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/backing-dev.h>
42 #include <linux/genhd.h>
43 #include <linux/idr.h>
45 #include <linux/lru_cache.h>
46 #include <linux/prefetch.h>
47 #include <linux/drbd_genl_api.h>
48 #include <linux/drbd.h>
49 #include "drbd_strings.h"
50 #include "drbd_state.h"
51 #include "drbd_protocol.h"
54 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
55 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
56 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
57 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
59 # define __protected_by(x)
60 # define __protected_read_by(x)
61 # define __protected_write_by(x)
62 # define __must_hold(x)
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 /* This is used to stop/restart our threads.
81 * Cannot use SIGTERM nor SIGKILL, since these
82 * are sent out by init on runlevel changes
83 * I choose SIGHUP for now.
85 #define DRBD_SIGKILL SIGHUP
87 #define ID_IN_SYNC (4711ULL)
88 #define ID_OUT_OF_SYNC (4712ULL)
89 #define ID_SYNCER (-1ULL)
91 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
94 struct drbd_connection
;
96 #define __drbd_printk_device(level, device, fmt, args...) \
97 dev_printk(level, disk_to_dev((device)->vdisk), fmt, ## args)
98 #define __drbd_printk_peer_device(level, peer_device, fmt, args...) \
99 dev_printk(level, disk_to_dev((peer_device)->device->vdisk), fmt, ## args)
100 #define __drbd_printk_resource(level, resource, fmt, args...) \
101 printk(level "drbd %s: " fmt, (resource)->name, ## args)
102 #define __drbd_printk_connection(level, connection, fmt, args...) \
103 printk(level "drbd %s: " fmt, (connection)->resource->name, ## args)
105 void drbd_printk_with_wrong_object_type(void);
107 #define __drbd_printk_if_same_type(obj, type, func, level, fmt, args...) \
108 (__builtin_types_compatible_p(typeof(obj), type) || \
109 __builtin_types_compatible_p(typeof(obj), const type)), \
110 func(level, (const type)(obj), fmt, ## args)
112 #define drbd_printk(level, obj, fmt, args...) \
113 __builtin_choose_expr( \
114 __drbd_printk_if_same_type(obj, struct drbd_device *, \
115 __drbd_printk_device, level, fmt, ## args), \
116 __builtin_choose_expr( \
117 __drbd_printk_if_same_type(obj, struct drbd_resource *, \
118 __drbd_printk_resource, level, fmt, ## args), \
119 __builtin_choose_expr( \
120 __drbd_printk_if_same_type(obj, struct drbd_connection *, \
121 __drbd_printk_connection, level, fmt, ## args), \
122 __builtin_choose_expr( \
123 __drbd_printk_if_same_type(obj, struct drbd_peer_device *, \
124 __drbd_printk_peer_device, level, fmt, ## args), \
125 drbd_printk_with_wrong_object_type()))))
127 #define drbd_dbg(obj, fmt, args...) \
128 drbd_printk(KERN_DEBUG, obj, fmt, ## args)
129 #define drbd_alert(obj, fmt, args...) \
130 drbd_printk(KERN_ALERT, obj, fmt, ## args)
131 #define drbd_err(obj, fmt, args...) \
132 drbd_printk(KERN_ERR, obj, fmt, ## args)
133 #define drbd_warn(obj, fmt, args...) \
134 drbd_printk(KERN_WARNING, obj, fmt, ## args)
135 #define drbd_info(obj, fmt, args...) \
136 drbd_printk(KERN_INFO, obj, fmt, ## args)
137 #define drbd_emerg(obj, fmt, args...) \
138 drbd_printk(KERN_EMERG, obj, fmt, ## args)
140 #define dynamic_drbd_dbg(device, fmt, args...) \
141 dynamic_dev_dbg(disk_to_dev(device->vdisk), fmt, ## args)
143 #define D_ASSERT(device, exp) do { \
145 drbd_err(device, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__); \
149 * expect - Make an assertion
151 * Unlike the assert macro, this macro returns a boolean result.
153 #define expect(exp) ({ \
154 bool _bool = (exp); \
156 drbd_err(device, "ASSERTION %s FAILED in %s\n", \
161 /* Defines to control fault insertion */
163 DRBD_FAULT_MD_WR
= 0, /* meta data write */
164 DRBD_FAULT_MD_RD
= 1, /* read */
165 DRBD_FAULT_RS_WR
= 2, /* resync */
166 DRBD_FAULT_RS_RD
= 3,
167 DRBD_FAULT_DT_WR
= 4, /* data */
168 DRBD_FAULT_DT_RD
= 5,
169 DRBD_FAULT_DT_RA
= 6, /* data read ahead */
170 DRBD_FAULT_BM_ALLOC
= 7, /* bitmap allocation */
171 DRBD_FAULT_AL_EE
= 8, /* alloc ee */
172 DRBD_FAULT_RECEIVE
= 9, /* Changes some bytes upon receiving a [rs]data block */
178 _drbd_insert_fault(struct drbd_device
*device
, unsigned int type
);
181 drbd_insert_fault(struct drbd_device
*device
, unsigned int type
) {
182 #ifdef CONFIG_DRBD_FAULT_INJECTION
184 (enable_faults
& (1<<type
)) &&
185 _drbd_insert_fault(device
, type
);
191 /* integer division, round _UP_ to the next integer */
192 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
193 /* usual integer division */
194 #define div_floor(A, B) ((A)/(B))
196 extern struct ratelimit_state drbd_ratelimit_state
;
197 extern struct idr drbd_devices
; /* RCU, updates: genl_lock() */
198 extern struct list_head drbd_resources
; /* RCU, updates: genl_lock() */
200 extern const char *cmdname(enum drbd_packet cmd
);
202 /* for sending/receiving the bitmap,
203 * possibly in some encoding scheme */
206 * stores total bits and long words
207 * of the bitmap, so we don't need to
208 * call the accessor functions over and again. */
209 unsigned long bm_bits
;
210 unsigned long bm_words
;
211 /* during xfer, current position within the bitmap */
212 unsigned long bit_offset
;
213 unsigned long word_offset
;
215 /* statistics; index: (h->command == P_BITMAP) */
220 extern void INFO_bm_xfer_stats(struct drbd_device
*device
,
221 const char *direction
, struct bm_xfer_ctx
*c
);
223 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx
*c
)
225 /* word_offset counts "native long words" (32 or 64 bit),
227 * Encoded packet may end at an unaligned bit offset.
228 * In case a fallback clear text packet is transmitted in
229 * between, we adjust this offset back to the last 64bit
230 * aligned "native long word", which makes coding and decoding
231 * the plain text bitmap much more convenient. */
232 #if BITS_PER_LONG == 64
233 c
->word_offset
= c
->bit_offset
>> 6;
234 #elif BITS_PER_LONG == 32
235 c
->word_offset
= c
->bit_offset
>> 5;
236 c
->word_offset
&= ~(1UL);
238 # error "unsupported BITS_PER_LONG"
242 extern unsigned int drbd_header_size(struct drbd_connection
*connection
);
244 /**********************************************************************/
245 enum drbd_thread_state
{
254 struct task_struct
*task
;
255 struct completion stop
;
256 enum drbd_thread_state t_state
;
257 int (*function
) (struct drbd_thread
*);
258 struct drbd_resource
*resource
;
259 struct drbd_connection
*connection
;
264 static inline enum drbd_thread_state
get_t_state(struct drbd_thread
*thi
)
266 /* THINK testing the t_state seems to be uncritical in all cases
267 * (but thread_{start,stop}), so we can read it *without* the lock.
275 struct list_head list
;
276 int (*cb
)(struct drbd_work
*, int cancel
);
279 struct drbd_device_work
{
281 struct drbd_device
*device
;
284 #include "drbd_interval.h"
286 extern int drbd_wait_misc(struct drbd_device
*, struct drbd_interval
*);
288 extern void lock_all_resources(void);
289 extern void unlock_all_resources(void);
291 struct drbd_request
{
293 struct drbd_device
*device
;
295 /* if local IO is not allowed, will be NULL.
296 * if local IO _is_ allowed, holds the locally submitted bio clone,
297 * or, after local IO completion, the ERR_PTR(error).
298 * see drbd_request_endio(). */
299 struct bio
*private_bio
;
301 struct drbd_interval i
;
303 /* epoch: used to check on "completion" whether this req was in
304 * the current epoch, and we therefore have to close it,
305 * causing a p_barrier packet to be send, starting a new epoch.
307 * This corresponds to "barrier" in struct p_barrier[_ack],
308 * and to "barrier_nr" in struct drbd_epoch (and various
309 * comments/function parameters/local variable names).
313 struct list_head tl_requests
; /* ring list in the transfer log */
314 struct bio
*master_bio
; /* master bio pointer */
316 /* see struct drbd_device */
317 struct list_head req_pending_master_completion
;
318 struct list_head req_pending_local
;
320 /* for generic IO accounting */
321 unsigned long start_jif
;
323 /* for DRBD internal statistics */
325 /* Minimal set of time stamps to determine if we wait for activity log
326 * transactions, local disk or peer. 32 bit "jiffies" are good enough,
327 * we don't expect a DRBD request to be stalled for several month.
330 /* before actual request processing */
331 unsigned long in_actlog_jif
;
334 unsigned long pre_submit_jif
;
337 unsigned long pre_send_jif
;
338 unsigned long acked_jif
;
339 unsigned long net_done_jif
;
341 /* Possibly even more detail to track each phase:
342 * master_completion_jif
343 * how long did it take to complete the master bio
344 * (application visible latency)
346 * how long the master bio was blocked until we finally allocated
349 * how long did we wait for activity log transactions
352 * when did we finally queue it for sending
354 * when did we start sending it
356 * how long did we block in the network stack trying to send it
358 * when did we receive (or fake, in protocol A) a remote ACK
360 * when did we receive final acknowledgement (P_BARRIER_ACK),
361 * or decide, e.g. on connection loss, that we do no longer expect
362 * anything from this peer for this request.
366 * when did we start submiting to the lower level device,
367 * and how long did we block in that submit function
368 * local_completion_jif
369 * how long did it take the lower level device to complete this request
373 /* once it hits 0, we may complete the master_bio */
374 atomic_t completion_ref
;
375 /* once it hits 0, we may destroy this drbd_request object */
378 unsigned rq_state
; /* see comments above _req_mod() */
382 struct drbd_connection
*connection
;
383 struct list_head list
;
384 unsigned int barrier_nr
;
385 atomic_t epoch_size
; /* increased on every request added. */
386 atomic_t active
; /* increased on every req. added, and dec on every finished. */
390 /* Prototype declaration of function defined in drbd_receiver.c */
391 int drbdd_init(struct drbd_thread
*);
392 int drbd_asender(struct drbd_thread
*);
394 /* drbd_epoch flag bits */
396 DE_HAVE_BARRIER_NUMBER
,
403 EV_CLEANUP
= 32, /* used as flag */
411 struct drbd_peer_request
{
413 struct drbd_peer_device
*peer_device
;
414 struct drbd_epoch
*epoch
; /* for writes */
416 atomic_t pending_bios
;
417 struct drbd_interval i
;
418 /* see comments on ee flag bits below */
420 unsigned long submit_jif
;
423 struct digest_info
*digest
;
428 * While corresponding bios are in flight, the only modification will be
429 * set_bit WAS_ERROR, which has to be atomic.
