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 /* I don't remember why XCPU ...
81 * This is used to wake the asender,
82 * and to interrupt sending the sending task
85 #define DRBD_SIG SIGXCPU
87 /* This is used to stop/restart our threads.
88 * Cannot use SIGTERM nor SIGKILL, since these
89 * are sent out by init on runlevel changes
90 * I choose SIGHUP for now.
92 #define DRBD_SIGKILL SIGHUP
94 #define ID_IN_SYNC (4711ULL)
95 #define ID_OUT_OF_SYNC (4712ULL)
96 #define ID_SYNCER (-1ULL)
98 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
101 struct drbd_connection;
103 #define __drbd_printk_device(level, device, fmt, args...) \
104 dev_printk(level, disk_to_dev((device)->vdisk), fmt, ## args)
105 #define __drbd_printk_peer_device(level, peer_device, fmt, args...) \
106 dev_printk(level, disk_to_dev((peer_device)->device->vdisk), fmt, ## args)
107 #define __drbd_printk_resource(level, resource, fmt, args...) \
108 printk(level "drbd %s: " fmt, (resource)->name, ## args)
109 #define __drbd_printk_connection(level, connection, fmt, args...) \
110 printk(level "drbd %s: " fmt, (connection)->resource->name, ## args)
112 void drbd_printk_with_wrong_object_type(void);
114 #define __drbd_printk_if_same_type(obj, type, func, level, fmt, args...) \
115 (__builtin_types_compatible_p(typeof(obj), type) || \
116 __builtin_types_compatible_p(typeof(obj), const type)), \
117 func(level, (const type)(obj), fmt, ## args)
119 #define drbd_printk(level, obj, fmt, args...) \
120 __builtin_choose_expr( \
121 __drbd_printk_if_same_type(obj, struct drbd_device *, \
122 __drbd_printk_device, level, fmt, ## args), \
123 __builtin_choose_expr( \
124 __drbd_printk_if_same_type(obj, struct drbd_resource *, \
125 __drbd_printk_resource, level, fmt, ## args), \
126 __builtin_choose_expr( \
127 __drbd_printk_if_same_type(obj, struct drbd_connection *, \
128 __drbd_printk_connection, level, fmt, ## args), \
129 __builtin_choose_expr( \
130 __drbd_printk_if_same_type(obj, struct drbd_peer_device *, \
131 __drbd_printk_peer_device, level, fmt, ## args), \
132 drbd_printk_with_wrong_object_type()))))
134 #define drbd_dbg(obj, fmt, args...) \
135 drbd_printk(KERN_DEBUG, obj, fmt, ## args)
136 #define drbd_alert(obj, fmt, args...) \
137 drbd_printk(KERN_ALERT, obj, fmt, ## args)
138 #define drbd_err(obj, fmt, args...) \
139 drbd_printk(KERN_ERR, obj, fmt, ## args)
140 #define drbd_warn(obj, fmt, args...) \
141 drbd_printk(KERN_WARNING, obj, fmt, ## args)
142 #define drbd_info(obj, fmt, args...) \
143 drbd_printk(KERN_INFO, obj, fmt, ## args)
144 #define drbd_emerg(obj, fmt, args...) \
145 drbd_printk(KERN_EMERG, obj, fmt, ## args)
147 #define dynamic_drbd_dbg(device, fmt, args...) \
148 dynamic_dev_dbg(disk_to_dev(device->vdisk), fmt, ## args)
150 #define D_ASSERT(device, exp) do { \
152 drbd_err(device, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__); \
156 * expect - Make an assertion
158 * Unlike the assert macro, this macro returns a boolean result.
160 #define expect(exp) ({ \
161 bool _bool = (exp); \
163 drbd_err(device, "ASSERTION %s FAILED in %s\n", \
168 /* Defines to control fault insertion */
170 DRBD_FAULT_MD_WR = 0, /* meta data write */
171 DRBD_FAULT_MD_RD = 1, /* read */
172 DRBD_FAULT_RS_WR = 2, /* resync */
173 DRBD_FAULT_RS_RD = 3,
174 DRBD_FAULT_DT_WR = 4, /* data */
175 DRBD_FAULT_DT_RD = 5,
176 DRBD_FAULT_DT_RA = 6, /* data read ahead */
177 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */
178 DRBD_FAULT_AL_EE = 8, /* alloc ee */
179 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
185 _drbd_insert_fault(struct drbd_device *device, unsigned int type);
188 drbd_insert_fault(struct drbd_device *device, unsigned int type) {
189 #ifdef CONFIG_DRBD_FAULT_INJECTION
191 (enable_faults & (1<<type)) &&
192 _drbd_insert_fault(device, type);
198 /* integer division, round _UP_ to the next integer */
199 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
200 /* usual integer division */
201 #define div_floor(A, B) ((A)/(B))
203 extern struct ratelimit_state drbd_ratelimit_state;
204 extern struct idr drbd_devices; /* RCU, updates: genl_lock() */
205 extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */
207 extern const char *cmdname(enum drbd_packet cmd);
209 /* for sending/receiving the bitmap,
210 * possibly in some encoding scheme */
213 * stores total bits and long words
214 * of the bitmap, so we don't need to
215 * call the accessor functions over and again. */
216 unsigned long bm_bits;
217 unsigned long bm_words;
218 /* during xfer, current position within the bitmap */
219 unsigned long bit_offset;
220 unsigned long word_offset;
222 /* statistics; index: (h->command == P_BITMAP) */
227 extern void INFO_bm_xfer_stats(struct drbd_device *device,
228 const char *direction, struct bm_xfer_ctx *c);
230 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
232 /* word_offset counts "native long words" (32 or 64 bit),
234 * Encoded packet may end at an unaligned bit offset.
235 * In case a fallback clear text packet is transmitted in
236 * between, we adjust this offset back to the last 64bit
237 * aligned "native long word", which makes coding and decoding
238 * the plain text bitmap much more convenient. */
239 #if BITS_PER_LONG == 64
240 c->word_offset = c->bit_offset >> 6;
241 #elif BITS_PER_LONG == 32
242 c->word_offset = c->bit_offset >> 5;
243 c->word_offset &= ~(1UL);
245 # error "unsupported BITS_PER_LONG"
249 extern unsigned int drbd_header_size(struct drbd_connection *connection);
251 /**********************************************************************/
252 enum drbd_thread_state {
261 struct task_struct *task;
262 struct completion stop;
263 enum drbd_thread_state t_state;
264 int (*function) (struct drbd_thread *);
265 struct drbd_resource *resource;
266 struct drbd_connection *connection;
271 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
273 /* THINK testing the t_state seems to be uncritical in all cases
274 * (but thread_{start,stop}), so we can read it *without* the lock.
282 struct list_head list;
283 int (*cb)(struct drbd_work *, int cancel);
286 struct drbd_device_work {
288 struct drbd_device *device;
291 #include "drbd_interval.h"
293 extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *);
295 extern void lock_all_resources(void);
296 extern void unlock_all_resources(void);
298 struct drbd_request {
300 struct drbd_device *device;
302 /* if local IO is not allowed, will be NULL.
303 * if local IO _is_ allowed, holds the locally submitted bio clone,
304 * or, after local IO completion, the ERR_PTR(error).
305 * see drbd_request_endio(). */
306 struct bio *private_bio;
308 struct drbd_interval i;
310 /* epoch: used to check on "completion" whether this req was in
311 * the current epoch, and we therefore have to close it,
312 * causing a p_barrier packet to be send, starting a new epoch.
314 * This corresponds to "barrier" in struct p_barrier[_ack],
315 * and to "barrier_nr" in struct drbd_epoch (and various
316 * comments/function parameters/local variable names).
320 struct list_head tl_requests; /* ring list in the transfer log */
321 struct bio *master_bio; /* master bio pointer */
323 /* see struct drbd_device */
324 struct list_head req_pending_master_completion;
325 struct list_head req_pending_local;
327 /* for generic IO accounting */
328 unsigned long start_jif;
330 /* for DRBD internal statistics */
332 /* Minimal set of time stamps to determine if we wait for activity log
333 * transactions, local disk or peer. 32 bit "jiffies" are good enough,
334 * we don't expect a DRBD request to be stalled for several month.
