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 D_ASSERT(device, exp) do { \
152 drbd_err(device, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__); \
156 * expect - Make an assertion
158 * Unlike the assert macro, this macro returns a boolean result.
160 #define expect(exp) ({ \
161 bool _bool = (exp); \
163 drbd_err(device, "ASSERTION %s FAILED in %s\n", \
168 /* Defines to control fault insertion */
170 DRBD_FAULT_MD_WR
= 0, /* meta data write */
171 DRBD_FAULT_MD_RD
= 1, /* read */
172 DRBD_FAULT_RS_WR
= 2, /* resync */
173 DRBD_FAULT_RS_RD
= 3,
174 DRBD_FAULT_DT_WR
= 4, /* data */
175 DRBD_FAULT_DT_RD
= 5,
176 DRBD_FAULT_DT_RA
= 6, /* data read ahead */
177 DRBD_FAULT_BM_ALLOC
= 7, /* bitmap allocation */
178 DRBD_FAULT_AL_EE
= 8, /* alloc ee */
179 DRBD_FAULT_RECEIVE
= 9, /* Changes some bytes upon receiving a [rs]data block */
185 _drbd_insert_fault(struct drbd_device
*device
, unsigned int type
);
188 drbd_insert_fault(struct drbd_device
*device
, unsigned int type
) {
189 #ifdef CONFIG_DRBD_FAULT_INJECTION
191 (enable_faults
& (1<<type
)) &&
192 _drbd_insert_fault(device
, type
);
198 /* integer division, round _UP_ to the next integer */
199 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
200 /* usual integer division */
201 #define div_floor(A, B) ((A)/(B))
203 extern struct ratelimit_state drbd_ratelimit_state
;
204 extern struct idr drbd_devices
; /* RCU, updates: genl_lock() */
205 extern struct list_head drbd_resources
; /* RCU, updates: genl_lock() */
207 extern const char *cmdname(enum drbd_packet cmd
);
209 /* for sending/receiving the bitmap,
210 * possibly in some encoding scheme */
213 * stores total bits and long words
214 * of the bitmap, so we don't need to
215 * call the accessor functions over and again. */
216 unsigned long bm_bits
;
217 unsigned long bm_words
;
218 /* during xfer, current position within the bitmap */
219 unsigned long bit_offset
;
220 unsigned long word_offset
;
222 /* statistics; index: (h->command == P_BITMAP) */
227 extern void INFO_bm_xfer_stats(struct drbd_device
*device
,
228 const char *direction
, struct bm_xfer_ctx
*c
);
230 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx
*c
)
232 /* word_offset counts "native long words" (32 or 64 bit),
234 * Encoded packet may end at an unaligned bit offset.
235 * In case a fallback clear text packet is transmitted in
236 * between, we adjust this offset back to the last 64bit
237 * aligned "native long word", which makes coding and decoding
238 * the plain text bitmap much more convenient. */
239 #if BITS_PER_LONG == 64
240 c
->word_offset
= c
->bit_offset
>> 6;
241 #elif BITS_PER_LONG == 32
242 c
->word_offset
= c
->bit_offset
>> 5;
243 c
->word_offset
&= ~(1UL);
245 # error "unsupported BITS_PER_LONG"
249 extern unsigned int drbd_header_size(struct drbd_connection
*connection
);
251 /**********************************************************************/
252 enum drbd_thread_state
{
261 struct task_struct
*task
;
262 struct completion stop
;
263 enum drbd_thread_state t_state
;
264 int (*function
) (struct drbd_thread
*);
265 struct drbd_connection
*connection
;
270 static inline enum drbd_thread_state
get_t_state(struct drbd_thread
*thi
)
272 /* THINK testing the t_state seems to be uncritical in all cases
273 * (but thread_{start,stop}), so we can read it *without* the lock.
281 struct list_head list
;
282 int (*cb
)(struct drbd_work
*, int cancel
);
284 struct drbd_device
*device
;
285 struct drbd_connection
*connection
;
289 #include "drbd_interval.h"
291 extern int drbd_wait_misc(struct drbd_device
*, struct drbd_interval
*);
293 struct drbd_request
{
296 /* if local IO is not allowed, will be NULL.
297 * if local IO _is_ allowed, holds the locally submitted bio clone,
298 * or, after local IO completion, the ERR_PTR(error).
299 * see drbd_request_endio(). */
300 struct bio
*private_bio
;
302 struct drbd_interval i
;
304 /* epoch: used to check on "completion" whether this req was in
305 * the current epoch, and we therefore have to close it,
306 * causing a p_barrier packet to be send, starting a new epoch.
308 * This corresponds to "barrier" in struct p_barrier[_ack],
309 * and to "barrier_nr" in struct drbd_epoch (and various
310 * comments/function parameters/local variable names).
314 struct list_head tl_requests
; /* ring list in the transfer log */
315 struct bio
*master_bio
; /* master bio pointer */
316 unsigned long start_time
;
318 /* once it hits 0, we may complete the master_bio */
319 atomic_t completion_ref
;
320 /* once it hits 0, we may destroy this drbd_request object */
323 unsigned rq_state
; /* see comments above _req_mod() */
327 struct drbd_connection
*connection
;
328 struct list_head list
;
329 unsigned int barrier_nr
;
330 atomic_t epoch_size
; /* increased on every request added. */
331 atomic_t active
; /* increased on every req. added, and dec on every finished. */
335 /* Prototype declaration of function defined in drbd_receiver.c */
336 int drbdd_init(struct drbd_thread
*);
337 int drbd_asender(struct drbd_thread
*);
339 /* drbd_epoch flag bits */
341 DE_HAVE_BARRIER_NUMBER
,
348 EV_CLEANUP
= 32, /* used as flag */
351 struct drbd_wq_barrier
{
353 struct completion done
;
361 struct drbd_peer_request
{
363 struct drbd_epoch
*epoch
; /* for writes */
365 atomic_t pending_bios
;
366 struct drbd_interval i
;
367 /* see comments on ee flag bits below */
371 struct digest_info
*digest
;
376 * While corresponding bios are in flight, the only modification will be
377 * set_bit WAS_ERROR, which has to be atomic.
378 * If no bios are in flight yet, or all have been completed,
379 * non-atomic modification to ee->flags is ok.
382 __EE_CALL_AL_COMPLETE_IO
,
383 __EE_MAY_SET_IN_SYNC
,
385 /* In case a barrier failed,
386 * we need to resubmit without the barrier flag. */
389 /* we may have several bios per peer request.
390 * if any of those fail, we set this flag atomically
391 * from the endio callback */
394 /* This ee has a pointer to a digest instead of a block id */
397 /* Conflicting local requests need to be restarted after this request */
398 __EE_RESTART_REQUESTS
,
400 /* The peer wants a write ACK for this (wire proto C) */
403 /* Is set when net_conf had two_primaries set while creating this peer_req */
404 __EE_IN_INTERVAL_TREE
,
406 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
407 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
408 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
409 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
410 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
411 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
412 #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK)
413 #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE)
415 /* flag bits per device */
417 UNPLUG_REMOTE
, /* sending a "UnplugRemote" could help */
418 MD_DIRTY
, /* current uuids and flags not yet on disk */
419 USE_DEGR_WFC_T
, /* degr-wfc-timeout instead of wfc-timeout. */
422 CRASHED_PRIMARY
, /* This node was a crashed primary.