430 * If no bios are in flight yet, or all have been completed,
431 * non-atomic modification to ee->flags is ok.
434 __EE_CALL_AL_COMPLETE_IO
,
435 __EE_MAY_SET_IN_SYNC
,
437 /* is this a TRIM aka REQ_DISCARD? */
439 /* our lower level cannot handle trim,
440 * and we want to fall back to zeroout instead */
441 __EE_IS_TRIM_USE_ZEROOUT
,
443 /* In case a barrier failed,
444 * we need to resubmit without the barrier flag. */
447 /* we may have several bios per peer request.
448 * if any of those fail, we set this flag atomically
449 * from the endio callback */
452 /* This ee has a pointer to a digest instead of a block id */
455 /* Conflicting local requests need to be restarted after this request */
456 __EE_RESTART_REQUESTS
,
458 /* The peer wants a write ACK for this (wire proto C) */
461 /* Is set when net_conf had two_primaries set while creating this peer_req */
462 __EE_IN_INTERVAL_TREE
,
465 /* has this been submitted, or does it still wait for something else? */
468 /* this is/was a write request */
471 /* this originates from application on peer
472 * (not some resync or verify or other DRBD internal request) */
475 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
476 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
477 #define EE_IS_TRIM (1<<__EE_IS_TRIM)
478 #define EE_IS_TRIM_USE_ZEROOUT (1<<__EE_IS_TRIM_USE_ZEROOUT)
479 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
480 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
481 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
482 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
483 #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK)
484 #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE)
485 #define EE_SUBMITTED (1<<__EE_SUBMITTED)
486 #define EE_WRITE (1<<__EE_WRITE)
487 #define EE_APPLICATION (1<<__EE_APPLICATION)
489 /* flag bits per device */
491 UNPLUG_REMOTE
, /* sending a "UnplugRemote" could help */
492 MD_DIRTY
, /* current uuids and flags not yet on disk */
493 USE_DEGR_WFC_T
, /* degr-wfc-timeout instead of wfc-timeout. */
496 CRASHED_PRIMARY
, /* This node was a crashed primary.
497 * Gets cleared when the state.conn
498 * goes into C_CONNECTED state. */
501 MD_NO_FUA
, /* Users wants us to not use FUA/FLUSH on meta data dev */
503 BITMAP_IO
, /* suspend application io;
504 once no more io in flight, start bitmap io */
505 BITMAP_IO_QUEUED
, /* Started bitmap IO */
506 WAS_IO_ERROR
, /* Local disk failed, returned IO error */
507 WAS_READ_ERROR
, /* Local disk READ failed (set additionally to the above) */
508 FORCE_DETACH
, /* Force-detach from local disk, aborting any pending local IO */
509 RESYNC_AFTER_NEG
, /* Resync after online grow after the attach&negotiate finished. */
510 RESIZE_PENDING
, /* Size change detected locally, waiting for the response from
511 * the peer, if it changed there as well. */
512 NEW_CUR_UUID
, /* Create new current UUID when thawing IO */
513 AL_SUSPENDED
, /* Activity logging is currently suspended. */
514 AHEAD_TO_SYNC_SOURCE
, /* Ahead -> SyncSource queued */
515 B_RS_H_DONE
, /* Before resync handler done (already executed) */
516 DISCARD_MY_DATA
, /* discard_my_data flag per volume */
519 FLUSH_PENDING
, /* if set, device->flush_jif is when we submitted that flush
520 * from drbd_flush_after_epoch() */
522 /* cleared only after backing device related structures have been destroyed. */
523 GOING_DISKLESS
, /* Disk is being detached, because of io-error, or admin request. */
525 /* to be used in drbd_device_post_work() */
526 GO_DISKLESS
, /* tell worker to schedule cleanup before detach */
527 DESTROY_DISK
, /* tell worker to close backing devices and destroy related structures. */
528 MD_SYNC
, /* tell worker to call drbd_md_sync() */
529 RS_START
, /* tell worker to start resync/OV */
530 RS_PROGRESS
, /* tell worker that resync made significant progress */
531 RS_DONE
, /* tell worker that resync is done */
534 struct drbd_bitmap
; /* opaque for drbd_device */
536 /* definition of bits in bm_flags to be used in drbd_bm_lock
537 * and drbd_bitmap_io and friends. */
539 /* do we need to kfree, or vfree bm_pages? */
540 BM_P_VMALLOCED
= 0x10000, /* internal use only, will be masked out */
542 /* currently locked for bulk operation */
543 BM_LOCKED_MASK
= 0xf,
545 /* in detail, that is: */
550 /* so we can mark it locked for bulk operation,
551 * and still allow all non-bulk operations */
554 /* (test bit, count bit) allowed (common case) */
555 BM_LOCKED_TEST_ALLOWED
= BM_DONT_CLEAR
| BM_DONT_SET
| BM_IS_LOCKED
,
557 /* testing bits, as well as setting new bits allowed, but clearing bits
558 * would be unexpected. Used during bitmap receive. Setting new bits
559 * requires sending of "out-of-sync" information, though. */
560 BM_LOCKED_SET_ALLOWED
= BM_DONT_CLEAR
| BM_IS_LOCKED
,
562 /* for drbd_bm_write_copy_pages, everything is allowed,
563 * only concurrent bulk operations are locked out. */
564 BM_LOCKED_CHANGE_ALLOWED
= BM_IS_LOCKED
,
567 struct drbd_work_queue
{
569 spinlock_t q_lock
; /* to protect the list. */
570 wait_queue_head_t q_wait
;
575 struct socket
*socket
;
576 /* this way we get our
577 * send/receive buffers off the stack */
583 u64 md_offset
; /* sector offset to 'super' block */
585 u64 la_size_sect
; /* last agreed size, unit sectors */
586 spinlock_t uuid_lock
;
592 s32 al_offset
; /* signed relative sector offset to activity log */
593 s32 bm_offset
; /* signed relative sector offset to bitmap */
595 /* cached value of bdev->disk_conf->meta_dev_idx (see below) */
598 /* see al_tr_number_to_on_disk_sector() */
600 u32 al_stripe_size_4k
;
601 u32 al_size_4k
; /* cached product of the above */
604 struct drbd_backing_dev
{
605 struct block_device
*backing_bdev
;
606 struct block_device
*md_bdev
;
608 struct disk_conf
*disk_conf
; /* RCU, for updates: resource->conf_update */
609 sector_t known_size
; /* last known size of that backing device */
614 unsigned long start_jif
; /* last call to drbd_md_get_buffer */
615 unsigned long submit_jif
; /* last _drbd_md_sync_page_io() submit */
616 const char *current_use
;
626 int (*io_fn
)(struct drbd_device
*device
);
627 void (*done
)(struct drbd_device
*device
, int rv
);
631 unsigned int head_index
;
633 int total
; /* sum of all values */
636 extern struct fifo_buffer
*fifo_alloc(int fifo_size
);
638 /* flag bits per connection */
640 NET_CONGESTED
, /* The data socket is congested */
641 RESOLVE_CONFLICTS
, /* Set on one node, cleared on the peer! */
643 GOT_PING_ACK
, /* set when we receive a ping_ack packet, ping_wait gets woken */
644 CONN_WD_ST_CHG_REQ
, /* A cluster wide state change on the connection is active */
647 CONN_DRY_RUN
, /* Expect disconnect after resync handshake. */
648 CREATE_BARRIER
, /* next P_DATA is preceded by a P_BARRIER */
649 STATE_SENT
, /* Do not change state/UUIDs while this is set */
650 CALLBACK_PENDING
, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
651 * pending, from drbd worker context.
652 * If set, bdi_write_congested() returns true,
653 * so shrink_page_list() would not recurse into,
654 * and potentially deadlock on, this drbd worker.
658 DEVICE_WORK_PENDING
, /* tell worker that some device has pending work */
661 enum which_state
{ NOW
, OLD
= NOW
, NEW
};
663 struct drbd_resource
{
665 #ifdef CONFIG_DEBUG_FS
666 struct dentry
*debugfs_res
;
667 struct dentry
*debugfs_res_volumes
;
668 struct dentry
*debugfs_res_connections
;
669 struct dentry
*debugfs_res_in_flight_summary
;
672 struct idr devices
; /* volume number to device mapping */
673 struct list_head connections
;
674 struct list_head resources
;
675 struct res_opts res_opts
;
676 struct mutex conf_update
; /* mutex for ready-copy-update of net_conf and disk_conf */
677 struct mutex adm_mutex
; /* mutex to serialize administrative requests */
680 unsigned susp
:1; /* IO suspended by user */
681 unsigned susp_nod
:1; /* IO suspended because no data */
682 unsigned susp_fen
:1; /* IO suspended because fence peer handler runs */
684 enum write_ordering_e write_ordering
;
686 cpumask_var_t cpu_mask
;
689 struct drbd_thread_timing_details
691 unsigned long start_jif
;
693 const char *caller_fn
;
698 struct drbd_connection
{
699 struct list_head connections
;
700 struct drbd_resource
*resource
;
701 #ifdef CONFIG_DEBUG_FS
702 struct dentry
*debugfs_conn
;
703 struct dentry
*debugfs_conn_callback_history
;
704 struct dentry
*debugfs_conn_oldest_requests
;
707 struct idr peer_devices
; /* volume number to peer device mapping */
708 enum drbd_conns cstate
; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
709 struct mutex cstate_mutex
; /* Protects graceful disconnects */
710 unsigned int connect_cnt
; /* Inc each time a connection is established */
713 struct net_conf
*net_conf
; /* content protected by rcu */
714 wait_queue_head_t ping_wait
; /* Woken upon reception of a ping, and a state change */
716 struct sockaddr_storage my_addr
;
718 struct sockaddr_storage peer_addr
;
721 struct drbd_socket data
; /* data/barrier/cstate/parameter packets */
722 struct drbd_socket meta
; /* ping/ack (metadata) packets */
723 int agreed_pro_version
; /* actually used protocol version */
725 unsigned long last_received
; /* in jiffies, either socket */
726 unsigned int ko_count
;
728 struct list_head transfer_log
; /* all requests not yet fully processed */
730 struct crypto_hash
*cram_hmac_tfm
;
731 struct crypto_hash
*integrity_tfm
; /* checksums we compute, updates protected by connection->data->mutex */
732 struct crypto_hash
*peer_integrity_tfm
; /* checksums we verify, only accessed from receiver thread */
733 struct crypto_hash
*csums_tfm
;
734 struct crypto_hash
*verify_tfm
;
739 struct drbd_epoch
*current_epoch
;
740 spinlock_t epoch_lock
;
742 atomic_t current_tle_nr
; /* transfer log epoch number */
743 unsigned current_tle_writes
; /* writes seen within this tl epoch */
745 unsigned long last_reconnect_jif
;
746 struct drbd_thread receiver
;
747 struct drbd_thread worker
;
748 struct drbd_thread ack_receiver
;
749 struct workqueue_struct
*ack_sender
;
752 * so we can look up the oldest pending requests more quickly.