337 /* before actual request processing */
338 unsigned long in_actlog_jif;
341 unsigned long pre_submit_jif;
344 unsigned long pre_send_jif;
345 unsigned long acked_jif;
346 unsigned long net_done_jif;
348 /* Possibly even more detail to track each phase:
349 * master_completion_jif
350 * how long did it take to complete the master bio
351 * (application visible latency)
353 * how long the master bio was blocked until we finally allocated
356 * how long did we wait for activity log transactions
359 * when did we finally queue it for sending
361 * when did we start sending it
363 * how long did we block in the network stack trying to send it
365 * when did we receive (or fake, in protocol A) a remote ACK
367 * when did we receive final acknowledgement (P_BARRIER_ACK),
368 * or decide, e.g. on connection loss, that we do no longer expect
369 * anything from this peer for this request.
373 * when did we start submiting to the lower level device,
374 * and how long did we block in that submit function
375 * local_completion_jif
376 * how long did it take the lower level device to complete this request
380 /* once it hits 0, we may complete the master_bio */
381 atomic_t completion_ref;
382 /* once it hits 0, we may destroy this drbd_request object */
385 unsigned rq_state; /* see comments above _req_mod() */
389 struct drbd_connection *connection;
390 struct list_head list;
391 unsigned int barrier_nr;
392 atomic_t epoch_size; /* increased on every request added. */
393 atomic_t active; /* increased on every req. added, and dec on every finished. */
397 /* Prototype declaration of function defined in drbd_receiver.c */
398 int drbdd_init(struct drbd_thread *);
399 int drbd_asender(struct drbd_thread *);
401 /* drbd_epoch flag bits */
403 DE_HAVE_BARRIER_NUMBER,
410 EV_CLEANUP = 32, /* used as flag */
418 struct drbd_peer_request {
420 struct drbd_peer_device *peer_device;
421 struct drbd_epoch *epoch; /* for writes */
423 atomic_t pending_bios;
424 struct drbd_interval i;
425 /* see comments on ee flag bits below */
427 unsigned long submit_jif;
430 struct digest_info *digest;
435 * While corresponding bios are in flight, the only modification will be
436 * set_bit WAS_ERROR, which has to be atomic.
437 * If no bios are in flight yet, or all have been completed,
438 * non-atomic modification to ee->flags is ok.
441 __EE_CALL_AL_COMPLETE_IO,
442 __EE_MAY_SET_IN_SYNC,
444 /* is this a TRIM aka REQ_DISCARD? */
446 /* our lower level cannot handle trim,
447 * and we want to fall back to zeroout instead */
448 __EE_IS_TRIM_USE_ZEROOUT,
450 /* In case a barrier failed,
451 * we need to resubmit without the barrier flag. */
454 /* we may have several bios per peer request.
455 * if any of those fail, we set this flag atomically
456 * from the endio callback */
459 /* This ee has a pointer to a digest instead of a block id */
462 /* Conflicting local requests need to be restarted after this request */
463 __EE_RESTART_REQUESTS,
465 /* The peer wants a write ACK for this (wire proto C) */
468 /* Is set when net_conf had two_primaries set while creating this peer_req */
469 __EE_IN_INTERVAL_TREE,
472 /* has this been submitted, or does it still wait for something else? */
475 /* this is/was a write request */
478 /* this originates from application on peer
479 * (not some resync or verify or other DRBD internal request) */
482 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
483 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
484 #define EE_IS_TRIM (1<<__EE_IS_TRIM)
485 #define EE_IS_TRIM_USE_ZEROOUT (1<<__EE_IS_TRIM_USE_ZEROOUT)
486 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
487 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
488 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
489 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
490 #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK)
491 #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE)
492 #define EE_SUBMITTED (1<<__EE_SUBMITTED)
493 #define EE_WRITE (1<<__EE_WRITE)
494 #define EE_APPLICATION (1<<__EE_APPLICATION)
496 /* flag bits per device */
498 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
499 MD_DIRTY, /* current uuids and flags not yet on disk */
500 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
503 CRASHED_PRIMARY, /* This node was a crashed primary.
504 * Gets cleared when the state.conn
505 * goes into C_CONNECTED state. */
508 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */
510 SUSPEND_IO, /* suspend application io */
511 BITMAP_IO, /* suspend application io;
512 once no more io in flight, start bitmap io */
513 BITMAP_IO_QUEUED, /* Started bitmap IO */
514 WAS_IO_ERROR, /* Local disk failed, returned IO error */
515 WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */
516 FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */
517 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
518 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
519 * the peer, if it changed there as well. */
520 NEW_CUR_UUID, /* Create new current UUID when thawing IO */
521 AL_SUSPENDED, /* Activity logging is currently suspended. */
522 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */
523 B_RS_H_DONE, /* Before resync handler done (already executed) */
524 DISCARD_MY_DATA, /* discard_my_data flag per volume */
527 FLUSH_PENDING, /* if set, device->flush_jif is when we submitted that flush
528 * from drbd_flush_after_epoch() */
530 /* cleared only after backing device related structures have been destroyed. */
531 GOING_DISKLESS, /* Disk is being detached, because of io-error, or admin request. */
533 /* to be used in drbd_device_post_work() */
534 GO_DISKLESS, /* tell worker to schedule cleanup before detach */
535 DESTROY_DISK, /* tell worker to close backing devices and destroy related structures. */
536 MD_SYNC, /* tell worker to call drbd_md_sync() */
537 RS_START, /* tell worker to start resync/OV */
538 RS_PROGRESS, /* tell worker that resync made significant progress */
539 RS_DONE, /* tell worker that resync is done */
542 struct drbd_bitmap; /* opaque for drbd_device */
544 /* definition of bits in bm_flags to be used in drbd_bm_lock
545 * and drbd_bitmap_io and friends. */
547 /* do we need to kfree, or vfree bm_pages? */
548 BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */
550 /* currently locked for bulk operation */
551 BM_LOCKED_MASK = 0xf,
553 /* in detail, that is: */
558 /* so we can mark it locked for bulk operation,
559 * and still allow all non-bulk operations */
562 /* (test bit, count bit) allowed (common case) */
563 BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED,
565 /* testing bits, as well as setting new bits allowed, but clearing bits
566 * would be unexpected. Used during bitmap receive. Setting new bits
567 * requires sending of "out-of-sync" information, though. */
568 BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED,
570 /* for drbd_bm_write_copy_pages, everything is allowed,
571 * only concurrent bulk operations are locked out. */
572 BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED,
575 struct drbd_work_queue {
577 spinlock_t q_lock; /* to protect the list. */
578 wait_queue_head_t q_wait;
583 struct socket *socket;
584 /* this way we get our
585 * send/receive buffers off the stack */
591 u64 md_offset; /* sector offset to 'super' block */
593 u64 la_size_sect; /* last agreed size, unit sectors */
594 spinlock_t uuid_lock;
600 s32 al_offset; /* signed relative sector offset to activity log */
601 s32 bm_offset; /* signed relative sector offset to bitmap */
603 /* cached value of bdev->disk_conf->meta_dev_idx (see below) */
606 /* see al_tr_number_to_on_disk_sector() */
608 u32 al_stripe_size_4k;
609 u32 al_size_4k; /* cached product of the above */
612 struct drbd_backing_dev {
613 struct block_device *backing_bdev;
614 struct block_device *md_bdev;
616 struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */
617 sector_t known_size; /* last known size of that backing device */
622 unsigned long start_jif; /* last call to drbd_md_get_buffer */
623 unsigned long submit_jif; /* last _drbd_md_sync_page_io() submit */
624 const char *current_use;
634 int (*io_fn)(struct drbd_device *device);
635 void (*done)(struct drbd_device *device, int rv);
639 unsigned int head_index;
641 int total; /* sum of all values */
644 extern struct fifo_buffer *fifo_alloc(int fifo_size);
646 /* flag bits per connection */
648 NET_CONGESTED, /* The data socket is congested */
649 RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */
650 SEND_PING, /* whether asender should send a ping asap */
651 SIGNAL_ASENDER, /* whether asender wants to be interrupted */
652 GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */
653 CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */
656 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
657 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */
658 STATE_SENT, /* Do not change state/UUIDs while this is set */
659 CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
660 * pending, from drbd worker context.
661 * If set, bdi_write_congested() returns true,
662 * so shrink_page_list() would not recurse into,
663 * and potentially deadlock on, this drbd worker.
