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_strings.h"
49 #include "drbd_state.h"
50 #include "drbd_protocol.h"
53 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
54 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
55 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
56 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
58 # define __protected_by(x)
59 # define __protected_read_by(x)
60 # define __protected_write_by(x)
61 # define __must_hold(x)
64 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
66 /* module parameter, defined in drbd_main.c */
67 extern unsigned int minor_count;
68 extern bool disable_sendpage;
69 extern bool allow_oos;
70 void tl_abort_disk_io(struct drbd_device *device);
72 #ifdef CONFIG_DRBD_FAULT_INJECTION
73 extern int enable_faults;
74 extern int fault_rate;
75 extern int fault_devs;
78 extern char usermode_helper[];
81 /* I don't remember why XCPU ...
82 * This is used to wake the asender,
83 * and to interrupt sending the sending task
86 #define DRBD_SIG SIGXCPU
88 /* This is used to stop/restart our threads.
89 * Cannot use SIGTERM nor SIGKILL, since these
90 * are sent out by init on runlevel changes
91 * I choose SIGHUP for now.
93 #define DRBD_SIGKILL SIGHUP
95 #define ID_IN_SYNC (4711ULL)
96 #define ID_OUT_OF_SYNC (4712ULL)
97 #define ID_SYNCER (-1ULL)
99 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
102 struct drbd_connection;
104 #define __drbd_printk_device(level, device, fmt, args...) \
105 dev_printk(level, disk_to_dev((device)->vdisk), fmt, ## args)
106 #define __drbd_printk_peer_device(level, peer_device, fmt, args...) \
107 dev_printk(level, disk_to_dev((peer_device)->device->vdisk), fmt, ## args)
108 #define __drbd_printk_resource(level, resource, fmt, args...) \
109 printk(level "drbd %s: " fmt, (resource)->name, ## args)
110 #define __drbd_printk_connection(level, connection, fmt, args...) \
111 printk(level "drbd %s: " fmt, (connection)->resource->name, ## args)
113 void drbd_printk_with_wrong_object_type(void);
115 #define __drbd_printk_if_same_type(obj, type, func, level, fmt, args...) \
116 (__builtin_types_compatible_p(typeof(obj), type) || \
117 __builtin_types_compatible_p(typeof(obj), const type)), \
118 func(level, (const type)(obj), fmt, ## args)
120 #define drbd_printk(level, obj, fmt, args...) \
121 __builtin_choose_expr( \
122 __drbd_printk_if_same_type(obj, struct drbd_device *, \
123 __drbd_printk_device, level, fmt, ## args), \
124 __builtin_choose_expr( \
125 __drbd_printk_if_same_type(obj, struct drbd_resource *, \
126 __drbd_printk_resource, level, fmt, ## args), \
127 __builtin_choose_expr( \
128 __drbd_printk_if_same_type(obj, struct drbd_connection *, \
129 __drbd_printk_connection, level, fmt, ## args), \
130 __builtin_choose_expr( \
131 __drbd_printk_if_same_type(obj, struct drbd_peer_device *, \
132 __drbd_printk_peer_device, level, fmt, ## args), \
133 drbd_printk_with_wrong_object_type()))))
135 #define drbd_dbg(obj, fmt, args...) \
136 drbd_printk(KERN_DEBUG, obj, fmt, ## args)
137 #define drbd_alert(obj, fmt, args...) \
138 drbd_printk(KERN_ALERT, obj, fmt, ## args)
139 #define drbd_err(obj, fmt, args...) \
140 drbd_printk(KERN_ERR, obj, fmt, ## args)
141 #define drbd_warn(obj, fmt, args...) \
142 drbd_printk(KERN_WARNING, obj, fmt, ## args)
143 #define drbd_info(obj, fmt, args...) \
144 drbd_printk(KERN_INFO, obj, fmt, ## args)
145 #define drbd_emerg(obj, fmt, args...) \
146 drbd_printk(KERN_EMERG, obj, fmt, ## args)
148 #define dynamic_drbd_dbg(device, fmt, args...) \
149 dynamic_dev_dbg(disk_to_dev(device->vdisk), fmt, ## args)
151 #define D_ASSERT(device, exp) do { \
153 drbd_err(device, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__); \
157 * expect - Make an assertion
159 * Unlike the assert macro, this macro returns a boolean result.
161 #define expect(exp) ({ \
162 bool _bool = (exp); \
164 drbd_err(device, "ASSERTION %s FAILED in %s\n", \
169 /* Defines to control fault insertion */
171 DRBD_FAULT_MD_WR = 0, /* meta data write */
172 DRBD_FAULT_MD_RD = 1, /* read */
173 DRBD_FAULT_RS_WR = 2, /* resync */
174 DRBD_FAULT_RS_RD = 3,
175 DRBD_FAULT_DT_WR = 4, /* data */
176 DRBD_FAULT_DT_RD = 5,
177 DRBD_FAULT_DT_RA = 6, /* data read ahead */
178 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */
179 DRBD_FAULT_AL_EE = 8, /* alloc ee */
180 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
186 _drbd_insert_fault(struct drbd_device *device, unsigned int type);
189 drbd_insert_fault(struct drbd_device *device, unsigned int type) {
190 #ifdef CONFIG_DRBD_FAULT_INJECTION
192 (enable_faults & (1<<type)) &&
193 _drbd_insert_fault(device, type);
199 /* integer division, round _UP_ to the next integer */
200 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
201 /* usual integer division */
202 #define div_floor(A, B) ((A)/(B))
204 extern struct ratelimit_state drbd_ratelimit_state;
205 extern struct idr drbd_devices; /* RCU, updates: genl_lock() */
206 extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */
208 extern const char *cmdname(enum drbd_packet cmd);
210 /* for sending/receiving the bitmap,
211 * possibly in some encoding scheme */
214 * stores total bits and long words
215 * of the bitmap, so we don't need to
216 * call the accessor functions over and again. */
217 unsigned long bm_bits;
218 unsigned long bm_words;
219 /* during xfer, current position within the bitmap */
220 unsigned long bit_offset;
221 unsigned long word_offset;
223 /* statistics; index: (h->command == P_BITMAP) */
228 extern void INFO_bm_xfer_stats(struct drbd_device *device,
229 const char *direction, struct bm_xfer_ctx *c);
231 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
233 /* word_offset counts "native long words" (32 or 64 bit),
235 * Encoded packet may end at an unaligned bit offset.
236 * In case a fallback clear text packet is transmitted in
237 * between, we adjust this offset back to the last 64bit
238 * aligned "native long word", which makes coding and decoding
239 * the plain text bitmap much more convenient. */
240 #if BITS_PER_LONG == 64
241 c->word_offset = c->bit_offset >> 6;
242 #elif BITS_PER_LONG == 32
243 c->word_offset = c->bit_offset >> 5;
244 c->word_offset &= ~(1UL);
246 # error "unsupported BITS_PER_LONG"
250 extern unsigned int drbd_header_size(struct drbd_connection *connection);
252 /**********************************************************************/
253 enum drbd_thread_state {
262 struct task_struct *task;
263 struct completion stop;
264 enum drbd_thread_state t_state;
265 int (*function) (struct drbd_thread *);
266 struct drbd_resource *resource;
267 struct drbd_connection *connection;
272 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
274 /* THINK testing the t_state seems to be uncritical in all cases
275 * (but thread_{start,stop}), so we can read it *without* the lock.
283 struct list_head list;
284 int (*cb)(struct drbd_work *, int cancel);
287 struct drbd_device_work {
289 struct drbd_device *device;
292 #include "drbd_interval.h"
294 extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *);
296 struct drbd_request {
298 struct drbd_device *device;
300 /* if local IO is not allowed, will be NULL.
301 * if local IO _is_ allowed, holds the locally submitted bio clone,
302 * or, after local IO completion, the ERR_PTR(error).
303 * see drbd_request_endio(). */
304 struct bio *private_bio;
306 struct drbd_interval i;
308 /* epoch: used to check on "completion" whether this req was in
309 * the current epoch, and we therefore have to close it,
310 * causing a p_barrier packet to be send, starting a new epoch.