753 * protected by resource->req_lock */
754 struct drbd_request
*req_next
; /* DRBD 9: todo.req_next */
755 struct drbd_request
*req_ack_pending
;
756 struct drbd_request
*req_not_net_done
;
759 struct drbd_work_queue sender_work
;
761 #define DRBD_THREAD_DETAILS_HIST 16
762 unsigned int w_cb_nr
; /* keeps counting up */
763 unsigned int r_cb_nr
; /* keeps counting up */
764 struct drbd_thread_timing_details w_timing_details
[DRBD_THREAD_DETAILS_HIST
];
765 struct drbd_thread_timing_details r_timing_details
[DRBD_THREAD_DETAILS_HIST
];
768 unsigned long last_sent_barrier_jif
;
770 /* whether this sender thread
771 * has processed a single write yet. */
772 bool seen_any_write_yet
;
774 /* Which barrier number to send with the next P_BARRIER */
775 int current_epoch_nr
;
777 /* how many write requests have been sent
778 * with req->epoch == current_epoch_nr.
779 * If none, no P_BARRIER will be sent. */
780 unsigned current_epoch_writes
;
784 static inline bool has_net_conf(struct drbd_connection
*connection
)
789 has_net_conf
= rcu_dereference(connection
->net_conf
);
795 void __update_timing_details(
796 struct drbd_thread_timing_details
*tdp
,
799 const char *fn
, const unsigned int line
);
801 #define update_worker_timing_details(c, cb) \
802 __update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ )
803 #define update_receiver_timing_details(c, cb) \
804 __update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ )
806 struct submit_worker
{
807 struct workqueue_struct
*wq
;
808 struct work_struct worker
;
810 /* protected by ..->resource->req_lock */
811 struct list_head writes
;
814 struct drbd_peer_device
{
815 struct list_head peer_devices
;
816 struct drbd_device
*device
;
817 struct drbd_connection
*connection
;
818 struct work_struct send_acks_work
;
819 #ifdef CONFIG_DEBUG_FS
820 struct dentry
*debugfs_peer_dev
;
825 struct drbd_resource
*resource
;
826 struct list_head peer_devices
;
827 struct list_head pending_bitmap_io
;
829 unsigned long flush_jif
;
830 #ifdef CONFIG_DEBUG_FS
831 struct dentry
*debugfs_minor
;
832 struct dentry
*debugfs_vol
;
833 struct dentry
*debugfs_vol_oldest_requests
;
834 struct dentry
*debugfs_vol_act_log_extents
;
835 struct dentry
*debugfs_vol_resync_extents
;
836 struct dentry
*debugfs_vol_data_gen_id
;
837 struct dentry
*debugfs_vol_ed_gen_id
;
840 unsigned int vnr
; /* volume number within the connection */
841 unsigned int minor
; /* device minor number */
845 /* things that are stored as / read from meta data on disk */
848 /* configured by drbdsetup */
849 struct drbd_backing_dev
*ldev
__protected_by(local
);
851 sector_t p_size
; /* partner's disk size */
852 struct request_queue
*rq_queue
;
853 struct block_device
*this_bdev
;
854 struct gendisk
*vdisk
;
856 unsigned long last_reattach_jif
;
857 struct drbd_work resync_work
;
858 struct drbd_work unplug_work
;
859 struct timer_list resync_timer
;
860 struct timer_list md_sync_timer
;
861 struct timer_list start_resync_timer
;
862 struct timer_list request_timer
;
864 /* Used after attach while negotiating new disk state. */
865 union drbd_state new_state_tmp
;
867 union drbd_dev_state state
;
868 wait_queue_head_t misc_wait
;
869 wait_queue_head_t state_wait
; /* upon each state change. */
870 unsigned int send_cnt
;
871 unsigned int recv_cnt
;
872 unsigned int read_cnt
;
873 unsigned int writ_cnt
;
874 unsigned int al_writ_cnt
;
875 unsigned int bm_writ_cnt
;
876 atomic_t ap_bio_cnt
; /* Requests we need to complete */
877 atomic_t ap_actlog_cnt
; /* Requests waiting for activity log */
878 atomic_t ap_pending_cnt
; /* AP data packets on the wire, ack expected */
879 atomic_t rs_pending_cnt
; /* RS request/data packets on the wire */
880 atomic_t unacked_cnt
; /* Need to send replies for */
881 atomic_t local_cnt
; /* Waiting for local completion */
882 atomic_t suspend_cnt
;
884 /* Interval tree of pending local requests */
885 struct rb_root read_requests
;
886 struct rb_root write_requests
;
888 /* for statistics and timeouts */
889 /* [0] read, [1] write */
890 struct list_head pending_master_completion
[2];
891 struct list_head pending_completion
[2];
893 /* use checksums for *this* resync */
895 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
896 unsigned long rs_total
;
897 /* number of resync blocks that failed in this run */
898 unsigned long rs_failed
;
899 /* Syncer's start time [unit jiffies] */
900 unsigned long rs_start
;
901 /* cumulated time in PausedSyncX state [unit jiffies] */
902 unsigned long rs_paused
;
903 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
904 unsigned long rs_same_csum
;
905 #define DRBD_SYNC_MARKS 8
906 #define DRBD_SYNC_MARK_STEP (3*HZ)
907 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
908 unsigned long rs_mark_left
[DRBD_SYNC_MARKS
];
909 /* marks's time [unit jiffies] */
910 unsigned long rs_mark_time
[DRBD_SYNC_MARKS
];
911 /* current index into rs_mark_{left,time} */
913 unsigned long rs_last_bcast
; /* [unit jiffies] */
915 /* where does the admin want us to start? (sector) */
916 sector_t ov_start_sector
;
917 sector_t ov_stop_sector
;
918 /* where are we now? (sector) */
919 sector_t ov_position
;
920 /* Start sector of out of sync range (to merge printk reporting). */
921 sector_t ov_last_oos_start
;
922 /* size of out-of-sync range in sectors. */
923 sector_t ov_last_oos_size
;
924 unsigned long ov_left
; /* in bits */
926 struct drbd_bitmap
*bitmap
;
927 unsigned long bm_resync_fo
; /* bit offset for drbd_bm_find_next */
929 /* Used to track operations of resync... */
930 struct lru_cache
*resync
;
931 /* Number of locked elements in resync LRU */
932 unsigned int resync_locked
;
933 /* resync extent number waiting for application requests */
934 unsigned int resync_wenr
;
939 struct list_head active_ee
; /* IO in progress (P_DATA gets written to disk) */
940 struct list_head sync_ee
; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
941 struct list_head done_ee
; /* need to send P_WRITE_ACK */
942 struct list_head read_ee
; /* [RS]P_DATA_REQUEST being read */
943 struct list_head net_ee
; /* zero-copy network send in progress */
946 struct list_head resync_reads
;
947 atomic_t pp_in_use
; /* allocated from page pool */
948 atomic_t pp_in_use_by_net
; /* sendpage()d, still referenced by tcp */
949 wait_queue_head_t ee_wait
;
950 struct drbd_md_io md_io
;
952 wait_queue_head_t al_wait
;
953 struct lru_cache
*act_log
; /* activity log */
954 unsigned int al_tr_number
;
956 wait_queue_head_t seq_wait
;
958 unsigned int peer_seq
;
959 spinlock_t peer_seq_lock
;
960 unsigned long comm_bm_set
; /* communicated number of set bits. */
961 struct bm_io_work bm_io_work
;
962 u64 ed_uuid
; /* UUID of the exposed data */
963 struct mutex own_state_mutex
;
964 struct mutex
*state_mutex
; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */
965 char congestion_reason
; /* Why we where congested... */
966 atomic_t rs_sect_in
; /* for incoming resync data rate, SyncTarget */
967 atomic_t rs_sect_ev
; /* for submitted resync data rate, both */
968 int rs_last_sect_ev
; /* counter to compare with */
969 int rs_last_events
; /* counter of read or write "events" (unit sectors)
970 * on the lower level device when we last looked. */
971 int c_sync_rate
; /* current resync rate after syncer throttle magic */
972 struct fifo_buffer
*rs_plan_s
; /* correction values of resync planer (RCU, connection->conn_update) */
973 int rs_in_flight
; /* resync sectors in flight (to proxy, in proxy and from proxy) */
974 atomic_t ap_in_flight
; /* App sectors in flight (waiting for ack) */
975 unsigned int peer_max_bio_size
;
976 unsigned int local_max_bio_size
;
978 /* any requests that would block in drbd_make_request()
979 * are deferred to this single-threaded work queue */
980 struct submit_worker submit
;
983 struct drbd_bm_aio_ctx
{
984 struct drbd_device
*device
;
985 struct list_head list
; /* on device->pending_bitmap_io */;
986 unsigned long start_jif
;
990 #define BM_AIO_COPY_PAGES 1
991 #define BM_AIO_WRITE_HINTED 2
992 #define BM_AIO_WRITE_ALL_PAGES 4
993 #define BM_AIO_READ 8
998 struct drbd_config_context
{
999 /* assigned from drbd_genlmsghdr */
1001 /* assigned from request attributes, if present */
1002 unsigned int volume
;
1003 #define VOLUME_UNSPECIFIED (-1U)
1004 /* pointer into the request skb,
1005 * limited lifetime! */
1006 char *resource_name
;
1007 struct nlattr
*my_addr
;
1008 struct nlattr
*peer_addr
;
1011 struct sk_buff
*reply_skb
;
1012 /* pointer into reply buffer */
1013 struct drbd_genlmsghdr
*reply_dh
;
1014 /* resolved from attributes, if possible */
1015 struct drbd_device
*device
;
1016 struct drbd_resource
*resource
;
1017 struct drbd_connection
*connection
;
1020 static inline struct drbd_device
*minor_to_device(unsigned int minor
)
1022 return (struct drbd_device
*)idr_find(&drbd_devices
, minor
);
1025 static inline struct drbd_peer_device
*first_peer_device(struct drbd_device
*device
)
1027 return list_first_entry_or_null(&device
->peer_devices
, struct drbd_peer_device
, peer_devices
);
1030 static inline struct drbd_peer_device
*
1031 conn_peer_device(struct drbd_connection
*connection
, int volume_number
)
1033 return idr_find(&connection
->peer_devices
, volume_number
);
1036 #define for_each_resource(resource, _resources) \
1037 list_for_each_entry(resource, _resources, resources)
1039 #define for_each_resource_rcu(resource, _resources) \
