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 /* module parameter, defined in drbd_main.c */
65 extern unsigned int minor_count;
66 extern bool disable_sendpage;
67 extern bool allow_oos;
68 void tl_abort_disk_io(struct drbd_device *device);
70 #ifdef CONFIG_DRBD_FAULT_INJECTION
71 extern int enable_faults;
72 extern int fault_rate;
73 extern int fault_devs;
76 extern char usermode_helper[];
79 /* I don't remember why XCPU ...
80 * This is used to wake the asender,
81 * and to interrupt sending the sending task
84 #define DRBD_SIG SIGXCPU
86 /* This is used to stop/restart our threads.
87 * Cannot use SIGTERM nor SIGKILL, since these
88 * are sent out by init on runlevel changes
89 * I choose SIGHUP for now.
91 #define DRBD_SIGKILL SIGHUP
93 #define ID_IN_SYNC (4711ULL)
94 #define ID_OUT_OF_SYNC (4712ULL)
95 #define ID_SYNCER (-1ULL)
97 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
100 struct drbd_connection;
102 #define __drbd_printk_device(level, device, fmt, args...) \
103 dev_printk(level, disk_to_dev((device)->vdisk), fmt, ## args)
104 #define __drbd_printk_peer_device(level, peer_device, fmt, args...) \
105 dev_printk(level, disk_to_dev((peer_device)->device->vdisk), fmt, ## args)
106 #define __drbd_printk_resource(level, resource, fmt, args...) \
107 printk(level "drbd %s: " fmt, (resource)->name, ## args)
108 #define __drbd_printk_connection(level, connection, fmt, args...) \
109 printk(level "drbd %s: " fmt, (connection)->resource->name, ## args)
111 void drbd_printk_with_wrong_object_type(void);
113 #define __drbd_printk_if_same_type(obj, type, func, level, fmt, args...) \
114 (__builtin_types_compatible_p(typeof(obj), type) || \
115 __builtin_types_compatible_p(typeof(obj), const type)), \
116 func(level, (const type)(obj), fmt, ## args)
118 #define drbd_printk(level, obj, fmt, args...) \
119 __builtin_choose_expr( \
120 __drbd_printk_if_same_type(obj, struct drbd_device *, \
121 __drbd_printk_device, level, fmt, ## args), \
122 __builtin_choose_expr( \
123 __drbd_printk_if_same_type(obj, struct drbd_resource *, \
124 __drbd_printk_resource, level, fmt, ## args), \
125 __builtin_choose_expr( \
126 __drbd_printk_if_same_type(obj, struct drbd_connection *, \
127 __drbd_printk_connection, level, fmt, ## args), \
128 __builtin_choose_expr( \
129 __drbd_printk_if_same_type(obj, struct drbd_peer_device *, \
130 __drbd_printk_peer_device, level, fmt, ## args), \
131 drbd_printk_with_wrong_object_type()))))
133 #define drbd_dbg(obj, fmt, args...) \
134 drbd_printk(KERN_DEBUG, obj, fmt, ## args)
135 #define drbd_alert(obj, fmt, args...) \
136 drbd_printk(KERN_ALERT, obj, fmt, ## args)
137 #define drbd_err(obj, fmt, args...) \
138 drbd_printk(KERN_ERR, obj, fmt, ## args)
139 #define drbd_warn(obj, fmt, args...) \
140 drbd_printk(KERN_WARNING, obj, fmt, ## args)
141 #define drbd_info(obj, fmt, args...) \
142 drbd_printk(KERN_INFO, obj, fmt, ## args)
143 #define drbd_emerg(obj, fmt, args...) \
144 drbd_printk(KERN_EMERG, obj, fmt, ## args)
146 #define dynamic_drbd_dbg(device, fmt, args...) \
147 dynamic_dev_dbg(disk_to_dev(device->vdisk), fmt, ## args)
149 #define D_ASSERT(device, exp) do { \
151 drbd_err(device, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__); \
155 * expect - Make an assertion
157 * Unlike the assert macro, this macro returns a boolean result.
159 #define expect(exp) ({ \
160 bool _bool = (exp); \
162 drbd_err(device, "ASSERTION %s FAILED in %s\n", \
167 /* Defines to control fault insertion */
169 DRBD_FAULT_MD_WR = 0, /* meta data write */
170 DRBD_FAULT_MD_RD = 1, /* read */
171 DRBD_FAULT_RS_WR = 2, /* resync */
172 DRBD_FAULT_RS_RD = 3,
173 DRBD_FAULT_DT_WR = 4, /* data */
174 DRBD_FAULT_DT_RD = 5,
175 DRBD_FAULT_DT_RA = 6, /* data read ahead */
176 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */
177 DRBD_FAULT_AL_EE = 8, /* alloc ee */
178 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
184 _drbd_insert_fault(struct drbd_device *device, unsigned int type);
187 drbd_insert_fault(struct drbd_device *device, unsigned int type) {
188 #ifdef CONFIG_DRBD_FAULT_INJECTION
190 (enable_faults & (1<<type)) &&
191 _drbd_insert_fault(device, type);
197 /* integer division, round _UP_ to the next integer */
198 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
199 /* usual integer division */
200 #define div_floor(A, B) ((A)/(B))
202 extern struct ratelimit_state drbd_ratelimit_state;
203 extern struct idr drbd_devices; /* RCU, updates: genl_lock() */
204 extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */
206 extern const char *cmdname(enum drbd_packet cmd);
208 /* for sending/receiving the bitmap,
209 * possibly in some encoding scheme */
212 * stores total bits and long words
213 * of the bitmap, so we don't need to
214 * call the accessor functions over and again. */
215 unsigned long bm_bits;
216 unsigned long bm_words;
217 /* during xfer, current position within the bitmap */
218 unsigned long bit_offset;
219 unsigned long word_offset;
221 /* statistics; index: (h->command == P_BITMAP) */
226 extern void INFO_bm_xfer_stats(struct drbd_device *device,
227 const char *direction, struct bm_xfer_ctx *c);
229 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
231 /* word_offset counts "native long words" (32 or 64 bit),
233 * Encoded packet may end at an unaligned bit offset.
234 * In case a fallback clear text packet is transmitted in
235 * between, we adjust this offset back to the last 64bit
236 * aligned "native long word", which makes coding and decoding
237 * the plain text bitmap much more convenient. */
238 #if BITS_PER_LONG == 64
239 c->word_offset = c->bit_offset >> 6;
240 #elif BITS_PER_LONG == 32
241 c->word_offset = c->bit_offset >> 5;
242 c->word_offset &= ~(1UL);
244 # error "unsupported BITS_PER_LONG"
248 extern unsigned int drbd_header_size(struct drbd_connection *connection);
250 /**********************************************************************/
251 enum drbd_thread_state {
260 struct task_struct *task;
261 struct completion stop;
262 enum drbd_thread_state t_state;
263 int (*function) (struct drbd_thread *);
264 struct drbd_resource *resource;
265 struct drbd_connection *connection;
270 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
272 /* THINK testing the t_state seems to be uncritical in all cases
273 * (but thread_{start,stop}), so we can read it *without* the lock.
281 struct list_head list;
282 int (*cb)(struct drbd_work *, int cancel);
285 struct drbd_device_work {
287 struct drbd_device *device;
290 #include "drbd_interval.h"
292 extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *);
294 struct drbd_request {
296 struct drbd_device *device;
298 /* if local IO is not allowed, will be NULL.
299 * if local IO _is_ allowed, holds the locally submitted bio clone,
300 * or, after local IO completion, the ERR_PTR(error).
301 * see drbd_request_endio(). */
302 struct bio *private_bio;
304 struct drbd_interval i;
306 /* epoch: used to check on "completion" whether this req was in
307 * the current epoch, and we therefore have to close it,
308 * causing a p_barrier packet to be send, starting a new epoch.
310 * This corresponds to "barrier" in struct p_barrier[_ack],
311 * and to "barrier_nr" in struct drbd_epoch (and various
312 * comments/function parameters/local variable names).