312 * This corresponds to "barrier" in struct p_barrier[_ack],
313 * and to "barrier_nr" in struct drbd_epoch (and various
314 * comments/function parameters/local variable names).
318 struct list_head tl_requests; /* ring list in the transfer log */
319 struct bio *master_bio; /* master bio pointer */
320 unsigned long start_time;
322 /* once it hits 0, we may complete the master_bio */
323 atomic_t completion_ref;
324 /* once it hits 0, we may destroy this drbd_request object */
327 unsigned rq_state; /* see comments above _req_mod() */
331 struct drbd_connection *connection;
332 struct list_head list;
333 unsigned int barrier_nr;
334 atomic_t epoch_size; /* increased on every request added. */
335 atomic_t active; /* increased on every req. added, and dec on every finished. */
339 /* Prototype declaration of function defined in drbd_receiver.c */
340 int drbdd_init(struct drbd_thread *);
341 int drbd_asender(struct drbd_thread *);
343 /* drbd_epoch flag bits */
345 DE_HAVE_BARRIER_NUMBER,
352 EV_CLEANUP = 32, /* used as flag */
360 struct drbd_peer_request {
362 struct drbd_peer_device *peer_device;
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 /* is this a TRIM aka REQ_DISCARD? */
387 /* our lower level cannot handle trim,
388 * and we want to fall back to zeroout instead */
389 __EE_IS_TRIM_USE_ZEROOUT,
391 /* In case a barrier failed,
392 * we need to resubmit without the barrier flag. */
395 /* we may have several bios per peer request.
396 * if any of those fail, we set this flag atomically
397 * from the endio callback */
400 /* This ee has a pointer to a digest instead of a block id */
403 /* Conflicting local requests need to be restarted after this request */
404 __EE_RESTART_REQUESTS,
406 /* The peer wants a write ACK for this (wire proto C) */
409 /* Is set when net_conf had two_primaries set while creating this peer_req */
410 __EE_IN_INTERVAL_TREE,
412 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
413 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
414 #define EE_IS_TRIM (1<<__EE_IS_TRIM)
415 #define EE_IS_TRIM_USE_ZEROOUT (1<<__EE_IS_TRIM_USE_ZEROOUT)
416 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
417 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
418 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
419 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
420 #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK)
421 #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE)
423 /* flag bits per device */
425 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
426 MD_DIRTY, /* current uuids and flags not yet on disk */
427 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
430 CRASHED_PRIMARY, /* This node was a crashed primary.
431 * Gets cleared when the state.conn
432 * goes into C_CONNECTED state. */
435 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */
437 SUSPEND_IO, /* suspend application io */
438 BITMAP_IO, /* suspend application io;
439 once no more io in flight, start bitmap io */
440 BITMAP_IO_QUEUED, /* Started bitmap IO */
441 WAS_IO_ERROR, /* Local disk failed, returned IO error */
442 WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */
443 FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */
444 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
445 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
446 * the peer, if it changed there as well. */
447 NEW_CUR_UUID, /* Create new current UUID when thawing IO */
448 AL_SUSPENDED, /* Activity logging is currently suspended. */
449 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */
450 B_RS_H_DONE, /* Before resync handler done (already executed) */
451 DISCARD_MY_DATA, /* discard_my_data flag per volume */
454 /* cleared only after backing device related structures have been destroyed. */
455 GOING_DISKLESS, /* Disk is being detached, because of io-error, or admin request. */
457 /* to be used in drbd_device_post_work() */
458 GO_DISKLESS, /* tell worker to schedule cleanup before detach */
459 DESTROY_DISK, /* tell worker to close backing devices and destroy related structures. */
460 MD_SYNC, /* tell worker to call drbd_md_sync() */
461 RS_START, /* tell worker to start resync/OV */
462 RS_PROGRESS, /* tell worker that resync made significant progress */
463 RS_DONE, /* tell worker that resync is done */
466 struct drbd_bitmap; /* opaque for drbd_device */
468 /* definition of bits in bm_flags to be used in drbd_bm_lock
469 * and drbd_bitmap_io and friends. */
471 /* do we need to kfree, or vfree bm_pages? */
472 BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */
474 /* currently locked for bulk operation */
475 BM_LOCKED_MASK = 0xf,
477 /* in detail, that is: */
482 /* so we can mark it locked for bulk operation,
483 * and still allow all non-bulk operations */
486 /* (test bit, count bit) allowed (common case) */
487 BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED,
489 /* testing bits, as well as setting new bits allowed, but clearing bits
490 * would be unexpected. Used during bitmap receive. Setting new bits
491 * requires sending of "out-of-sync" information, though. */
492 BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED,
494 /* for drbd_bm_write_copy_pages, everything is allowed,
495 * only concurrent bulk operations are locked out. */
496 BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED,
499 struct drbd_work_queue {
501 spinlock_t q_lock; /* to protect the list. */
502 wait_queue_head_t q_wait;
507 struct socket *socket;
508 /* this way we get our
509 * send/receive buffers off the stack */
515 u64 md_offset; /* sector offset to 'super' block */
517 u64 la_size_sect; /* last agreed size, unit sectors */
518 spinlock_t uuid_lock;
524 s32 al_offset; /* signed relative sector offset to activity log */
525 s32 bm_offset; /* signed relative sector offset to bitmap */
527 /* cached value of bdev->disk_conf->meta_dev_idx (see below) */
530 /* see al_tr_number_to_on_disk_sector() */
532 u32 al_stripe_size_4k;
533 u32 al_size_4k; /* cached product of the above */
536 struct drbd_backing_dev {
537 struct block_device *backing_bdev;
538 struct block_device *md_bdev;
540 struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */
541 sector_t known_size; /* last known size of that backing device */
553 int (*io_fn)(struct drbd_device *device);
554 void (*done)(struct drbd_device *device, int rv);
557 enum write_ordering_e {
564 unsigned int head_index;
566 int total; /* sum of all values */
569 extern struct fifo_buffer *fifo_alloc(int fifo_size);
571 /* flag bits per connection */
573 NET_CONGESTED, /* The data socket is congested */
574 RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */
575 SEND_PING, /* whether asender should send a ping asap */
576 SIGNAL_ASENDER, /* whether asender wants to be interrupted */
577 GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */
578 CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */
581 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
582 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */
583 STATE_SENT, /* Do not change state/UUIDs while this is set */
584 CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
585 * pending, from drbd worker context.
586 * If set, bdi_write_congested() returns true,
587 * so shrink_page_list() would not recurse into,
588 * and potentially deadlock on, this drbd worker.
592 DEVICE_WORK_PENDING, /* tell worker that some device has pending work */
595 struct drbd_resource {
598 struct idr devices; /* volume number to device mapping */
599 struct list_head connections;
600 struct list_head resources;
601 struct res_opts res_opts;
602 struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */
603 struct mutex adm_mutex; /* mutex to serialize administrative requests */
606 unsigned susp:1; /* IO suspended by user */
607 unsigned susp_nod:1; /* IO suspended because no data */
608 unsigned susp_fen:1; /* IO suspended because fence peer handler runs */
610 enum write_ordering_e write_ordering;
612 cpumask_var_t cpu_mask;
615 struct drbd_connection {
616 struct list_head connections;
617 struct drbd_resource *resource;
619 struct idr peer_devices; /* volume number to peer device mapping */
620 enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
621 struct mutex cstate_mutex; /* Protects graceful disconnects */
622 unsigned int connect_cnt; /* Inc each time a connection is established */
625 struct net_conf *net_conf; /* content protected by rcu */
626 wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */
628 struct sockaddr_storage my_addr;
630 struct sockaddr_storage peer_addr;
633 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
634 struct drbd_socket meta; /* ping/ack (metadata) packets */
635 int agreed_pro_version; /* actually used protocol version */
637 unsigned long last_received; /* in jiffies, either socket */
638 unsigned int ko_count;
640 struct list_head transfer_log; /* all requests not yet fully processed */
642 struct crypto_hash *cram_hmac_tfm;
643 struct crypto_hash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */
644 struct crypto_hash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */
645 struct crypto_hash *csums_tfm;
646 struct crypto_hash *verify_tfm;
651 struct drbd_epoch *current_epoch;
652 spinlock_t epoch_lock;
654 atomic_t current_tle_nr; /* transfer log epoch number */
655 unsigned current_tle_writes; /* writes seen within this tl epoch */
657 unsigned long last_reconnect_jif;
658 struct drbd_thread receiver;
659 struct drbd_thread worker;
660 struct drbd_thread asender;
663 struct drbd_work_queue sender_work;
666 /* whether this sender thread
667 * has processed a single write yet. */
668 bool seen_any_write_yet;
670 /* Which barrier number to send with the next P_BARRIER */
671 int current_epoch_nr;
673 /* how many write requests have been sent
674 * with req->epoch == current_epoch_nr.