423 * Gets cleared when the state.conn
424 * goes into C_CONNECTED state. */
427 MD_NO_FUA
, /* Users wants us to not use FUA/FLUSH on meta data dev */
428 SUSPEND_IO
, /* suspend application io */
429 BITMAP_IO
, /* suspend application io;
430 once no more io in flight, start bitmap io */
431 BITMAP_IO_QUEUED
, /* Started bitmap IO */
432 GO_DISKLESS
, /* Disk is being detached, on io-error or admin request. */
433 WAS_IO_ERROR
, /* Local disk failed, returned IO error */
434 WAS_READ_ERROR
, /* Local disk READ failed (set additionally to the above) */
435 FORCE_DETACH
, /* Force-detach from local disk, aborting any pending local IO */
436 RESYNC_AFTER_NEG
, /* Resync after online grow after the attach&negotiate finished. */
437 RESIZE_PENDING
, /* Size change detected locally, waiting for the response from
438 * the peer, if it changed there as well. */
439 NEW_CUR_UUID
, /* Create new current UUID when thawing IO */
440 AL_SUSPENDED
, /* Activity logging is currently suspended. */
441 AHEAD_TO_SYNC_SOURCE
, /* Ahead -> SyncSource queued */
442 B_RS_H_DONE
, /* Before resync handler done (already executed) */
443 DISCARD_MY_DATA
, /* discard_my_data flag per volume */
447 struct drbd_bitmap
; /* opaque for drbd_device */
449 /* definition of bits in bm_flags to be used in drbd_bm_lock
450 * and drbd_bitmap_io and friends. */
452 /* do we need to kfree, or vfree bm_pages? */
453 BM_P_VMALLOCED
= 0x10000, /* internal use only, will be masked out */
455 /* currently locked for bulk operation */
456 BM_LOCKED_MASK
= 0xf,
458 /* in detail, that is: */
463 /* so we can mark it locked for bulk operation,
464 * and still allow all non-bulk operations */
467 /* (test bit, count bit) allowed (common case) */
468 BM_LOCKED_TEST_ALLOWED
= BM_DONT_CLEAR
| BM_DONT_SET
| BM_IS_LOCKED
,
470 /* testing bits, as well as setting new bits allowed, but clearing bits
471 * would be unexpected. Used during bitmap receive. Setting new bits
472 * requires sending of "out-of-sync" information, though. */
473 BM_LOCKED_SET_ALLOWED
= BM_DONT_CLEAR
| BM_IS_LOCKED
,
475 /* for drbd_bm_write_copy_pages, everything is allowed,
476 * only concurrent bulk operations are locked out. */
477 BM_LOCKED_CHANGE_ALLOWED
= BM_IS_LOCKED
,
480 struct drbd_work_queue
{
482 spinlock_t q_lock
; /* to protect the list. */
483 wait_queue_head_t q_wait
;
488 struct socket
*socket
;
489 /* this way we get our
490 * send/receive buffers off the stack */
496 u64 md_offset
; /* sector offset to 'super' block */
498 u64 la_size_sect
; /* last agreed size, unit sectors */
499 spinlock_t uuid_lock
;
505 s32 al_offset
; /* signed relative sector offset to activity log */
506 s32 bm_offset
; /* signed relative sector offset to bitmap */
508 /* cached value of bdev->disk_conf->meta_dev_idx (see below) */
511 /* see al_tr_number_to_on_disk_sector() */
513 u32 al_stripe_size_4k
;
514 u32 al_size_4k
; /* cached product of the above */
517 struct drbd_backing_dev
{
518 struct block_device
*backing_bdev
;
519 struct block_device
*md_bdev
;
521 struct disk_conf
*disk_conf
; /* RCU, for updates: first_peer_device(device)->connection->conf_update */
522 sector_t known_size
; /* last known size of that backing device */
534 int (*io_fn
)(struct drbd_device
*device
);
535 void (*done
)(struct drbd_device
*device
, int rv
);
538 enum write_ordering_e
{
545 unsigned int head_index
;
547 int total
; /* sum of all values */
550 extern struct fifo_buffer
*fifo_alloc(int fifo_size
);
552 /* flag bits per connection */
554 NET_CONGESTED
, /* The data socket is congested */
555 RESOLVE_CONFLICTS
, /* Set on one node, cleared on the peer! */
556 SEND_PING
, /* whether asender should send a ping asap */
557 SIGNAL_ASENDER
, /* whether asender wants to be interrupted */
558 GOT_PING_ACK
, /* set when we receive a ping_ack packet, ping_wait gets woken */
559 CONN_WD_ST_CHG_REQ
, /* A cluster wide state change on the connection is active */
562 CONN_DRY_RUN
, /* Expect disconnect after resync handshake. */
563 CREATE_BARRIER
, /* next P_DATA is preceded by a P_BARRIER */
564 STATE_SENT
, /* Do not change state/UUIDs while this is set */
565 CALLBACK_PENDING
, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
566 * pending, from drbd worker context.
567 * If set, bdi_write_congested() returns true,
568 * so shrink_page_list() would not recurse into,
569 * and potentially deadlock on, this drbd worker.
574 struct drbd_resource
{
577 struct idr devices
; /* volume number to device mapping */
578 struct list_head connections
;
579 struct list_head resources
;
580 struct res_opts res_opts
;
583 struct drbd_connection
{
584 struct list_head connections
;
585 struct drbd_resource
*resource
;
587 struct idr peer_devices
; /* volume number to peer device mapping */
588 enum drbd_conns cstate
; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
589 unsigned susp
:1; /* IO suspended by user */
590 unsigned susp_nod
:1; /* IO suspended because no data */
591 unsigned susp_fen
:1; /* IO suspended because fence peer handler runs */
592 struct mutex cstate_mutex
; /* Protects graceful disconnects */
593 unsigned int connect_cnt
; /* Inc each time a connection is established */
596 struct net_conf
*net_conf
; /* content protected by rcu */
597 struct mutex conf_update
; /* mutex for ready-copy-update of net_conf and disk_conf */
598 wait_queue_head_t ping_wait
; /* Woken upon reception of a ping, and a state change */
600 struct sockaddr_storage my_addr
;
602 struct sockaddr_storage peer_addr
;
605 struct drbd_socket data
; /* data/barrier/cstate/parameter packets */
606 struct drbd_socket meta
; /* ping/ack (metadata) packets */
607 int agreed_pro_version
; /* actually used protocol version */
608 unsigned long last_received
; /* in jiffies, either socket */
609 unsigned int ko_count
;
613 struct list_head transfer_log
; /* all requests not yet fully processed */
615 struct crypto_hash
*cram_hmac_tfm
;
616 struct crypto_hash
*integrity_tfm
; /* checksums we compute, updates protected by connection->data->mutex */
617 struct crypto_hash
*peer_integrity_tfm
; /* checksums we verify, only accessed from receiver thread */
618 struct crypto_hash
*csums_tfm
;
619 struct crypto_hash
*verify_tfm
;
624 struct drbd_epoch
*current_epoch
;
625 spinlock_t epoch_lock
;
627 enum write_ordering_e write_ordering
;
628 atomic_t current_tle_nr
; /* transfer log epoch number */
629 unsigned current_tle_writes
; /* writes seen within this tl epoch */
631 unsigned long last_reconnect_jif
;
632 struct drbd_thread receiver
;
633 struct drbd_thread worker
;
634 struct drbd_thread asender
;
635 cpumask_var_t cpu_mask
;
638 struct drbd_work_queue sender_work
;
641 /* whether this sender thread
642 * has processed a single write yet. */
643 bool seen_any_write_yet
;
645 /* Which barrier number to send with the next P_BARRIER */
646 int current_epoch_nr
;
648 /* how many write requests have been sent
649 * with req->epoch == current_epoch_nr.
650 * If none, no P_BARRIER will be sent. */
651 unsigned current_epoch_writes
;
655 struct submit_worker
{
656 struct workqueue_struct
*wq
;
657 struct work_struct worker
;
660 struct list_head writes
;
663 struct drbd_peer_device
{
664 struct list_head peer_devices
;
665 struct drbd_device
*device
;
666 struct drbd_connection
*connection
;
670 struct drbd_resource
*resource
;
671 struct list_head peer_devices
;
672 int vnr
; /* volume number within the connection */
675 /* things that are stored as / read from meta data on disk */
678 /* configured by drbdsetup */
679 struct drbd_backing_dev
*ldev
__protected_by(local
);
681 sector_t p_size
; /* partner's disk size */
682 struct request_queue
*rq_queue
;
683 struct block_device
*this_bdev
;
684 struct gendisk
*vdisk
;
686 unsigned long last_reattach_jif
;
687 struct drbd_work resync_work
,
692 struct timer_list resync_timer
;
693 struct timer_list md_sync_timer
;
694 struct timer_list start_resync_timer
;
695 struct timer_list request_timer
;
696 #ifdef DRBD_DEBUG_MD_SYNC
700 } last_md_mark_dirty
;
703 /* Used after attach while negotiating new disk state. */
704 union drbd_state new_state_tmp
;
706 union drbd_dev_state state
;
707 wait_queue_head_t misc_wait
;
708 wait_queue_head_t state_wait
; /* upon each state change. */
709 unsigned int send_cnt
;
710 unsigned int recv_cnt
;
711 unsigned int read_cnt
;
712 unsigned int writ_cnt
;
713 unsigned int al_writ_cnt
;
714 unsigned int bm_writ_cnt
;
715 atomic_t ap_bio_cnt
; /* Requests we need to complete */
716 atomic_t ap_pending_cnt
; /* AP data packets on the wire, ack expected */
717 atomic_t rs_pending_cnt
; /* RS request/data packets on the wire */
718 atomic_t unacked_cnt
; /* Need to send replies for */
719 atomic_t local_cnt
; /* Waiting for local completion */
721 /* Interval tree of pending local requests */
722 struct rb_root read_requests
;
723 struct rb_root write_requests
;
725 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
726 unsigned long rs_total
;
727 /* number of resync blocks that failed in this run */
728 unsigned long rs_failed
;
729 /* Syncer's start time [unit jiffies] */
730 unsigned long rs_start
;
731 /* cumulated time in PausedSyncX state [unit jiffies] */
732 unsigned long rs_paused
;
733 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
734 unsigned long rs_same_csum
;
735 #define DRBD_SYNC_MARKS 8
736 #define DRBD_SYNC_MARK_STEP (3*HZ)
737 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
738 unsigned long rs_mark_left
[DRBD_SYNC_MARKS
];
739 /* marks's time [unit jiffies] */
740 unsigned long rs_mark_time
[DRBD_SYNC_MARKS
];
741 /* current index into rs_mark_{left,time} */
743 unsigned long rs_last_bcast
; /* [unit jiffies] */
745 /* where does the admin want us to start? (sector) */
746 sector_t ov_start_sector
;
747 sector_t ov_stop_sector
;
748 /* where are we now? (sector) */
749 sector_t ov_position
;
750 /* Start sector of out of sync range (to merge printk reporting). */
751 sector_t ov_last_oos_start
;
752 /* size of out-of-sync range in sectors. */
753 sector_t ov_last_oos_size
;
754 unsigned long ov_left