667 DEVICE_WORK_PENDING, /* tell worker that some device has pending work */
670 struct drbd_resource {
672 #ifdef CONFIG_DEBUG_FS
673 struct dentry *debugfs_res;
674 struct dentry *debugfs_res_volumes;
675 struct dentry *debugfs_res_connections;
676 struct dentry *debugfs_res_in_flight_summary;
679 struct idr devices; /* volume number to device mapping */
680 struct list_head connections;
681 struct list_head resources;
682 struct res_opts res_opts;
683 struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */
684 struct mutex adm_mutex; /* mutex to serialize administrative requests */
687 unsigned susp:1; /* IO suspended by user */
688 unsigned susp_nod:1; /* IO suspended because no data */
689 unsigned susp_fen:1; /* IO suspended because fence peer handler runs */
691 enum write_ordering_e write_ordering;
693 cpumask_var_t cpu_mask;
696 struct drbd_thread_timing_details
698 unsigned long start_jif;
700 const char *caller_fn;
705 struct drbd_connection {
706 struct list_head connections;
707 struct drbd_resource *resource;
708 #ifdef CONFIG_DEBUG_FS
709 struct dentry *debugfs_conn;
710 struct dentry *debugfs_conn_callback_history;
711 struct dentry *debugfs_conn_oldest_requests;
714 struct idr peer_devices; /* volume number to peer device mapping */
715 enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
716 struct mutex cstate_mutex; /* Protects graceful disconnects */
717 unsigned int connect_cnt; /* Inc each time a connection is established */
720 struct net_conf *net_conf; /* content protected by rcu */
721 wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */
723 struct sockaddr_storage my_addr;
725 struct sockaddr_storage peer_addr;
728 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
729 struct drbd_socket meta; /* ping/ack (metadata) packets */
730 int agreed_pro_version; /* actually used protocol version */
732 unsigned long last_received; /* in jiffies, either socket */
733 unsigned int ko_count;
735 struct list_head transfer_log; /* all requests not yet fully processed */
737 struct crypto_hash *cram_hmac_tfm;
738 struct crypto_hash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */
739 struct crypto_hash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */
740 struct crypto_hash *csums_tfm;
741 struct crypto_hash *verify_tfm;
746 struct drbd_epoch *current_epoch;
747 spinlock_t epoch_lock;
749 atomic_t current_tle_nr; /* transfer log epoch number */
750 unsigned current_tle_writes; /* writes seen within this tl epoch */
752 unsigned long last_reconnect_jif;
753 struct drbd_thread receiver;
754 struct drbd_thread worker;
755 struct drbd_thread asender;
758 * so we can look up the oldest pending requests more quickly.
759 * protected by resource->req_lock */
760 struct drbd_request *req_next; /* DRBD 9: todo.req_next */
761 struct drbd_request *req_ack_pending;
762 struct drbd_request *req_not_net_done;
765 struct drbd_work_queue sender_work;
767 #define DRBD_THREAD_DETAILS_HIST 16
768 unsigned int w_cb_nr; /* keeps counting up */
769 unsigned int r_cb_nr; /* keeps counting up */
770 struct drbd_thread_timing_details w_timing_details[DRBD_THREAD_DETAILS_HIST];
771 struct drbd_thread_timing_details r_timing_details[DRBD_THREAD_DETAILS_HIST];
774 /* whether this sender thread
775 * has processed a single write yet. */
776 bool seen_any_write_yet;
778 /* Which barrier number to send with the next P_BARRIER */
779 int current_epoch_nr;
781 /* how many write requests have been sent
782 * with req->epoch == current_epoch_nr.
783 * If none, no P_BARRIER will be sent. */
784 unsigned current_epoch_writes;
788 void __update_timing_details(
789 struct drbd_thread_timing_details *tdp,
792 const char *fn, const unsigned int line);
794 #define update_worker_timing_details(c, cb) \
795 __update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ )
796 #define update_receiver_timing_details(c, cb) \
797 __update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ )
799 struct submit_worker {
800 struct workqueue_struct *wq;
801 struct work_struct worker;
803 /* protected by ..->resource->req_lock */
804 struct list_head writes;
807 struct drbd_peer_device {
808 struct list_head peer_devices;
809 struct drbd_device *device;
810 struct drbd_connection *connection;
811 #ifdef CONFIG_DEBUG_FS
812 struct dentry *debugfs_peer_dev;
817 struct drbd_resource *resource;
818 struct list_head peer_devices;
819 struct list_head pending_bitmap_io;
821 unsigned long flush_jif;
822 #ifdef CONFIG_DEBUG_FS
823 struct dentry *debugfs_minor;
824 struct dentry *debugfs_vol;
825 struct dentry *debugfs_vol_oldest_requests;
826 struct dentry *debugfs_vol_act_log_extents;
827 struct dentry *debugfs_vol_resync_extents;
828 struct dentry *debugfs_vol_data_gen_id;
831 unsigned int vnr; /* volume number within the connection */
832 unsigned int minor; /* device minor number */
836 /* things that are stored as / read from meta data on disk */
839 /* configured by drbdsetup */
840 struct drbd_backing_dev *ldev __protected_by(local);
842 sector_t p_size; /* partner's disk size */
843 struct request_queue *rq_queue;
844 struct block_device *this_bdev;
845 struct gendisk *vdisk;
847 unsigned long last_reattach_jif;
848 struct drbd_work resync_work;
849 struct drbd_work unplug_work;
850 struct timer_list resync_timer;
851 struct timer_list md_sync_timer;
852 struct timer_list start_resync_timer;
853 struct timer_list request_timer;
855 /* Used after attach while negotiating new disk state. */
856 union drbd_state new_state_tmp;
858 union drbd_dev_state state;
859 wait_queue_head_t misc_wait;
860 wait_queue_head_t state_wait; /* upon each state change. */
861 unsigned int send_cnt;
862 unsigned int recv_cnt;
863 unsigned int read_cnt;
864 unsigned int writ_cnt;
865 unsigned int al_writ_cnt;
866 unsigned int bm_writ_cnt;
867 atomic_t ap_bio_cnt; /* Requests we need to complete */
868 atomic_t ap_actlog_cnt; /* Requests waiting for activity log */
869 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
870 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
871 atomic_t unacked_cnt; /* Need to send replies for */
872 atomic_t local_cnt; /* Waiting for local completion */
874 /* Interval tree of pending local requests */
875 struct rb_root read_requests;
876 struct rb_root write_requests;
878 /* for statistics and timeouts */
879 /* [0] read, [1] write */
880 struct list_head pending_master_completion[2];
881 struct list_head pending_completion[2];
883 /* use checksums for *this* resync */
885 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
886 unsigned long rs_total;
887 /* number of resync blocks that failed in this run */
888 unsigned long rs_failed;
889 /* Syncer's start time [unit jiffies] */
890 unsigned long rs_start;
891 /* cumulated time in PausedSyncX state [unit jiffies] */
892 unsigned long rs_paused;
893 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
894 unsigned long rs_same_csum;
895 #define DRBD_SYNC_MARKS 8
896 #define DRBD_SYNC_MARK_STEP (3*HZ)
897 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
898 unsigned long rs_mark_left[DRBD_SYNC_MARKS];
899 /* marks's time [unit jiffies] */
900 unsigned long rs_mark_time[DRBD_SYNC_MARKS];
901 /* current index into rs_mark_{left,time} */
903 unsigned long rs_last_bcast; /* [unit jiffies] */
905 /* where does the admin want us to start? (sector) */
906 sector_t ov_start_sector;
907 sector_t ov_stop_sector;
908 /* where are we now? (sector) */
909 sector_t ov_position;
910 /* Start sector of out of sync range (to merge printk reporting). */
911 sector_t ov_last_oos_start;
912 /* size of out-of-sync range in sectors. */
913 sector_t ov_last_oos_size;
914 unsigned long ov_left; /* in bits */
916 struct drbd_bitmap *bitmap;
917 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
919 /* Used to track operations of resync... */
920 struct lru_cache *resync;
921 /* Number of locked elements in resync LRU */
922 unsigned int resync_locked;
923 /* resync extent number waiting for application requests */
924 unsigned int resync_wenr;
929 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
930 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
931 struct list_head done_ee; /* need to send P_WRITE_ACK */
932 struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */
933 struct list_head net_ee; /* zero-copy network send in progress */
936 struct list_head resync_reads;
937 atomic_t pp_in_use; /* allocated from page pool */
938 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */
939 wait_queue_head_t ee_wait;
940 struct drbd_md_io md_io;
942 wait_queue_head_t al_wait;
943 struct lru_cache *act_log; /* activity log */
944 unsigned int al_tr_number;
946 wait_queue_head_t seq_wait;
948 unsigned int peer_seq;
949 spinlock_t peer_seq_lock;
950 unsigned long comm_bm_set; /* communicated number of set bits. */
951 struct bm_io_work bm_io_work;
952 u64 ed_uuid; /* UUID of the exposed data */
953 struct mutex own_state_mutex;
954 struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */
955 char congestion_reason; /* Why we where congested... */
956 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
957 atomic_t rs_sect_ev; /* for submitted resync data rate, both */
958 int rs_last_sect_ev; /* counter to compare with */
959 int rs_last_events; /* counter of read or write "events" (unit sectors)
960 * on the lower level device when we last looked. */
961 int c_sync_rate; /* current resync rate after syncer throttle magic */
962 struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */
963 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