675 * If none, no P_BARRIER will be sent. */
676 unsigned current_epoch_writes;
680 struct submit_worker {
681 struct workqueue_struct *wq;
682 struct work_struct worker;
685 struct list_head writes;
688 struct drbd_peer_device {
689 struct list_head peer_devices;
690 struct drbd_device *device;
691 struct drbd_connection *connection;
695 struct drbd_resource *resource;
696 struct list_head peer_devices;
697 int vnr; /* volume number within the connection */
700 /* things that are stored as / read from meta data on disk */
703 /* configured by drbdsetup */
704 struct drbd_backing_dev *ldev __protected_by(local);
706 sector_t p_size; /* partner's disk size */
707 struct request_queue *rq_queue;
708 struct block_device *this_bdev;
709 struct gendisk *vdisk;
711 unsigned long last_reattach_jif;
712 struct drbd_work resync_work;
713 struct drbd_work unplug_work;
714 struct timer_list resync_timer;
715 struct timer_list md_sync_timer;
716 struct timer_list start_resync_timer;
717 struct timer_list request_timer;
719 /* Used after attach while negotiating new disk state. */
720 union drbd_state new_state_tmp;
722 union drbd_dev_state state;
723 wait_queue_head_t misc_wait;
724 wait_queue_head_t state_wait; /* upon each state change. */
725 unsigned int send_cnt;
726 unsigned int recv_cnt;
727 unsigned int read_cnt;
728 unsigned int writ_cnt;
729 unsigned int al_writ_cnt;
730 unsigned int bm_writ_cnt;
731 atomic_t ap_bio_cnt; /* Requests we need to complete */
732 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
733 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
734 atomic_t unacked_cnt; /* Need to send replies for */
735 atomic_t local_cnt; /* Waiting for local completion */
737 /* Interval tree of pending local requests */
738 struct rb_root read_requests;
739 struct rb_root write_requests;
741 /* use checksums for *this* resync */
743 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
744 unsigned long rs_total;
745 /* number of resync blocks that failed in this run */
746 unsigned long rs_failed;
747 /* Syncer's start time [unit jiffies] */
748 unsigned long rs_start;
749 /* cumulated time in PausedSyncX state [unit jiffies] */
750 unsigned long rs_paused;
751 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
752 unsigned long rs_same_csum;
753 #define DRBD_SYNC_MARKS 8
754 #define DRBD_SYNC_MARK_STEP (3*HZ)
755 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
756 unsigned long rs_mark_left[DRBD_SYNC_MARKS];
757 /* marks's time [unit jiffies] */
758 unsigned long rs_mark_time[DRBD_SYNC_MARKS];
759 /* current index into rs_mark_{left,time} */
761 unsigned long rs_last_bcast; /* [unit jiffies] */
763 /* where does the admin want us to start? (sector) */
764 sector_t ov_start_sector;
765 sector_t ov_stop_sector;
766 /* where are we now? (sector) */
767 sector_t ov_position;
768 /* Start sector of out of sync range (to merge printk reporting). */
769 sector_t ov_last_oos_start;
770 /* size of out-of-sync range in sectors. */
771 sector_t ov_last_oos_size;
772 unsigned long ov_left; /* in bits */
774 struct drbd_bitmap *bitmap;
775 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
777 /* Used to track operations of resync... */
778 struct lru_cache *resync;
779 /* Number of locked elements in resync LRU */
780 unsigned int resync_locked;
781 /* resync extent number waiting for application requests */
782 unsigned int resync_wenr;
787 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
788 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
789 struct list_head done_ee; /* need to send P_WRITE_ACK */
790 struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */
791 struct list_head net_ee; /* zero-copy network send in progress */
794 struct list_head resync_reads;
795 atomic_t pp_in_use; /* allocated from page pool */
796 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */
797 wait_queue_head_t ee_wait;
798 struct page *md_io_page; /* one page buffer for md_io */
799 struct drbd_md_io md_io;
800 atomic_t md_io_in_use; /* protects the md_io, md_io_page and md_io_tmpp */
802 wait_queue_head_t al_wait;
803 struct lru_cache *act_log; /* activity log */
804 unsigned int al_tr_number;
806 wait_queue_head_t seq_wait;
808 unsigned int peer_seq;
809 spinlock_t peer_seq_lock;
811 unsigned long comm_bm_set; /* communicated number of set bits. */
812 struct bm_io_work bm_io_work;
813 u64 ed_uuid; /* UUID of the exposed data */
814 struct mutex own_state_mutex;
815 struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */
816 char congestion_reason; /* Why we where congested... */
817 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
818 atomic_t rs_sect_ev; /* for submitted resync data rate, both */
819 int rs_last_sect_ev; /* counter to compare with */
820 int rs_last_events; /* counter of read or write "events" (unit sectors)
821 * on the lower level device when we last looked. */
822 int c_sync_rate; /* current resync rate after syncer throttle magic */
823 struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */
824 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
825 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
826 unsigned int peer_max_bio_size;
827 unsigned int local_max_bio_size;
829 /* any requests that would block in drbd_make_request()
830 * are deferred to this single-threaded work queue */
831 struct submit_worker submit;
834 struct drbd_config_context {
835 /* assigned from drbd_genlmsghdr */
837 /* assigned from request attributes, if present */
839 #define VOLUME_UNSPECIFIED (-1U)
840 /* pointer into the request skb,
841 * limited lifetime! */
843 struct nlattr *my_addr;
844 struct nlattr *peer_addr;
847 struct sk_buff *reply_skb;
848 /* pointer into reply buffer */
849 struct drbd_genlmsghdr *reply_dh;
850 /* resolved from attributes, if possible */
851 struct drbd_device *device;
852 struct drbd_resource *resource;
853 struct drbd_connection *connection;
856 static inline struct drbd_device *minor_to_device(unsigned int minor)
858 return (struct drbd_device *)idr_find(&drbd_devices, minor);
861 static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device)
863 return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices);
866 #define for_each_resource(resource, _resources) \
867 list_for_each_entry(resource, _resources, resources)
869 #define for_each_resource_rcu(resource, _resources) \
870 list_for_each_entry_rcu(resource, _resources, resources)
872 #define for_each_resource_safe(resource, tmp, _resources) \
873 list_for_each_entry_safe(resource, tmp, _resources, resources)
875 #define for_each_connection(connection, resource) \
876 list_for_each_entry(connection, &resource->connections, connections)
878 #define for_each_connection_rcu(connection, resource) \
879 list_for_each_entry_rcu(connection, &resource->connections, connections)
881 #define for_each_connection_safe(connection, tmp, resource) \
882 list_for_each_entry_safe(connection, tmp, &resource->connections, connections)
884 #define for_each_peer_device(peer_device, device) \
885 list_for_each_entry(peer_device, &device->peer_devices, peer_devices)
887 #define for_each_peer_device_rcu(peer_device, device) \
888 list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices)
890 #define for_each_peer_device_safe(peer_device, tmp, device) \
891 list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices)
893 static inline unsigned int device_to_minor(struct drbd_device *device)
895 return device->minor;
899 * function declarations
900 *************************/
906 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
909 extern void drbd_init_set_defaults(struct drbd_device *device);
910 extern int drbd_thread_start(struct drbd_thread *thi);
911 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
913 extern void drbd_thread_current_set_cpu(struct drbd_thread *thi);
915 #define drbd_thread_current_set_cpu(A) ({})
917 extern void tl_release(struct drbd_connection *, unsigned int barrier_nr,
918 unsigned int set_size);
919 extern void tl_clear(struct drbd_connection *);
920 extern void drbd_free_sock(struct drbd_connection *connection);
921 extern int drbd_send(struct drbd_connection *connection, struct socket *sock,
922 void *buf, size_t size, unsigned msg_flags);
923 extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t,
926 extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd);
927 extern int drbd_send_protocol(struct drbd_connection *connection);
928 extern int drbd_send_uuids(struct drbd_peer_device *);
929 extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *);
930 extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *);
931 extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags);
932 extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s);
933 extern int drbd_send_current_state(struct drbd_peer_device *);
934 extern int drbd_send_sync_param(struct drbd_peer_device *);
935 extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr,
937 extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet,
938 struct drbd_peer_request *);
939 extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet,
940 struct p_block_req *rp);
941 extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet,
942 struct p_data *dp, int data_size);