; /* in bits */
756 struct drbd_bitmap
*bitmap
;
757 unsigned long bm_resync_fo
; /* bit offset for drbd_bm_find_next */
759 /* Used to track operations of resync... */
760 struct lru_cache
*resync
;
761 /* Number of locked elements in resync LRU */
762 unsigned int resync_locked
;
763 /* resync extent number waiting for application requests */
764 unsigned int resync_wenr
;
769 struct list_head active_ee
; /* IO in progress (P_DATA gets written to disk) */
770 struct list_head sync_ee
; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
771 struct list_head done_ee
; /* need to send P_WRITE_ACK */
772 struct list_head read_ee
; /* [RS]P_DATA_REQUEST being read */
773 struct list_head net_ee
; /* zero-copy network send in progress */
776 struct list_head resync_reads
;
777 atomic_t pp_in_use
; /* allocated from page pool */
778 atomic_t pp_in_use_by_net
; /* sendpage()d, still referenced by tcp */
779 wait_queue_head_t ee_wait
;
780 struct page
*md_io_page
; /* one page buffer for md_io */
781 struct drbd_md_io md_io
;
782 atomic_t md_io_in_use
; /* protects the md_io, md_io_page and md_io_tmpp */
784 wait_queue_head_t al_wait
;
785 struct lru_cache
*act_log
; /* activity log */
786 unsigned int al_tr_number
;
788 wait_queue_head_t seq_wait
;
790 unsigned int peer_seq
;
791 spinlock_t peer_seq_lock
;
793 unsigned long comm_bm_set
; /* communicated number of set bits. */
794 struct bm_io_work bm_io_work
;
795 u64 ed_uuid
; /* UUID of the exposed data */
796 struct mutex own_state_mutex
;
797 struct mutex
*state_mutex
; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */
798 char congestion_reason
; /* Why we where congested... */
799 atomic_t rs_sect_in
; /* for incoming resync data rate, SyncTarget */
800 atomic_t rs_sect_ev
; /* for submitted resync data rate, both */
801 int rs_last_sect_ev
; /* counter to compare with */
802 int rs_last_events
; /* counter of read or write "events" (unit sectors)
803 * on the lower level device when we last looked. */
804 int c_sync_rate
; /* current resync rate after syncer throttle magic */
805 struct fifo_buffer
*rs_plan_s
; /* correction values of resync planer (RCU, connection->conn_update) */
806 int rs_in_flight
; /* resync sectors in flight (to proxy, in proxy and from proxy) */
807 atomic_t ap_in_flight
; /* App sectors in flight (waiting for ack) */
808 unsigned int peer_max_bio_size
;
809 unsigned int local_max_bio_size
;
811 /* any requests that would block in drbd_make_request()
812 * are deferred to this single-threaded work queue */
813 struct submit_worker submit
;
816 static inline struct drbd_device
*minor_to_device(unsigned int minor
)
818 return (struct drbd_device
*)idr_find(&drbd_devices
, minor
);
821 static inline struct drbd_peer_device
*first_peer_device(struct drbd_device
*device
)
823 return list_first_entry(&device
->peer_devices
, struct drbd_peer_device
, peer_devices
);
826 #define for_each_resource(resource, _resources) \
827 list_for_each_entry(resource, _resources, resources)
829 #define for_each_resource_rcu(resource, _resources) \
830 list_for_each_entry_rcu(resource, _resources, resources)
832 #define for_each_resource_safe(resource, tmp, _resources) \
833 list_for_each_entry_safe(resource, tmp, _resources, resources)
835 #define for_each_connection(connection, resource) \
836 list_for_each_entry(connection, &resource->connections, connections)
838 #define for_each_connection_rcu(connection, resource) \
839 list_for_each_entry_rcu(connection, &resource->connections, connections)
841 #define for_each_connection_safe(connection, tmp, resource) \
842 list_for_each_entry_safe(connection, tmp, &resource->connections, connections)
844 #define for_each_peer_device(peer_device, device) \
845 list_for_each_entry(peer_device, &device->peer_devices, peer_devices)
847 #define for_each_peer_device_rcu(peer_device, device) \
848 list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices)
850 #define for_each_peer_device_safe(peer_device, tmp, device) \
851 list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices)
853 static inline unsigned int device_to_minor(struct drbd_device
*device
)
855 return device
->minor
;
858 static inline struct drbd_device
*vnr_to_device(struct drbd_connection
*connection
, int vnr
)
860 struct drbd_peer_device
*peer_device
;
862 peer_device
= idr_find(&connection
->peer_devices
, vnr
);
863 return peer_device
? peer_device
->device
: NULL
;
867 * function declarations
868 *************************/
874 DDSF_NO_RESYNC
= 2, /* Do not run a resync for the new space */
877 extern void drbd_init_set_defaults(struct drbd_device
*device
);
878 extern int drbd_thread_start(struct drbd_thread
*thi
);
879 extern void _drbd_thread_stop(struct drbd_thread
*thi
, int restart
, int wait
);
880 extern char *drbd_task_to_thread_name(struct drbd_connection
*connection
, struct task_struct
*task
);
882 extern void drbd_thread_current_set_cpu(struct drbd_thread
*thi
);
883 extern void drbd_calc_cpu_mask(struct drbd_connection
*connection
);
885 #define drbd_thread_current_set_cpu(A) ({})
886 #define drbd_calc_cpu_mask(A) ({})
888 extern void tl_release(struct drbd_connection
*, unsigned int barrier_nr
,
889 unsigned int set_size
);
890 extern void tl_clear(struct drbd_connection
*);
891 extern void drbd_free_sock(struct drbd_connection
*connection
);
892 extern int drbd_send(struct drbd_connection
*connection
, struct socket
*sock
,
893 void *buf
, size_t size
, unsigned msg_flags
);
894 extern int drbd_send_all(struct drbd_connection
*, struct socket
*, void *, size_t,
897 extern int __drbd_send_protocol(struct drbd_connection
*connection
, enum drbd_packet cmd
);
898 extern int drbd_send_protocol(struct drbd_connection
*connection
);
899 extern int drbd_send_uuids(struct drbd_device
*device
);
900 extern int drbd_send_uuids_skip_initial_sync(struct drbd_device
*device
);
901 extern void drbd_gen_and_send_sync_uuid(struct drbd_device
*device
);
902 extern int drbd_send_sizes(struct drbd_device
*device
, int trigger_reply
, enum dds_flags flags
);
903 extern int drbd_send_state(struct drbd_device
*device
, union drbd_state s
);
904 extern int drbd_send_current_state(struct drbd_device
*device
);
905 extern int drbd_send_sync_param(struct drbd_device
*device
);
906 extern void drbd_send_b_ack(struct drbd_connection
*connection
, u32 barrier_nr
,
908 extern int drbd_send_ack(struct drbd_device
*, enum drbd_packet
,
909 struct drbd_peer_request
*);
910 extern void drbd_send_ack_rp(struct drbd_device
*device
, enum drbd_packet cmd
,
911 struct p_block_req
*rp
);
912 extern void drbd_send_ack_dp(struct drbd_device
*device
, enum drbd_packet cmd
,
913 struct p_data
*dp
, int data_size
);
914 extern int drbd_send_ack_ex(struct drbd_device
*device
, enum drbd_packet cmd
,
915 sector_t sector
, int blksize
, u64 block_id
);
916 extern int drbd_send_out_of_sync(struct drbd_device
*, struct drbd_request
*);
917 extern int drbd_send_block(struct drbd_device
*, enum drbd_packet
,
918 struct drbd_peer_request
*);
919 extern int drbd_send_dblock(struct drbd_device
*device
, struct drbd_request
*req
);
920 extern int drbd_send_drequest(struct drbd_device
*device
, int cmd
,
921 sector_t sector
, int size
, u64 block_id
);
922 extern int drbd_send_drequest_csum(struct drbd_device
*device
, sector_t sector
,
923 int size
, void *digest
, int digest_size
,
924 enum drbd_packet cmd
);
925 extern int drbd_send_ov_request(struct drbd_device
*device
, sector_t sector
, int size
);
927 extern int drbd_send_bitmap(struct drbd_device
*device
);
928 extern void drbd_send_sr_reply(struct drbd_device
*device
, enum drbd_state_rv retcode
);
929 extern void conn_send_sr_reply(struct drbd_connection
*connection
, enum drbd_state_rv retcode
);
930 extern void drbd_free_bc(struct drbd_backing_dev
*ldev
);
931 extern void drbd_device_cleanup(struct drbd_device
*device
);
932 void drbd_print_uuids(struct drbd_device
*device
, const char *text
);
934 extern void conn_md_sync(struct drbd_connection
*connection
);
935 extern void drbd_md_write(struct drbd_device
*device
, void *buffer
);
936 extern void drbd_md_sync(struct drbd_device
*device
);
937 extern int drbd_md_read(struct drbd_device
*device
, struct drbd_backing_dev
*bdev
);
938 extern void drbd_uuid_set(struct drbd_device
*device
, int idx
, u64 val
) __must_hold(local
);
939 extern void _drbd_uuid_set(struct drbd_device
*device
, int idx
, u64 val
) __must_hold(local
);
940 extern void drbd_uuid_new_current(struct drbd_device
*device
) __must_hold(local
);
941 extern void drbd_uuid_set_bm(struct drbd_device
*device
, u64 val
) __must_hold(local
);
942 extern void drbd_uuid_move_history(struct drbd_device
*device
) __must_hold(local
);
943 extern void __drbd_uuid_set(struct drbd_device
*device
, int idx
, u64 val
) __must_hold(local
);
944 extern void drbd_md_set_flag(struct drbd_device
*device
, int flags
) __must_hold(local
);
945 extern void drbd_md_clear_flag(struct drbd_device
*device
, int flags
)__must_hold(local
);
946 extern int drbd_md_test_flag(struct drbd_backing_dev
*, int);
947 #ifndef DRBD_DEBUG_MD_SYNC
948 extern void drbd_md_mark_dirty(struct drbd_device
*device
);
950 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
951 extern void drbd_md_mark_dirty_(struct drbd_device
*device
,
952 unsigned int line
, const char *func
);
954 extern void drbd_queue_bitmap_io(struct drbd_device
*device
,
955 int (*io_fn
)(struct drbd_device
*),
956 void (*done
)(struct drbd_device
*, int),
957 char *why
, enum bm_flag flags
);
958 extern int drbd_bitmap_io(struct drbd_device
*device
,
959 int (*io_fn
)(struct drbd_device
*),
960 char *why
, enum bm_flag flags
);
961 extern int drbd_bitmap_io_from_worker(struct drbd_device
*device
,
962 int (*io_fn
)(struct drbd_device
*),
963 char *why
, enum bm_flag flags
);
964 extern int drbd_bmio_set_n_write(struct drbd_device
*device
);
965 extern int drbd_bmio_clear_n_write(struct drbd_device
*device
);
966 extern void drbd_ldev_destroy(struct drbd_device
*device
);
970 * We currently have two possible layouts.