1040 list_for_each_entry_rcu(resource, _resources, resources)
1042 #define for_each_resource_safe(resource, tmp, _resources) \
1043 list_for_each_entry_safe(resource, tmp, _resources, resources)
1045 #define for_each_connection(connection, resource) \
1046 list_for_each_entry(connection, &resource->connections, connections)
1048 #define for_each_connection_rcu(connection, resource) \
1049 list_for_each_entry_rcu(connection, &resource->connections, connections)
1051 #define for_each_connection_safe(connection, tmp, resource) \
1052 list_for_each_entry_safe(connection, tmp, &resource->connections, connections)
1054 #define for_each_peer_device(peer_device, device) \
1055 list_for_each_entry(peer_device, &device->peer_devices, peer_devices)
1057 #define for_each_peer_device_rcu(peer_device, device) \
1058 list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices)
1060 #define for_each_peer_device_safe(peer_device, tmp, device) \
1061 list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices)
1063 static inline unsigned int device_to_minor(struct drbd_device
*device
)
1065 return device
->minor
;
1069 * function declarations
1070 *************************/
1076 DDSF_NO_RESYNC
= 2, /* Do not run a resync for the new space */
1079 extern void drbd_init_set_defaults(struct drbd_device
*device
);
1080 extern int drbd_thread_start(struct drbd_thread
*thi
);
1081 extern void _drbd_thread_stop(struct drbd_thread
*thi
, int restart
, int wait
);
1083 extern void drbd_thread_current_set_cpu(struct drbd_thread
*thi
);
1085 #define drbd_thread_current_set_cpu(A) ({})
1087 extern void tl_release(struct drbd_connection
*, unsigned int barrier_nr
,
1088 unsigned int set_size
);
1089 extern void tl_clear(struct drbd_connection
*);
1090 extern void drbd_free_sock(struct drbd_connection
*connection
);
1091 extern int drbd_send(struct drbd_connection
*connection
, struct socket
*sock
,
1092 void *buf
, size_t size
, unsigned msg_flags
);
1093 extern int drbd_send_all(struct drbd_connection
*, struct socket
*, void *, size_t,
1096 extern int __drbd_send_protocol(struct drbd_connection
*connection
, enum drbd_packet cmd
);
1097 extern int drbd_send_protocol(struct drbd_connection
*connection
);
1098 extern int drbd_send_uuids(struct drbd_peer_device
*);
1099 extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device
*);
1100 extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device
*);
1101 extern int drbd_send_sizes(struct drbd_peer_device
*, int trigger_reply
, enum dds_flags flags
);
1102 extern int drbd_send_state(struct drbd_peer_device
*, union drbd_state s
);
1103 extern int drbd_send_current_state(struct drbd_peer_device
*);
1104 extern int drbd_send_sync_param(struct drbd_peer_device
*);
1105 extern void drbd_send_b_ack(struct drbd_connection
*connection
, u32 barrier_nr
,
1107 extern int drbd_send_ack(struct drbd_peer_device
*, enum drbd_packet
,
1108 struct drbd_peer_request
*);
1109 extern void drbd_send_ack_rp(struct drbd_peer_device
*, enum drbd_packet
,
1110 struct p_block_req
*rp
);
1111 extern void drbd_send_ack_dp(struct drbd_peer_device
*, enum drbd_packet
,
1112 struct p_data
*dp
, int data_size
);
1113 extern int drbd_send_ack_ex(struct drbd_peer_device
*, enum drbd_packet
,
1114 sector_t sector
, int blksize
, u64 block_id
);
1115 extern int drbd_send_out_of_sync(struct drbd_peer_device
*, struct drbd_request
*);
1116 extern int drbd_send_block(struct drbd_peer_device
*, enum drbd_packet
,
1117 struct drbd_peer_request
*);
1118 extern int drbd_send_dblock(struct drbd_peer_device
*, struct drbd_request
*req
);
1119 extern int drbd_send_drequest(struct drbd_peer_device
*, int cmd
,
1120 sector_t sector
, int size
, u64 block_id
);
1121 extern int drbd_send_drequest_csum(struct drbd_peer_device
*, sector_t sector
,
1122 int size
, void *digest
, int digest_size
,
1123 enum drbd_packet cmd
);
1124 extern int drbd_send_ov_request(struct drbd_peer_device
*, sector_t sector
, int size
);
1126 extern int drbd_send_bitmap(struct drbd_device
*device
);
1127 extern void drbd_send_sr_reply(struct drbd_peer_device
*, enum drbd_state_rv retcode
);
1128 extern void conn_send_sr_reply(struct drbd_connection
*connection
, enum drbd_state_rv retcode
);
1129 extern void drbd_backing_dev_free(struct drbd_device
*device
, struct drbd_backing_dev
*ldev
);
1130 extern void drbd_device_cleanup(struct drbd_device
*device
);
1131 void drbd_print_uuids(struct drbd_device
*device
, const char *text
);
1133 extern void conn_md_sync(struct drbd_connection
*connection
);
1134 extern void drbd_md_write(struct drbd_device
*device
, void *buffer
);
1135 extern void drbd_md_sync(struct drbd_device
*device
);
1136 extern int drbd_md_read(struct drbd_device
*device
, struct drbd_backing_dev
*bdev
);
1137 extern void drbd_uuid_set(struct drbd_device
*device
, int idx
, u64 val
) __must_hold(local
);
1138 extern void _drbd_uuid_set(struct drbd_device
*device
, int idx
, u64 val
) __must_hold(local
);
1139 extern void drbd_uuid_new_current(struct drbd_device
*device
) __must_hold(local
);
1140 extern void drbd_uuid_set_bm(struct drbd_device
*device
, u64 val
) __must_hold(local
);
1141 extern void drbd_uuid_move_history(struct drbd_device
*device
) __must_hold(local
);
1142 extern void __drbd_uuid_set(struct drbd_device
*device
, int idx
, u64 val
) __must_hold(local
);
1143 extern void drbd_md_set_flag(struct drbd_device
*device
, int flags
) __must_hold(local
);
1144 extern void drbd_md_clear_flag(struct drbd_device
*device
, int flags
)__must_hold(local
);
1145 extern int drbd_md_test_flag(struct drbd_backing_dev
*, int);
1146 extern void drbd_md_mark_dirty(struct drbd_device
*device
);
1147 extern void drbd_queue_bitmap_io(struct drbd_device
*device
,
1148 int (*io_fn
)(struct drbd_device
*),
1149 void (*done
)(struct drbd_device
*, int),
1150 char *why
, enum bm_flag flags
);
1151 extern int drbd_bitmap_io(struct drbd_device
*device
,
1152 int (*io_fn
)(struct drbd_device
*),
1153 char *why
, enum bm_flag flags
);
1154 extern int drbd_bitmap_io_from_worker(struct drbd_device
*device
,
1155 int (*io_fn
)(struct drbd_device
*),
1156 char *why
, enum bm_flag flags
);
1157 extern int drbd_bmio_set_n_write(struct drbd_device
*device
) __must_hold(local
);
1158 extern int drbd_bmio_clear_n_write(struct drbd_device
*device
) __must_hold(local
);
1162 * We currently have two possible layouts.
1163 * Offsets in (512 byte) sectors.
1165 * |----------- md_size_sect ------------------|
1166 * [ 4k superblock ][ activity log ][ Bitmap ]
1167 * | al_offset == 8 |
1168 * | bm_offset = al_offset + X |
1169 * ==> bitmap sectors = md_size_sect - bm_offset
1172 * old, indexed fixed size meta data:
1175 * |----------- md_size_sect ------------------|
1176 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*]
1178 * | bm_offset = al_offset - Y |
1179 * ==> bitmap sectors = Y = al_offset - bm_offset
1181 * [padding*] are zero or up to 7 unused 512 Byte sectors to the
1182 * end of the device, so that the [4k superblock] will be 4k aligned.
1184 * The activity log consists of 4k transaction blocks,
1185 * which are written in a ring-buffer, or striped ring-buffer like fashion,
1186 * which are writtensize used to be fixed 32kB,
1187 * but is about to become configurable.
1190 /* Our old fixed size meta data layout
1191 * allows up to about 3.8TB, so if you want more,
1192 * you need to use the "flexible" meta data format. */
1193 #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */
1194 #define MD_4kB_SECT 8
1195 #define MD_32kB_SECT 64
1197 /* One activity log extent represents 4M of storage */
1198 #define AL_EXTENT_SHIFT 22
1199 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1201 /* We could make these currently hardcoded constants configurable
1202 * variables at create-md time (or even re-configurable at runtime?).
1203 * Which will require some more changes to the DRBD "super block"
1206 * updates per transaction:
1207 * This many changes to the active set can be logged with one transaction.
1208 * This number is arbitrary.
1209 * context per transaction:
1210 * This many context extent numbers are logged with each transaction.
1211 * This number is resulting from the transaction block size (4k), the layout
1212 * of the transaction header, and the number of updates per transaction.
1213 * See drbd_actlog.c:struct al_transaction_on_disk
1215 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1216 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1218 #if BITS_PER_LONG == 32
1220 #define cpu_to_lel(A) cpu_to_le32(A)
1221 #define lel_to_cpu(A) le32_to_cpu(A)
1222 #elif BITS_PER_LONG == 64
1224 #define cpu_to_lel(A) cpu_to_le64(A)
1225 #define lel_to_cpu(A) le64_to_cpu(A)
1227 #error "LN2 of BITS_PER_LONG unknown!"
1231 /* 16MB sized 'bitmap extent' to track syncer usage */
1233 int rs_left
; /* number of bits set (out of sync) in this extent. */
1234 int rs_failed
; /* number of failed resync requests in this extent. */
1235 unsigned long flags
;
1236 struct lc_element lce
;
1239 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1240 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1241 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1245 * We need to store one bit for a block.
1246 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1247 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1248 * Bit 1 ==> local node thinks this block needs to be synced.