943 extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet,
944 sector_t sector, int blksize, u64 block_id);
945 extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *);
946 extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet,
947 struct drbd_peer_request *);
948 extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req);
949 extern int drbd_send_drequest(struct drbd_peer_device *, int cmd,
950 sector_t sector, int size, u64 block_id);
951 extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector,
952 int size, void *digest, int digest_size,
953 enum drbd_packet cmd);
954 extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size);
956 extern int drbd_send_bitmap(struct drbd_device *device);
957 extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode);
958 extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode);
959 extern void drbd_free_ldev(struct drbd_backing_dev *ldev);
960 extern void drbd_device_cleanup(struct drbd_device *device);
961 void drbd_print_uuids(struct drbd_device *device, const char *text);
963 extern void conn_md_sync(struct drbd_connection *connection);
964 extern void drbd_md_write(struct drbd_device *device, void *buffer);
965 extern void drbd_md_sync(struct drbd_device *device);
966 extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev);
967 extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
968 extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
969 extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local);
970 extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local);
971 extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local);
972 extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
973 extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local);
974 extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local);
975 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
976 extern void drbd_md_mark_dirty(struct drbd_device *device);
977 extern void drbd_queue_bitmap_io(struct drbd_device *device,
978 int (*io_fn)(struct drbd_device *),
979 void (*done)(struct drbd_device *, int),
980 char *why, enum bm_flag flags);
981 extern int drbd_bitmap_io(struct drbd_device *device,
982 int (*io_fn)(struct drbd_device *),
983 char *why, enum bm_flag flags);
984 extern int drbd_bitmap_io_from_worker(struct drbd_device *device,
985 int (*io_fn)(struct drbd_device *),
986 char *why, enum bm_flag flags);
987 extern int drbd_bmio_set_n_write(struct drbd_device *device) __must_hold(local);
988 extern int drbd_bmio_clear_n_write(struct drbd_device *device) __must_hold(local);
992 * We currently have two possible layouts.
993 * Offsets in (512 byte) sectors.
995 * |----------- md_size_sect ------------------|
996 * [ 4k superblock ][ activity log ][ Bitmap ]
998 * | bm_offset = al_offset + X |
999 * ==> bitmap sectors = md_size_sect - bm_offset
1002 * old, indexed fixed size meta data:
1005 * |----------- md_size_sect ------------------|
1006 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*]
1008 * | bm_offset = al_offset - Y |
1009 * ==> bitmap sectors = Y = al_offset - bm_offset
1011 * [padding*] are zero or up to 7 unused 512 Byte sectors to the
1012 * end of the device, so that the [4k superblock] will be 4k aligned.
1014 * The activity log consists of 4k transaction blocks,
1015 * which are written in a ring-buffer, or striped ring-buffer like fashion,
1016 * which are writtensize used to be fixed 32kB,
1017 * but is about to become configurable.
1020 /* Our old fixed size meta data layout
1021 * allows up to about 3.8TB, so if you want more,
1022 * you need to use the "flexible" meta data format. */
1023 #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */
1024 #define MD_4kB_SECT 8
1025 #define MD_32kB_SECT 64
1027 /* One activity log extent represents 4M of storage */
1028 #define AL_EXTENT_SHIFT 22
1029 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1031 /* We could make these currently hardcoded constants configurable
1032 * variables at create-md time (or even re-configurable at runtime?).
1033 * Which will require some more changes to the DRBD "super block"
1036 * updates per transaction:
1037 * This many changes to the active set can be logged with one transaction.
1038 * This number is arbitrary.
1039 * context per transaction:
1040 * This many context extent numbers are logged with each transaction.
1041 * This number is resulting from the transaction block size (4k), the layout
1042 * of the transaction header, and the number of updates per transaction.
1043 * See drbd_actlog.c:struct al_transaction_on_disk
1045 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1046 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1048 #if BITS_PER_LONG == 32
1050 #define cpu_to_lel(A) cpu_to_le32(A)
1051 #define lel_to_cpu(A) le32_to_cpu(A)
1052 #elif BITS_PER_LONG == 64
1054 #define cpu_to_lel(A) cpu_to_le64(A)
1055 #define lel_to_cpu(A) le64_to_cpu(A)
1057 #error "LN2 of BITS_PER_LONG unknown!"
1061 /* 16MB sized 'bitmap extent' to track syncer usage */
1063 int rs_left; /* number of bits set (out of sync) in this extent. */
1064 int rs_failed; /* number of failed resync requests in this extent. */
1065 unsigned long flags;
1066 struct lc_element lce;
1069 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1070 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1071 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1075 * We need to store one bit for a block.
1076 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1077 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1078 * Bit 1 ==> local node thinks this block needs to be synced.
1081 #define SLEEP_TIME (HZ/10)
1083 /* We do bitmap IO in units of 4k blocks.
1084 * We also still have a hardcoded 4k per bit relation. */
1085 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1086 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1087 /* mostly arbitrarily set the represented size of one bitmap extent,
1088 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1089 * at 4k per bit resolution) */
1090 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
1091 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1093 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1094 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1097 /* thus many _storage_ sectors are described by one bit */
1098 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1099 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1100 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1102 /* bit to represented kilo byte conversion */
1103 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1105 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1106 * _storage_ sector is located in */
1107 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1108 #define BM_BIT_TO_EXT(x) ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1110 /* first storage sector a bitmap extent corresponds to */
1111 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1112 /* how much _storage_ sectors we have per bitmap extent */
1113 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1114 /* how many bits are covered by one bitmap extent (resync extent) */
1115 #define BM_BITS_PER_EXT (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1117 #define BM_BLOCKS_PER_BM_EXT_MASK (BM_BITS_PER_EXT - 1)
1120 /* in one sector of the bitmap, we have this many activity_log extents. */
1121 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1123 /* the extent in "PER_EXTENT" below is an activity log extent
1124 * we need that many (long words/bytes) to store the bitmap
1125 * of one AL_EXTENT_SIZE chunk of storage.
1126 * we can store the bitmap for that many AL_EXTENTS within
1127 * one sector of the _on_disk_ bitmap:
1128 * bit 0 bit 37 bit 38 bit (512*8)-1
1129 * ...|........|........|.. // ..|........|
1130 * sect. 0 `296 `304 ^(512*8*8)-1
1132 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1133 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1134 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1137 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1138 /* we have a certain meta data variant that has a fixed on-disk size of 128
1139 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity
1140 * log, leaving this many sectors for the bitmap.
1143 #define DRBD_MAX_SECTORS_FIXED_BM \
1144 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9)))
1145 #if !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1146 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1147 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1149 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM
1150 /* 16 TB in units of sectors */
1151 #if BITS_PER_LONG == 32
1152 /* adjust by one page worth of bitmap,
1153 * so we won't wrap around in drbd_bm_find_next_bit.
1154 * you should use 64bit OS for that much storage, anyways. */
1155 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1157 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1158 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1159 /* corresponds to (1UL << 38) bits right now. */
1163 /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE,
1164 * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte.