971 * Offsets in (512 byte) sectors.
973 * |----------- md_size_sect ------------------|
974 * [ 4k superblock ][ activity log ][ Bitmap ]
976 * | bm_offset = al_offset + X |
977 * ==> bitmap sectors = md_size_sect - bm_offset
980 * old, indexed fixed size meta data:
983 * |----------- md_size_sect ------------------|
984 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*]
986 * | bm_offset = al_offset - Y |
987 * ==> bitmap sectors = Y = al_offset - bm_offset
989 * [padding*] are zero or up to 7 unused 512 Byte sectors to the
990 * end of the device, so that the [4k superblock] will be 4k aligned.
992 * The activity log consists of 4k transaction blocks,
993 * which are written in a ring-buffer, or striped ring-buffer like fashion,
994 * which are writtensize used to be fixed 32kB,
995 * but is about to become configurable.
998 /* Our old fixed size meta data layout
999 * allows up to about 3.8TB, so if you want more,
1000 * you need to use the "flexible" meta data format. */
1001 #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */
1002 #define MD_4kB_SECT 8
1003 #define MD_32kB_SECT 64
1005 /* One activity log extent represents 4M of storage */
1006 #define AL_EXTENT_SHIFT 22
1007 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1009 /* We could make these currently hardcoded constants configurable
1010 * variables at create-md time (or even re-configurable at runtime?).
1011 * Which will require some more changes to the DRBD "super block"
1014 * updates per transaction:
1015 * This many changes to the active set can be logged with one transaction.
1016 * This number is arbitrary.
1017 * context per transaction:
1018 * This many context extent numbers are logged with each transaction.
1019 * This number is resulting from the transaction block size (4k), the layout
1020 * of the transaction header, and the number of updates per transaction.
1021 * See drbd_actlog.c:struct al_transaction_on_disk
1023 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1024 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1026 #if BITS_PER_LONG == 32
1028 #define cpu_to_lel(A) cpu_to_le32(A)
1029 #define lel_to_cpu(A) le32_to_cpu(A)
1030 #elif BITS_PER_LONG == 64
1032 #define cpu_to_lel(A) cpu_to_le64(A)
1033 #define lel_to_cpu(A) le64_to_cpu(A)
1035 #error "LN2 of BITS_PER_LONG unknown!"
1039 /* 16MB sized 'bitmap extent' to track syncer usage */
1041 int rs_left
; /* number of bits set (out of sync) in this extent. */
1042 int rs_failed
; /* number of failed resync requests in this extent. */
1043 unsigned long flags
;
1044 struct lc_element lce
;
1047 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1048 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1049 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1053 * We need to store one bit for a block.
1054 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1055 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1056 * Bit 1 ==> local node thinks this block needs to be synced.
1059 #define SLEEP_TIME (HZ/10)
1061 /* We do bitmap IO in units of 4k blocks.
1062 * We also still have a hardcoded 4k per bit relation. */
1063 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1064 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1065 /* mostly arbitrarily set the represented size of one bitmap extent,
1066 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1067 * at 4k per bit resolution) */
1068 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
1069 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1071 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1072 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1075 /* thus many _storage_ sectors are described by one bit */
1076 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1077 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1078 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1080 /* bit to represented kilo byte conversion */
1081 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1083 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1084 * _storage_ sector is located in */
1085 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1087 /* how much _storage_ sectors we have per bitmap sector */
1088 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1089 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1091 /* in one sector of the bitmap, we have this many activity_log extents. */
1092 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1094 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1095 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1097 /* the extent in "PER_EXTENT" below is an activity log extent
1098 * we need that many (long words/bytes) to store the bitmap
1099 * of one AL_EXTENT_SIZE chunk of storage.
1100 * we can store the bitmap for that many AL_EXTENTS within
1101 * one sector of the _on_disk_ bitmap:
1102 * bit 0 bit 37 bit 38 bit (512*8)-1
1103 * ...|........|........|.. // ..|........|
1104 * sect. 0 `296 `304 ^(512*8*8)-1
1106 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1107 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1108 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1111 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1112 /* we have a certain meta data variant that has a fixed on-disk size of 128
1113 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity
1114 * log, leaving this many sectors for the bitmap.
1117 #define DRBD_MAX_SECTORS_FIXED_BM \
1118 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9)))
1119 #if !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1120 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1121 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1123 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM
1124 /* 16 TB in units of sectors */
1125 #if BITS_PER_LONG == 32
1126 /* adjust by one page worth of bitmap,
1127 * so we won't wrap around in drbd_bm_find_next_bit.
1128 * you should use 64bit OS for that much storage, anyways. */
1129 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1131 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1132 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1133 /* corresponds to (1UL << 38) bits right now. */
1137 /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE,
1138 * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte.
1139 * Since we may live in a mixed-platform cluster,
1140 * we limit us to a platform agnostic constant here for now.
1141 * A followup commit may allow even bigger BIO sizes,
1142 * once we thought that through. */
1143 #define DRBD_MAX_BIO_SIZE (1U << 20)
1144 #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1145 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1147 #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */
1149 #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */
1150 #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */
1152 extern int drbd_bm_init(struct drbd_device
*device
);
1153 extern int drbd_bm_resize(struct drbd_device
*device
, sector_t sectors
, int set_new_bits
);
1154 extern void drbd_bm_cleanup(struct drbd_device
*device
);
1155 extern void drbd_bm_set_all(struct drbd_device
*device
);
1156 extern void drbd_bm_clear_all(struct drbd_device
*device
);
1157 /* set/clear/test only a few bits at a time */
1158 extern int drbd_bm_set_bits(
1159 struct drbd_device
*device
, unsigned long s
, unsigned long e
);
1160 extern int drbd_bm_clear_bits(
1161 struct drbd_device
*device
, unsigned long s
, unsigned long e
);
1162 extern int drbd_bm_count_bits(
1163 struct drbd_device
*device
, const unsigned long s
, const unsigned long e
);
1164 /* bm_set_bits variant for use while holding drbd_bm_lock,
1165 * may process the whole bitmap in one go */
1166 extern void _drbd_bm_set_bits(struct drbd_device
*device
,
1167 const unsigned long s
, const unsigned long e
);
1168 extern int drbd_bm_test_bit(struct drbd_device
*device
, unsigned long bitnr
);
1169 extern int drbd_bm_e_weight(struct drbd_device
*device
, unsigned long enr
);
1170 extern int drbd_bm_write_page(struct drbd_device
*device
, unsigned int idx
) __must_hold(local
);
1171 extern int drbd_bm_read(struct drbd_device
*device
) __must_hold(local
);
1172 extern void drbd_bm_mark_for_writeout(struct drbd_device
*device
, int page_nr
);
1173 extern int drbd_bm_write(struct drbd_device
*device
) __must_hold(local
);
1174 extern int drbd_bm_write_hinted(struct drbd_device
*device
) __must_hold(local
);
1175 extern int drbd_bm_write_all(struct drbd_device
*device
) __must_hold(local
);
1176 extern int drbd_bm_write_copy_pages(struct drbd_device
*device
) __must_hold(local
);
1177 extern size_t drbd_bm_words(struct drbd_device
*device
);
1178 extern unsigned long drbd_bm_bits(struct drbd_device
*device
);
1179 extern sector_t
drbd_bm_capacity(struct drbd_device
*device
);
1181 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1182 extern unsigned long drbd_bm_find_next(struct drbd_device
*device
, unsigned long bm_fo
);
1183 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1184 extern unsigned long _drbd_bm_find_next(struct drbd_device
*device
, unsigned long bm_fo
);
1185 extern unsigned long _drbd_bm_find_next_zero(struct drbd_device
*device
, unsigned long bm_fo
);
1186 extern unsigned long _drbd_bm_total_weight(struct drbd_device
*device
);
1187 extern unsigned long drbd_bm_total_weight(struct drbd_device
*device
);
1188 extern int drbd_bm_rs_done(struct drbd_device
*device
);
1189 /* for receive_bitmap */
1190 extern void drbd_bm_merge_lel(struct drbd_device
*device
, size_t offset
,
1191 size_t number
, unsigned long *buffer
);
1192 /* for _drbd_send_bitmap */
1193 extern void drbd_bm_get_lel(struct drbd_device
*device
, size_t offset
,
1194 size_t number
, unsigned long *buffer
);
1196 extern void drbd_bm_lock(struct drbd_device
*device
, char *why
, enum bm_flag flags
);
1197 extern void drbd_bm_unlock(struct drbd_device
*device
);
1200 extern struct kmem_cache
*drbd_request_cache
;
1201 extern struct kmem_cache
*drbd_ee_cache
; /* peer requests */
1202 extern struct kmem_cache
*drbd_bm_ext_cache
; /* bitmap extents */
1203 extern struct kmem_cache
*drbd_al_ext_cache
; /* activity log extents */
1204 extern mempool_t
*drbd_request_mempool
;
1205 extern mempool_t
*drbd_ee_mempool
;
1207 /* drbd's page pool, used to buffer data received from the peer,
1208 * or data requested by the peer.