1251 #define SLEEP_TIME (HZ/10)
1253 /* We do bitmap IO in units of 4k blocks.
1254 * We also still have a hardcoded 4k per bit relation. */
1255 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1256 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1257 /* mostly arbitrarily set the represented size of one bitmap extent,
1258 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1259 * at 4k per bit resolution) */
1260 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
1261 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1263 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1264 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1267 /* thus many _storage_ sectors are described by one bit */
1268 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1269 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1270 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1272 /* bit to represented kilo byte conversion */
1273 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1275 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1276 * _storage_ sector is located in */
1277 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1278 #define BM_BIT_TO_EXT(x) ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1280 /* first storage sector a bitmap extent corresponds to */
1281 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1282 /* how much _storage_ sectors we have per bitmap extent */
1283 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1284 /* how many bits are covered by one bitmap extent (resync extent) */
1285 #define BM_BITS_PER_EXT (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1287 #define BM_BLOCKS_PER_BM_EXT_MASK (BM_BITS_PER_EXT - 1)
1290 /* in one sector of the bitmap, we have this many activity_log extents. */
1291 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1293 /* the extent in "PER_EXTENT" below is an activity log extent
1294 * we need that many (long words/bytes) to store the bitmap
1295 * of one AL_EXTENT_SIZE chunk of storage.
1296 * we can store the bitmap for that many AL_EXTENTS within
1297 * one sector of the _on_disk_ bitmap:
1298 * bit 0 bit 37 bit 38 bit (512*8)-1
1299 * ...|........|........|.. // ..|........|
1300 * sect. 0 `296 `304 ^(512*8*8)-1
1302 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1303 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1304 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1307 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1308 /* we have a certain meta data variant that has a fixed on-disk size of 128
1309 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity
1310 * log, leaving this many sectors for the bitmap.
1313 #define DRBD_MAX_SECTORS_FIXED_BM \
1314 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9)))
1315 #if !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1316 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1317 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1319 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM
1320 /* 16 TB in units of sectors */
1321 #if BITS_PER_LONG == 32
1322 /* adjust by one page worth of bitmap,
1323 * so we won't wrap around in drbd_bm_find_next_bit.
1324 * you should use 64bit OS for that much storage, anyways. */
1325 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1327 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1328 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1329 /* corresponds to (1UL << 38) bits right now. */
1333 /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE,
1334 * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte.
1335 * Since we may live in a mixed-platform cluster,
1336 * we limit us to a platform agnostic constant here for now.
1337 * A followup commit may allow even bigger BIO sizes,
1338 * once we thought that through. */
1339 #define DRBD_MAX_BIO_SIZE (1U << 20)
1340 #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1341 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1343 #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */
1345 #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */
1346 #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */
1348 /* For now, don't allow more than one activity log extent worth of data
1349 * to be discarded in one go. We may need to rework drbd_al_begin_io()
1350 * to allow for even larger discard ranges */
1351 #define DRBD_MAX_DISCARD_SIZE AL_EXTENT_SIZE
1352 #define DRBD_MAX_DISCARD_SECTORS (DRBD_MAX_DISCARD_SIZE >> 9)
1354 extern int drbd_bm_init(struct drbd_device
*device
);
1355 extern int drbd_bm_resize(struct drbd_device
*device
, sector_t sectors
, int set_new_bits
);
1356 extern void drbd_bm_cleanup(struct drbd_device
*device
);
1357 extern void drbd_bm_set_all(struct drbd_device
*device
);
1358 extern void drbd_bm_clear_all(struct drbd_device
*device
);
1359 /* set/clear/test only a few bits at a time */
1360 extern int drbd_bm_set_bits(
1361 struct drbd_device
*device
, unsigned long s
, unsigned long e
);
1362 extern int drbd_bm_clear_bits(
1363 struct drbd_device
*device
, unsigned long s
, unsigned long e
);
1364 extern int drbd_bm_count_bits(
1365 struct drbd_device
*device
, const unsigned long s
, const unsigned long e
);
1366 /* bm_set_bits variant for use while holding drbd_bm_lock,
1367 * may process the whole bitmap in one go */
1368 extern void _drbd_bm_set_bits(struct drbd_device
*device
,
1369 const unsigned long s
, const unsigned long e
);
1370 extern int drbd_bm_test_bit(struct drbd_device
*device
, unsigned long bitnr
);
1371 extern int drbd_bm_e_weight(struct drbd_device
*device
, unsigned long enr
);
1372 extern int drbd_bm_read(struct drbd_device
*device
) __must_hold(local
);
1373 extern void drbd_bm_mark_for_writeout(struct drbd_device
*device
, int page_nr
);
1374 extern int drbd_bm_write(struct drbd_device
*device
) __must_hold(local
);
1375 extern int drbd_bm_write_hinted(struct drbd_device
*device
) __must_hold(local
);
1376 extern int drbd_bm_write_lazy(struct drbd_device
*device
, unsigned upper_idx
) __must_hold(local
);
1377 extern int drbd_bm_write_all(struct drbd_device
*device
) __must_hold(local
);
1378 extern int drbd_bm_write_copy_pages(struct drbd_device
*device
) __must_hold(local
);
1379 extern size_t drbd_bm_words(struct drbd_device
*device
);
1380 extern unsigned long drbd_bm_bits(struct drbd_device
*device
);
1381 extern sector_t
drbd_bm_capacity(struct drbd_device
*device
);
1383 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1384 extern unsigned long drbd_bm_find_next(struct drbd_device
*device
, unsigned long bm_fo
);
1385 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1386 extern unsigned long _drbd_bm_find_next(struct drbd_device
*device
, unsigned long bm_fo
);
1387 extern unsigned long _drbd_bm_find_next_zero(struct drbd_device
*device
, unsigned long bm_fo
);
1388 extern unsigned long _drbd_bm_total_weight(struct drbd_device
*device
);
1389 extern unsigned long drbd_bm_total_weight(struct drbd_device
*device
);
1390 /* for receive_bitmap */
1391 extern void drbd_bm_merge_lel(struct drbd_device
*device
, size_t offset
,
1392 size_t number
, unsigned long *buffer
);
1393 /* for _drbd_send_bitmap */
1394 extern void drbd_bm_get_lel(struct drbd_device
*device
, size_t offset
,
1395 size_t number
, unsigned long *buffer
);
1397 extern void drbd_bm_lock(struct drbd_device
*device
, char *why
, enum bm_flag flags
);
1398 extern void drbd_bm_unlock(struct drbd_device
*device
);
1401 extern struct kmem_cache
*drbd_request_cache
;
1402 extern struct kmem_cache
*drbd_ee_cache
; /* peer requests */
1403 extern struct kmem_cache
*drbd_bm_ext_cache
; /* bitmap extents */
1404 extern struct kmem_cache
*drbd_al_ext_cache
; /* activity log extents */
1405 extern mempool_t
*drbd_request_mempool
;
1406 extern mempool_t
*drbd_ee_mempool
;
1408 /* drbd's page pool, used to buffer data received from the peer,
1409 * or data requested by the peer.
1411 * This does not have an emergency reserve.
1413 * When allocating from this pool, it first takes pages from the pool.
1414 * Only if the pool is depleted will try to allocate from the system.
1416 * The assumption is that pages taken from this pool will be processed,
1417 * and given back, "quickly", and then can be recycled, so we can avoid
1418 * frequent calls to alloc_page(), and still will be able to make progress even
1419 * under memory pressure.
1421 extern struct page
*drbd_pp_pool
;
1422 extern spinlock_t drbd_pp_lock
;
1423 extern int drbd_pp_vacant
;
1424 extern wait_queue_head_t drbd_pp_wait
;
1426 /* We also need a standard (emergency-reserve backed) page pool
1427 * for meta data IO (activity log, bitmap).
1428 * We can keep it global, as long as it is used as "N pages at a time".
1429 * 128 should be plenty, currently we probably can get away with as few as 1.