1165 * Since we may live in a mixed-platform cluster,
1166 * we limit us to a platform agnostic constant here for now.
1167 * A followup commit may allow even bigger BIO sizes,
1168 * once we thought that through. */
1169 #define DRBD_MAX_BIO_SIZE (1U << 20)
1170 #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1171 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1173 #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */
1175 #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */
1176 #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */
1178 /* For now, don't allow more than one activity log extent worth of data
1179 * to be discarded in one go. We may need to rework drbd_al_begin_io()
1180 * to allow for even larger discard ranges */
1181 #define DRBD_MAX_DISCARD_SIZE AL_EXTENT_SIZE
1182 #define DRBD_MAX_DISCARD_SECTORS (DRBD_MAX_DISCARD_SIZE >> 9)
1184 extern int drbd_bm_init(struct drbd_device *device);
1185 extern int drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits);
1186 extern void drbd_bm_cleanup(struct drbd_device *device);
1187 extern void drbd_bm_set_all(struct drbd_device *device);
1188 extern void drbd_bm_clear_all(struct drbd_device *device);
1189 /* set/clear/test only a few bits at a time */
1190 extern int drbd_bm_set_bits(
1191 struct drbd_device *device, unsigned long s, unsigned long e);
1192 extern int drbd_bm_clear_bits(
1193 struct drbd_device *device, unsigned long s, unsigned long e);
1194 extern int drbd_bm_count_bits(
1195 struct drbd_device *device, const unsigned long s, const unsigned long e);
1196 /* bm_set_bits variant for use while holding drbd_bm_lock,
1197 * may process the whole bitmap in one go */
1198 extern void _drbd_bm_set_bits(struct drbd_device *device,
1199 const unsigned long s, const unsigned long e);
1200 extern int drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr);
1201 extern int drbd_bm_e_weight(struct drbd_device *device, unsigned long enr);
1202 extern int drbd_bm_read(struct drbd_device *device) __must_hold(local);
1203 extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr);
1204 extern int drbd_bm_write(struct drbd_device *device) __must_hold(local);
1205 extern int drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local);
1206 extern int drbd_bm_write_lazy(struct drbd_device *device, unsigned upper_idx) __must_hold(local);
1207 extern int drbd_bm_write_all(struct drbd_device *device) __must_hold(local);
1208 extern int drbd_bm_write_copy_pages(struct drbd_device *device) __must_hold(local);
1209 extern size_t drbd_bm_words(struct drbd_device *device);
1210 extern unsigned long drbd_bm_bits(struct drbd_device *device);
1211 extern sector_t drbd_bm_capacity(struct drbd_device *device);
1213 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1214 extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo);
1215 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1216 extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo);
1217 extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo);
1218 extern unsigned long _drbd_bm_total_weight(struct drbd_device *device);
1219 extern unsigned long drbd_bm_total_weight(struct drbd_device *device);
1220 /* for receive_bitmap */
1221 extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset,
1222 size_t number, unsigned long *buffer);
1223 /* for _drbd_send_bitmap */
1224 extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset,
1225 size_t number, unsigned long *buffer);
1227 extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags);
1228 extern void drbd_bm_unlock(struct drbd_device *device);
1231 extern struct kmem_cache *drbd_request_cache;
1232 extern struct kmem_cache *drbd_ee_cache; /* peer requests */
1233 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1234 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1235 extern mempool_t *drbd_request_mempool;
1236 extern mempool_t *drbd_ee_mempool;
1238 /* drbd's page pool, used to buffer data received from the peer,
1239 * or data requested by the peer.
1241 * This does not have an emergency reserve.
1243 * When allocating from this pool, it first takes pages from the pool.
1244 * Only if the pool is depleted will try to allocate from the system.
1246 * The assumption is that pages taken from this pool will be processed,
1247 * and given back, "quickly", and then can be recycled, so we can avoid
1248 * frequent calls to alloc_page(), and still will be able to make progress even
1249 * under memory pressure.
1251 extern struct page *drbd_pp_pool;
1252 extern spinlock_t drbd_pp_lock;
1253 extern int drbd_pp_vacant;
1254 extern wait_queue_head_t drbd_pp_wait;
1256 /* We also need a standard (emergency-reserve backed) page pool
1257 * for meta data IO (activity log, bitmap).
1258 * We can keep it global, as long as it is used as "N pages at a time".
1259 * 128 should be plenty, currently we probably can get away with as few as 1.
1261 #define DRBD_MIN_POOL_PAGES 128
1262 extern mempool_t *drbd_md_io_page_pool;
1264 /* We also need to make sure we get a bio
1265 * when we need it for housekeeping purposes */
1266 extern struct bio_set *drbd_md_io_bio_set;
1267 /* to allocate from that set */
1268 extern struct bio *bio_alloc_drbd(gfp_t gfp_mask);
1270 extern rwlock_t global_state_lock;
1272 extern int conn_lowest_minor(struct drbd_connection *connection);
1273 extern enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor);
1274 extern void drbd_destroy_device(struct kref *kref);
1275 extern void drbd_delete_device(struct drbd_device *device);
1277 extern struct drbd_resource *drbd_create_resource(const char *name);
1278 extern void drbd_free_resource(struct drbd_resource *resource);
1280 extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts);
1281 extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts);
1282 extern void drbd_destroy_connection(struct kref *kref);
1283 extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len,
1284 void *peer_addr, int peer_addr_len);
1285 extern struct drbd_resource *drbd_find_resource(const char *name);
1286 extern void drbd_destroy_resource(struct kref *kref);
1287 extern void conn_free_crypto(struct drbd_connection *connection);
1289 extern int proc_details;
1292 extern void do_submit(struct work_struct *ws);
1293 extern void __drbd_make_request(struct drbd_device *, struct bio *, unsigned long);
1294 extern void drbd_make_request(struct request_queue *q, struct bio *bio);
1295 extern int drbd_read_remote(struct drbd_device *device, struct drbd_request *req);
1296 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1297 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1301 extern int drbd_msg_put_info(struct sk_buff *skb, const char *info);
1302 extern void drbd_suspend_io(struct drbd_device *device);
1303 extern void drbd_resume_io(struct drbd_device *device);
1304 extern char *ppsize(char *buf, unsigned long long size);
1305 extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int);
1306 enum determine_dev_size {
1307 DS_ERROR_SHRINK = -3,
1308 DS_ERROR_SPACE_MD = -2,
1313 DS_GREW_FROM_ZERO = 3,
1315 extern enum determine_dev_size
1316 drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local);
1317 extern void resync_after_online_grow(struct drbd_device *);
1318 extern void drbd_reconsider_max_bio_size(struct drbd_device *device, struct drbd_backing_dev *bdev);
1319 extern enum drbd_state_rv drbd_set_role(struct drbd_device *device,
1320 enum drbd_role new_role,
1322 extern bool conn_try_outdate_peer(struct drbd_connection *connection);
1323 extern void conn_try_outdate_peer_async(struct drbd_connection *connection);
1324 extern int drbd_khelper(struct drbd_device *device, char *cmd);
1327 /* bi_end_io handlers */
1328 extern void drbd_md_io_complete(struct bio *bio, int error);
1329 extern void drbd_peer_request_endio(struct bio *bio, int error);
1330 extern void drbd_request_endio(struct bio *bio, int error);
1331 extern int drbd_worker(struct drbd_thread *thi);
1332 enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor);
1333 void drbd_resync_after_changed(struct drbd_device *device);
1334 extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side);
1335 extern void resume_next_sg(struct drbd_device *device);
1336 extern void suspend_other_sg(struct drbd_device *device);
1337 extern int drbd_resync_finished(struct drbd_device *device);
1338 /* maybe rather drbd_main.c ? */
1339 extern void *drbd_md_get_buffer(struct drbd_device *device);
1340 extern void drbd_md_put_buffer(struct drbd_device *device);
1341 extern int drbd_md_sync_page_io(struct drbd_device *device,
1342 struct drbd_backing_dev *bdev, sector_t sector, int rw);
1343 extern void drbd_ov_out_of_sync_found(struct drbd_device *, sector_t, int);
1344 extern void wait_until_done_or_force_detached(struct drbd_device *device,