1210 * This does not have an emergency reserve.
1212 * When allocating from this pool, it first takes pages from the pool.
1213 * Only if the pool is depleted will try to allocate from the system.
1215 * The assumption is that pages taken from this pool will be processed,
1216 * and given back, "quickly", and then can be recycled, so we can avoid
1217 * frequent calls to alloc_page(), and still will be able to make progress even
1218 * under memory pressure.
1220 extern struct page
*drbd_pp_pool
;
1221 extern spinlock_t drbd_pp_lock
;
1222 extern int drbd_pp_vacant
;
1223 extern wait_queue_head_t drbd_pp_wait
;
1225 /* We also need a standard (emergency-reserve backed) page pool
1226 * for meta data IO (activity log, bitmap).
1227 * We can keep it global, as long as it is used as "N pages at a time".
1228 * 128 should be plenty, currently we probably can get away with as few as 1.
1230 #define DRBD_MIN_POOL_PAGES 128
1231 extern mempool_t
*drbd_md_io_page_pool
;
1233 /* We also need to make sure we get a bio
1234 * when we need it for housekeeping purposes */
1235 extern struct bio_set
*drbd_md_io_bio_set
;
1236 /* to allocate from that set */
1237 extern struct bio
*bio_alloc_drbd(gfp_t gfp_mask
);
1239 extern rwlock_t global_state_lock
;
1241 extern int conn_lowest_minor(struct drbd_connection
*connection
);
1242 enum drbd_ret_code
drbd_create_device(struct drbd_resource
*resource
, unsigned int minor
, int vnr
);
1243 extern void drbd_destroy_device(struct kref
*kref
);
1244 extern void drbd_delete_device(struct drbd_device
*mdev
);
1246 extern struct drbd_resource
*drbd_create_resource(const char *name
);
1247 extern void drbd_free_resource(struct drbd_resource
*resource
);
1249 extern int set_resource_options(struct drbd_resource
*resource
, struct res_opts
*res_opts
);
1250 extern struct drbd_connection
*conn_create(const char *name
, struct res_opts
*res_opts
);
1251 extern void drbd_destroy_connection(struct kref
*kref
);
1252 extern struct drbd_connection
*conn_get_by_addrs(void *my_addr
, int my_addr_len
,
1253 void *peer_addr
, int peer_addr_len
);
1254 extern struct drbd_resource
*drbd_find_resource(const char *name
);
1255 extern void drbd_destroy_resource(struct kref
*kref
);
1256 extern void conn_free_crypto(struct drbd_connection
*connection
);
1258 extern int proc_details
;
1261 extern void do_submit(struct work_struct
*ws
);
1262 extern void __drbd_make_request(struct drbd_device
*, struct bio
*, unsigned long);
1263 extern void drbd_make_request(struct request_queue
*q
, struct bio
*bio
);
1264 extern int drbd_read_remote(struct drbd_device
*device
, struct drbd_request
*req
);
1265 extern int drbd_merge_bvec(struct request_queue
*q
, struct bvec_merge_data
*bvm
, struct bio_vec
*bvec
);
1266 extern int is_valid_ar_handle(struct drbd_request
*, sector_t
);
1270 extern int drbd_msg_put_info(const char *info
);
1271 extern void drbd_suspend_io(struct drbd_device
*device
);
1272 extern void drbd_resume_io(struct drbd_device
*device
);
1273 extern char *ppsize(char *buf
, unsigned long long size
);
1274 extern sector_t
drbd_new_dev_size(struct drbd_device
*, struct drbd_backing_dev
*, sector_t
, int);
1275 enum determine_dev_size
{
1276 DS_ERROR_SHRINK
= -3,
1277 DS_ERROR_SPACE_MD
= -2,
1282 DS_GREW_FROM_ZERO
= 3,
1284 extern enum determine_dev_size
1285 drbd_determine_dev_size(struct drbd_device
*, enum dds_flags
, struct resize_parms
*) __must_hold(local
);
1286 extern void resync_after_online_grow(struct drbd_device
*);
1287 extern void drbd_reconsider_max_bio_size(struct drbd_device
*device
);
1288 extern enum drbd_state_rv
drbd_set_role(struct drbd_device
*device
,
1289 enum drbd_role new_role
,
1291 extern bool conn_try_outdate_peer(struct drbd_connection
*connection
);
1292 extern void conn_try_outdate_peer_async(struct drbd_connection
*connection
);
1293 extern int drbd_khelper(struct drbd_device
*device
, char *cmd
);
1296 extern int drbd_worker(struct drbd_thread
*thi
);
1297 enum drbd_ret_code
drbd_resync_after_valid(struct drbd_device
*device
, int o_minor
);
1298 void drbd_resync_after_changed(struct drbd_device
*device
);
1299 extern void drbd_start_resync(struct drbd_device
*device
, enum drbd_conns side
);
1300 extern void resume_next_sg(struct drbd_device
*device
);
1301 extern void suspend_other_sg(struct drbd_device
*device
);
1302 extern int drbd_resync_finished(struct drbd_device
*device
);
1303 /* maybe rather drbd_main.c ? */
1304 extern void *drbd_md_get_buffer(struct drbd_device
*device
);
1305 extern void drbd_md_put_buffer(struct drbd_device
*device
);
1306 extern int drbd_md_sync_page_io(struct drbd_device
*device
,
1307 struct drbd_backing_dev
*bdev
, sector_t sector
, int rw
);
1308 extern void drbd_ov_out_of_sync_found(struct drbd_device
*, sector_t
, int);
1309 extern void wait_until_done_or_force_detached(struct drbd_device
*device
,
1310 struct drbd_backing_dev
*bdev
, unsigned int *done
);
1311 extern void drbd_rs_controller_reset(struct drbd_device
*device
);
1313 static inline void ov_out_of_sync_print(struct drbd_device
*device
)
1315 if (device
->ov_last_oos_size
) {
1316 drbd_err(device
, "Out of sync: start=%llu, size=%lu (sectors)\n",
1317 (unsigned long long)device
->ov_last_oos_start
,
1318 (unsigned long)device
->ov_last_oos_size
);
1320 device
->ov_last_oos_size
= 0;
1324 extern void drbd_csum_bio(struct drbd_device
*, struct crypto_hash
*, struct bio
*, void *);
1325 extern void drbd_csum_ee(struct drbd_device
*, struct crypto_hash
*,
1326 struct drbd_peer_request
*, void *);
1327 /* worker callbacks */
1328 extern int w_e_end_data_req(struct drbd_work
*, int);
1329 extern int w_e_end_rsdata_req(struct drbd_work
*, int);
1330 extern int w_e_end_csum_rs_req(struct drbd_work
*, int);
1331 extern int w_e_end_ov_reply(struct drbd_work
*, int);
1332 extern int w_e_end_ov_req(struct drbd_work
*, int);
1333 extern int w_ov_finished(struct drbd_work
*, int);
1334 extern int w_resync_timer(struct drbd_work
*, int);
1335 extern int w_send_write_hint(struct drbd_work
*, int);
1336 extern int w_make_resync_request(struct drbd_work
*, int);
1337 extern int w_send_dblock(struct drbd_work
*, int);
1338 extern int w_send_read_req(struct drbd_work
*, int);
1339 extern int w_prev_work_done(struct drbd_work
*, int);
1340 extern int w_e_reissue(struct drbd_work
*, int);
1341 extern int w_restart_disk_io(struct drbd_work
*, int);
1342 extern int w_send_out_of_sync(struct drbd_work
*, int);
1343 extern int w_start_resync(struct drbd_work
*, int);
1345 extern void resync_timer_fn(unsigned long data
);
1346 extern void start_resync_timer_fn(unsigned long data
);
1348 /* drbd_receiver.c */
1349 extern int drbd_rs_should_slow_down(struct drbd_device
*device
, sector_t sector
);
1350 extern int drbd_submit_peer_request(struct drbd_device
*,
1351 struct drbd_peer_request
*, const unsigned,
1353 extern int drbd_free_peer_reqs(struct drbd_device
*, struct list_head
*);
1354 extern struct drbd_peer_request
*drbd_alloc_peer_req(struct drbd_device
*, u64
,
1355 sector_t
, unsigned int,
1356 gfp_t
) __must_hold(local
);
1357 extern void __drbd_free_peer_req(struct drbd_device
*, struct drbd_peer_request
*,
1359 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1360 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
1361 extern struct page
*drbd_alloc_pages(struct drbd_device
*, unsigned int, bool);
1362 extern void drbd_set_recv_tcq(struct drbd_device
*device
, int tcq_enabled
);
1363 extern void _drbd_clear_done_ee(struct drbd_device
*device
, struct list_head
*to_be_freed
);
1364 extern void conn_flush_workqueue(struct drbd_connection
*connection
);
1365 extern int drbd_connected(struct drbd_device
*device
);
1366 static inline void drbd_flush_workqueue(struct drbd_device
*device
)
1368 conn_flush_workqueue(first_peer_device(device
)->connection
);
1371 /* Yes, there is kernel_setsockopt, but only since 2.6.18.
