *bits_left
= drbd_bm_total_weight(device
) - device
->rs_failed
;
1925 /* >> 10 to prevent overflow,
1926 * +1 to prevent division by zero */
1927 if (*bits_left
> device
->rs_total
) {
1928 /* doh. maybe a logic bug somewhere.
1929 * may also be just a race condition
1930 * between this and a disconnect during sync.
1931 * for now, just prevent in-kernel buffer overflow.
1934 drbd_warn(device
, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
1935 drbd_conn_str(device
->state
.conn
),
1936 *bits_left
, device
->rs_total
, device
->rs_failed
);
1939 /* Make sure the division happens in long context.
1940 * We allow up to one petabyte storage right now,
1941 * at a granularity of 4k per bit that is 2**38 bits.
1942 * After shift right and multiplication by 1000,
1943 * this should still fit easily into a 32bit long,
1944 * so we don't need a 64bit division on 32bit arch.
1945 * Note: currently we don't support such large bitmaps on 32bit
1946 * arch anyways, but no harm done to be prepared for it here.
1948 unsigned int shift
= device
->rs_total
> UINT_MAX
? 16 : 10;
1949 unsigned long left
= *bits_left
>> shift
;
1950 unsigned long total
= 1UL + (device
->rs_total
>> shift
);
1951 unsigned long tmp
= 1000UL - left
* 1000UL/total
;
1952 *per_mil_done
= tmp
;
1957 /* this throttles on-the-fly application requests
1958 * according to max_buffers settings;
1959 * maybe re-implement using semaphores? */
1960 static inline int drbd_get_max_buffers(struct drbd_device
*device
)
1962 struct net_conf
*nc
;
1966 nc
= rcu_dereference(first_peer_device(device
)->connection
->net_conf
);
1967 mxb
= nc
? nc
->max_buffers
: 1000000; /* arbitrary limit on open requests */
1973 static inline int drbd_state_is_stable(struct drbd_device
*device
)
1975 union drbd_dev_state s
= device
->state
;
1977 /* DO NOT add a default clause, we want the compiler to warn us
1978 * for any newly introduced state we may have forgotten to add here */
1980 switch ((enum drbd_conns
)s
.conn
) {
1981 /* new io only accepted when there is no connection, ... */
1983 case C_WF_CONNECTION
:
1984 /* ... or there is a well established connection. */
1990 case C_PAUSED_SYNC_S
:
1991 case C_PAUSED_SYNC_T
:
1994 /* transitional states, IO allowed */
1995 case C_DISCONNECTING
:
1999 case C_NETWORK_FAILURE
:
2000 case C_PROTOCOL_ERROR
:
2002 case C_WF_REPORT_PARAMS
:
2003 case C_STARTING_SYNC_S
:
2004 case C_STARTING_SYNC_T
:
2007 /* Allow IO in BM exchange states with new protocols */
2009 if (first_peer_device(device
)->connection
->agreed_pro_version
< 96)
2013 /* no new io accepted in these states */
2015 case C_WF_SYNC_UUID
:
2021 switch ((enum drbd_disk_state
)s
.disk
) {
2023 case D_INCONSISTENT
:
2028 /* disk state is stable as well. */
2031 /* no new io accepted during transitional states */
2043 static inline int drbd_suspended(struct drbd_device
*device
)
2045 struct drbd_connection
*connection
= first_peer_device(device
)->connection
;
2047 return connection
->susp
|| connection
->susp_fen
|| connection
->susp_nod
;
2050 static inline bool may_inc_ap_bio(struct drbd_device
*device
)
2052 int mxb
= drbd_get_max_buffers(device
);
2054 if (drbd_suspended(device
))
2056 if (test_bit(SUSPEND_IO
, &device
->flags
))
2059 /* to avoid potential deadlock or bitmap corruption,
2060 * in various places, we only allow new application io
2061 * to start during "stable" states. */
2063 /* no new io accepted when attaching or detaching the disk */
2064 if (!drbd_state_is_stable(device
))
2067 /* since some older kernels don't have atomic_add_unless,
2068 * and we are within the spinlock anyways, we have this workaround. */
2069 if (atomic_read(&device
->ap_bio_cnt
) > mxb
)
2071 if (test_bit(BITMAP_IO
, &device
->flags
))
2076 static inline bool inc_ap_bio_cond(struct drbd_device
*device
)
2080 spin_lock_irq(&first_peer_device(device
)->connection
->req_lock
);
2081 rv
= may_inc_ap_bio(device
);
2083 atomic_inc(&device
->ap_bio_cnt
);
2084 spin_unlock_irq(&first_peer_device(device
)->connection
->req_lock
);
2089 static inline void inc_ap_bio(struct drbd_device
*device
)
2092 * as long as the device is suspended
2093 * until the bitmap is no longer on the fly during connection
2094 * handshake as long as we would exceed the max_buffer limit.
2096 * to avoid races with the reconnect code,
2097 * we need to atomic_inc within the spinlock. */
2099 wait_event(device
->misc_wait
, inc_ap_bio_cond(device
));
2102 static inline void dec_ap_bio(struct drbd_device
*device
)
2104 int mxb
= drbd_get_max_buffers(device
);
2105 int ap_bio
= atomic_dec_return(&device
->ap_bio_cnt
);
2107 D_ASSERT(ap_bio
>= 0);
2109 if (ap_bio
== 0 && test_bit(BITMAP_IO
, &device
->flags
)) {
2110 if (!test_and_set_bit(BITMAP_IO_QUEUED
, &device
->flags
))
2111 drbd_queue_work(&first_peer_device(device
)->connection
->sender_work
, &device
->bm_io_work
.w
);
2114 /* this currently does wake_up for every dec_ap_bio!
2115 * maybe rather introduce some type of hysteresis?
2116 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2118 wake_up(&device
->misc_wait
);
2121 static inline bool verify_can_do_stop_sector(struct drbd_device
*device
)
2123 return first_peer_device(device
)->connection
->agreed_pro_version
>= 97 &&
2124 first_peer_device(device
)->connection
->agreed_pro_version
!= 100;
2127 static inline int drbd_set_ed_uuid(struct drbd_device
*device
, u64 val
)
2129 int changed
= device
->ed_uuid
!= val
;
2130 device
->ed_uuid
= val
;
2134 static inline int drbd_queue_order_type(struct drbd_device
*device
)
2136 /* sorry, we currently have no working implementation
2137 * of distributed TCQ stuff */
2138 #ifndef QUEUE_ORDERED_NONE
2139 #define QUEUE_ORDERED_NONE 0
2141 return QUEUE_ORDERED_NONE
;
2144 static inline void drbd_md_flush(struct drbd_device
*device
)
2148 if (device
->ldev
== NULL
) {
2149 drbd_warn(device
, "device->ldev == NULL in drbd_md_flush\n");
2153 if (test_bit(MD_NO_FUA
, &device
->flags
))
2156 r
= blkdev_issue_flush(device
->ldev
->md_bdev
, GFP_NOIO
, NULL
);
2158 set_bit(MD_NO_FUA
, &device
->flags
);
2159 drbd_err(device
, "meta data flush failed with status %d, disabling md-flushes\n", r
);
2163 static inline struct drbd_connection
*first_connection(struct drbd_resource
*resource
)
2165 return list_first_entry(&resource
->connections
,
2166 struct drbd_connection
, connections
);