1345 struct drbd_backing_dev *bdev, unsigned int *done);
1346 extern void drbd_rs_controller_reset(struct drbd_device *device);
1348 static inline void ov_out_of_sync_print(struct drbd_device *device)
1350 if (device->ov_last_oos_size) {
1351 drbd_err(device, "Out of sync: start=%llu, size=%lu (sectors)\n",
1352 (unsigned long long)device->ov_last_oos_start,
1353 (unsigned long)device->ov_last_oos_size);
1355 device->ov_last_oos_size = 0;
1359 extern void drbd_csum_bio(struct crypto_hash *, struct bio *, void *);
1360 extern void drbd_csum_ee(struct crypto_hash *, struct drbd_peer_request *, void *);
1361 /* worker callbacks */
1362 extern int w_e_end_data_req(struct drbd_work *, int);
1363 extern int w_e_end_rsdata_req(struct drbd_work *, int);
1364 extern int w_e_end_csum_rs_req(struct drbd_work *, int);
1365 extern int w_e_end_ov_reply(struct drbd_work *, int);
1366 extern int w_e_end_ov_req(struct drbd_work *, int);
1367 extern int w_ov_finished(struct drbd_work *, int);
1368 extern int w_resync_timer(struct drbd_work *, int);
1369 extern int w_send_write_hint(struct drbd_work *, int);
1370 extern int w_send_dblock(struct drbd_work *, int);
1371 extern int w_send_read_req(struct drbd_work *, int);
1372 extern int w_e_reissue(struct drbd_work *, int);
1373 extern int w_restart_disk_io(struct drbd_work *, int);
1374 extern int w_send_out_of_sync(struct drbd_work *, int);
1375 extern int w_start_resync(struct drbd_work *, int);
1377 extern void resync_timer_fn(unsigned long data);
1378 extern void start_resync_timer_fn(unsigned long data);
1380 extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req);
1382 /* drbd_receiver.c */
1383 extern int drbd_receiver(struct drbd_thread *thi);
1384 extern int drbd_asender(struct drbd_thread *thi);
1385 extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device);
1386 extern bool drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector);
1387 extern int drbd_submit_peer_request(struct drbd_device *,
1388 struct drbd_peer_request *, const unsigned,
1390 extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *);
1391 extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64,
1392 sector_t, unsigned int,
1394 gfp_t) __must_hold(local);
1395 extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *,
1397 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1398 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
1399 extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool);
1400 extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
1401 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
1402 extern int drbd_connected(struct drbd_peer_device *);
1404 /* Yes, there is kernel_setsockopt, but only since 2.6.18.
1405 * So we have our own copy of it here. */
1406 static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1407 char *optval, int optlen)
1409 mm_segment_t oldfs = get_fs();
1410 char __user *uoptval;
1413 uoptval = (char __user __force *)optval;
1416 if (level == SOL_SOCKET)
1417 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
1419 err = sock->ops->setsockopt(sock, level, optname, uoptval,
1425 static inline void drbd_tcp_cork(struct socket *sock)
1428 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1429 (char*)&val, sizeof(val));
1432 static inline void drbd_tcp_uncork(struct socket *sock)
1435 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1436 (char*)&val, sizeof(val));
1439 static inline void drbd_tcp_nodelay(struct socket *sock)
1442 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1443 (char*)&val, sizeof(val));
1446 static inline void drbd_tcp_quickack(struct socket *sock)
1449 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1450 (char*)&val, sizeof(val));
1453 /* sets the number of 512 byte sectors of our virtual device */
1454 static inline void drbd_set_my_capacity(struct drbd_device *device,
1457 /* set_capacity(device->this_bdev->bd_disk, size); */
1458 set_capacity(device->vdisk, size);
1459 device->this_bdev->bd_inode->i_size = (loff_t)size << 9;
1463 * used to submit our private bio
1465 static inline void drbd_generic_make_request(struct drbd_device *device,
1466 int fault_type, struct bio *bio)
1469 if (!bio->bi_bdev) {
1470 printk(KERN_ERR "drbd%d: drbd_generic_make_request: "
1471 "bio->bi_bdev == NULL\n",
1472 device_to_minor(device));
1473 bio_endio(bio, -ENODEV);
1477 if (drbd_insert_fault(device, fault_type))
1478 bio_endio(bio, -EIO);
1480 generic_make_request(bio);
1483 void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev,
1484 enum write_ordering_e wo);
1487 extern struct proc_dir_entry *drbd_proc;
1488 extern const struct file_operations drbd_proc_fops;
1489 extern const char *drbd_conn_str(enum drbd_conns s);
1490 extern const char *drbd_role_str(enum drbd_role s);
1493 extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i);
1494 extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i);
1495 extern void drbd_al_begin_io_commit(struct drbd_device *device);
1496 extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i);
1497 extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i);
1498 extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i);
1499 extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector);
1500 extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector);
1501 extern int drbd_try_rs_begin_io(struct drbd_device *device, sector_t sector);
1502 extern void drbd_rs_cancel_all(struct drbd_device *device);
1503 extern int drbd_rs_del_all(struct drbd_device *device);
1504 extern void drbd_rs_failed_io(struct drbd_device *device,
1505 sector_t sector, int size);
1506 extern void drbd_advance_rs_marks(struct drbd_device *device, unsigned long still_to_go);
1508 enum update_sync_bits_mode { RECORD_RS_FAILED, SET_OUT_OF_SYNC, SET_IN_SYNC };
1509 extern int __drbd_change_sync(struct drbd_device *device, sector_t sector, int size,
1510 enum update_sync_bits_mode mode,
1511 const char *file, const unsigned int line);
1512 #define drbd_set_in_sync(device, sector, size) \
1513 __drbd_change_sync(device, sector, size, SET_IN_SYNC, __FILE__, __LINE__)
1514 #define drbd_set_out_of_sync(device, sector, size) \
1515 __drbd_change_sync(device, sector, size, SET_OUT_OF_SYNC, __FILE__, __LINE__)
1516 #define drbd_rs_failed_io(device, sector, size) \
1517 __drbd_change_sync(device, sector, size, RECORD_RS_FAILED, __FILE__, __LINE__)
1518 extern void drbd_al_shrink(struct drbd_device *device);
1519 extern int drbd_initialize_al(struct drbd_device *, void *);
1522 /* state info broadcast */
1524 enum drbd_state_info_bcast_reason sib_reason;
1528 unsigned helper_exit_code;
1531 union drbd_state os;
1532 union drbd_state ns;
1536 void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib);
1539 * inline helper functions
1540 *************************/
1542 /* see also page_chain_add and friends in drbd_receiver.c */
1543 static inline struct page *page_chain_next(struct page *page)
1545 return (struct page *)page_private(page);
1547 #define page_chain_for_each(page) \
1548 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1549 page = page_chain_next(page))
1550 #define page_chain_for_each_safe(page, n) \
1551 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1554 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req)
1556 struct page *page = peer_req->pages;
1557 page_chain_for_each(page) {
1558 if (page_count(page) > 1)
1564 static inline enum drbd_state_rv
1565 _drbd_set_state(struct drbd_device *device, union drbd_state ns,
1566 enum chg_state_flags flags, struct completion *done)
1568 enum drbd_state_rv rv;
1570 read_lock(&global_state_lock);
1571 rv = __drbd_set_state(device, ns, flags, done);
1572 read_unlock(&global_state_lock);
1577 static inline union drbd_state drbd_read_state(struct drbd_device *device)
1579 struct drbd_resource *resource = device->resource;
1580 union drbd_state rv;
1582 rv.i = device->state.i;
1583 rv.susp = resource->susp;
1584 rv.susp_nod = resource->susp_nod;
1585 rv.susp_fen = resource->susp_fen;
1590 enum drbd_force_detach_flags {
1597 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1598 static inline void __drbd_chk_io_error_(struct drbd_device *device,
1599 enum drbd_force_detach_flags df,
1602 enum drbd_io_error_p ep;
1605 ep = rcu_dereference(device->ldev->disk_conf)->on_io_error;
1608 case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */
1609 if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) {
1610 if (__ratelimit(&drbd_ratelimit_state))
1611 drbd_err(device, "Local IO failed in %s.\n", where);
1612 if (device->state.disk > D_INCONSISTENT)
1613 _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL);
1616 /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */
1618 case EP_CALL_HELPER:
1619 /* Remember whether we saw a READ or WRITE error.