964 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
965 unsigned int peer_max_bio_size;
966 unsigned int local_max_bio_size;
968 /* any requests that would block in drbd_make_request()
969 * are deferred to this single-threaded work queue */
970 struct submit_worker submit;
973 struct drbd_bm_aio_ctx {
974 struct drbd_device *device;
975 struct list_head list; /* on device->pending_bitmap_io */;
976 unsigned long start_jif;
980 #define BM_AIO_COPY_PAGES 1
981 #define BM_AIO_WRITE_HINTED 2
982 #define BM_AIO_WRITE_ALL_PAGES 4
983 #define BM_AIO_READ 8
988 struct drbd_config_context {
989 /* assigned from drbd_genlmsghdr */
991 /* assigned from request attributes, if present */
993 #define VOLUME_UNSPECIFIED (-1U)
994 /* pointer into the request skb,
995 * limited lifetime! */
997 struct nlattr *my_addr;
998 struct nlattr *peer_addr;
1001 struct sk_buff *reply_skb;
1002 /* pointer into reply buffer */
1003 struct drbd_genlmsghdr *reply_dh;
1004 /* resolved from attributes, if possible */
1005 struct drbd_device *device;
1006 struct drbd_resource *resource;
1007 struct drbd_connection *connection;
1010 static inline struct drbd_device *minor_to_device(unsigned int minor)
1012 return (struct drbd_device *)idr_find(&drbd_devices, minor);
1015 static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device)
1017 return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices);
1020 #define for_each_resource(resource, _resources) \
1021 list_for_each_entry(resource, _resources, resources)
1023 #define for_each_resource_rcu(resource, _resources) \
1024 list_for_each_entry_rcu(resource, _resources, resources)
1026 #define for_each_resource_safe(resource, tmp, _resources) \
1027 list_for_each_entry_safe(resource, tmp, _resources, resources)
1029 #define for_each_connection(connection, resource) \
1030 list_for_each_entry(connection, &resource->connections, connections)
1032 #define for_each_connection_rcu(connection, resource) \
1033 list_for_each_entry_rcu(connection, &resource->connections, connections)
1035 #define for_each_connection_safe(connection, tmp, resource) \
1036 list_for_each_entry_safe(connection, tmp, &resource->connections, connections)
1038 #define for_each_peer_device(peer_device, device) \
1039 list_for_each_entry(peer_device, &device->peer_devices, peer_devices)
1041 #define for_each_peer_device_rcu(peer_device, device) \
1042 list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices)
1044 #define for_each_peer_device_safe(peer_device, tmp, device) \
1045 list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices)
1047 static inline unsigned int device_to_minor(struct drbd_device *device)
1049 return device->minor;
1053 * function declarations
1054 *************************/
1060 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1063 extern void drbd_init_set_defaults(struct drbd_device *device);
1064 extern int drbd_thread_start(struct drbd_thread *thi);
1065 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1067 extern void drbd_thread_current_set_cpu(struct drbd_thread *thi);
1069 #define drbd_thread_current_set_cpu(A) ({})
1071 extern void tl_release(struct drbd_connection *, unsigned int barrier_nr,
1072 unsigned int set_size);
1073 extern void tl_clear(struct drbd_connection *);
1074 extern void drbd_free_sock(struct drbd_connection *connection);
1075 extern int drbd_send(struct drbd_connection *connection, struct socket *sock,
1076 void *buf, size_t size, unsigned msg_flags);
1077 extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t,
1080 extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd);
1081 extern int drbd_send_protocol(struct drbd_connection *connection);
1082 extern int drbd_send_uuids(struct drbd_peer_device *);
1083 extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *);
1084 extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *);
1085 extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags);
1086 extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s);
1087 extern int drbd_send_current_state(struct drbd_peer_device *);
1088 extern int drbd_send_sync_param(struct drbd_peer_device *);
1089 extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr,
1091 extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet,
1092 struct drbd_peer_request *);
1093 extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet,
1094 struct p_block_req *rp);
1095 extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet,
1096 struct p_data *dp, int data_size);
1097 extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet,
1098 sector_t sector, int blksize, u64 block_id);
1099 extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *);
1100 extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet,
1101 struct drbd_peer_request *);
1102 extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req);
1103 extern int drbd_send_drequest(struct drbd_peer_device *, int cmd,
1104 sector_t sector, int size, u64 block_id);
1105 extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector,
1106 int size, void *digest, int digest_size,
1107 enum drbd_packet cmd);
1108 extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size);
1110 extern int drbd_send_bitmap(struct drbd_device *device);
1111 extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode);
1112 extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode);
1113 extern void drbd_free_ldev(struct drbd_backing_dev *ldev);
1114 extern void drbd_device_cleanup(struct drbd_device *device);
1115 void drbd_print_uuids(struct drbd_device *device, const char *text);
1117 extern void conn_md_sync(struct drbd_connection *connection);
1118 extern void drbd_md_write(struct drbd_device *device, void *buffer);
1119 extern void drbd_md_sync(struct drbd_device *device);
1120 extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev);
1121 extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1122 extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1123 extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local);
1124 extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local);
1125 extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local);
1126 extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1127 extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local);
1128 extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local);
1129 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1130 extern void drbd_md_mark_dirty(struct drbd_device *device);
1131 extern void drbd_queue_bitmap_io(struct drbd_device *device,
1132 int (*io_fn)(struct drbd_device *),
1133 void (*done)(struct drbd_device *, int),
1134 char *why, enum bm_flag flags);
1135 extern int drbd_bitmap_io(struct drbd_device *device,
1136 int (*io_fn)(struct drbd_device *),
1137 char *why, enum bm_flag flags);
1138 extern int drbd_bitmap_io_from_worker(struct drbd_device *device,
1139 int (*io_fn)(struct drbd_device *),
1140 char *why, enum bm_flag flags);
1141 extern int drbd_bmio_set_n_write(struct drbd_device *device) __must_hold(local);
1142 extern int drbd_bmio_clear_n_write(struct drbd_device *device) __must_hold(local);
1146 * We currently have two possible layouts.
1147 * Offsets in (512 byte) sectors.
1149 * |----------- md_size_sect ------------------|
1150 * [ 4k superblock ][ activity log ][ Bitmap ]
1151 * | al_offset == 8 |
1152 * | bm_offset = al_offset + X |
1153 * ==> bitmap sectors = md_size_sect - bm_offset
1156 * old, indexed fixed size meta data:
1159 * |----------- md_size_sect ------------------|
1160 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*]
1162 * | bm_offset = al_offset - Y |
1163 * ==> bitmap sectors = Y = al_offset - bm_offset
1165 * [padding*] are zero or up to 7 unused 512 Byte sectors to the
1166 * end of the device, so that the [4k superblock] will be 4k aligned.
1168 * The activity log consists of 4k transaction blocks,
1169 * which are written in a ring-buffer, or striped ring-buffer like fashion,
1170 * which are writtensize used to be fixed 32kB,
1171 * but is about to become configurable.
1174 /* Our old fixed size meta data layout
1175 * allows up to about 3.8TB, so if you want more,
1176 * you need to use the "flexible" meta data format. */
1177 #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */
1178 #define MD_4kB_SECT 8
1179 #define MD_32kB_SECT 64
1181 /* One activity log extent represents 4M of storage */
1182 #define AL_EXTENT_SHIFT 22
1183 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1185 /* We could make these currently hardcoded constants configurable
1186 * variables at create-md time (or even re-configurable at runtime?).
1187 * Which will require some more changes to the DRBD "super block"
1190 * updates per transaction:
1191 * This many changes to the active set can be logged with one transaction.
1192 * This number is arbitrary.
1193 * context per transaction:
1194 * This many context extent numbers are logged with each transaction.
1195 * This number is resulting from the transaction block size (4k), the layout
1196 * of the transaction header, and the number of updates per transaction.
1197 * See drbd_actlog.c:struct al_transaction_on_disk
1199 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1200 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1202 #if BITS_PER_LONG == 32
1204 #define cpu_to_lel(A) cpu_to_le32(A)
1205 #define lel_to_cpu(A) le32_to_cpu(A)
1206 #elif BITS_PER_LONG == 64
1208 #define cpu_to_lel(A) cpu_to_le64(A)
1209 #define lel_to_cpu(A) le64_to_cpu(A)
1211 #error "LN2 of BITS_PER_LONG unknown!"