316 struct list_head tl_requests; /* ring list in the transfer log */
317 struct bio *master_bio; /* master bio pointer */
319 /* see struct drbd_device */
320 struct list_head req_pending_master_completion;
321 struct list_head req_pending_local;
323 /* for generic IO accounting */
324 unsigned long start_jif;
326 /* for DRBD internal statistics */
328 /* Minimal set of time stamps to determine if we wait for activity log
329 * transactions, local disk or peer. 32 bit "jiffies" are good enough,
330 * we don't expect a DRBD request to be stalled for several month.
333 /* before actual request processing */
334 unsigned long in_actlog_jif;
337 unsigned long pre_submit_jif;
340 unsigned long pre_send_jif;
341 unsigned long acked_jif;
342 unsigned long net_done_jif;
344 /* Possibly even more detail to track each phase:
345 * master_completion_jif
346 * how long did it take to complete the master bio
347 * (application visible latency)
349 * how long the master bio was blocked until we finally allocated
352 * how long did we wait for activity log transactions
355 * when did we finally queue it for sending
357 * when did we start sending it
359 * how long did we block in the network stack trying to send it
361 * when did we receive (or fake, in protocol A) a remote ACK
363 * when did we receive final acknowledgement (P_BARRIER_ACK),
364 * or decide, e.g. on connection loss, that we do no longer expect
365 * anything from this peer for this request.
369 * when did we start submiting to the lower level device,
370 * and how long did we block in that submit function
371 * local_completion_jif
372 * how long did it take the lower level device to complete this request
376 /* once it hits 0, we may complete the master_bio */
377 atomic_t completion_ref;
378 /* once it hits 0, we may destroy this drbd_request object */
381 unsigned rq_state; /* see comments above _req_mod() */
385 struct drbd_connection *connection;
386 struct list_head list;
387 unsigned int barrier_nr;
388 atomic_t epoch_size; /* increased on every request added. */
389 atomic_t active; /* increased on every req. added, and dec on every finished. */
393 /* Prototype declaration of function defined in drbd_receiver.c */
394 int drbdd_init(struct drbd_thread *);
395 int drbd_asender(struct drbd_thread *);
397 /* drbd_epoch flag bits */
399 DE_HAVE_BARRIER_NUMBER,
406 EV_CLEANUP = 32, /* used as flag */
414 struct drbd_peer_request {
416 struct drbd_peer_device *peer_device;
417 struct drbd_epoch *epoch; /* for writes */
419 atomic_t pending_bios;
420 struct drbd_interval i;
421 /* see comments on ee flag bits below */
423 unsigned long submit_jif;
426 struct digest_info *digest;
431 * While corresponding bios are in flight, the only modification will be
432 * set_bit WAS_ERROR, which has to be atomic.
433 * If no bios are in flight yet, or all have been completed,
434 * non-atomic modification to ee->flags is ok.
437 __EE_CALL_AL_COMPLETE_IO,
438 __EE_MAY_SET_IN_SYNC,
440 /* is this a TRIM aka REQ_DISCARD? */
442 /* our lower level cannot handle trim,
443 * and we want to fall back to zeroout instead */
444 __EE_IS_TRIM_USE_ZEROOUT,
446 /* In case a barrier failed,
447 * we need to resubmit without the barrier flag. */
450 /* we may have several bios per peer request.
451 * if any of those fail, we set this flag atomically
452 * from the endio callback */
455 /* This ee has a pointer to a digest instead of a block id */
458 /* Conflicting local requests need to be restarted after this request */
459 __EE_RESTART_REQUESTS,
461 /* The peer wants a write ACK for this (wire proto C) */
464 /* Is set when net_conf had two_primaries set while creating this peer_req */
465 __EE_IN_INTERVAL_TREE,
468 /* has this been submitted, or does it still wait for something else? */
471 /* this is/was a write request */
474 /* this originates from application on peer
475 * (not some resync or verify or other DRBD internal request) */
478 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
479 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
480 #define EE_IS_TRIM (1<<__EE_IS_TRIM)
481 #define EE_IS_TRIM_USE_ZEROOUT (1<<__EE_IS_TRIM_USE_ZEROOUT)
482 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
483 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
484 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
485 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
486 #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK)
487 #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE)
488 #define EE_SUBMITTED (1<<__EE_SUBMITTED)
489 #define EE_WRITE (1<<__EE_WRITE)
490 #define EE_APPLICATION (1<<__EE_APPLICATION)
492 /* flag bits per device */
494 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
495 MD_DIRTY, /* current uuids and flags not yet on disk */
496 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
499 CRASHED_PRIMARY, /* This node was a crashed primary.
500 * Gets cleared when the state.conn
501 * goes into C_CONNECTED state. */
504 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */
506 SUSPEND_IO, /* suspend application io */
507 BITMAP_IO, /* suspend application io;
508 once no more io in flight, start bitmap io */
509 BITMAP_IO_QUEUED, /* Started bitmap IO */
510 WAS_IO_ERROR, /* Local disk failed, returned IO error */
511 WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */
512 FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */
513 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
514 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
515 * the peer, if it changed there as well. */
516 NEW_CUR_UUID, /* Create new current UUID when thawing IO */
517 AL_SUSPENDED, /* Activity logging is currently suspended. */
518 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */
519 B_RS_H_DONE, /* Before resync handler done (already executed) */
520 DISCARD_MY_DATA, /* discard_my_data flag per volume */
523 FLUSH_PENDING, /* if set, device->flush_jif is when we submitted that flush
524 * from drbd_flush_after_epoch() */
526 /* cleared only after backing device related structures have been destroyed. */
527 GOING_DISKLESS, /* Disk is being detached, because of io-error, or admin request. */
529 /* to be used in drbd_device_post_work() */
530 GO_DISKLESS, /* tell worker to schedule cleanup before detach */
531 DESTROY_DISK, /* tell worker to close backing devices and destroy related structures. */
532 MD_SYNC, /* tell worker to call drbd_md_sync() */
533 RS_START, /* tell worker to start resync/OV */
534 RS_PROGRESS, /* tell worker that resync made significant progress */
535 RS_DONE, /* tell worker that resync is done */
538 struct drbd_bitmap; /* opaque for drbd_device */
540 /* definition of bits in bm_flags to be used in drbd_bm_lock
541 * and drbd_bitmap_io and friends. */
543 /* do we need to kfree, or vfree bm_pages? */
544 BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */
546 /* currently locked for bulk operation */
547 BM_LOCKED_MASK = 0xf,
549 /* in detail, that is: */
554 /* so we can mark it locked for bulk operation,
555 * and still allow all non-bulk operations */
558 /* (test bit, count bit) allowed (common case) */
559 BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED,
561 /* testing bits, as well as setting new bits allowed, but clearing bits
562 * would be unexpected. Used during bitmap receive. Setting new bits
563 * requires sending of "out-of-sync" information, though. */
564 BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED,
566 /* for drbd_bm_write_copy_pages, everything is allowed,
567 * only concurrent bulk operations are locked out. */
568 BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED,
571 struct drbd_work_queue {
573 spinlock_t q_lock; /* to protect the list. */
574 wait_queue_head_t q_wait;
579 struct socket *socket;
580 /* this way we get our
581 * send/receive buffers off the stack */
587 u64 md_offset; /* sector offset to 'super' block */
589 u64 la_size_sect; /* last agreed size, unit sectors */
590 spinlock_t uuid_lock;
596 s32 al_offset; /* signed relative sector offset to activity log */
597 s32 bm_offset; /* signed relative sector offset to bitmap */
599 /* cached value of bdev->disk_conf->meta_dev_idx (see below) */
602 /* see al_tr_number_to_on_disk_sector() */
604 u32 al_stripe_size_4k;
605 u32 al_size_4k; /* cached product of the above */
608 struct drbd_backing_dev {
609 struct block_device *backing_bdev;
610 struct block_device *md_bdev;
612 struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */
613 sector_t known_size; /* last known size of that backing device */
618 unsigned long start_jif; /* last call to drbd_md_get_buffer */
619 unsigned long submit_jif; /* last _drbd_md_sync_page_io() submit */
620 const char *current_use;
630 int (*io_fn)(struct drbd_device *device);
631 void (*done)(struct drbd_device *device, int rv);
634 enum write_ordering_e {
641 unsigned int head_index;
643 int total; /* sum of all values */
646 extern struct fifo_buffer *fifo_alloc(int fifo_size);
648 /* flag bits per connection */
650 NET_CONGESTED, /* The data socket is congested */
651 RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */
652 SEND_PING, /* whether asender should send a ping asap */
653 SIGNAL_ASENDER, /* whether asender wants to be interrupted */
654 GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */
655 CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */
658 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
659 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */
660 STATE_SENT, /* Do not change state/UUIDs while this is set */
661 CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
662 * pending, from drbd worker context.