1431 #define DRBD_MIN_POOL_PAGES 128
1432 extern mempool_t
*drbd_md_io_page_pool
;
1434 /* We also need to make sure we get a bio
1435 * when we need it for housekeeping purposes */
1436 extern struct bio_set
*drbd_md_io_bio_set
;
1437 /* to allocate from that set */
1438 extern struct bio
*bio_alloc_drbd(gfp_t gfp_mask
);
1440 extern struct mutex resources_mutex
;
1442 extern int conn_lowest_minor(struct drbd_connection
*connection
);
1443 extern enum drbd_ret_code
drbd_create_device(struct drbd_config_context
*adm_ctx
, unsigned int minor
);
1444 extern void drbd_destroy_device(struct kref
*kref
);
1445 extern void drbd_delete_device(struct drbd_device
*device
);
1447 extern struct drbd_resource
*drbd_create_resource(const char *name
);
1448 extern void drbd_free_resource(struct drbd_resource
*resource
);
1450 extern int set_resource_options(struct drbd_resource
*resource
, struct res_opts
*res_opts
);
1451 extern struct drbd_connection
*conn_create(const char *name
, struct res_opts
*res_opts
);
1452 extern void drbd_destroy_connection(struct kref
*kref
);
1453 extern struct drbd_connection
*conn_get_by_addrs(void *my_addr
, int my_addr_len
,
1454 void *peer_addr
, int peer_addr_len
);
1455 extern struct drbd_resource
*drbd_find_resource(const char *name
);
1456 extern void drbd_destroy_resource(struct kref
*kref
);
1457 extern void conn_free_crypto(struct drbd_connection
*connection
);
1459 extern int proc_details
;
1462 extern void do_submit(struct work_struct
*ws
);
1463 extern void __drbd_make_request(struct drbd_device
*, struct bio
*, unsigned long);
1464 extern blk_qc_t
drbd_make_request(struct request_queue
*q
, struct bio
*bio
);
1465 extern int drbd_read_remote(struct drbd_device
*device
, struct drbd_request
*req
);
1466 extern int is_valid_ar_handle(struct drbd_request
*, sector_t
);
1471 extern struct mutex notification_mutex
;
1473 extern void drbd_suspend_io(struct drbd_device
*device
);
1474 extern void drbd_resume_io(struct drbd_device
*device
);
1475 extern char *ppsize(char *buf
, unsigned long long size
);
1476 extern sector_t
drbd_new_dev_size(struct drbd_device
*, struct drbd_backing_dev
*, sector_t
, int);
1477 enum determine_dev_size
{
1478 DS_ERROR_SHRINK
= -3,
1479 DS_ERROR_SPACE_MD
= -2,
1484 DS_GREW_FROM_ZERO
= 3,
1486 extern enum determine_dev_size
1487 drbd_determine_dev_size(struct drbd_device
*, enum dds_flags
, struct resize_parms
*) __must_hold(local
);
1488 extern void resync_after_online_grow(struct drbd_device
*);
1489 extern void drbd_reconsider_max_bio_size(struct drbd_device
*device
, struct drbd_backing_dev
*bdev
);
1490 extern enum drbd_state_rv
drbd_set_role(struct drbd_device
*device
,
1491 enum drbd_role new_role
,
1493 extern bool conn_try_outdate_peer(struct drbd_connection
*connection
);
1494 extern void conn_try_outdate_peer_async(struct drbd_connection
*connection
);
1495 extern int drbd_khelper(struct drbd_device
*device
, char *cmd
);
1498 /* bi_end_io handlers */
1499 extern void drbd_md_endio(struct bio
*bio
);
1500 extern void drbd_peer_request_endio(struct bio
*bio
);
1501 extern void drbd_request_endio(struct bio
*bio
);
1502 extern int drbd_worker(struct drbd_thread
*thi
);
1503 enum drbd_ret_code
drbd_resync_after_valid(struct drbd_device
*device
, int o_minor
);
1504 void drbd_resync_after_changed(struct drbd_device
*device
);
1505 extern void drbd_start_resync(struct drbd_device
*device
, enum drbd_conns side
);
1506 extern void resume_next_sg(struct drbd_device
*device
);
1507 extern void suspend_other_sg(struct drbd_device
*device
);
1508 extern int drbd_resync_finished(struct drbd_device
*device
);
1509 /* maybe rather drbd_main.c ? */
1510 extern void *drbd_md_get_buffer(struct drbd_device
*device
, const char *intent
);
1511 extern void drbd_md_put_buffer(struct drbd_device
*device
);
1512 extern int drbd_md_sync_page_io(struct drbd_device
*device
,
1513 struct drbd_backing_dev
*bdev
, sector_t sector
, int rw
);
1514 extern void drbd_ov_out_of_sync_found(struct drbd_device
*, sector_t
, int);
1515 extern void wait_until_done_or_force_detached(struct drbd_device
*device
,
1516 struct drbd_backing_dev
*bdev
, unsigned int *done
);
1517 extern void drbd_rs_controller_reset(struct drbd_device
*device
);
1519 static inline void ov_out_of_sync_print(struct drbd_device
*device
)
1521 if (device
->ov_last_oos_size
) {
1522 drbd_err(device
, "Out of sync: start=%llu, size=%lu (sectors)\n",
1523 (unsigned long long)device
->ov_last_oos_start
,
1524 (unsigned long)device
->ov_last_oos_size
);
1526 device
->ov_last_oos_size
= 0;
1530 extern void drbd_csum_bio(struct crypto_hash
*, struct bio
*, void *);
1531 extern void drbd_csum_ee(struct crypto_hash
*, struct drbd_peer_request
*, void *);
1532 /* worker callbacks */
1533 extern int w_e_end_data_req(struct drbd_work
*, int);
1534 extern int w_e_end_rsdata_req(struct drbd_work
*, int);
1535 extern int w_e_end_csum_rs_req(struct drbd_work
*, int);
1536 extern int w_e_end_ov_reply(struct drbd_work
*, int);
1537 extern int w_e_end_ov_req(struct drbd_work
*, int);
1538 extern int w_ov_finished(struct drbd_work
*, int);
1539 extern int w_resync_timer(struct drbd_work
*, int);
1540 extern int w_send_write_hint(struct drbd_work
*, int);
1541 extern int w_send_dblock(struct drbd_work
*, int);
1542 extern int w_send_read_req(struct drbd_work
*, int);
1543 extern int w_e_reissue(struct drbd_work
*, int);
1544 extern int w_restart_disk_io(struct drbd_work
*, int);
1545 extern int w_send_out_of_sync(struct drbd_work
*, int);
1546 extern int w_start_resync(struct drbd_work
*, int);
1548 extern void resync_timer_fn(unsigned long data
);
1549 extern void start_resync_timer_fn(unsigned long data
);
1551 extern void drbd_endio_write_sec_final(struct drbd_peer_request
*peer_req
);
1553 /* drbd_receiver.c */
1554 extern int drbd_receiver(struct drbd_thread
*thi
);
1555 extern int drbd_ack_receiver(struct drbd_thread
*thi
);
1556 extern void drbd_send_ping_wf(struct work_struct
*ws
);
1557 extern void drbd_send_acks_wf(struct work_struct
*ws
);
1558 extern bool drbd_rs_c_min_rate_throttle(struct drbd_device
*device
);
1559 extern bool drbd_rs_should_slow_down(struct drbd_device
*device
, sector_t sector
,
1560 bool throttle_if_app_is_waiting
);
1561 extern int drbd_submit_peer_request(struct drbd_device
*,
1562 struct drbd_peer_request
*, const unsigned,
1564 extern int drbd_free_peer_reqs(struct drbd_device
*, struct list_head
*);
1565 extern struct drbd_peer_request
*drbd_alloc_peer_req(struct drbd_peer_device
*, u64
,
1566 sector_t
, unsigned int,
1568 gfp_t
) __must_hold(local
);
1569 extern void __drbd_free_peer_req(struct drbd_device
*, struct drbd_peer_request
*,
1571 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1572 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
1573 extern struct page
*drbd_alloc_pages(struct drbd_peer_device
*, unsigned int, bool);
1574 extern void drbd_set_recv_tcq(struct drbd_device
*device
, int tcq_enabled
);
1575 extern void _drbd_clear_done_ee(struct drbd_device
*device
, struct list_head
*to_be_freed
);
1576 extern int drbd_connected(struct drbd_peer_device
*);
1578 static inline void drbd_tcp_cork(struct socket
*sock
)
1581 (void) kernel_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1582 (char*)&val
, sizeof(val
));
1585 static inline void drbd_tcp_uncork(struct socket
*sock
)
1588 (void) kernel_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1589 (char*)&val
, sizeof(val
));
1592 static inline void drbd_tcp_nodelay(struct socket
*sock
)
1595 (void) kernel_setsockopt(sock
, SOL_TCP
, TCP_NODELAY
,
1596 (char*)&val
, sizeof(val
));
1599 static inline void drbd_tcp_quickack(struct socket
*sock
)
1602 (void) kernel_setsockopt(sock
, SOL_TCP
, TCP_QUICKACK
,
1603 (char*)&val
, sizeof(val
));
1606 /* sets the number of 512 byte sectors of our virtual device */
1607 static inline void drbd_set_my_capacity(struct drbd_device
*device
,
1610 /* set_capacity(device->this_bdev->bd_disk, size); */
1611 set_capacity(device
->vdisk
, size
);
1612 device
->this_bdev
->bd_inode
->i_size
= (loff_t
)size
<< 9;
1616 * used to submit our private bio
1618 static inline void drbd_generic_make_request(struct drbd_device
*device
,
1619 int fault_type
, struct bio
*bio
)
1622 if (!bio
->bi_bdev
) {
1623 drbd_err(device
, "drbd_generic_make_request: bio->bi_bdev == NULL\n");
1624 bio
->bi_error
= -ENODEV
;
1629 if (drbd_insert_fault(device
, fault_type
))
1632 generic_make_request(bio
);
1635 void drbd_bump_write_ordering(struct drbd_resource
*resource
, struct drbd_backing_dev
*bdev
,
1636 enum write_ordering_e wo
);
1639 extern struct proc_dir_entry
*drbd_proc
;
1640 extern const struct file_operations drbd_proc_fops
;
1641 extern const char *drbd_conn_str(enum drbd_conns s
);
1642 extern const char *drbd_role_str(enum drbd_role s
);
1645 extern bool drbd_al_begin_io_prepare(struct drbd_device
*device
, struct drbd_interval
*i
);
1646 extern int drbd_al_begin_io_nonblock(struct drbd_device
*device
, struct drbd_interval
*i
);
1647 extern void drbd_al_begin_io_commit(struct drbd_device
*device
);
1648 extern bool drbd_al_begin_io_fastpath(struct drbd_device
*device
, struct drbd_interval
*i
);
1649 extern void drbd_al_begin_io(struct drbd_device
*device
, struct drbd_interval
*i
);
1650 extern void drbd_al_complete_io(struct drbd_device
*device
, struct drbd_interval
*i
);
1651 extern void drbd_rs_complete_io(struct drbd_device
*device
, sector_t sector
);
1652 extern int drbd_rs_begin_io(struct drbd_device
*device
, sector_t sector
);
1653 extern int drbd_try_rs_begin_io(struct drbd_device
*device
, sector_t sector
);
1654 extern void drbd_rs_cancel_all(struct drbd_device
*device
);
1655 extern int drbd_rs_del_all(struct drbd_device
*device
);
1656 extern void drbd_rs_failed_io(struct drbd_device
*device
,
1657 sector_t sector
, int size
);
1658 extern void drbd_advance_rs_marks(struct drbd_device
*device
, unsigned long still_to_go
);
1660 enum update_sync_bits_mode
{ RECORD_RS_FAILED
, SET_OUT_OF_SYNC
, SET_IN_SYNC
};
1661 extern int __drbd_change_sync(struct drbd_device
*device
, sector_t sector
, int size
,
1662 enum update_sync_bits_mode mode
);
1663 #define drbd_set_in_sync(device, sector, size) \
1664 __drbd_change_sync(device, sector, size, SET_IN_SYNC)
1665 #define drbd_set_out_of_sync(device, sector, size) \
1666 __drbd_change_sync(device, sector, size, SET_OUT_OF_SYNC)
1667 #define drbd_rs_failed_io(device, sector, size) \
1668 __drbd_change_sync(device, sector, size, RECORD_RS_FAILED)
1669 extern void drbd_al_shrink(struct drbd_device
*device
);
1670 extern int drbd_al_initialize(struct drbd_device
*, void *);
1673 /* state info broadcast */
1675 enum drbd_state_info_bcast_reason sib_reason
;
1679 unsigned helper_exit_code
;
1682 union drbd_state os
;
1683 union drbd_state ns
;
1687 void drbd_bcast_event(struct drbd_device
*device
, const struct sib_info
*sib
);
1689 extern void notify_resource_state(struct sk_buff
*,
1691 struct drbd_resource
*,
1692 struct resource_info
*,
1693 enum drbd_notification_type
);
1694 extern void notify_device_state(struct sk_buff
*,
1696 struct drbd_device
*,
1697 struct device_info
*,
1698 enum drbd_notification_type
);
1699 extern void notify_connection_state(struct sk_buff
*,
1701 struct drbd_connection
*,
1702 struct connection_info
*,
1703 enum drbd_notification_type
);
1704 extern void notify_peer_device_state(struct sk_buff
*,
1706 struct drbd_peer_device
*,
1707 struct peer_device_info
*,
1708 enum drbd_notification_type
);
1709 extern void notify_helper(enum drbd_notification_type
, struct drbd_device
*,
1710 struct drbd_connection
*, const char *, int);
1713 * inline helper functions
1714 *************************/
1716 /* see also page_chain_add and friends in drbd_receiver.c */
1717 static inline struct page
*page_chain_next(struct page
*page
)
1719 return (struct page
*)page_private(page
);
1721 #define page_chain_for_each(page) \
1722 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1723 page = page_chain_next(page))
1724 #define page_chain_for_each_safe(page, n) \
1725 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1728 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request
*peer_req
)
1730 struct page
*page
= peer_req
->pages
;
1731 page_chain_for_each(page
) {
1732 if (page_count(page
) > 1)
1738 static inline union drbd_state
drbd_read_state(struct drbd_device
*device
)
1740 struct drbd_resource
*resource
= device
->resource
;
1741 union drbd_state rv
;
1743 rv
.i
= device
->state
.i
;
1744 rv
.susp
= resource
->susp
;
1745 rv
.susp_nod
= resource
->susp_nod
;
1746 rv
.susp_fen
= resource
->susp_fen
;
1751 enum drbd_force_detach_flags
{
1758 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1759 static inline void __drbd_chk_io_error_(struct drbd_device
*device
,
1760 enum drbd_force_detach_flags df
,
1763 enum drbd_io_error_p ep
;
1766 ep
= rcu_dereference(device
->ldev
->disk_conf
)->on_io_error
;
1769 case EP_PASS_ON
: /* FIXME would this be better named "Ignore"? */
1770 if (df
== DRBD_READ_ERROR
|| df
== DRBD_WRITE_ERROR
) {
1771 if (__ratelimit(&drbd_ratelimit_state
))
1772 drbd_err(device
, "Local IO failed in %s.\n", where
);
1773 if (device
->state
.disk
> D_INCONSISTENT
)
1774 _drbd_set_state(_NS(device
, disk
, D_INCONSISTENT
), CS_HARD
, NULL
);
1777 /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */
1779 case EP_CALL_HELPER
:
1780 /* Remember whether we saw a READ or WRITE error.