1215 /* 16MB sized 'bitmap extent' to track syncer usage */
1217 int rs_left; /* number of bits set (out of sync) in this extent. */
1218 int rs_failed; /* number of failed resync requests in this extent. */
1219 unsigned long flags;
1220 struct lc_element lce;
1223 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1224 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1225 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1229 * We need to store one bit for a block.
1230 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1231 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1232 * Bit 1 ==> local node thinks this block needs to be synced.
1235 #define SLEEP_TIME (HZ/10)
1237 /* We do bitmap IO in units of 4k blocks.
1238 * We also still have a hardcoded 4k per bit relation. */
1239 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1240 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1241 /* mostly arbitrarily set the represented size of one bitmap extent,
1242 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1243 * at 4k per bit resolution) */
1244 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
1245 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1247 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1248 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1251 /* thus many _storage_ sectors are described by one bit */
1252 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1253 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1254 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1256 /* bit to represented kilo byte conversion */
1257 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1259 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1260 * _storage_ sector is located in */
1261 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1262 #define BM_BIT_TO_EXT(x) ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1264 /* first storage sector a bitmap extent corresponds to */
1265 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1266 /* how much _storage_ sectors we have per bitmap extent */
1267 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1268 /* how many bits are covered by one bitmap extent (resync extent) */
1269 #define BM_BITS_PER_EXT (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1271 #define BM_BLOCKS_PER_BM_EXT_MASK (BM_BITS_PER_EXT - 1)
1274 /* in one sector of the bitmap, we have this many activity_log extents. */
1275 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1277 /* the extent in "PER_EXTENT" below is an activity log extent
1278 * we need that many (long words/bytes) to store the bitmap
1279 * of one AL_EXTENT_SIZE chunk of storage.
1280 * we can store the bitmap for that many AL_EXTENTS within
1281 * one sector of the _on_disk_ bitmap:
1282 * bit 0 bit 37 bit 38 bit (512*8)-1
1283 * ...|........|........|.. // ..|........|
1284 * sect. 0 `296 `304 ^(512*8*8)-1
1286 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1287 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1288 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1291 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1292 /* we have a certain meta data variant that has a fixed on-disk size of 128
1293 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity
1294 * log, leaving this many sectors for the bitmap.
1297 #define DRBD_MAX_SECTORS_FIXED_BM \
1298 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9)))
1299 #if !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1300 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1301 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1303 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM
1304 /* 16 TB in units of sectors */
1305 #if BITS_PER_LONG == 32
1306 /* adjust by one page worth of bitmap,
1307 * so we won't wrap around in drbd_bm_find_next_bit.
1308 * you should use 64bit OS for that much storage, anyways. */
1309 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1311 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1312 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1313 /* corresponds to (1UL << 38) bits right now. */
1317 /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE,
1318 * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte.
1319 * Since we may live in a mixed-platform cluster,
1320 * we limit us to a platform agnostic constant here for now.
1321 * A followup commit may allow even bigger BIO sizes,
1322 * once we thought that through. */
1323 #define DRBD_MAX_BIO_SIZE (1U << 20)
1324 #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1325 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1327 #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */
1329 #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */
1330 #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */
1332 /* For now, don't allow more than one activity log extent worth of data
1333 * to be discarded in one go. We may need to rework drbd_al_begin_io()
1334 * to allow for even larger discard ranges */
1335 #define DRBD_MAX_DISCARD_SIZE AL_EXTENT_SIZE
1336 #define DRBD_MAX_DISCARD_SECTORS (DRBD_MAX_DISCARD_SIZE >> 9)
1338 extern int drbd_bm_init(struct drbd_device *device);
1339 extern int drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits);
1340 extern void drbd_bm_cleanup(struct drbd_device *device);
1341 extern void drbd_bm_set_all(struct drbd_device *device);
1342 extern void drbd_bm_clear_all(struct drbd_device *device);
1343 /* set/clear/test only a few bits at a time */
1344 extern int drbd_bm_set_bits(
1345 struct drbd_device *device, unsigned long s, unsigned long e);
1346 extern int drbd_bm_clear_bits(
1347 struct drbd_device *device, unsigned long s, unsigned long e);
1348 extern int drbd_bm_count_bits(
1349 struct drbd_device *device, const unsigned long s, const unsigned long e);
1350 /* bm_set_bits variant for use while holding drbd_bm_lock,
1351 * may process the whole bitmap in one go */
1352 extern void _drbd_bm_set_bits(struct drbd_device *device,
1353 const unsigned long s, const unsigned long e);
1354 extern int drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr);
1355 extern int drbd_bm_e_weight(struct drbd_device *device, unsigned long enr);
1356 extern int drbd_bm_read(struct drbd_device *device) __must_hold(local);
1357 extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr);
1358 extern int drbd_bm_write(struct drbd_device *device) __must_hold(local);
1359 extern int drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local);
1360 extern int drbd_bm_write_lazy(struct drbd_device *device, unsigned upper_idx) __must_hold(local);
1361 extern int drbd_bm_write_all(struct drbd_device *device) __must_hold(local);
1362 extern int drbd_bm_write_copy_pages(struct drbd_device *device) __must_hold(local);
1363 extern size_t drbd_bm_words(struct drbd_device *device);
1364 extern unsigned long drbd_bm_bits(struct drbd_device *device);
1365 extern sector_t drbd_bm_capacity(struct drbd_device *device);
1367 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1368 extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo);
1369 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1370 extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo);
1371 extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo);
1372 extern unsigned long _drbd_bm_total_weight(struct drbd_device *device);
1373 extern unsigned long drbd_bm_total_weight(struct drbd_device *device);
1374 /* for receive_bitmap */
1375 extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset,
1376 size_t number, unsigned long *buffer);
1377 /* for _drbd_send_bitmap */
1378 extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset,
1379 size_t number, unsigned long *buffer);
1381 extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags);
1382 extern void drbd_bm_unlock(struct drbd_device *device);
1385 extern struct kmem_cache *drbd_request_cache;
1386 extern struct kmem_cache *drbd_ee_cache; /* peer requests */
1387 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1388 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1389 extern mempool_t *drbd_request_mempool;
1390 extern mempool_t *drbd_ee_mempool;
1392 /* drbd's page pool, used to buffer data received from the peer,
1393 * or data requested by the peer.
1395 * This does not have an emergency reserve.
1397 * When allocating from this pool, it first takes pages from the pool.
1398 * Only if the pool is depleted will try to allocate from the system.
1400 * The assumption is that pages taken from this pool will be processed,
1401 * and given back, "quickly", and then can be recycled, so we can avoid
1402 * frequent calls to alloc_page(), and still will be able to make progress even
1403 * under memory pressure.
1405 extern struct page *drbd_pp_pool;
1406 extern spinlock_t drbd_pp_lock;
1407 extern int drbd_pp_vacant;
1408 extern wait_queue_head_t drbd_pp_wait;
1410 /* We also need a standard (emergency-reserve backed) page pool
1411 * for meta data IO (activity log, bitmap).
1412 * We can keep it global, as long as it is used as "N pages at a time".
1413 * 128 should be plenty, currently we probably can get away with as few as 1.