663 * If set, bdi_write_congested() returns true,
664 * so shrink_page_list() would not recurse into,
665 * and potentially deadlock on, this drbd worker.
669 DEVICE_WORK_PENDING, /* tell worker that some device has pending work */
672 struct drbd_resource {
674 #ifdef CONFIG_DEBUG_FS
675 struct dentry *debugfs_res;
676 struct dentry *debugfs_res_volumes;
677 struct dentry *debugfs_res_connections;
678 struct dentry *debugfs_res_in_flight_summary;
681 struct idr devices; /* volume number to device mapping */
682 struct list_head connections;
683 struct list_head resources;
684 struct res_opts res_opts;
685 struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */
686 struct mutex adm_mutex; /* mutex to serialize administrative requests */
689 unsigned susp:1; /* IO suspended by user */
690 unsigned susp_nod:1; /* IO suspended because no data */
691 unsigned susp_fen:1; /* IO suspended because fence peer handler runs */
693 enum write_ordering_e write_ordering;
695 cpumask_var_t cpu_mask;
698 struct drbd_thread_timing_details
700 unsigned long start_jif;
702 const char *caller_fn;
707 struct drbd_connection {
708 struct list_head connections;
709 struct drbd_resource *resource;
710 #ifdef CONFIG_DEBUG_FS
711 struct dentry *debugfs_conn;
712 struct dentry *debugfs_conn_callback_history;
713 struct dentry *debugfs_conn_oldest_requests;
716 struct idr peer_devices; /* volume number to peer device mapping */
717 enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
718 struct mutex cstate_mutex; /* Protects graceful disconnects */
719 unsigned int connect_cnt; /* Inc each time a connection is established */
722 struct net_conf *net_conf; /* content protected by rcu */
723 wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */
725 struct sockaddr_storage my_addr;
727 struct sockaddr_storage peer_addr;
730 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
731 struct drbd_socket meta; /* ping/ack (metadata) packets */
732 int agreed_pro_version; /* actually used protocol version */
734 unsigned long last_received; /* in jiffies, either socket */
735 unsigned int ko_count;
737 struct list_head transfer_log; /* all requests not yet fully processed */
739 struct crypto_hash *cram_hmac_tfm;
740 struct crypto_hash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */
741 struct crypto_hash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */
742 struct crypto_hash *csums_tfm;
743 struct crypto_hash *verify_tfm;
748 struct drbd_epoch *current_epoch;
749 spinlock_t epoch_lock;
751 atomic_t current_tle_nr; /* transfer log epoch number */
752 unsigned current_tle_writes; /* writes seen within this tl epoch */
754 unsigned long last_reconnect_jif;
755 struct drbd_thread receiver;
756 struct drbd_thread worker;
757 struct drbd_thread asender;
760 * so we can look up the oldest pending requests more quickly.
761 * protected by resource->req_lock */
762 struct drbd_request *req_next; /* DRBD 9: todo.req_next */
763 struct drbd_request *req_ack_pending;
764 struct drbd_request *req_not_net_done;
767 struct drbd_work_queue sender_work;
769 #define DRBD_THREAD_DETAILS_HIST 16
770 unsigned int w_cb_nr; /* keeps counting up */
771 unsigned int r_cb_nr; /* keeps counting up */
772 struct drbd_thread_timing_details w_timing_details[DRBD_THREAD_DETAILS_HIST];
773 struct drbd_thread_timing_details r_timing_details[DRBD_THREAD_DETAILS_HIST];
776 /* whether this sender thread
777 * has processed a single write yet. */
778 bool seen_any_write_yet;
780 /* Which barrier number to send with the next P_BARRIER */
781 int current_epoch_nr;
783 /* how many write requests have been sent
784 * with req->epoch == current_epoch_nr.
785 * If none, no P_BARRIER will be sent. */
786 unsigned current_epoch_writes;
790 void __update_timing_details(
791 struct drbd_thread_timing_details *tdp,
794 const char *fn, const unsigned int line);
796 #define update_worker_timing_details(c, cb) \
797 __update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ )
798 #define update_receiver_timing_details(c, cb) \
799 __update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ )
801 struct submit_worker {
802 struct workqueue_struct *wq;
803 struct work_struct worker;
805 /* protected by ..->resource->req_lock */
806 struct list_head writes;
809 struct drbd_peer_device {
810 struct list_head peer_devices;
811 struct drbd_device *device;
812 struct drbd_connection *connection;
813 #ifdef CONFIG_DEBUG_FS
814 struct dentry *debugfs_peer_dev;
819 struct drbd_resource *resource;
820 struct list_head peer_devices;
821 struct list_head pending_bitmap_io;
823 unsigned long flush_jif;
824 #ifdef CONFIG_DEBUG_FS
825 struct dentry *debugfs_minor;
826 struct dentry *debugfs_vol;
827 struct dentry *debugfs_vol_oldest_requests;
828 struct dentry *debugfs_vol_act_log_extents;
829 struct dentry *debugfs_vol_resync_extents;
830 struct dentry *debugfs_vol_data_gen_id;
833 unsigned int vnr; /* volume number within the connection */
834 unsigned int minor; /* device minor number */
838 /* things that are stored as / read from meta data on disk */
841 /* configured by drbdsetup */
842 struct drbd_backing_dev *ldev __protected_by(local);
844 sector_t p_size; /* partner's disk size */
845 struct request_queue *rq_queue;
846 struct block_device *this_bdev;
847 struct gendisk *vdisk;
849 unsigned long last_reattach_jif;
850 struct drbd_work resync_work;
851 struct drbd_work unplug_work;
852 struct timer_list resync_timer;
853 struct timer_list md_sync_timer;
854 struct timer_list start_resync_timer;
855 struct timer_list request_timer;
857 /* Used after attach while negotiating new disk state. */
858 union drbd_state new_state_tmp;
860 union drbd_dev_state state;
861 wait_queue_head_t misc_wait;
862 wait_queue_head_t state_wait; /* upon each state change. */
863 unsigned int send_cnt;
864 unsigned int recv_cnt;
865 unsigned int read_cnt;
866 unsigned int writ_cnt;
867 unsigned int al_writ_cnt;
868 unsigned int bm_writ_cnt;
869 atomic_t ap_bio_cnt; /* Requests we need to complete */
870 atomic_t ap_actlog_cnt; /* Requests waiting for activity log */
871 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
872 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
873 atomic_t unacked_cnt; /* Need to send replies for */
874 atomic_t local_cnt; /* Waiting for local completion */
876 /* Interval tree of pending local requests */
877 struct rb_root read_requests;
878 struct rb_root write_requests;
880 /* for statistics and timeouts */
881 /* [0] read, [1] write */
882 struct list_head pending_master_completion[2];
883 struct list_head pending_completion[2];
885 /* use checksums for *this* resync */
887 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
888 unsigned long rs_total;
889 /* number of resync blocks that failed in this run */
890 unsigned long rs_failed;
891 /* Syncer's start time [unit jiffies] */
892 unsigned long rs_start;
893 /* cumulated time in PausedSyncX state [unit jiffies] */
894 unsigned long rs_paused;
895 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
896 unsigned long rs_same_csum;
897 #define DRBD_SYNC_MARKS 8
898 #define DRBD_SYNC_MARK_STEP (3*HZ)
899 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
900 unsigned long rs_mark_left[DRBD_SYNC_MARKS];
901 /* marks's time [unit jiffies] */
902 unsigned long rs_mark_time[DRBD_SYNC_MARKS];
903 /* current index into rs_mark_{left,time} */
905 unsigned long rs_last_bcast; /* [unit jiffies] */
907 /* where does the admin want us to start? (sector) */
908 sector_t ov_start_sector;
909 sector_t ov_stop_sector;
910 /* where are we now? (sector) */
911 sector_t ov_position;
912 /* Start sector of out of sync range (to merge printk reporting). */
913 sector_t ov_last_oos_start;
914 /* size of out-of-sync range in sectors. */
915 sector_t ov_last_oos_size;
916 unsigned long ov_left; /* in bits */
918 struct drbd_bitmap *bitmap;
919 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
921 /* Used to track operations of resync... */
922 struct lru_cache *resync;
923 /* Number of locked elements in resync LRU */
924 unsigned int resync_locked;
925 /* resync extent number waiting for application requests */
926 unsigned int resync_wenr;
931 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
932 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
933 struct list_head done_ee; /* need to send P_WRITE_ACK */
934 struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */
935 struct list_head net_ee; /* zero-copy network send in progress */
938 struct list_head resync_reads;
939 atomic_t pp_in_use; /* allocated from page pool */
940 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */
941 wait_queue_head_t ee_wait;
942 struct drbd_md_io md_io;
944 wait_queue_head_t al_wait;
945 struct lru_cache *act_log; /* activity log */
946 unsigned int al_tr_number;
948 wait_queue_head_t seq_wait;
950 unsigned int peer_seq;
951 spinlock_t peer_seq_lock;
952 unsigned long comm_bm_set; /* communicated number of set bits. */
953 struct bm_io_work bm_io_work;
954 u64 ed_uuid; /* UUID of the exposed data */
955 struct mutex own_state_mutex;
956 struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */
957 char congestion_reason; /* Why we where congested... */
958 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
959 atomic_t rs_sect_ev; /* for submitted resync data rate, both */
960 int rs_last_sect_ev; /* counter to compare with */
961 int rs_last_events; /* counter of read or write "events" (unit sectors)
962 * on the lower level device when we last looked. */
963 int c_sync_rate; /* current resync rate after syncer throttle magic */
964 struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */
965 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
966 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
967 unsigned int peer_max_bio_size;
968 unsigned int local_max_bio_size;
970 /* any requests that would block in drbd_make_request()
971 * are deferred to this single-threaded work queue */
972 struct submit_worker submit;
975 struct drbd_bm_aio_ctx {
976 struct drbd_device *device;
977 struct list_head list; /* on device->pending_bitmap_io */;
978 unsigned long start_jif;
982 #define BM_AIO_COPY_PAGES 1
983 #define BM_AIO_WRITE_HINTED 2
984 #define BM_AIO_WRITE_ALL_PAGES 4
985 #define BM_AIO_READ 8
990 struct drbd_config_context {
991 /* assigned from drbd_genlmsghdr */
993 /* assigned from request attributes, if present */
995 #define VOLUME_UNSPECIFIED (-1U)
996 /* pointer into the request skb,
997 * limited lifetime! */
999 struct nlattr *my_addr;
1000 struct nlattr *peer_addr;
1003 struct sk_buff *reply_skb;
1004 /* pointer into reply buffer */
1005 struct drbd_genlmsghdr *reply_dh;
1006 /* resolved from attributes, if possible */
1007 struct drbd_device *device;
1008 struct drbd_resource *resource;
1009 struct drbd_connection *connection;
1012 static inline struct drbd_device *minor_to_device(unsigned int minor)
1014 return (struct drbd_device *)idr_find(&drbd_devices, minor);
1017 static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device)
1019 return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices);
1022 #define for_each_resource(resource, _resources) \
1023 list_for_each_entry(resource, _resources, resources)
1025 #define for_each_resource_rcu(resource, _resources) \
1026 list_for_each_entry_rcu(resource, _resources, resources)
1028 #define for_each_resource_safe(resource, tmp, _resources) \
1029 list_for_each_entry_safe(resource, tmp, _resources, resources)
1031 #define for_each_connection(connection, resource) \
1032 list_for_each_entry(connection, &resource->connections, connections)
1034 #define for_each_connection_rcu(connection, resource) \
1035 list_for_each_entry_rcu(connection, &resource->connections, connections)
1037 #define for_each_connection_safe(connection, tmp, resource) \
1038 list_for_each_entry_safe(connection, tmp, &resource->connections, connections)
1040 #define for_each_peer_device(peer_device, device) \
1041 list_for_each_entry(peer_device, &device->peer_devices, peer_devices)
1043 #define for_each_peer_device_rcu(peer_device, device) \
1044 list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices)
1046 #define for_each_peer_device_safe(peer_device, tmp, device) \
1047 list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices)
1049 static inline unsigned int device_to_minor(struct drbd_device *device)
1051 return device->minor;
1055 * function declarations
1056 *************************/
1062 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1065 extern void drbd_init_set_defaults(struct drbd_device *device);
1066 extern int drbd_thread_start(struct drbd_thread *thi);
1067 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1069 extern void drbd_thread_current_set_cpu(struct drbd_thread *thi);
1071 #define drbd_thread_current_set_cpu(A) ({})
1073 extern void tl_release(struct drbd_connection *, unsigned int barrier_nr,
1074 unsigned int set_size);
1075 extern void tl_clear(struct drbd_connection *);
1076 extern void drbd_free_sock(struct drbd_connection *connection);
1077 extern int drbd_send(struct drbd_connection *connection, struct socket *sock,
1078 void *buf, size_t size, unsigned msg_flags);
1079 extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t,
1082 extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd);
1083 extern int drbd_send_protocol(struct drbd_connection *connection);
1084 extern int drbd_send_uuids(struct drbd_peer_device *);
1085 extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *);
1086 extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *);
1087 extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags);
1088 extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s);
1089 extern int drbd_send_current_state(struct drbd_peer_device *);
1090 extern int drbd_send_sync_param(struct drbd_peer_device *);
1091 extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr,
1093 extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet,
1094 struct drbd_peer_request *);
1095 extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet,
1096 struct p_block_req *rp);
1097 extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet,
1098 struct p_data *dp, int data_size);
1099 extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet,
1100 sector_t sector, int blksize, u64 block_id);
1101 extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *);
1102 extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet,
1103 struct drbd_peer_request *);
1104 extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req);
1105 extern int drbd_send_drequest(struct drbd_peer_device *, int cmd,
1106 sector_t sector, int size, u64 block_id);
1107 extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector,
1108 int size, void *digest, int digest_size,
1109 enum drbd_packet cmd);
1110 extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size);
1112 extern int drbd_send_bitmap(struct drbd_device *device);
1113 extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode);
1114 extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode);
1115 extern void drbd_free_ldev(struct drbd_backing_dev *ldev);
1116 extern void drbd_device_cleanup(struct drbd_device *device);
1117 void drbd_print_uuids(struct drbd_device *device, const char *text);
1119 extern void conn_md_sync(struct drbd_connection *connection);
1120 extern void drbd_md_write(struct drbd_device *device, void *buffer);
1121 extern void drbd_md_sync(struct drbd_device *device);
1122 extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev);
1123 extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1124 extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1125 extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local);
1126 extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local);
1127 extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local);
1128 extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1129 extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local);
1130 extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local);
1131 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1132 extern void drbd_md_mark_dirty(struct drbd_device *device);
1133 extern void drbd_queue_bitmap_io(struct drbd_device *device,
1134 int (*io_fn)(struct drbd_device *),
1135 void (*done)(struct drbd_device *, int),
1136 char *why, enum bm_flag flags);
1137 extern int drbd_bitmap_io(struct drbd_device *device,
1138 int (*io_fn)(struct drbd_device *),
1139 char *why, enum bm_flag flags);
1140 extern int drbd_bitmap_io_from_worker(struct drbd_device *device,
1141 int (*io_fn)(struct drbd_device *),
1142 char *why, enum bm_flag flags);
1143 extern int drbd_bmio_set_n_write(struct drbd_device *device) __must_hold(local);
1144 extern int drbd_bmio_clear_n_write(struct drbd_device *device) __must_hold(local);
1148 * We currently have two possible layouts.