1782 * Recovery of the affected area for WRITE failure is covered
1783 * by the activity log.
1784 * READ errors may fall outside that area though. Certain READ
1785 * errors can be "healed" by writing good data to the affected
1786 * blocks, which triggers block re-allocation in lower layers.
1788 * If we can not write the bitmap after a READ error,
1789 * we may need to trigger a full sync (see w_go_diskless()).
1791 * Force-detach is not really an IO error, but rather a
1792 * desperate measure to try to deal with a completely
1793 * unresponsive lower level IO stack.
1794 * Still it should be treated as a WRITE error.
1796 * Meta IO error is always WRITE error:
1797 * we read meta data only once during attach,
1798 * which will fail in case of errors.
1800 set_bit(WAS_IO_ERROR
, &device
->flags
);
1801 if (df
== DRBD_READ_ERROR
)
1802 set_bit(WAS_READ_ERROR
, &device
->flags
);
1803 if (df
== DRBD_FORCE_DETACH
)
1804 set_bit(FORCE_DETACH
, &device
->flags
);
1805 if (device
->state
.disk
> D_FAILED
) {
1806 _drbd_set_state(_NS(device
, disk
, D_FAILED
), CS_HARD
, NULL
);
1808 "Local IO failed in %s. Detaching...\n", where
);
1815 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1816 * @device: DRBD device.
1817 * @error: Error code passed to the IO completion callback
1818 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1820 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1822 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1823 static inline void drbd_chk_io_error_(struct drbd_device
*device
,
1824 int error
, enum drbd_force_detach_flags forcedetach
, const char *where
)
1827 unsigned long flags
;
1828 spin_lock_irqsave(&device
->resource
->req_lock
, flags
);
1829 __drbd_chk_io_error_(device
, forcedetach
, where
);
1830 spin_unlock_irqrestore(&device
->resource
->req_lock
, flags
);
1836 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1837 * @bdev: Meta data block device.
1839 * BTW, for internal meta data, this happens to be the maximum capacity
1840 * we could agree upon with our peer node.
1842 static inline sector_t
drbd_md_first_sector(struct drbd_backing_dev
*bdev
)
1844 switch (bdev
->md
.meta_dev_idx
) {
1845 case DRBD_MD_INDEX_INTERNAL
:
1846 case DRBD_MD_INDEX_FLEX_INT
:
1847 return bdev
->md
.md_offset
+ bdev
->md
.bm_offset
;
1848 case DRBD_MD_INDEX_FLEX_EXT
:
1850 return bdev
->md
.md_offset
;
1855 * drbd_md_last_sector() - Return the last sector number of the meta data area
1856 * @bdev: Meta data block device.
1858 static inline sector_t
drbd_md_last_sector(struct drbd_backing_dev
*bdev
)
1860 switch (bdev
->md
.meta_dev_idx
) {
1861 case DRBD_MD_INDEX_INTERNAL
:
1862 case DRBD_MD_INDEX_FLEX_INT
:
1863 return bdev
->md
.md_offset
+ MD_4kB_SECT
-1;
1864 case DRBD_MD_INDEX_FLEX_EXT
:
1866 return bdev
->md
.md_offset
+ bdev
->md
.md_size_sect
-1;
1870 /* Returns the number of 512 byte sectors of the device */
1871 static inline sector_t
drbd_get_capacity(struct block_device
*bdev
)
1873 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1874 return bdev
? i_size_read(bdev
->bd_inode
) >> 9 : 0;
1878 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1879 * @bdev: Meta data block device.
1881 * returns the capacity we announce to out peer. we clip ourselves at the
1882 * various MAX_SECTORS, because if we don't, current implementation will
1883 * oops sooner or later
1885 static inline sector_t
drbd_get_max_capacity(struct drbd_backing_dev
*bdev
)
1889 switch (bdev
->md
.meta_dev_idx
) {
1890 case DRBD_MD_INDEX_INTERNAL
:
1891 case DRBD_MD_INDEX_FLEX_INT
:
1892 s
= drbd_get_capacity(bdev
->backing_bdev
)
1893 ? min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1894 drbd_md_first_sector(bdev
))
1897 case DRBD_MD_INDEX_FLEX_EXT
:
1898 s
= min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1899 drbd_get_capacity(bdev
->backing_bdev
));
1900 /* clip at maximum size the meta device can support */
1901 s
= min_t(sector_t
, s
,
1902 BM_EXT_TO_SECT(bdev
->md
.md_size_sect
1903 - bdev
->md
.bm_offset
));
1906 s
= min_t(sector_t
, DRBD_MAX_SECTORS
,
1907 drbd_get_capacity(bdev
->backing_bdev
));
1913 * drbd_md_ss() - Return the sector number of our meta data super block
1914 * @bdev: Meta data block device.
1916 static inline sector_t
drbd_md_ss(struct drbd_backing_dev
*bdev
)
1918 const int meta_dev_idx
= bdev
->md
.meta_dev_idx
;
1920 if (meta_dev_idx
== DRBD_MD_INDEX_FLEX_EXT
)
1923 /* Since drbd08, internal meta data is always "flexible".
1924 * position: last 4k aligned block of 4k size */
1925 if (meta_dev_idx
== DRBD_MD_INDEX_INTERNAL
||
1926 meta_dev_idx
== DRBD_MD_INDEX_FLEX_INT
)
1927 return (drbd_get_capacity(bdev
->backing_bdev
) & ~7ULL) - 8;
1929 /* external, some index; this is the old fixed size layout */
1930 return MD_128MB_SECT
* bdev
->md
.meta_dev_idx
;
1934 drbd_queue_work(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1936 unsigned long flags
;
1937 spin_lock_irqsave(&q
->q_lock
, flags
);
1938 list_add_tail(&w
->list
, &q
->q
);
1939 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1940 wake_up(&q
->q_wait
);
1944 drbd_queue_work_if_unqueued(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1946 unsigned long flags
;
1947 spin_lock_irqsave(&q
->q_lock
, flags
);
1948 if (list_empty_careful(&w
->list
))
1949 list_add_tail(&w
->list
, &q
->q
);
1950 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1951 wake_up(&q
->q_wait
);
1955 drbd_device_post_work(struct drbd_device
*device
, int work_bit
)
1957 if (!test_and_set_bit(work_bit
, &device
->flags
)) {
1958 struct drbd_connection
*connection
=
1959 first_peer_device(device
)->connection
;
1960 struct drbd_work_queue
*q
= &connection
->sender_work
;
1961 if (!test_and_set_bit(DEVICE_WORK_PENDING
, &connection
->flags
))
1962 wake_up(&q
->q_wait
);
1966 extern void drbd_flush_workqueue(struct drbd_work_queue
*work_queue
);
1968 /* To get the ack_receiver out of the blocking network stack,
1969 * so it can change its sk_rcvtimeo from idle- to ping-timeout,
1970 * and send a ping, we need to send a signal.
1971 * Which signal we send is irrelevant. */
1972 static inline void wake_ack_receiver(struct drbd_connection
*connection
)
1974 struct task_struct
*task
= connection
->ack_receiver
.task
;
1975 if (task
&& get_t_state(&connection
->ack_receiver
) == RUNNING
)
1976 force_sig(SIGXCPU
, task
);
1979 static inline void request_ping(struct drbd_connection
*connection
)
1981 set_bit(SEND_PING
, &connection
->flags
);
1982 wake_ack_receiver(connection
);
1985 extern void *conn_prepare_command(struct drbd_connection
*, struct drbd_socket
*);
1986 extern void *drbd_prepare_command(struct drbd_peer_device
*, struct drbd_socket
*);
1987 extern int conn_send_command(struct drbd_connection
*, struct drbd_socket
*,
1988 enum drbd_packet
, unsigned int, void *,
1990 extern int drbd_send_command(struct drbd_peer_device
*, struct drbd_socket
*,
1991 enum drbd_packet
, unsigned int, void *,
1994 extern int drbd_send_ping(struct drbd_connection
*connection
);
1995 extern int drbd_send_ping_ack(struct drbd_connection
*connection
);
1996 extern int drbd_send_state_req(struct drbd_peer_device
*, union drbd_state
, union drbd_state
);
1997 extern int conn_send_state_req(struct drbd_connection
*, union drbd_state
, union drbd_state
);
1999 static inline void drbd_thread_stop(struct drbd_thread
*thi
)
2001 _drbd_thread_stop(thi
, false, true);
2004 static inline void drbd_thread_stop_nowait(struct drbd_thread
*thi
)
2006 _drbd_thread_stop(thi
, false, false);
2009 static inline void drbd_thread_restart_nowait(struct drbd_thread
*thi
)
2011 _drbd_thread_stop(thi
, true, false);
2014 /* counts how many answer packets packets we expect from our peer,
2015 * for either explicit application requests,
2016 * or implicit barrier packets as necessary.