1621 * Recovery of the affected area for WRITE failure is covered
1622 * by the activity log.
1623 * READ errors may fall outside that area though. Certain READ
1624 * errors can be "healed" by writing good data to the affected
1625 * blocks, which triggers block re-allocation in lower layers.
1627 * If we can not write the bitmap after a READ error,
1628 * we may need to trigger a full sync (see w_go_diskless()).
1630 * Force-detach is not really an IO error, but rather a
1631 * desperate measure to try to deal with a completely
1632 * unresponsive lower level IO stack.
1633 * Still it should be treated as a WRITE error.
1635 * Meta IO error is always WRITE error:
1636 * we read meta data only once during attach,
1637 * which will fail in case of errors.
1639 set_bit(WAS_IO_ERROR, &device->flags);
1640 if (df == DRBD_READ_ERROR)
1641 set_bit(WAS_READ_ERROR, &device->flags);
1642 if (df == DRBD_FORCE_DETACH)
1643 set_bit(FORCE_DETACH, &device->flags);
1644 if (device->state.disk > D_FAILED) {
1645 _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL);
1647 "Local IO failed in %s. Detaching...\n", where);
1654 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1655 * @device: DRBD device.
1656 * @error: Error code passed to the IO completion callback
1657 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1659 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1661 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1662 static inline void drbd_chk_io_error_(struct drbd_device *device,
1663 int error, enum drbd_force_detach_flags forcedetach, const char *where)
1666 unsigned long flags;
1667 spin_lock_irqsave(&device->resource->req_lock, flags);
1668 __drbd_chk_io_error_(device, forcedetach, where);
1669 spin_unlock_irqrestore(&device->resource->req_lock, flags);
1675 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1676 * @bdev: Meta data block device.
1678 * BTW, for internal meta data, this happens to be the maximum capacity
1679 * we could agree upon with our peer node.
1681 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1683 switch (bdev->md.meta_dev_idx) {
1684 case DRBD_MD_INDEX_INTERNAL:
1685 case DRBD_MD_INDEX_FLEX_INT:
1686 return bdev->md.md_offset + bdev->md.bm_offset;
1687 case DRBD_MD_INDEX_FLEX_EXT:
1689 return bdev->md.md_offset;
1694 * drbd_md_last_sector() - Return the last sector number of the meta data area
1695 * @bdev: Meta data block device.
1697 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1699 switch (bdev->md.meta_dev_idx) {
1700 case DRBD_MD_INDEX_INTERNAL:
1701 case DRBD_MD_INDEX_FLEX_INT:
1702 return bdev->md.md_offset + MD_4kB_SECT -1;
1703 case DRBD_MD_INDEX_FLEX_EXT:
1705 return bdev->md.md_offset + bdev->md.md_size_sect -1;
1709 /* Returns the number of 512 byte sectors of the device */
1710 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1712 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1713 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0;
1717 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1718 * @bdev: Meta data block device.
1720 * returns the capacity we announce to out peer. we clip ourselves at the
1721 * various MAX_SECTORS, because if we don't, current implementation will
1722 * oops sooner or later
1724 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1728 switch (bdev->md.meta_dev_idx) {
1729 case DRBD_MD_INDEX_INTERNAL:
1730 case DRBD_MD_INDEX_FLEX_INT:
1731 s = drbd_get_capacity(bdev->backing_bdev)
1732 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1733 drbd_md_first_sector(bdev))
1736 case DRBD_MD_INDEX_FLEX_EXT:
1737 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1738 drbd_get_capacity(bdev->backing_bdev));
1739 /* clip at maximum size the meta device can support */
1740 s = min_t(sector_t, s,
1741 BM_EXT_TO_SECT(bdev->md.md_size_sect
1742 - bdev->md.bm_offset));
1745 s = min_t(sector_t, DRBD_MAX_SECTORS,
1746 drbd_get_capacity(bdev->backing_bdev));
1752 * drbd_md_ss() - Return the sector number of our meta data super block
1753 * @bdev: Meta data block device.
1755 static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev)
1757 const int meta_dev_idx = bdev->md.meta_dev_idx;
1759 if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT)
1762 /* Since drbd08, internal meta data is always "flexible".
1763 * position: last 4k aligned block of 4k size */
1764 if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL ||
1765 meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)
1766 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8;
1768 /* external, some index; this is the old fixed size layout */
1769 return MD_128MB_SECT * bdev->md.meta_dev_idx;
1773 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1775 unsigned long flags;
1776 spin_lock_irqsave(&q->q_lock, flags);
1777 list_add_tail(&w->list, &q->q);
1778 spin_unlock_irqrestore(&q->q_lock, flags);
1779 wake_up(&q->q_wait);
1783 drbd_device_post_work(struct drbd_device *device, int work_bit)
1785 if (!test_and_set_bit(work_bit, &device->flags)) {
1786 struct drbd_connection *connection =
1787 first_peer_device(device)->connection;
1788 struct drbd_work_queue *q = &connection->sender_work;
1789 if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags))
1790 wake_up(&q->q_wait);
1794 extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue);
1796 static inline void wake_asender(struct drbd_connection *connection)
1798 if (test_bit(SIGNAL_ASENDER, &connection->flags))
1799 force_sig(DRBD_SIG, connection->asender.task);
1802 static inline void request_ping(struct drbd_connection *connection)
1804 set_bit(SEND_PING, &connection->flags);
1805 wake_asender(connection);
1808 extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *);
1809 extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *);
1810 extern int conn_send_command(struct drbd_connection *, struct drbd_socket *,
1811 enum drbd_packet, unsigned int, void *,
1813 extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *,
1814 enum drbd_packet, unsigned int, void *,
1817 extern int drbd_send_ping(struct drbd_connection *connection);
1818 extern int drbd_send_ping_ack(struct drbd_connection *connection);
1819 extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state);
1820 extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state);
1822 static inline void drbd_thread_stop(struct drbd_thread *thi)
1824 _drbd_thread_stop(thi, false, true);
1827 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
1829 _drbd_thread_stop(thi, false, false);
1832 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
1834 _drbd_thread_stop(thi, true, false);
1837 /* counts how many answer packets packets we expect from our peer,
1838 * for either explicit application requests,
1839 * or implicit barrier packets as necessary.
1842 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1843 * it is much easier and equally valid to count what we queue for the
1844 * worker, even before it actually was queued or send.
1845 * (drbd_make_request_common; recovery path on read io-error)
1847 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1848 * _req_mod(req, DATA_RECEIVED)
1849 * [from receive_DataReply]
1850 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1851 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1852 * for some reason it is NOT decreased in got_NegAck,
1853 * but in the resulting cleanup code from report_params.
1854 * we should try to remember the reason for that...
1855 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
1856 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
1857 * [from tl_clear_barrier]
1859 static inline void inc_ap_pending(struct drbd_device *device)
1861 atomic_inc(&device->ap_pending_cnt);
1864 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \
1865 if (atomic_read(&device->which) < 0) \
1866 drbd_err(device, "in %s:%d: " #which " = %d < 0 !\n", \
1868 atomic_read(&device->which))
1870 #define dec_ap_pending(device) _dec_ap_pending(device, __func__, __LINE__)
1871 static inline void _dec_ap_pending(struct drbd_device *device, const char *func, int line)
1873 if (atomic_dec_and_test(&device->ap_pending_cnt))
1874 wake_up(&device->misc_wait);
1875 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line);
1878 /* counts how many resync-related answers we still expect from the peer
1880 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1881 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
1882 * (or P_NEG_ACK with ID_SYNCER)
1884 static inline void inc_rs_pending(struct drbd_device *device)
1886 atomic_inc(&device->rs_pending_cnt);
1889 #define dec_rs_pending(device) _dec_rs_pending(device, __func__, __LINE__)
1890 static inline void _dec_rs_pending(struct drbd_device *device, const char *func, int line)
1892 atomic_dec(&device->rs_pending_cnt);
1893 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line);
1896 /* counts how many answers we still need to send to the peer.