1149 * Offsets in (512 byte) sectors.
1151 * |----------- md_size_sect ------------------|
1152 * [ 4k superblock ][ activity log ][ Bitmap ]
1153 * | al_offset == 8 |
1154 * | bm_offset = al_offset + X |
1155 * ==> bitmap sectors = md_size_sect - bm_offset
1158 * old, indexed fixed size meta data:
1161 * |----------- md_size_sect ------------------|
1162 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*]
1164 * | bm_offset = al_offset - Y |
1165 * ==> bitmap sectors = Y = al_offset - bm_offset
1167 * [padding*] are zero or up to 7 unused 512 Byte sectors to the
1168 * end of the device, so that the [4k superblock] will be 4k aligned.
1170 * The activity log consists of 4k transaction blocks,
1171 * which are written in a ring-buffer, or striped ring-buffer like fashion,
1172 * which are writtensize used to be fixed 32kB,
1173 * but is about to become configurable.
1176 /* Our old fixed size meta data layout
1177 * allows up to about 3.8TB, so if you want more,
1178 * you need to use the "flexible" meta data format. */
1179 #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */
1180 #define MD_4kB_SECT 8
1181 #define MD_32kB_SECT 64
1183 /* One activity log extent represents 4M of storage */
1184 #define AL_EXTENT_SHIFT 22
1185 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1187 /* We could make these currently hardcoded constants configurable
1188 * variables at create-md time (or even re-configurable at runtime?).
1189 * Which will require some more changes to the DRBD "super block"
1192 * updates per transaction:
1193 * This many changes to the active set can be logged with one transaction.
1194 * This number is arbitrary.
1195 * context per transaction:
1196 * This many context extent numbers are logged with each transaction.
1197 * This number is resulting from the transaction block size (4k), the layout
1198 * of the transaction header, and the number of updates per transaction.
1199 * See drbd_actlog.c:struct al_transaction_on_disk
1201 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1202 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1204 #if BITS_PER_LONG == 32
1206 #define cpu_to_lel(A) cpu_to_le32(A)
1207 #define lel_to_cpu(A) le32_to_cpu(A)
1208 #elif BITS_PER_LONG == 64
1210 #define cpu_to_lel(A) cpu_to_le64(A)
1211 #define lel_to_cpu(A) le64_to_cpu(A)
1213 #error "LN2 of BITS_PER_LONG unknown!"
1217 /* 16MB sized 'bitmap extent' to track syncer usage */
1219 int rs_left; /* number of bits set (out of sync) in this extent. */
1220 int rs_failed; /* number of failed resync requests in this extent. */
1221 unsigned long flags;
1222 struct lc_element lce;
1225 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1226 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1227 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1231 * We need to store one bit for a block.
1232 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1233 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1234 * Bit 1 ==> local node thinks this block needs to be synced.
1237 #define SLEEP_TIME (HZ/10)
1239 /* We do bitmap IO in units of 4k blocks.
1240 * We also still have a hardcoded 4k per bit relation. */
1241 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1242 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1243 /* mostly arbitrarily set the represented size of one bitmap extent,
1244 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1245 * at 4k per bit resolution) */
1246 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
1247 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1249 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1250 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1253 /* thus many _storage_ sectors are described by one bit */
1254 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1255 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1256 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1258 /* bit to represented kilo byte conversion */
1259 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1261 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1262 * _storage_ sector is located in */
1263 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1264 #define BM_BIT_TO_EXT(x) ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1266 /* first storage sector a bitmap extent corresponds to */
1267 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1268 /* how much _storage_ sectors we have per bitmap extent */
1269 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1270 /* how many bits are covered by one bitmap extent (resync extent) */
1271 #define BM_BITS_PER_EXT (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1273 #define BM_BLOCKS_PER_BM_EXT_MASK (BM_BITS_PER_EXT - 1)
1276 /* in one sector of the bitmap, we have this many activity_log extents. */
1277 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1279 /* the extent in "PER_EXTENT" below is an activity log extent
1280 * we need that many (long words/bytes) to store the bitmap
1281 * of one AL_EXTENT_SIZE chunk of storage.
1282 * we can store the bitmap for that many AL_EXTENTS within
1283 * one sector of the _on_disk_ bitmap:
1284 * bit 0 bit 37 bit 38 bit (512*8)-1
1285 * ...|........|........|.. // ..|........|
1286 * sect. 0 `296 `304 ^(512*8*8)-1
1288 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1289 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1290 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1293 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1294 /* we have a certain meta data variant that has a fixed on-disk size of 128
1295 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity
1296 * log, leaving this many sectors for the bitmap.
1299 #define DRBD_MAX_SECTORS_FIXED_BM \
1300 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9)))
1301 #if !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1302 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1303 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1305 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM
1306 /* 16 TB in units of sectors */
1307 #if BITS_PER_LONG == 32
1308 /* adjust by one page worth of bitmap,
1309 * so we won't wrap around in drbd_bm_find_next_bit.
1310 * you should use 64bit OS for that much storage, anyways. */
1311 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1313 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1314 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1315 /* corresponds to (1UL << 38) bits right now. */
1319 /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE,
1320 * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte.
1321 * Since we may live in a mixed-platform cluster,
1322 * we limit us to a platform agnostic constant here for now.
1323 * A followup commit may allow even bigger BIO sizes,
1324 * once we thought that through. */
1325 #define DRBD_MAX_BIO_SIZE (1U << 20)
1326 #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1327 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1329 #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */
1331 #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */
1332 #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */
1334 /* For now, don't allow more than one activity log extent worth of data
1335 * to be discarded in one go. We may need to rework drbd_al_begin_io()
1336 * to allow for even larger discard ranges */
1337 #define DRBD_MAX_DISCARD_SIZE AL_EXTENT_SIZE
1338 #define DRBD_MAX_DISCARD_SECTORS (DRBD_MAX_DISCARD_SIZE >> 9)
1340 extern int drbd_bm_init(struct drbd_device *device);
1341 extern int drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits);
1342 extern void drbd_bm_cleanup(struct drbd_device *device);
1343 extern void drbd_bm_set_all(struct drbd_device *device);
1344 extern void drbd_bm_clear_all(struct drbd_device *device);
1345 /* set/clear/test only a few bits at a time */
1346 extern int drbd_bm_set_bits(
1347 struct drbd_device *device, unsigned long s, unsigned long e);
1348 extern int drbd_bm_clear_bits(
1349 struct drbd_device *device, unsigned long s, unsigned long e);
1350 extern int drbd_bm_count_bits(
1351 struct drbd_device *device, const unsigned long s, const unsigned long e);
1352 /* bm_set_bits variant for use while holding drbd_bm_lock,
1353 * may process the whole bitmap in one go */
1354 extern void _drbd_bm_set_bits(struct drbd_device *device,
1355 const unsigned long s, const unsigned long e);
1356 extern int drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr);
1357 extern int drbd_bm_e_weight(struct drbd_device *device, unsigned long enr);
1358 extern int drbd_bm_read(struct drbd_device *device) __must_hold(local);
1359 extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr);
1360 extern int drbd_bm_write(struct drbd_device *device) __must_hold(local);
1361 extern int drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local);
1362 extern int drbd_bm_write_lazy(struct drbd_device *device, unsigned upper_idx) __must_hold(local);
1363 extern int drbd_bm_write_all(struct drbd_device *device) __must_hold(local);
1364 extern int drbd_bm_write_copy_pages(struct drbd_device *device) __must_hold(local);
1365 extern size_t drbd_bm_words(struct drbd_device *device);
1366 extern unsigned long drbd_bm_bits(struct drbd_device *device);
1367 extern sector_t drbd_bm_capacity(struct drbd_device *device);
1369 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1370 extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo);
1371 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1372 extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo);
1373 extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo);
1374 extern unsigned long _drbd_bm_total_weight(struct drbd_device *device);
1375 extern unsigned long drbd_bm_total_weight(struct drbd_device *device);
1376 /* for receive_bitmap */
1377 extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset,
1378 size_t number, unsigned long *buffer);
1379 /* for _drbd_send_bitmap */
1380 extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset,
1381 size_t number, unsigned long *buffer);
1383 extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags);
1384 extern void drbd_bm_unlock(struct drbd_device *device);
1387 extern struct kmem_cache *drbd_request_cache;
1388 extern struct kmem_cache *drbd_ee_cache; /* peer requests */
1389 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1390 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1391 extern mempool_t *drbd_request_mempool;
1392 extern mempool_t *drbd_ee_mempool;
1394 /* drbd's page pool, used to buffer data received from the peer,
1395 * or data requested by the peer.