1372 * So we have our own copy of it here. */
1373 static inline int drbd_setsockopt(struct socket
*sock
, int level
, int optname
,
1374 char *optval
, int optlen
)
1376 mm_segment_t oldfs
= get_fs();
1377 char __user
*uoptval
;
1380 uoptval
= (char __user __force
*)optval
;
1383 if (level
== SOL_SOCKET
)
1384 err
= sock_setsockopt(sock
, level
, optname
, uoptval
, optlen
);
1386 err
= sock
->ops
->setsockopt(sock
, level
, optname
, uoptval
,
1392 static inline void drbd_tcp_cork(struct socket
*sock
)
1395 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1396 (char*)&val
, sizeof(val
));
1399 static inline void drbd_tcp_uncork(struct socket
*sock
)
1402 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1403 (char*)&val
, sizeof(val
));
1406 static inline void drbd_tcp_nodelay(struct socket
*sock
)
1409 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_NODELAY
,
1410 (char*)&val
, sizeof(val
));
1413 static inline void drbd_tcp_quickack(struct socket
*sock
)
1416 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_QUICKACK
,
1417 (char*)&val
, sizeof(val
));
1420 void drbd_bump_write_ordering(struct drbd_connection
*connection
, enum write_ordering_e wo
);
1423 extern struct proc_dir_entry
*drbd_proc
;
1424 extern const struct file_operations drbd_proc_fops
;
1425 extern const char *drbd_conn_str(enum drbd_conns s
);
1426 extern const char *drbd_role_str(enum drbd_role s
);
1429 extern int drbd_al_begin_io_nonblock(struct drbd_device
*device
, struct drbd_interval
*i
);
1430 extern void drbd_al_begin_io_commit(struct drbd_device
*device
, bool delegate
);
1431 extern bool drbd_al_begin_io_fastpath(struct drbd_device
*device
, struct drbd_interval
*i
);
1432 extern void drbd_al_begin_io(struct drbd_device
*device
, struct drbd_interval
*i
, bool delegate
);
1433 extern void drbd_al_complete_io(struct drbd_device
*device
, struct drbd_interval
*i
);
1434 extern void drbd_rs_complete_io(struct drbd_device
*device
, sector_t sector
);
1435 extern int drbd_rs_begin_io(struct drbd_device
*device
, sector_t sector
);
1436 extern int drbd_try_rs_begin_io(struct drbd_device
*device
, sector_t sector
);
1437 extern void drbd_rs_cancel_all(struct drbd_device
*device
);
1438 extern int drbd_rs_del_all(struct drbd_device
*device
);
1439 extern void drbd_rs_failed_io(struct drbd_device
*device
,
1440 sector_t sector
, int size
);
1441 extern void drbd_advance_rs_marks(struct drbd_device
*device
, unsigned long still_to_go
);
1442 extern void __drbd_set_in_sync(struct drbd_device
*device
, sector_t sector
,
1443 int size
, const char *file
, const unsigned int line
);
1444 #define drbd_set_in_sync(device, sector, size) \
1445 __drbd_set_in_sync(device, sector, size, __FILE__, __LINE__)
1446 extern int __drbd_set_out_of_sync(struct drbd_device
*device
, sector_t sector
,
1447 int size
, const char *file
, const unsigned int line
);
1448 #define drbd_set_out_of_sync(device, sector, size) \
1449 __drbd_set_out_of_sync(device, sector, size, __FILE__, __LINE__)
1450 extern void drbd_al_shrink(struct drbd_device
*device
);
1451 extern int drbd_initialize_al(struct drbd_device
*, void *);
1454 /* state info broadcast */
1456 enum drbd_state_info_bcast_reason sib_reason
;
1460 unsigned helper_exit_code
;
1463 union drbd_state os
;
1464 union drbd_state ns
;
1468 void drbd_bcast_event(struct drbd_device
*device
, const struct sib_info
*sib
);
1471 * inline helper functions
1472 *************************/
1474 /* see also page_chain_add and friends in drbd_receiver.c */
1475 static inline struct page
*page_chain_next(struct page
*page
)
1477 return (struct page
*)page_private(page
);
1479 #define page_chain_for_each(page) \
1480 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1481 page = page_chain_next(page))
1482 #define page_chain_for_each_safe(page, n) \
1483 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1486 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request
*peer_req
)
1488 struct page
*page
= peer_req
->pages
;
1489 page_chain_for_each(page
) {
1490 if (page_count(page
) > 1)
1496 static inline enum drbd_state_rv
1497 _drbd_set_state(struct drbd_device
*device
, union drbd_state ns
,
1498 enum chg_state_flags flags
, struct completion
*done
)
1500 enum drbd_state_rv rv
;
1502 read_lock(&global_state_lock
);
1503 rv
= __drbd_set_state(device
, ns
, flags
, done
);
1504 read_unlock(&global_state_lock
);
1509 static inline union drbd_state
drbd_read_state(struct drbd_device
*device
)
1511 union drbd_state rv
;
1513 rv
.i
= device
->state
.i
;
1514 rv
.susp
= first_peer_device(device
)->connection
->susp
;
1515 rv
.susp_nod
= first_peer_device(device
)->connection
->susp_nod
;
1516 rv
.susp_fen
= first_peer_device(device
)->connection
->susp_fen
;
1521 enum drbd_force_detach_flags
{
1528 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1529 static inline void __drbd_chk_io_error_(struct drbd_device
*device
,
1530 enum drbd_force_detach_flags df
,
1533 enum drbd_io_error_p ep
;
1536 ep
= rcu_dereference(device
->ldev
->disk_conf
)->on_io_error
;
1539 case EP_PASS_ON
: /* FIXME would this be better named "Ignore"? */
1540 if (df
== DRBD_READ_ERROR
|| df
== DRBD_WRITE_ERROR
) {
1541 if (__ratelimit(&drbd_ratelimit_state
))
1542 drbd_err(device
, "Local IO failed in %s.\n", where
);
1543 if (device
->state
.disk
> D_INCONSISTENT
)
1544 _drbd_set_state(_NS(device
, disk
, D_INCONSISTENT
), CS_HARD
, NULL
);
1547 /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */
1549 case EP_CALL_HELPER
:
1550 /* Remember whether we saw a READ or WRITE error.
1552 * Recovery of the affected area for WRITE failure is covered
1553 * by the activity log.
1554 * READ errors may fall outside that area though. Certain READ
1555 * errors can be "healed" by writing good data to the affected
1556 * blocks, which triggers block re-allocation in lower layers.
1558 * If we can not write the bitmap after a READ error,
1559 * we may need to trigger a full sync (see w_go_diskless()).
1561 * Force-detach is not really an IO error, but rather a
1562 * desperate measure to try to deal with a completely
1563 * unresponsive lower level IO stack.
1564 * Still it should be treated as a WRITE error.
1566 * Meta IO error is always WRITE error:
1567 * we read meta data only once during attach,
1568 * which will fail in case of errors.
1570 set_bit(WAS_IO_ERROR
, &device
->flags
);
1571 if (df
== DRBD_READ_ERROR
)
1572 set_bit(WAS_READ_ERROR
, &device
->flags
);
1573 if (df
== DRBD_FORCE_DETACH
)
1574 set_bit(FORCE_DETACH
, &device
->flags
);
1575 if (device
->state
.disk
> D_FAILED
) {
1576 _drbd_set_state(_NS(device
, disk
, D_FAILED
), CS_HARD
, NULL
);
1578 "Local IO failed in %s. Detaching...\n", where
);
1585 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1586 * @device: DRBD device.
1587 * @error: Error code passed to the IO completion callback
1588 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1590 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1592 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1593 static inline void drbd_chk_io_error_(struct drbd_device
*device
,
1594 int error
, enum drbd_force_detach_flags forcedetach
, const char *where
)
1597 unsigned long flags
;
1598 spin_lock_irqsave(&first_peer_device(device
)->connection
->req_lock
, flags
);
1599 __drbd_chk_io_error_(device
, forcedetach
, where
);
1600 spin_unlock_irqrestore(&first_peer_device(device
)->connection
->req_lock
, flags
);
1606 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1607 * @bdev: Meta data block device.