1415 #define DRBD_MIN_POOL_PAGES 128
1416 extern mempool_t *drbd_md_io_page_pool;
1418 /* We also need to make sure we get a bio
1419 * when we need it for housekeeping purposes */
1420 extern struct bio_set *drbd_md_io_bio_set;
1421 /* to allocate from that set */
1422 extern struct bio *bio_alloc_drbd(gfp_t gfp_mask);
1424 extern struct mutex resources_mutex;
1426 extern int conn_lowest_minor(struct drbd_connection *connection);
1427 extern enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor);
1428 extern void drbd_destroy_device(struct kref *kref);
1429 extern void drbd_delete_device(struct drbd_device *device);
1431 extern struct drbd_resource *drbd_create_resource(const char *name);
1432 extern void drbd_free_resource(struct drbd_resource *resource);
1434 extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts);
1435 extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts);
1436 extern void drbd_destroy_connection(struct kref *kref);
1437 extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len,
1438 void *peer_addr, int peer_addr_len);
1439 extern struct drbd_resource *drbd_find_resource(const char *name);
1440 extern void drbd_destroy_resource(struct kref *kref);
1441 extern void conn_free_crypto(struct drbd_connection *connection);
1443 extern int proc_details;
1446 extern void do_submit(struct work_struct *ws);
1447 extern void __drbd_make_request(struct drbd_device *, struct bio *, unsigned long);
1448 extern blk_qc_t drbd_make_request(struct request_queue *q, struct bio *bio);
1449 extern int drbd_read_remote(struct drbd_device *device, struct drbd_request *req);
1450 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1454 extern void drbd_suspend_io(struct drbd_device *device);
1455 extern void drbd_resume_io(struct drbd_device *device);
1456 extern char *ppsize(char *buf, unsigned long long size);
1457 extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int);
1458 enum determine_dev_size {
1459 DS_ERROR_SHRINK = -3,
1460 DS_ERROR_SPACE_MD = -2,
1465 DS_GREW_FROM_ZERO = 3,
1467 extern enum determine_dev_size
1468 drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local);
1469 extern void resync_after_online_grow(struct drbd_device *);
1470 extern void drbd_reconsider_max_bio_size(struct drbd_device *device, struct drbd_backing_dev *bdev);
1471 extern enum drbd_state_rv drbd_set_role(struct drbd_device *device,
1472 enum drbd_role new_role,
1474 extern bool conn_try_outdate_peer(struct drbd_connection *connection);
1475 extern void conn_try_outdate_peer_async(struct drbd_connection *connection);
1476 extern int drbd_khelper(struct drbd_device *device, char *cmd);
1479 /* bi_end_io handlers */
1480 extern void drbd_md_endio(struct bio *bio);
1481 extern void drbd_peer_request_endio(struct bio *bio);
1482 extern void drbd_request_endio(struct bio *bio);
1483 extern int drbd_worker(struct drbd_thread *thi);
1484 enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor);
1485 void drbd_resync_after_changed(struct drbd_device *device);
1486 extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side);
1487 extern void resume_next_sg(struct drbd_device *device);
1488 extern void suspend_other_sg(struct drbd_device *device);
1489 extern int drbd_resync_finished(struct drbd_device *device);
1490 /* maybe rather drbd_main.c ? */
1491 extern void *drbd_md_get_buffer(struct drbd_device *device, const char *intent);
1492 extern void drbd_md_put_buffer(struct drbd_device *device);
1493 extern int drbd_md_sync_page_io(struct drbd_device *device,
1494 struct drbd_backing_dev *bdev, sector_t sector, int rw);
1495 extern void drbd_ov_out_of_sync_found(struct drbd_device *, sector_t, int);
1496 extern void wait_until_done_or_force_detached(struct drbd_device *device,
1497 struct drbd_backing_dev *bdev, unsigned int *done);
1498 extern void drbd_rs_controller_reset(struct drbd_device *device);
1500 static inline void ov_out_of_sync_print(struct drbd_device *device)
1502 if (device->ov_last_oos_size) {
1503 drbd_err(device, "Out of sync: start=%llu, size=%lu (sectors)\n",
1504 (unsigned long long)device->ov_last_oos_start,
1505 (unsigned long)device->ov_last_oos_size);
1507 device->ov_last_oos_size = 0;
1511 extern void drbd_csum_bio(struct crypto_hash *, struct bio *, void *);
1512 extern void drbd_csum_ee(struct crypto_hash *, struct drbd_peer_request *, void *);
1513 /* worker callbacks */
1514 extern int w_e_end_data_req(struct drbd_work *, int);
1515 extern int w_e_end_rsdata_req(struct drbd_work *, int);
1516 extern int w_e_end_csum_rs_req(struct drbd_work *, int);
1517 extern int w_e_end_ov_reply(struct drbd_work *, int);
1518 extern int w_e_end_ov_req(struct drbd_work *, int);
1519 extern int w_ov_finished(struct drbd_work *, int);
1520 extern int w_resync_timer(struct drbd_work *, int);
1521 extern int w_send_write_hint(struct drbd_work *, int);
1522 extern int w_send_dblock(struct drbd_work *, int);
1523 extern int w_send_read_req(struct drbd_work *, int);
1524 extern int w_e_reissue(struct drbd_work *, int);
1525 extern int w_restart_disk_io(struct drbd_work *, int);
1526 extern int w_send_out_of_sync(struct drbd_work *, int);
1527 extern int w_start_resync(struct drbd_work *, int);
1529 extern void resync_timer_fn(unsigned long data);
1530 extern void start_resync_timer_fn(unsigned long data);
1532 extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req);
1534 /* drbd_receiver.c */
1535 extern int drbd_receiver(struct drbd_thread *thi);
1536 extern int drbd_asender(struct drbd_thread *thi);
1537 extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device);
1538 extern bool drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector,
1539 bool throttle_if_app_is_waiting);
1540 extern int drbd_submit_peer_request(struct drbd_device *,
1541 struct drbd_peer_request *, const unsigned,
1543 extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *);
1544 extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64,
1545 sector_t, unsigned int,
1547 gfp_t) __must_hold(local);
1548 extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *,
1550 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1551 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
1552 extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool);
1553 extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
1554 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
1555 extern int drbd_connected(struct drbd_peer_device *);
1557 static inline void drbd_tcp_cork(struct socket *sock)
1560 (void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
1561 (char*)&val, sizeof(val));
1564 static inline void drbd_tcp_uncork(struct socket *sock)
1567 (void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
1568 (char*)&val, sizeof(val));
1571 static inline void drbd_tcp_nodelay(struct socket *sock)
1574 (void) kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1575 (char*)&val, sizeof(val));
1578 static inline void drbd_tcp_quickack(struct socket *sock)
1581 (void) kernel_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1582 (char*)&val, sizeof(val));
1585 /* sets the number of 512 byte sectors of our virtual device */
1586 static inline void drbd_set_my_capacity(struct drbd_device *device,
1589 /* set_capacity(device->this_bdev->bd_disk, size); */
1590 set_capacity(device->vdisk, size);
1591 device->this_bdev->bd_inode->i_size = (loff_t)size << 9;
1595 * used to submit our private bio
1597 static inline void drbd_generic_make_request(struct drbd_device *device,
1598 int fault_type, struct bio *bio)
1601 if (!bio->bi_bdev) {
1602 drbd_err(device, "drbd_generic_make_request: bio->bi_bdev == NULL\n");
1603 bio->bi_error = -ENODEV;
1608 if (drbd_insert_fault(device, fault_type))
1611 generic_make_request(bio);
1614 void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev,
1615 enum write_ordering_e wo);
1618 extern struct proc_dir_entry *drbd_proc;
1619 extern const struct file_operations drbd_proc_fops;
1620 extern const char *drbd_conn_str(enum drbd_conns s);
1621 extern const char *drbd_role_str(enum drbd_role s);
1624 extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i);
1625 extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i);
1626 extern void drbd_al_begin_io_commit(struct drbd_device *device);
1627 extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i);
1628 extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i);
1629 extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i);
1630 extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector);
1631 extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector);
1632 extern int drbd_try_rs_begin_io(struct drbd_device *device, sector_t sector);
1633 extern void drbd_rs_cancel_all(struct drbd_device *device);
1634 extern int drbd_rs_del_all(struct drbd_device *device);
1635 extern void drbd_rs_failed_io(struct drbd_device *device,
1636 sector_t sector, int size);
1637 extern void drbd_advance_rs_marks(struct drbd_device *device, unsigned long still_to_go);
1639 enum update_sync_bits_mode { RECORD_RS_FAILED, SET_OUT_OF_SYNC, SET_IN_SYNC };
1640 extern int __drbd_change_sync(struct drbd_device *device, sector_t sector, int size,
1641 enum update_sync_bits_mode mode);
1642 #define drbd_set_in_sync(device, sector, size) \
1643 __drbd_change_sync(device, sector, size, SET_IN_SYNC)
1644 #define drbd_set_out_of_sync(device, sector, size) \
1645 __drbd_change_sync(device, sector, size, SET_OUT_OF_SYNC)
1646 #define drbd_rs_failed_io(device, sector, size) \
1647 __drbd_change_sync(device, sector, size, RECORD_RS_FAILED)
1648 extern void drbd_al_shrink(struct drbd_device *device);
1649 extern int drbd_initialize_al(struct drbd_device *, void *);
1652 /* state info broadcast */
1654 enum drbd_state_info_bcast_reason sib_reason;
1658 unsigned helper_exit_code;
1661 union drbd_state os;
1662 union drbd_state ns;
1666 void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib);
1669 * inline helper functions
1670 *************************/
1672 /* see also page_chain_add and friends in drbd_receiver.c */
1673 static inline struct page *page_chain_next(struct page *page)
1675 return (struct page *)page_private(page);
1677 #define page_chain_for_each(page) \
1678 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1679 page = page_chain_next(page))
1680 #define page_chain_for_each_safe(page, n) \
1681 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1684 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req)
1686 struct page *page = peer_req->pages;
1687 page_chain_for_each(page) {
1688 if (page_count(page) > 1)
1694 static inline union drbd_state drbd_read_state(struct drbd_device *device)
1696 struct drbd_resource *resource = device->resource;
1697 union drbd_state rv;
1699 rv.i = device->state.i;
1700 rv.susp = resource->susp;
1701 rv.susp_nod = resource->susp_nod;
1702 rv.susp_fen = resource->susp_fen;
1707 enum drbd_force_detach_flags {
1714 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1715 static inline void __drbd_chk_io_error_(struct drbd_device *device,
1716 enum drbd_force_detach_flags df,
1719 enum drbd_io_error_p ep;
1722 ep = rcu_dereference(device->ldev->disk_conf)->on_io_error;
1725 case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */
1726 if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) {
1727 if (__ratelimit(&drbd_ratelimit_state))
1728 drbd_err(device, "Local IO failed in %s.\n", where);
1729 if (device->state.disk > D_INCONSISTENT)
1730 _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL);
1733 /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */
1735 case EP_CALL_HELPER:
1736 /* Remember whether we saw a READ or WRITE error.