2019 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
2020 * it is much easier and equally valid to count what we queue for the
2021 * worker, even before it actually was queued or send.
2022 * (drbd_make_request_common; recovery path on read io-error)
2024 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
2025 * _req_mod(req, DATA_RECEIVED)
2026 * [from receive_DataReply]
2027 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
2028 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
2029 * for some reason it is NOT decreased in got_NegAck,
2030 * but in the resulting cleanup code from report_params.
2031 * we should try to remember the reason for that...
2032 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
2033 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
2034 * [from tl_clear_barrier]
2036 static inline void inc_ap_pending(struct drbd_device
*device
)
2038 atomic_inc(&device
->ap_pending_cnt
);
2041 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \
2042 if (atomic_read(&device->which) < 0) \
2043 drbd_err(device, "in %s:%d: " #which " = %d < 0 !\n", \
2045 atomic_read(&device->which))
2047 #define dec_ap_pending(device) _dec_ap_pending(device, __func__, __LINE__)
2048 static inline void _dec_ap_pending(struct drbd_device
*device
, const char *func
, int line
)
2050 if (atomic_dec_and_test(&device
->ap_pending_cnt
))
2051 wake_up(&device
->misc_wait
);
2052 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt
, func
, line
);
2055 /* counts how many resync-related answers we still expect from the peer
2057 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
2058 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
2059 * (or P_NEG_ACK with ID_SYNCER)
2061 static inline void inc_rs_pending(struct drbd_device
*device
)
2063 atomic_inc(&device
->rs_pending_cnt
);
2066 #define dec_rs_pending(device) _dec_rs_pending(device, __func__, __LINE__)
2067 static inline void _dec_rs_pending(struct drbd_device
*device
, const char *func
, int line
)
2069 atomic_dec(&device
->rs_pending_cnt
);
2070 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt
, func
, line
);
2073 /* counts how many answers we still need to send to the peer.
2075 * receive_Data unless protocol A;
2076 * we need to send a P_RECV_ACK (proto B)
2077 * or P_WRITE_ACK (proto C)
2078 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2079 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2080 * receive_Barrier_* we need to send a P_BARRIER_ACK
2082 static inline void inc_unacked(struct drbd_device
*device
)
2084 atomic_inc(&device
->unacked_cnt
);
2087 #define dec_unacked(device) _dec_unacked(device, __func__, __LINE__)
2088 static inline void _dec_unacked(struct drbd_device
*device
, const char *func
, int line
)
2090 atomic_dec(&device
->unacked_cnt
);
2091 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt
, func
, line
);
2094 #define sub_unacked(device, n) _sub_unacked(device, n, __func__, __LINE__)
2095 static inline void _sub_unacked(struct drbd_device
*device
, int n
, const char *func
, int line
)
2097 atomic_sub(n
, &device
->unacked_cnt
);
2098 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt
, func
, line
);
2101 static inline bool is_sync_state(enum drbd_conns connection_state
)
2104 (connection_state
== C_SYNC_SOURCE
2105 || connection_state
== C_SYNC_TARGET
2106 || connection_state
== C_PAUSED_SYNC_S
2107 || connection_state
== C_PAUSED_SYNC_T
);
2111 * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev
2112 * @_device: DRBD device.
2113 * @_min_state: Minimum device state required for success.
2115 * You have to call put_ldev() when finished working with device->ldev.
2117 #define get_ldev_if_state(_device, _min_state) \
2118 (_get_ldev_if_state((_device), (_min_state)) ? \
2119 ({ __acquire(x); true; }) : false)
2120 #define get_ldev(_device) get_ldev_if_state(_device, D_INCONSISTENT)
2122 static inline void put_ldev(struct drbd_device
*device
)
2124 enum drbd_disk_state disk_state
= device
->state
.disk
;
2125 /* We must check the state *before* the atomic_dec becomes visible,
2126 * or we have a theoretical race where someone hitting zero,
2127 * while state still D_FAILED, will then see D_DISKLESS in the
2128 * condition below and calling into destroy, where he must not, yet. */
2129 int i
= atomic_dec_return(&device
->local_cnt
);
2131 /* This may be called from some endio handler,
2132 * so we must not sleep here. */
2135 D_ASSERT(device
, i
>= 0);
2137 if (disk_state
== D_DISKLESS
)
2138 /* even internal references gone, safe to destroy */
2139 drbd_device_post_work(device
, DESTROY_DISK
);
2140 if (disk_state
== D_FAILED
)
2141 /* all application IO references gone. */
2142 if (!test_and_set_bit(GOING_DISKLESS
, &device
->flags
))
2143 drbd_device_post_work(device
, GO_DISKLESS
);
2144 wake_up(&device
->misc_wait
);
2149 static inline int _get_ldev_if_state(struct drbd_device
*device
, enum drbd_disk_state mins
)
2153 /* never get a reference while D_DISKLESS */
2154 if (device
->state
.disk
== D_DISKLESS
)
2157 atomic_inc(&device
->local_cnt
);
2158 io_allowed
= (device
->state
.disk
>= mins
);
2164 extern int _get_ldev_if_state(struct drbd_device
*device
, enum drbd_disk_state mins
);
2167 /* this throttles on-the-fly application requests
2168 * according to max_buffers settings;
2169 * maybe re-implement using semaphores? */
2170 static inline int drbd_get_max_buffers(struct drbd_device
*device
)
2172 struct net_conf
*nc
;
2176 nc
= rcu_dereference(first_peer_device(device
)->connection
->net_conf
);
2177 mxb
= nc
? nc
->max_buffers
: 1000000; /* arbitrary limit on open requests */
2183 static inline int drbd_state_is_stable(struct drbd_device
*device
)
2185 union drbd_dev_state s
= device
->state
;
2187 /* DO NOT add a default clause, we want the compiler to warn us
2188 * for any newly introduced state we may have forgotten to add here */
2190 switch ((enum drbd_conns
)s
.conn
) {
2191 /* new io only accepted when there is no connection, ... */
2193 case C_WF_CONNECTION
:
2194 /* ... or there is a well established connection. */
2200 case C_PAUSED_SYNC_S
:
2201 case C_PAUSED_SYNC_T
:
2204 /* transitional states, IO allowed */
2205 case C_DISCONNECTING
:
2209 case C_NETWORK_FAILURE
:
2210 case C_PROTOCOL_ERROR
:
2212 case C_WF_REPORT_PARAMS
:
2213 case C_STARTING_SYNC_S
:
2214 case C_STARTING_SYNC_T
:
2217 /* Allow IO in BM exchange states with new protocols */
2219 if (first_peer_device(device
)->connection
->agreed_pro_version
< 96)
2223 /* no new io accepted in these states */
2225 case C_WF_SYNC_UUID
:
2231 switch ((enum drbd_disk_state
)s
.disk
) {
2233 case D_INCONSISTENT
:
2238 /* disk state is stable as well. */
2241 /* no new io accepted during transitional states */
2253 static inline int drbd_suspended(struct drbd_device
*device
)
2255 struct drbd_resource
*resource
= device
->resource
;
2257 return resource
->susp
|| resource
->susp_fen
|| resource
->susp_nod
;
2260 static inline bool may_inc_ap_bio(struct drbd_device
*device
)
2262 int mxb
= drbd_get_max_buffers(device
);
2264 if (drbd_suspended(device
))
2266 if (atomic_read(&device
->suspend_cnt
))
2269 /* to avoid potential deadlock or bitmap corruption,
2270 * in various places, we only allow new application io
2271 * to start during "stable" states. */
2273 /* no new io accepted when attaching or detaching the disk */
2274 if (!drbd_state_is_stable(device
))
2277 /* since some older kernels don't have atomic_add_unless,
2278 * and we are within the spinlock anyways, we have this workaround. */
2279 if (atomic_read(&device
->ap_bio_cnt
) > mxb
)
2281 if (test_bit(BITMAP_IO
, &device
->flags
))
2286 static inline bool inc_ap_bio_cond(struct drbd_device
*device
)
2290 spin_lock_irq(&device
->resource
->req_lock
);
2291 rv
= may_inc_ap_bio(device
);
2293 atomic_inc(&device
->ap_bio_cnt
);
2294 spin_unlock_irq(&device
->resource
->req_lock
);
2299 static inline void inc_ap_bio(struct drbd_device
*device
)
2302 * as long as the device is suspended
2303 * until the bitmap is no longer on the fly during connection
2304 * handshake as long as we would exceed the max_buffer limit.
2306 * to avoid races with the reconnect code,
2307 * we need to atomic_inc within the spinlock. */
2309 wait_event(device
->misc_wait
, inc_ap_bio_cond(device
));
2312 static inline void dec_ap_bio(struct drbd_device
*device
)
2314 int mxb
= drbd_get_max_buffers(device
);
2315 int ap_bio
= atomic_dec_return(&device
->ap_bio_cnt
);
2317 D_ASSERT(device
, ap_bio
>= 0);
2319 if (ap_bio
== 0 && test_bit(BITMAP_IO
, &device
->flags
)) {
2320 if (!test_and_set_bit(BITMAP_IO_QUEUED
, &device
->flags
))
2321 drbd_queue_work(&first_peer_device(device
)->
2322 connection
->sender_work
,
2323 &device
->bm_io_work
.w
);
2326 /* this currently does wake_up for every dec_ap_bio!
2327 * maybe rather introduce some type of hysteresis?
2328 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2330 wake_up(&device
->misc_wait
);
2333 static inline bool verify_can_do_stop_sector(struct drbd_device
*device
)
2335 return first_peer_device(device
)->connection
->agreed_pro_version
>= 97 &&
2336 first_peer_device(device
)->connection
->agreed_pro_version
!= 100;
2339 static inline int drbd_set_ed_uuid(struct drbd_device
*device
, u64 val
)
2341 int changed
= device
->ed_uuid
!= val
;
2342 device
->ed_uuid
= val
;
2346 static inline int drbd_queue_order_type(struct drbd_device
*device
)
2348 /* sorry, we currently have no working implementation
2349 * of distributed TCQ stuff */
2350 #ifndef QUEUE_ORDERED_NONE
2351 #define QUEUE_ORDERED_NONE 0
2353 return QUEUE_ORDERED_NONE
;
2356 static inline struct drbd_connection
*first_connection(struct drbd_resource
*resource
)
2358 return list_first_entry_or_null(&resource
->connections
,
2359 struct drbd_connection
, connections
);