1397 * This does not have an emergency reserve.
1399 * When allocating from this pool, it first takes pages from the pool.
1400 * Only if the pool is depleted will try to allocate from the system.
1402 * The assumption is that pages taken from this pool will be processed,
1403 * and given back, "quickly", and then can be recycled, so we can avoid
1404 * frequent calls to alloc_page(), and still will be able to make progress even
1405 * under memory pressure.
1407 extern struct page *drbd_pp_pool;
1408 extern spinlock_t drbd_pp_lock;
1409 extern int drbd_pp_vacant;
1410 extern wait_queue_head_t drbd_pp_wait;
1412 /* We also need a standard (emergency-reserve backed) page pool
1413 * for meta data IO (activity log, bitmap).
1414 * We can keep it global, as long as it is used as "N pages at a time".
1415 * 128 should be plenty, currently we probably can get away with as few as 1.
1417 #define DRBD_MIN_POOL_PAGES 128
1418 extern mempool_t *drbd_md_io_page_pool;
1420 /* We also need to make sure we get a bio
1421 * when we need it for housekeeping purposes */
1422 extern struct bio_set *drbd_md_io_bio_set;
1423 /* to allocate from that set */
1424 extern struct bio *bio_alloc_drbd(gfp_t gfp_mask);
1426 extern rwlock_t global_state_lock;
1428 extern int conn_lowest_minor(struct drbd_connection *connection);
1429 extern enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor);
1430 extern void drbd_destroy_device(struct kref *kref);
1431 extern void drbd_delete_device(struct drbd_device *device);
1433 extern struct drbd_resource *drbd_create_resource(const char *name);
1434 extern void drbd_free_resource(struct drbd_resource *resource);
1436 extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts);
1437 extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts);
1438 extern void drbd_destroy_connection(struct kref *kref);
1439 extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len,
1440 void *peer_addr, int peer_addr_len);
1441 extern struct drbd_resource *drbd_find_resource(const char *name);
1442 extern void drbd_destroy_resource(struct kref *kref);
1443 extern void conn_free_crypto(struct drbd_connection *connection);
1445 extern int proc_details;
1448 extern void do_submit(struct work_struct *ws);
1449 extern void __drbd_make_request(struct drbd_device *, struct bio *, unsigned long);
1450 extern void drbd_make_request(struct request_queue *q, struct bio *bio);
1451 extern int drbd_read_remote(struct drbd_device *device, struct drbd_request *req);
1452 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1453 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1457 extern void drbd_suspend_io(struct drbd_device *device);
1458 extern void drbd_resume_io(struct drbd_device *device);
1459 extern char *ppsize(char *buf, unsigned long long size);
1460 extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int);
1461 enum determine_dev_size {
1462 DS_ERROR_SHRINK = -3,
1463 DS_ERROR_SPACE_MD = -2,
1468 DS_GREW_FROM_ZERO = 3,
1470 extern enum determine_dev_size
1471 drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local);
1472 extern void resync_after_online_grow(struct drbd_device *);
1473 extern void drbd_reconsider_max_bio_size(struct drbd_device *device, struct drbd_backing_dev *bdev);
1474 extern enum drbd_state_rv drbd_set_role(struct drbd_device *device,
1475 enum drbd_role new_role,
1477 extern bool conn_try_outdate_peer(struct drbd_connection *connection);
1478 extern void conn_try_outdate_peer_async(struct drbd_connection *connection);
1479 extern int drbd_khelper(struct drbd_device *device, char *cmd);
1482 /* bi_end_io handlers */
1483 extern void drbd_md_endio(struct bio *bio, int error);
1484 extern void drbd_peer_request_endio(struct bio *bio, int error);
1485 extern void drbd_request_endio(struct bio *bio, int error);
1486 extern int drbd_worker(struct drbd_thread *thi);
1487 enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor);
1488 void drbd_resync_after_changed(struct drbd_device *device);
1489 extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side);
1490 extern void resume_next_sg(struct drbd_device *device);
1491 extern void suspend_other_sg(struct drbd_device *device);
1492 extern int drbd_resync_finished(struct drbd_device *device);
1493 /* maybe rather drbd_main.c ? */
1494 extern void *drbd_md_get_buffer(struct drbd_device *device, const char *intent);
1495 extern void drbd_md_put_buffer(struct drbd_device *device);
1496 extern int drbd_md_sync_page_io(struct drbd_device *device,
1497 struct drbd_backing_dev *bdev, sector_t sector, int rw);
1498 extern void drbd_ov_out_of_sync_found(struct drbd_device *, sector_t, int);
1499 extern void wait_until_done_or_force_detached(struct drbd_device *device,
1500 struct drbd_backing_dev *bdev, unsigned int *done);
1501 extern void drbd_rs_controller_reset(struct drbd_device *device);
1503 static inline void ov_out_of_sync_print(struct drbd_device *device)
1505 if (device->ov_last_oos_size) {
1506 drbd_err(device, "Out of sync: start=%llu, size=%lu (sectors)\n",
1507 (unsigned long long)device->ov_last_oos_start,
1508 (unsigned long)device->ov_last_oos_size);
1510 device->ov_last_oos_size = 0;
1514 extern void drbd_csum_bio(struct crypto_hash *, struct bio *, void *);
1515 extern void drbd_csum_ee(struct crypto_hash *, struct drbd_peer_request *, void *);
1516 /* worker callbacks */
1517 extern int w_e_end_data_req(struct drbd_work *, int);
1518 extern int w_e_end_rsdata_req(struct drbd_work *, int);
1519 extern int w_e_end_csum_rs_req(struct drbd_work *, int);
1520 extern int w_e_end_ov_reply(struct drbd_work *, int);
1521 extern int w_e_end_ov_req(struct drbd_work *, int);
1522 extern int w_ov_finished(struct drbd_work *, int);
1523 extern int w_resync_timer(struct drbd_work *, int);
1524 extern int w_send_write_hint(struct drbd_work *, int);
1525 extern int w_send_dblock(struct drbd_work *, int);
1526 extern int w_send_read_req(struct drbd_work *, int);
1527 extern int w_e_reissue(struct drbd_work *, int);
1528 extern int w_restart_disk_io(struct drbd_work *, int);
1529 extern int w_send_out_of_sync(struct drbd_work *, int);
1530 extern int w_start_resync(struct drbd_work *, int);
1532 extern void resync_timer_fn(unsigned long data);
1533 extern void start_resync_timer_fn(unsigned long data);
1535 extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req);
1537 /* drbd_receiver.c */
1538 extern int drbd_receiver(struct drbd_thread *thi);
1539 extern int drbd_asender(struct drbd_thread *thi);
1540 extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device);
1541 extern bool drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector,
1542 bool throttle_if_app_is_waiting);
1543 extern int drbd_submit_peer_request(struct drbd_device *,
1544 struct drbd_peer_request *, const unsigned,
1546 extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *);
1547 extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64,
1548 sector_t, unsigned int,
1550 gfp_t) __must_hold(local);
1551 extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *,
1553 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1554 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
1555 extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool);
1556 extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
1557 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
1558 extern int drbd_connected(struct drbd_peer_device *);
1560 static inline void drbd_tcp_cork(struct socket *sock)
1563 (void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
1564 (char*)&val, sizeof(val));
1567 static inline void drbd_tcp_uncork(struct socket *sock)
1570 (void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