1738 * Recovery of the affected area for WRITE failure is covered
1739 * by the activity log.
1740 * READ errors may fall outside that area though. Certain READ
1741 * errors can be "healed" by writing good data to the affected
1742 * blocks, which triggers block re-allocation in lower layers.
1744 * If we can not write the bitmap after a READ error,
1745 * we may need to trigger a full sync (see w_go_diskless()).
1747 * Force-detach is not really an IO error, but rather a
1748 * desperate measure to try to deal with a completely
1749 * unresponsive lower level IO stack.
1750 * Still it should be treated as a WRITE error.
1752 * Meta IO error is always WRITE error:
1753 * we read meta data only once during attach,
1754 * which will fail in case of errors.
1756 set_bit(WAS_IO_ERROR, &device->flags);
1757 if (df == DRBD_READ_ERROR)
1758 set_bit(WAS_READ_ERROR, &device->flags);
1759 if (df == DRBD_FORCE_DETACH)
1760 set_bit(FORCE_DETACH, &device->flags);
1761 if (device->state.disk > D_FAILED) {
1762 _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL);
1764 "Local IO failed in %s. Detaching...\n", where);
1771 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1772 * @device: DRBD device.
1773 * @error: Error code passed to the IO completion callback
1774 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1776 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1778 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1779 static inline void drbd_chk_io_error_(struct drbd_device *device,
1780 int error, enum drbd_force_detach_flags forcedetach, const char *where)
1783 unsigned long flags;
1784 spin_lock_irqsave(&device->resource->req_lock, flags);
1785 __drbd_chk_io_error_(device, forcedetach, where);
1786 spin_unlock_irqrestore(&device->resource->req_lock, flags);
1792 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1793 * @bdev: Meta data block device.
1795 * BTW, for internal meta data, this happens to be the maximum capacity
1796 * we could agree upon with our peer node.
1798 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1800 switch (bdev->md.meta_dev_idx) {
1801 case DRBD_MD_INDEX_INTERNAL:
1802 case DRBD_MD_INDEX_FLEX_INT:
1803 return bdev->md.md_offset + bdev->md.bm_offset;
1804 case DRBD_MD_INDEX_FLEX_EXT:
1806 return bdev->md.md_offset;
1811 * drbd_md_last_sector() - Return the last sector number of the meta data area
1812 * @bdev: Meta data block device.
1814 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1816 switch (bdev->md.meta_dev_idx) {
1817 case DRBD_MD_INDEX_INTERNAL:
1818 case DRBD_MD_INDEX_FLEX_INT:
1819 return bdev->md.md_offset + MD_4kB_SECT -1;
1820 case DRBD_MD_INDEX_FLEX_EXT:
1822 return bdev->md.md_offset + bdev->md.md_size_sect -1;
1826 /* Returns the number of 512 byte sectors of the device */
1827 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1829 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1830 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0;
1834 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1835 * @bdev: Meta data block device.
1837 * returns the capacity we announce to out peer. we clip ourselves at the
1838 * various MAX_SECTORS, because if we don't, current implementation will
1839 * oops sooner or later
1841 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1845 switch (bdev->md.meta_dev_idx) {
1846 case DRBD_MD_INDEX_INTERNAL:
1847 case DRBD_MD_INDEX_FLEX_INT:
1848 s = drbd_get_capacity(bdev->backing_bdev)
1849 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1850 drbd_md_first_sector(bdev))
1853 case DRBD_MD_INDEX_FLEX_EXT:
1854 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1855 drbd_get_capacity(bdev->backing_bdev));
1856 /* clip at maximum size the meta device can support */
1857 s = min_t(sector_t, s,
1858 BM_EXT_TO_SECT(bdev->md.md_size_sect
1859 - bdev->md.bm_offset));
1862 s = min_t(sector_t, DRBD_MAX_SECTORS,
1863 drbd_get_capacity(bdev->backing_bdev));
1869 * drbd_md_ss() - Return the sector number of our meta data super block
1870 * @bdev: Meta data block device.
1872 static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev)
1874 const int meta_dev_idx = bdev->md.meta_dev_idx;
1876 if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT)
1879 /* Since drbd08, internal meta data is always "flexible".
1880 * position: last 4k aligned block of 4k size */
1881 if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL ||
1882 meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)
1883 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8;
1885 /* external, some index; this is the old fixed size layout */
1886 return MD_128MB_SECT * bdev->md.meta_dev_idx;
1890 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1892 unsigned long flags;
1893 spin_lock_irqsave(&q->q_lock, flags);
1894 list_add_tail(&w->list, &q->q);
1895 spin_unlock_irqrestore(&q->q_lock, flags);
1896 wake_up(&q->q_wait);
1900 drbd_queue_work_if_unqueued(struct drbd_work_queue *q, struct drbd_work *w)
1902 unsigned long flags;
1903 spin_lock_irqsave(&q->q_lock, flags);
1904 if (list_empty_careful(&w->list))
1905 list_add_tail(&w->list, &q->q);
1906 spin_unlock_irqrestore(&q->q_lock, flags);
1907 wake_up(&q->q_wait);
1911 drbd_device_post_work(struct drbd_device *device, int work_bit)
1913 if (!test_and_set_bit(work_bit, &device->flags)) {
1914 struct drbd_connection *connection =
1915 first_peer_device(device)->connection;
1916 struct drbd_work_queue *q = &connection->sender_work;
1917 if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags))
1918 wake_up(&q->q_wait);
1922 extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue);
1924 static inline void wake_asender(struct drbd_connection *connection)
1926 if (test_bit(SIGNAL_ASENDER, &connection->flags))
1927 force_sig(DRBD_SIG, connection->asender.task);
1930 static inline void request_ping(struct drbd_connection *connection)
1932 set_bit(SEND_PING, &connection->flags);
1933 wake_asender(connection);
1936 extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *);
1937 extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *);
1938 extern int conn_send_command(struct drbd_connection *, struct drbd_socket *,
1939 enum drbd_packet, unsigned int, void *,
1941 extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *,
1942 enum drbd_packet, unsigned int, void *,
1945 extern int drbd_send_ping(struct drbd_connection *connection);
1946 extern int drbd_send_ping_ack(struct drbd_connection *connection);
1947 extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state);
1948 extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state);
1950 static inline void drbd_thread_stop(struct drbd_thread *thi)
1952 _drbd_thread_stop(thi, false, true);
1955 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
1957 _drbd_thread_stop(thi, false, false);
1960 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
1962 _drbd_thread_stop(thi, true, false);
1965 /* counts how many answer packets packets we expect from our peer,
1966 * for either explicit application requests,
1967 * or implicit barrier packets as necessary.
1970 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1971 * it is much easier and equally valid to count what we queue for the
1972 * worker, even before it actually was queued or send.
1973 * (drbd_make_request_common; recovery path on read io-error)
1975 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1976 * _req_mod(req, DATA_RECEIVED)
1977 * [from receive_DataReply]
1978 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1979 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1980 * for some reason it is NOT decreased in got_NegAck,
1981 * but in the resulting cleanup code from report_params.
1982 * we should try to remember the reason for that...