1609 * BTW, for internal meta data, this happens to be the maximum capacity
1610 * we could agree upon with our peer node.
1612 static inline sector_t
drbd_md_first_sector(struct drbd_backing_dev
*bdev
)
1614 switch (bdev
->md
.meta_dev_idx
) {
1615 case DRBD_MD_INDEX_INTERNAL
:
1616 case DRBD_MD_INDEX_FLEX_INT
:
1617 return bdev
->md
.md_offset
+ bdev
->md
.bm_offset
;
1618 case DRBD_MD_INDEX_FLEX_EXT
:
1620 return bdev
->md
.md_offset
;
1625 * drbd_md_last_sector() - Return the last sector number of the meta data area
1626 * @bdev: Meta data block device.
1628 static inline sector_t
drbd_md_last_sector(struct drbd_backing_dev
*bdev
)
1630 switch (bdev
->md
.meta_dev_idx
) {
1631 case DRBD_MD_INDEX_INTERNAL
:
1632 case DRBD_MD_INDEX_FLEX_INT
:
1633 return bdev
->md
.md_offset
+ MD_4kB_SECT
-1;
1634 case DRBD_MD_INDEX_FLEX_EXT
:
1636 return bdev
->md
.md_offset
+ bdev
->md
.md_size_sect
-1;
1640 /* Returns the number of 512 byte sectors of the device */
1641 static inline sector_t
drbd_get_capacity(struct block_device
*bdev
)
1643 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1644 return bdev
? i_size_read(bdev
->bd_inode
) >> 9 : 0;
1648 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1649 * @bdev: Meta data block device.
1651 * returns the capacity we announce to out peer. we clip ourselves at the
1652 * various MAX_SECTORS, because if we don't, current implementation will
1653 * oops sooner or later
1655 static inline sector_t
drbd_get_max_capacity(struct drbd_backing_dev
*bdev
)
1659 switch (bdev
->md
.meta_dev_idx
) {
1660 case DRBD_MD_INDEX_INTERNAL
:
1661 case DRBD_MD_INDEX_FLEX_INT
:
1662 s
= drbd_get_capacity(bdev
->backing_bdev
)
1663 ? min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1664 drbd_md_first_sector(bdev
))
1667 case DRBD_MD_INDEX_FLEX_EXT
:
1668 s
= min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1669 drbd_get_capacity(bdev
->backing_bdev
));
1670 /* clip at maximum size the meta device can support */
1671 s
= min_t(sector_t
, s
,
1672 BM_EXT_TO_SECT(bdev
->md
.md_size_sect
1673 - bdev
->md
.bm_offset
));
1676 s
= min_t(sector_t
, DRBD_MAX_SECTORS
,
1677 drbd_get_capacity(bdev
->backing_bdev
));
1683 * drbd_md_ss() - Return the sector number of our meta data super block
1684 * @bdev: Meta data block device.
1686 static inline sector_t
drbd_md_ss(struct drbd_backing_dev
*bdev
)
1688 const int meta_dev_idx
= bdev
->md
.meta_dev_idx
;
1690 if (meta_dev_idx
== DRBD_MD_INDEX_FLEX_EXT
)
1693 /* Since drbd08, internal meta data is always "flexible".
1694 * position: last 4k aligned block of 4k size */
1695 if (meta_dev_idx
== DRBD_MD_INDEX_INTERNAL
||
1696 meta_dev_idx
== DRBD_MD_INDEX_FLEX_INT
)
1697 return (drbd_get_capacity(bdev
->backing_bdev
) & ~7ULL) - 8;
1699 /* external, some index; this is the old fixed size layout */
1700 return MD_128MB_SECT
* bdev
->md
.meta_dev_idx
;
1704 drbd_queue_work_front(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1706 unsigned long flags
;
1707 spin_lock_irqsave(&q
->q_lock
, flags
);
1708 list_add(&w
->list
, &q
->q
);
1709 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1710 wake_up(&q
->q_wait
);
1714 drbd_queue_work(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1716 unsigned long flags
;
1717 spin_lock_irqsave(&q
->q_lock
, flags
);
1718 list_add_tail(&w
->list
, &q
->q
);
1719 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1720 wake_up(&q
->q_wait
);
1723 static inline void wake_asender(struct drbd_connection
*connection
)
1725 if (test_bit(SIGNAL_ASENDER
, &connection
->flags
))
1726 force_sig(DRBD_SIG
, connection
->asender
.task
);
1729 static inline void request_ping(struct drbd_connection
*connection
)
1731 set_bit(SEND_PING
, &connection
->flags
);
1732 wake_asender(connection
);
1735 extern void *conn_prepare_command(struct drbd_connection
*, struct drbd_socket
*);
1736 extern void *drbd_prepare_command(struct drbd_device
*, struct drbd_socket
*);
1737 extern int conn_send_command(struct drbd_connection
*, struct drbd_socket
*,
1738 enum drbd_packet
, unsigned int, void *,
1740 extern int drbd_send_command(struct drbd_device
*, struct drbd_socket
*,
1741 enum drbd_packet
, unsigned int, void *,
1744 extern int drbd_send_ping(struct drbd_connection
*connection
);
1745 extern int drbd_send_ping_ack(struct drbd_connection
*connection
);
1746 extern int drbd_send_state_req(struct drbd_device
*, union drbd_state
, union drbd_state
);
1747 extern int conn_send_state_req(struct drbd_connection
*, union drbd_state
, union drbd_state
);
1749 static inline void drbd_thread_stop(struct drbd_thread
*thi
)
1751 _drbd_thread_stop(thi
, false, true);
1754 static inline void drbd_thread_stop_nowait(struct drbd_thread
*thi
)
1756 _drbd_thread_stop(thi
, false, false);
1759 static inline void drbd_thread_restart_nowait(struct drbd_thread
*thi
)
1761 _drbd_thread_stop(thi
, true, false);
1764 /* counts how many answer packets packets we expect from our peer,
1765 * for either explicit application requests,
1766 * or implicit barrier packets as necessary.
1769 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1770 * it is much easier and equally valid to count what we queue for the
1771 * worker, even before it actually was queued or send.
1772 * (drbd_make_request_common; recovery path on read io-error)
1774 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1775 * _req_mod(req, DATA_RECEIVED)
1776 * [from receive_DataReply]
1777 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1778 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1779 * for some reason it is NOT decreased in got_NegAck,
1780 * but in the resulting cleanup code from report_params.
1781 * we should try to remember the reason for that...
1782 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
1783 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
1784 * [from tl_clear_barrier]
1786 static inline void inc_ap_pending(struct drbd_device
*device
)
1788 atomic_inc(&device
->ap_pending_cnt
);
1791 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \
1792 if (atomic_read(&device->which) < 0) \
1793 drbd_err(device, "in %s:%d: " #which " = %d < 0 !\n", \
1795 atomic_read(&device->which))
1797 #define dec_ap_pending(device) _dec_ap_pending(device, __FUNCTION__, __LINE__)
1798 static inline void _dec_ap_pending(struct drbd_device
*device
, const char *func
, int line
)
1800 if (atomic_dec_and_test(&device
->ap_pending_cnt
))
1801 wake_up(&device
->misc_wait
);
1802 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt
, func
, line
);
1805 /* counts how many resync-related answers we still expect from the peer
1807 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1808 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
1809 * (or P_NEG_ACK with ID_SYNCER)
1811 static inline void inc_rs_pending(struct drbd_device
*device
)
1813 atomic_inc(&device
->rs_pending_cnt
);
1816 #define dec_rs_pending(device) _dec_rs_pending(device, __FUNCTION__, __LINE__)
1817 static inline void _dec_rs_pending(struct drbd_device
*device
, const char *func
, int line
)
1819 atomic_dec(&device
->rs_pending_cnt
);
1820 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt
, func
, line
);
1823 /* counts how many answers we still need to send to the peer.
1825 * receive_Data unless protocol A;
1826 * we need to send a P_RECV_ACK (proto B)
1827 * or P_WRITE_ACK (proto C)
1828 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1829 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1830 * receive_Barrier_* we need to send a P_BARRIER_ACK
1832 static inline void inc_unacked(struct drbd_device
*device
)
1834 atomic_inc(&device
->unacked_cnt
);
1837 #define dec_unacked(device) _dec_unacked(device, __FUNCTION__, __LINE__)
1838 static inline void _dec_unacked(struct drbd_device
*device
, const char *func
, int line
)
1840 atomic_dec(&device
->unacked_cnt
);
1841 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt
, func
, line
);
1844 #define sub_unacked(device, n) _sub_unacked(device, n, __FUNCTION__, __LINE__)
1845 static inline void _sub_unacked(struct drbd_device
*device
, int n
, const char *func
, int line
)
1847 atomic_sub(n
, &device
->unacked_cnt
);
1848 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt
, func
, line
);
1852 * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev
1855 * You have to call put_ldev() when finished working with device->ldev.