1571 (char*)&val, sizeof(val));
1574 static inline void drbd_tcp_nodelay(struct socket *sock)
1577 (void) kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1578 (char*)&val, sizeof(val));
1581 static inline void drbd_tcp_quickack(struct socket *sock)
1584 (void) kernel_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1585 (char*)&val, sizeof(val));
1588 /* sets the number of 512 byte sectors of our virtual device */
1589 static inline void drbd_set_my_capacity(struct drbd_device *device,
1592 /* set_capacity(device->this_bdev->bd_disk, size); */
1593 set_capacity(device->vdisk, size);
1594 device->this_bdev->bd_inode->i_size = (loff_t)size << 9;
1598 * used to submit our private bio
1600 static inline void drbd_generic_make_request(struct drbd_device *device,
1601 int fault_type, struct bio *bio)
1604 if (!bio->bi_bdev) {
1605 drbd_err(device, "drbd_generic_make_request: bio->bi_bdev == NULL\n");
1606 bio_endio(bio, -ENODEV);
1610 if (drbd_insert_fault(device, fault_type))
1611 bio_endio(bio, -EIO);
1613 generic_make_request(bio);
1616 void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev,
1617 enum write_ordering_e wo);
1620 extern struct proc_dir_entry *drbd_proc;
1621 extern const struct file_operations drbd_proc_fops;
1622 extern const char *drbd_conn_str(enum drbd_conns s);
1623 extern const char *drbd_role_str(enum drbd_role s);
1626 extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i);
1627 extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i);
1628 extern void drbd_al_begin_io_commit(struct drbd_device *device);
1629 extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i);
1630 extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i);
1631 extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i);
1632 extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector);
1633 extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector);
1634 extern int drbd_try_rs_begin_io(struct drbd_device *device, sector_t sector);
1635 extern void drbd_rs_cancel_all(struct drbd_device *device);
1636 extern int drbd_rs_del_all(struct drbd_device *device);
1637 extern void drbd_rs_failed_io(struct drbd_device *device,
1638 sector_t sector, int size);
1639 extern void drbd_advance_rs_marks(struct drbd_device *device, unsigned long still_to_go);
1641 enum update_sync_bits_mode { RECORD_RS_FAILED, SET_OUT_OF_SYNC, SET_IN_SYNC };
1642 extern int __drbd_change_sync(struct drbd_device *device, sector_t sector, int size,
1643 enum update_sync_bits_mode mode);
1644 #define drbd_set_in_sync(device, sector, size) \
1645 __drbd_change_sync(device, sector, size, SET_IN_SYNC)
1646 #define drbd_set_out_of_sync(device, sector, size) \
1647 __drbd_change_sync(device, sector, size, SET_OUT_OF_SYNC)
1648 #define drbd_rs_failed_io(device, sector, size) \
1649 __drbd_change_sync(device, sector, size, RECORD_RS_FAILED)
1650 extern void drbd_al_shrink(struct drbd_device *device);
1651 extern int drbd_initialize_al(struct drbd_device *, void *);
1654 /* state info broadcast */
1656 enum drbd_state_info_bcast_reason sib_reason;
1660 unsigned helper_exit_code;
1663 union drbd_state os;
1664 union drbd_state ns;
1668 void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib);
1671 * inline helper functions
1672 *************************/
1674 /* see also page_chain_add and friends in drbd_receiver.c */
1675 static inline struct page *page_chain_next(struct page *page)
1677 return (struct page *)page_private(page);
1679 #define page_chain_for_each(page) \
1680 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1681 page = page_chain_next(page))
1682 #define page_chain_for_each_safe(page, n) \
1683 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1686 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req)
1688 struct page *page = peer_req->pages;
1689 page_chain_for_each(page) {
1690 if (page_count(page) > 1)
1696 static inline enum drbd_state_rv
1697 _drbd_set_state(struct drbd_device *device, union drbd_state ns,
1698 enum chg_state_flags flags, struct completion *done)
1700 enum drbd_state_rv rv;
1702 read_lock(&global_state_lock);
1703 rv = __drbd_set_state(device, ns, flags, done);
1704 read_unlock(&global_state_lock);
1709 static inline union drbd_state drbd_read_state(struct drbd_device *device)
1711 struct drbd_resource *resource = device->resource;
1712 union drbd_state rv;
1714 rv.i = device->state.i;
1715 rv.susp = resource->susp;
1716 rv.susp_nod = resource->susp_nod;
1717 rv.susp_fen = resource->susp_fen;
1722 enum drbd_force_detach_flags {
1729 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1730 static inline void __drbd_chk_io_error_(struct drbd_device *device,
1731 enum drbd_force_detach_flags df,
1734 enum drbd_io_error_p ep;
1737 ep = rcu_dereference(device->ldev->disk_conf)->on_io_error;
1740 case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */
1741 if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) {
1742 if (__ratelimit(&drbd_ratelimit_state))
1743 drbd_err(device, "Local IO failed in %s.\n", where);
1744 if (device->state.disk > D_INCONSISTENT)
1745 _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL);
1748 /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */
1750 case EP_CALL_HELPER:
1751 /* Remember whether we saw a READ or WRITE error.
1753 * Recovery of the affected area for WRITE failure is covered
1754 * by the activity log.
1755 * READ errors may fall outside that area though. Certain READ
1756 * errors can be "healed" by writing good data to the affected
1757 * blocks, which triggers block re-allocation in lower layers.
1759 * If we can not write the bitmap after a READ error,
1760 * we may need to trigger a full sync (see w_go_diskless()).
1762 * Force-detach is not really an IO error, but rather a
1763 * desperate measure to try to deal with a completely
1764 * unresponsive lower level IO stack.
1765 * Still it should be treated as a WRITE error.
1767 * Meta IO error is always WRITE error:
1768 * we read meta data only once during attach,
1769 * which will fail in case of errors.
1771 set_bit(WAS_IO_ERROR, &device->flags);
1772 if (df == DRBD_READ_ERROR)
1773 set_bit(WAS_READ_ERROR, &device->flags);
1774 if (df == DRBD_FORCE_DETACH)
1775 set_bit(FORCE_DETACH, &device->flags);
1776 if (device->state.disk > D_FAILED) {
1777 _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL);
1779 "Local IO failed in %s. Detaching...\n", where);
1786 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1787 * @device: DRBD device.
1788 * @error: Error code passed to the IO completion callback
1789 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1791 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1793 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1794 static inline void drbd_chk_io_error_(struct drbd_device *device,
1795 int error, enum drbd_force_detach_flags forcedetach, const char *where)
1798 unsigned long flags;
1799 spin_lock_irqsave(&device->resource->req_lock, flags);
1800 __drbd_chk_io_error_(device, forcedetach, where);
1801 spin_unlock_irqrestore(&device->resource->req_lock, flags);
1807 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1808 * @bdev: Meta data block device.
1810 * BTW, for internal meta data, this happens to be the maximum capacity
1811 * we could agree upon with our peer node.
1813 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1815 switch (bdev->md.meta_dev_idx) {
1816 case DRBD_MD_INDEX_INTERNAL:
1817 case DRBD_MD_INDEX_FLEX_INT:
1818 return bdev->md.md_offset + bdev->md.bm_offset;
1819 case DRBD_MD_INDEX_FLEX_EXT:
1821 return bdev->md.md_offset;
1826 * drbd_md_last_sector() - Return the last sector number of the meta data area
1827 * @bdev: Meta data block device.