1983 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
1984 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
1985 * [from tl_clear_barrier]
1987 static inline void inc_ap_pending(struct drbd_device *device)
1989 atomic_inc(&device->ap_pending_cnt);
1992 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \
1993 if (atomic_read(&device->which) < 0) \
1994 drbd_err(device, "in %s:%d: " #which " = %d < 0 !\n", \
1996 atomic_read(&device->which))
1998 #define dec_ap_pending(device) _dec_ap_pending(device, __func__, __LINE__)
1999 static inline void _dec_ap_pending(struct drbd_device *device, const char *func, int line)
2001 if (atomic_dec_and_test(&device->ap_pending_cnt))
2002 wake_up(&device->misc_wait);
2003 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line);
2006 /* counts how many resync-related answers we still expect from the peer
2008 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
2009 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
2010 * (or P_NEG_ACK with ID_SYNCER)
2012 static inline void inc_rs_pending(struct drbd_device *device)
2014 atomic_inc(&device->rs_pending_cnt);
2017 #define dec_rs_pending(device) _dec_rs_pending(device, __func__, __LINE__)
2018 static inline void _dec_rs_pending(struct drbd_device *device, const char *func, int line)
2020 atomic_dec(&device->rs_pending_cnt);
2021 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line);
2024 /* counts how many answers we still need to send to the peer.
2026 * receive_Data unless protocol A;
2027 * we need to send a P_RECV_ACK (proto B)
2028 * or P_WRITE_ACK (proto C)
2029 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2030 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2031 * receive_Barrier_* we need to send a P_BARRIER_ACK
2033 static inline void inc_unacked(struct drbd_device *device)
2035 atomic_inc(&device->unacked_cnt);
2038 #define dec_unacked(device) _dec_unacked(device, __func__, __LINE__)
2039 static inline void _dec_unacked(struct drbd_device *device, const char *func, int line)
2041 atomic_dec(&device->unacked_cnt);
2042 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
2045 #define sub_unacked(device, n) _sub_unacked(device, n, __func__, __LINE__)
2046 static inline void _sub_unacked(struct drbd_device *device, int n, const char *func, int line)
2048 atomic_sub(n, &device->unacked_cnt);
2049 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
2052 static inline bool is_sync_state(enum drbd_conns connection_state)
2055 (connection_state == C_SYNC_SOURCE
2056 || connection_state == C_SYNC_TARGET
2057 || connection_state == C_PAUSED_SYNC_S
2058 || connection_state == C_PAUSED_SYNC_T);
2062 * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev
2063 * @_device: DRBD device.
2064 * @_min_state: Minimum device state required for success.
2066 * You have to call put_ldev() when finished working with device->ldev.
2068 #define get_ldev_if_state(_device, _min_state) \
2069 (_get_ldev_if_state((_device), (_min_state)) ? \
2070 ({ __acquire(x); true; }) : false)
2071 #define get_ldev(_device) get_ldev_if_state(_device, D_INCONSISTENT)
2073 static inline void put_ldev(struct drbd_device *device)
2075 enum drbd_disk_state disk_state = device->state.disk;
2076 /* We must check the state *before* the atomic_dec becomes visible,
2077 * or we have a theoretical race where someone hitting zero,
2078 * while state still D_FAILED, will then see D_DISKLESS in the
2079 * condition below and calling into destroy, where he must not, yet. */
2080 int i = atomic_dec_return(&device->local_cnt);
2082 /* This may be called from some endio handler,
2083 * so we must not sleep here. */
2086 D_ASSERT(device, i >= 0);
2088 if (disk_state == D_DISKLESS)
2089 /* even internal references gone, safe to destroy */
2090 drbd_device_post_work(device, DESTROY_DISK);
2091 if (disk_state == D_FAILED)
2092 /* all application IO references gone. */
2093 if (!test_and_set_bit(GOING_DISKLESS, &device->flags))
2094 drbd_device_post_work(device, GO_DISKLESS);
2095 wake_up(&device->misc_wait);
2100 static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins)
2104 /* never get a reference while D_DISKLESS */
2105 if (device->state.disk == D_DISKLESS)
2108 atomic_inc(&device->local_cnt);
2109 io_allowed = (device->state.disk >= mins);
2115 extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins);
2118 /* this throttles on-the-fly application requests
2119 * according to max_buffers settings;
2120 * maybe re-implement using semaphores? */
2121 static inline int drbd_get_max_buffers(struct drbd_device *device)
2123 struct net_conf *nc;
2127 nc = rcu_dereference(first_peer_device(device)->connection->net_conf);
2128 mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */
2134 static inline int drbd_state_is_stable(struct drbd_device *device)
2136 union drbd_dev_state s = device->state;
2138 /* DO NOT add a default clause, we want the compiler to warn us
2139 * for any newly introduced state we may have forgotten to add here */
2141 switch ((enum drbd_conns)s.conn) {
2142 /* new io only accepted when there is no connection, ... */
2144 case C_WF_CONNECTION:
2145 /* ... or there is a well established connection. */
2151 case C_PAUSED_SYNC_S:
2152 case C_PAUSED_SYNC_T:
2155 /* transitional states, IO allowed */
2156 case C_DISCONNECTING:
2160 case C_NETWORK_FAILURE:
2161 case C_PROTOCOL_ERROR:
2163 case C_WF_REPORT_PARAMS:
2164 case C_STARTING_SYNC_S:
2165 case C_STARTING_SYNC_T:
2168 /* Allow IO in BM exchange states with new protocols */
2170 if (first_peer_device(device)->connection->agreed_pro_version < 96)
2174 /* no new io accepted in these states */
2176 case C_WF_SYNC_UUID:
2182 switch ((enum drbd_disk_state)s.disk) {
2184 case D_INCONSISTENT:
2189 /* disk state is stable as well. */
2192 /* no new io accepted during transitional states */
2204 static inline int drbd_suspended(struct drbd_device *device)
2206 struct drbd_resource *resource = device->resource;
2208 return resource->susp || resource->susp_fen || resource->susp_nod;
2211 static inline bool may_inc_ap_bio(struct drbd_device *device)
2213 int mxb = drbd_get_max_buffers(device);
2215 if (drbd_suspended(device))
2217 if (test_bit(SUSPEND_IO, &device->flags))
2220 /* to avoid potential deadlock or bitmap corruption,
2221 * in various places, we only allow new application io
2222 * to start during "stable" states. */
2224 /* no new io accepted when attaching or detaching the disk */
2225 if (!drbd_state_is_stable(device))
2228 /* since some older kernels don't have atomic_add_unless,
2229 * and we are within the spinlock anyways, we have this workaround. */
2230 if (atomic_read(&device->ap_bio_cnt) > mxb)
2232 if (test_bit(BITMAP_IO, &device->flags))
2237 static inline bool inc_ap_bio_cond(struct drbd_device *device)
2241 spin_lock_irq(&device->resource->req_lock);
2242 rv = may_inc_ap_bio(device);
2244 atomic_inc(&device->ap_bio_cnt);
2245 spin_unlock_irq(&device->resource->req_lock);
2250 static inline void inc_ap_bio(struct drbd_device *device)
2253 * as long as the device is suspended
2254 * until the bitmap is no longer on the fly during connection
2255 * handshake as long as we would exceed the max_buffer limit.
2257 * to avoid races with the reconnect code,
2258 * we need to atomic_inc within the spinlock. */
2260 wait_event(device->misc_wait, inc_ap_bio_cond(device));
2263 static inline void dec_ap_bio(struct drbd_device *device)
2265 int mxb = drbd_get_max_buffers(device);
2266 int ap_bio = atomic_dec_return(&device->ap_bio_cnt);
2268 D_ASSERT(device, ap_bio >= 0);
2270 if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) {
2271 if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags))
2272 drbd_queue_work(&first_peer_device(device)->
2273 connection->sender_work,
2274 &device->bm_io_work.w);
2277 /* this currently does wake_up for every dec_ap_bio!
2278 * maybe rather introduce some type of hysteresis?
2279 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2281 wake_up(&device->misc_wait);
2284 static inline bool verify_can_do_stop_sector(struct drbd_device *device)
2286 return first_peer_device(device)->connection->agreed_pro_version >= 97 &&
2287 first_peer_device(device)->connection->agreed_pro_version != 100;
2290 static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val)
2292 int changed = device->ed_uuid != val;
2293 device->ed_uuid = val;
2297 static inline int drbd_queue_order_type(struct drbd_device *device)
2299 /* sorry, we currently have no working implementation
2300 * of distributed TCQ stuff */
2301 #ifndef QUEUE_ORDERED_NONE
2302 #define QUEUE_ORDERED_NONE 0
2304 return QUEUE_ORDERED_NONE;
2307 static inline struct drbd_connection *first_connection(struct drbd_resource *resource)
2309 return list_first_entry_or_null(&resource->connections,
2310 struct drbd_connection, connections);