1857 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
1858 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
1860 static inline void put_ldev(struct drbd_device
*device
)
1862 int i
= atomic_dec_return(&device
->local_cnt
);
1864 /* This may be called from some endio handler,
1865 * so we must not sleep here. */
1868 D_ASSERT(device
, i
>= 0);
1870 if (device
->state
.disk
== D_DISKLESS
)
1871 /* even internal references gone, safe to destroy */
1872 drbd_ldev_destroy(device
);
1873 if (device
->state
.disk
== D_FAILED
) {
1874 /* all application IO references gone. */
1875 if (!test_and_set_bit(GO_DISKLESS
, &device
->flags
))
1876 drbd_queue_work(&first_peer_device(device
)->connection
->sender_work
, &device
->go_diskless
);
1878 wake_up(&device
->misc_wait
);
1883 static inline int _get_ldev_if_state(struct drbd_device
*device
, enum drbd_disk_state mins
)
1887 /* never get a reference while D_DISKLESS */
1888 if (device
->state
.disk
== D_DISKLESS
)
1891 atomic_inc(&device
->local_cnt
);
1892 io_allowed
= (device
->state
.disk
>= mins
);
1898 extern int _get_ldev_if_state(struct drbd_device
*device
, enum drbd_disk_state mins
);
1901 /* you must have an "get_ldev" reference */
1902 static inline void drbd_get_syncer_progress(struct drbd_device
*device
,
1903 unsigned long *bits_left
, unsigned int *per_mil_done
)
1905 /* this is to break it at compile time when we change that, in case we
1906 * want to support more than (1<<32) bits on a 32bit arch. */
1907 typecheck(unsigned long, device
->rs_total
);
1909 /* note: both rs_total and rs_left are in bits, i.e. in
1910 * units of BM_BLOCK_SIZE.
1911 * for the percentage, we don't care. */
1913 if (device
->state
.conn
== C_VERIFY_S
|| device
->state
.conn
== C_VERIFY_T
)
1914 *bits_left
= device
->ov_left
;
1916 *bits_left
= drbd_bm_total_weight(device
) - device
->rs_failed
;
1917 /* >> 10 to prevent overflow,
1918 * +1 to prevent division by zero */
1919 if (*bits_left
> device
->rs_total
) {
1920 /* doh. maybe a logic bug somewhere.
1921 * may also be just a race condition
1922 * between this and a disconnect during sync.
1923 * for now, just prevent in-kernel buffer overflow.
1926 drbd_warn(device
, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
1927 drbd_conn_str(device
->state
.conn
),
1928 *bits_left
, device
->rs_total
, device
->rs_failed
);
1931 /* Make sure the division happens in long context.
1932 * We allow up to one petabyte storage right now,
1933 * at a granularity of 4k per bit that is 2**38 bits.
1934 * After shift right and multiplication by 1000,
1935 * this should still fit easily into a 32bit long,
1936 * so we don't need a 64bit division on 32bit arch.
1937 * Note: currently we don't support such large bitmaps on 32bit
1938 * arch anyways, but no harm done to be prepared for it here.
1940 unsigned int shift
= device
->rs_total
> UINT_MAX
? 16 : 10;
1941 unsigned long left
= *bits_left
>> shift
;
1942 unsigned long total
= 1UL + (device
->rs_total
>> shift
);
1943 unsigned long tmp
= 1000UL - left
* 1000UL/total
;
1944 *per_mil_done
= tmp
;
1949 /* this throttles on-the-fly application requests
1950 * according to max_buffers settings;
1951 * maybe re-implement using semaphores? */
1952 static inline int drbd_get_max_buffers(struct drbd_device
*device
)
1954 struct net_conf
*nc
;
1958 nc
= rcu_dereference(first_peer_device(device
)->connection
->net_conf
);
1959 mxb
= nc
? nc
->max_buffers
: 1000000; /* arbitrary limit on open requests */
1965 static inline int drbd_state_is_stable(struct drbd_device
*device
)
1967 union drbd_dev_state s
= device
->state
;
1969 /* DO NOT add a default clause, we want the compiler to warn us
1970 * for any newly introduced state we may have forgotten to add here */
1972 switch ((enum drbd_conns
)s
.conn
) {
1973 /* new io only accepted when there is no connection, ... */
1975 case C_WF_CONNECTION
:
1976 /* ... or there is a well established connection. */
1982 case C_PAUSED_SYNC_S
:
1983 case C_PAUSED_SYNC_T
:
1986 /* transitional states, IO allowed */
1987 case C_DISCONNECTING
:
1991 case C_NETWORK_FAILURE
:
1992 case C_PROTOCOL_ERROR
:
1994 case C_WF_REPORT_PARAMS
:
1995 case C_STARTING_SYNC_S
:
1996 case C_STARTING_SYNC_T
:
1999 /* Allow IO in BM exchange states with new protocols */
2001 if (first_peer_device(device
)->connection
->agreed_pro_version
< 96)
2005 /* no new io accepted in these states */
2007 case C_WF_SYNC_UUID
:
2013 switch ((enum drbd_disk_state
)s
.disk
) {
2015 case D_INCONSISTENT
:
2020 /* disk state is stable as well. */
2023 /* no new io accepted during transitional states */
2035 static inline int drbd_suspended(struct drbd_device
*device
)
2037 struct drbd_connection
*connection
= first_peer_device(device
)->connection
;
2039 return connection
->susp
|| connection
->susp_fen
|| connection
->susp_nod
;
2042 static inline bool may_inc_ap_bio(struct drbd_device
*device
)
2044 int mxb
= drbd_get_max_buffers(device
);
2046 if (drbd_suspended(device
))
2048 if (test_bit(SUSPEND_IO
, &device
->flags
))
2051 /* to avoid potential deadlock or bitmap corruption,
2052 * in various places, we only allow new application io
2053 * to start during "stable" states. */
2055 /* no new io accepted when attaching or detaching the disk */
2056 if (!drbd_state_is_stable(device
))
2059 /* since some older kernels don't have atomic_add_unless,
2060 * and we are within the spinlock anyways, we have this workaround. */
2061 if (atomic_read(&device
->ap_bio_cnt
) > mxb
)
2063 if (test_bit(BITMAP_IO
, &device
->flags
))
2068 static inline bool inc_ap_bio_cond(struct drbd_device
*device
)
2072 spin_lock_irq(&first_peer_device(device
)->connection
->req_lock
);
2073 rv
= may_inc_ap_bio(device
);
2075 atomic_inc(&device
->ap_bio_cnt
);
2076 spin_unlock_irq(&first_peer_device(device
)->connection
->req_lock
);
2081 static inline void inc_ap_bio(struct drbd_device
*device
)
2084 * as long as the device is suspended
2085 * until the bitmap is no longer on the fly during connection
2086 * handshake as long as we would exceed the max_buffer limit.
2088 * to avoid races with the reconnect code,
2089 * we need to atomic_inc within the spinlock. */
2091 wait_event(device
->misc_wait
, inc_ap_bio_cond(device
));
2094 static inline void dec_ap_bio(struct drbd_device
*device
)
2096 int mxb
= drbd_get_max_buffers(device
);
2097 int ap_bio
= atomic_dec_return(&device
->ap_bio_cnt
);
2099 D_ASSERT(device
, ap_bio
>= 0);
2101 if (ap_bio
== 0 && test_bit(BITMAP_IO
, &device
->flags
)) {
2102 if (!test_and_set_bit(BITMAP_IO_QUEUED
, &device
->flags
))
2103 drbd_queue_work(&first_peer_device(device
)->connection
->sender_work
, &device
->bm_io_work
.w
);
2106 /* this currently does wake_up for every dec_ap_bio!
2107 * maybe rather introduce some type of hysteresis?
2108 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2110 wake_up(&device
->misc_wait
);
2113 static inline bool verify_can_do_stop_sector(struct drbd_device
*device
)
2115 return first_peer_device(device
)->connection
->agreed_pro_version
>= 97 &&
2116 first_peer_device(device
)->connection
->agreed_pro_version
!= 100;
2119 static inline int drbd_set_ed_uuid(struct drbd_device
*device
, u64 val
)
2121 int changed
= device
->ed_uuid
!= val
;
2122 device
->ed_uuid
= val
;
2126 static inline int drbd_queue_order_type(struct drbd_device
*device
)
2128 /* sorry, we currently have no working implementation
2129 * of distributed TCQ stuff */
2130 #ifndef QUEUE_ORDERED_NONE
2131 #define QUEUE_ORDERED_NONE 0
2133 return QUEUE_ORDERED_NONE
;
2136 static inline void drbd_md_flush(struct drbd_device
*device
)
2140 if (device
->ldev
== NULL
) {
2141 drbd_warn(device
, "device->ldev == NULL in drbd_md_flush\n");
2145 if (test_bit(MD_NO_FUA
, &device
->flags
))
2148 r
= blkdev_issue_flush(device
->ldev
->md_bdev
, GFP_NOIO
, NULL
);
2150 set_bit(MD_NO_FUA
, &device
->flags
);
2151 drbd_err(device
, "meta data flush failed with status %d, disabling md-flushes\n", r
);
2155 static inline struct drbd_connection
*first_connection(struct drbd_resource
*resource
)
2157 return list_first_entry(&resource
->connections
,
2158 struct drbd_connection
, connections
);