1898 * receive_Data unless protocol A;
1899 * we need to send a P_RECV_ACK (proto B)
1900 * or P_WRITE_ACK (proto C)
1901 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1902 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1903 * receive_Barrier_* we need to send a P_BARRIER_ACK
1905 static inline void inc_unacked(struct drbd_device *device)
1907 atomic_inc(&device->unacked_cnt);
1910 #define dec_unacked(device) _dec_unacked(device, __func__, __LINE__)
1911 static inline void _dec_unacked(struct drbd_device *device, const char *func, int line)
1913 atomic_dec(&device->unacked_cnt);
1914 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
1917 #define sub_unacked(device, n) _sub_unacked(device, n, __func__, __LINE__)
1918 static inline void _sub_unacked(struct drbd_device *device, int n, const char *func, int line)
1920 atomic_sub(n, &device->unacked_cnt);
1921 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
1924 static inline bool is_sync_state(enum drbd_conns connection_state)
1927 (connection_state == C_SYNC_SOURCE
1928 || connection_state == C_SYNC_TARGET
1929 || connection_state == C_PAUSED_SYNC_S
1930 || connection_state == C_PAUSED_SYNC_T);
1934 * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev
1937 * You have to call put_ldev() when finished working with device->ldev.
1939 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
1940 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
1942 static inline void put_ldev(struct drbd_device *device)
1944 enum drbd_disk_state ds = device->state.disk;
1945 /* We must check the state *before* the atomic_dec becomes visible,
1946 * or we have a theoretical race where someone hitting zero,
1947 * while state still D_FAILED, will then see D_DISKLESS in the
1948 * condition below and calling into destroy, where he must not, yet. */
1949 int i = atomic_dec_return(&device->local_cnt);
1951 /* This may be called from some endio handler,
1952 * so we must not sleep here. */
1955 D_ASSERT(device, i >= 0);
1957 if (ds == D_DISKLESS)
1958 /* even internal references gone, safe to destroy */
1959 drbd_device_post_work(device, DESTROY_DISK);
1961 /* all application IO references gone. */
1962 if (!test_and_set_bit(GOING_DISKLESS, &device->flags))
1963 drbd_device_post_work(device, GO_DISKLESS);
1964 wake_up(&device->misc_wait);
1969 static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins)
1973 /* never get a reference while D_DISKLESS */
1974 if (device->state.disk == D_DISKLESS)
1977 atomic_inc(&device->local_cnt);
1978 io_allowed = (device->state.disk >= mins);
1984 extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins);
1987 /* this throttles on-the-fly application requests
1988 * according to max_buffers settings;
1989 * maybe re-implement using semaphores? */
1990 static inline int drbd_get_max_buffers(struct drbd_device *device)
1992 struct net_conf *nc;
1996 nc = rcu_dereference(first_peer_device(device)->connection->net_conf);
1997 mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */
2003 static inline int drbd_state_is_stable(struct drbd_device *device)
2005 union drbd_dev_state s = device->state;
2007 /* DO NOT add a default clause, we want the compiler to warn us
2008 * for any newly introduced state we may have forgotten to add here */
2010 switch ((enum drbd_conns)s.conn) {
2011 /* new io only accepted when there is no connection, ... */
2013 case C_WF_CONNECTION:
2014 /* ... or there is a well established connection. */
2020 case C_PAUSED_SYNC_S:
2021 case C_PAUSED_SYNC_T:
2024 /* transitional states, IO allowed */
2025 case C_DISCONNECTING:
2029 case C_NETWORK_FAILURE:
2030 case C_PROTOCOL_ERROR:
2032 case C_WF_REPORT_PARAMS:
2033 case C_STARTING_SYNC_S:
2034 case C_STARTING_SYNC_T:
2037 /* Allow IO in BM exchange states with new protocols */
2039 if (first_peer_device(device)->connection->agreed_pro_version < 96)
2043 /* no new io accepted in these states */
2045 case C_WF_SYNC_UUID:
2051 switch ((enum drbd_disk_state)s.disk) {
2053 case D_INCONSISTENT:
2058 /* disk state is stable as well. */
2061 /* no new io accepted during transitional states */
2073 static inline int drbd_suspended(struct drbd_device *device)
2075 struct drbd_resource *resource = device->resource;
2077 return resource->susp || resource->susp_fen || resource->susp_nod;
2080 static inline bool may_inc_ap_bio(struct drbd_device *device)
2082 int mxb = drbd_get_max_buffers(device);
2084 if (drbd_suspended(device))
2086 if (test_bit(SUSPEND_IO, &device->flags))
2089 /* to avoid potential deadlock or bitmap corruption,
2090 * in various places, we only allow new application io
2091 * to start during "stable" states. */
2093 /* no new io accepted when attaching or detaching the disk */
2094 if (!drbd_state_is_stable(device))
2097 /* since some older kernels don't have atomic_add_unless,
2098 * and we are within the spinlock anyways, we have this workaround. */
2099 if (atomic_read(&device->ap_bio_cnt) > mxb)
2101 if (test_bit(BITMAP_IO, &device->flags))
2106 static inline bool inc_ap_bio_cond(struct drbd_device *device)
2110 spin_lock_irq(&device->resource->req_lock);
2111 rv = may_inc_ap_bio(device);
2113 atomic_inc(&device->ap_bio_cnt);
2114 spin_unlock_irq(&device->resource->req_lock);
2119 static inline void inc_ap_bio(struct drbd_device *device)
2122 * as long as the device is suspended
2123 * until the bitmap is no longer on the fly during connection
2124 * handshake as long as we would exceed the max_buffer limit.
2126 * to avoid races with the reconnect code,
2127 * we need to atomic_inc within the spinlock. */
2129 wait_event(device->misc_wait, inc_ap_bio_cond(device));
2132 static inline void dec_ap_bio(struct drbd_device *device)
2134 int mxb = drbd_get_max_buffers(device);
2135 int ap_bio = atomic_dec_return(&device->ap_bio_cnt);
2137 D_ASSERT(device, ap_bio >= 0);
2139 if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) {
2140 if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags))
2141 drbd_queue_work(&first_peer_device(device)->
2142 connection->sender_work,
2143 &device->bm_io_work.w);
2146 /* this currently does wake_up for every dec_ap_bio!
2147 * maybe rather introduce some type of hysteresis?
2148 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2150 wake_up(&device->misc_wait);
2153 static inline bool verify_can_do_stop_sector(struct drbd_device *device)
2155 return first_peer_device(device)->connection->agreed_pro_version >= 97 &&
2156 first_peer_device(device)->connection->agreed_pro_version != 100;
2159 static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val)
2161 int changed = device->ed_uuid != val;
2162 device->ed_uuid = val;
2166 static inline int drbd_queue_order_type(struct drbd_device *device)
2168 /* sorry, we currently have no working implementation
2169 * of distributed TCQ stuff */
2170 #ifndef QUEUE_ORDERED_NONE
2171 #define QUEUE_ORDERED_NONE 0
2173 return QUEUE_ORDERED_NONE;
2176 static inline void drbd_md_flush(struct drbd_device *device)
2180 if (device->ldev == NULL) {
2181 drbd_warn(device, "device->ldev == NULL in drbd_md_flush\n");
2185 if (test_bit(MD_NO_FUA, &device->flags))
2188 r = blkdev_issue_flush(device->ldev->md_bdev, GFP_NOIO, NULL);
2190 set_bit(MD_NO_FUA, &device->flags);
2191 drbd_err(device, "meta data flush failed with status %d, disabling md-flushes\n", r);
2195 static inline struct drbd_connection *first_connection(struct drbd_resource *resource)
2197 return list_first_entry_or_null(&resource->connections,
2198 struct drbd_connection, connections);