1829 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1831 switch (bdev->md.meta_dev_idx) {
1832 case DRBD_MD_INDEX_INTERNAL:
1833 case DRBD_MD_INDEX_FLEX_INT:
1834 return bdev->md.md_offset + MD_4kB_SECT -1;
1835 case DRBD_MD_INDEX_FLEX_EXT:
1837 return bdev->md.md_offset + bdev->md.md_size_sect -1;
1841 /* Returns the number of 512 byte sectors of the device */
1842 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1844 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1845 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0;
1849 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1850 * @bdev: Meta data block device.
1852 * returns the capacity we announce to out peer. we clip ourselves at the
1853 * various MAX_SECTORS, because if we don't, current implementation will
1854 * oops sooner or later
1856 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1860 switch (bdev->md.meta_dev_idx) {
1861 case DRBD_MD_INDEX_INTERNAL:
1862 case DRBD_MD_INDEX_FLEX_INT:
1863 s = drbd_get_capacity(bdev->backing_bdev)
1864 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1865 drbd_md_first_sector(bdev))
1868 case DRBD_MD_INDEX_FLEX_EXT:
1869 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1870 drbd_get_capacity(bdev->backing_bdev));
1871 /* clip at maximum size the meta device can support */
1872 s = min_t(sector_t, s,
1873 BM_EXT_TO_SECT(bdev->md.md_size_sect
1874 - bdev->md.bm_offset));
1877 s = min_t(sector_t, DRBD_MAX_SECTORS,
1878 drbd_get_capacity(bdev->backing_bdev));
1884 * drbd_md_ss() - Return the sector number of our meta data super block
1885 * @bdev: Meta data block device.
1887 static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev)
1889 const int meta_dev_idx = bdev->md.meta_dev_idx;
1891 if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT)
1894 /* Since drbd08, internal meta data is always "flexible".
1895 * position: last 4k aligned block of 4k size */
1896 if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL ||
1897 meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)
1898 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8;
1900 /* external, some index; this is the old fixed size layout */
1901 return MD_128MB_SECT * bdev->md.meta_dev_idx;
1905 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1907 unsigned long flags;
1908 spin_lock_irqsave(&q->q_lock, flags);
1909 list_add_tail(&w->list, &q->q);
1910 spin_unlock_irqrestore(&q->q_lock, flags);
1911 wake_up(&q->q_wait);
1915 drbd_queue_work_if_unqueued(struct drbd_work_queue *q, struct drbd_work *w)
1917 unsigned long flags;
1918 spin_lock_irqsave(&q->q_lock, flags);
1919 if (list_empty_careful(&w->list))
1920 list_add_tail(&w->list, &q->q);
1921 spin_unlock_irqrestore(&q->q_lock, flags);
1922 wake_up(&q->q_wait);
1926 drbd_device_post_work(struct drbd_device *device, int work_bit)
1928 if (!test_and_set_bit(work_bit, &device->flags)) {
1929 struct drbd_connection *connection =
1930 first_peer_device(device)->connection;
1931 struct drbd_work_queue *q = &connection->sender_work;
1932 if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags))
1933 wake_up(&q->q_wait);
1937 extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue);
1939 static inline void wake_asender(struct drbd_connection *connection)
1941 if (test_bit(SIGNAL_ASENDER, &connection->flags))
1942 force_sig(DRBD_SIG, connection->asender.task);
1945 static inline void request_ping(struct drbd_connection *connection)
1947 set_bit(SEND_PING, &connection->flags);
1948 wake_asender(connection);
1951 extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *);
1952 extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *);
1953 extern int conn_send_command(struct drbd_connection *, struct drbd_socket *,
1954 enum drbd_packet, unsigned int, void *,
1956 extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *,
1957 enum drbd_packet, unsigned int, void *,
1960 extern int drbd_send_ping(struct drbd_connection *connection);
1961 extern int drbd_send_ping_ack(struct drbd_connection *connection);
1962 extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state);
1963 extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state);
1965 static inline void drbd_thread_stop(struct drbd_thread *thi)
1967 _drbd_thread_stop(thi, false, true);
1970 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
1972 _drbd_thread_stop(thi, false, false);
1975 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
1977 _drbd_thread_stop(thi, true, false);
1980 /* counts how many answer packets packets we expect from our peer,
1981 * for either explicit application requests,
1982 * or implicit barrier packets as necessary.
1985 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1986 * it is much easier and equally valid to count what we queue for the
1987 * worker, even before it actually was queued or send.
1988 * (drbd_make_request_common; recovery path on read io-error)
1990 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1991 * _req_mod(req, DATA_RECEIVED)
1992 * [from receive_DataReply]
1993 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1994 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1995 * for some reason it is NOT decreased in got_NegAck,
1996 * but in the resulting cleanup code from report_params.
1997 * we should try to remember the reason for that...
1998 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
1999 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
2000 * [from tl_clear_barrier]
2002 static inline void inc_ap_pending(struct drbd_device *device)
2004 atomic_inc(&device->ap_pending_cnt);
2007 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \
2008 if (atomic_read(&device->which) < 0) \
2009 drbd_err(device, "in %s:%d: " #which " = %d < 0 !\n", \
2011 atomic_read(&device->which))
2013 #define dec_ap_pending(device) _dec_ap_pending(device, __func__, __LINE__)
2014 static inline void _dec_ap_pending(struct drbd_device *device, const char *func, int line)
2016 if (atomic_dec_and_test(&device->ap_pending_cnt))
2017 wake_up(&device->misc_wait);
2018 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line);
2021 /* counts how many resync-related answers we still expect from the peer
2023 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
2024 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
2025 * (or P_NEG_ACK with ID_SYNCER)
2027 static inline void inc_rs_pending(struct drbd_device *device)
2029 atomic_inc(&device->rs_pending_cnt);
2032 #define dec_rs_pending(device) _dec_rs_pending(device, __func__, __LINE__)
2033 static inline void _dec_rs_pending(struct drbd_device *device, const char *func, int line)
2035 atomic_dec(&device->rs_pending_cnt);
2036 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line);
2039 /* counts how many answers we still need to send to the peer.
2041 * receive_Data unless protocol A;
2042 * we need to send a P_RECV_ACK (proto B)
2043 * or P_WRITE_ACK (proto C)
2044 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2045 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2046 * receive_Barrier_* we need to send a P_BARRIER_ACK
2048 static inline void inc_unacked(struct drbd_device *device)
2050 atomic_inc(&device->unacked_cnt);
2053 #define dec_unacked(device) _dec_unacked(device, __func__, __LINE__)
2054 static inline void _dec_unacked(struct drbd_device *device, const char *func, int line)
2056 atomic_dec(&device->unacked_cnt);
2057 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
2060 #define sub_unacked(device, n) _sub_unacked(device, n, __func__, __LINE__)
2061 static inline void _sub_unacked(struct drbd_device *device, int n, const char *func, int line)
2063 atomic_sub(n, &device->unacked_cnt);
2064 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
2067 static inline bool is_sync_state(enum drbd_conns connection_state)
2070 (connection_state == C_SYNC_SOURCE
2071 || connection_state == C_SYNC_TARGET
2072 || connection_state == C_PAUSED_SYNC_S
2073 || connection_state == C_PAUSED_SYNC_T);
2077 * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev
2078 * @_device: DRBD device.
2079 * @_min_state: Minimum device state required for success.
2081 * You have to call put_ldev() when finished working with device->ldev.
2083 #define get_ldev_if_state(_device, _min_state) \
2084 (_get_ldev_if_state((_device), (_min_state)) ? \
2085 ({ __acquire(x); true; }) : false)
2086 #define get_ldev(_device) get_ldev_if_state(_device, D_INCONSISTENT)
2088 static inline void put_ldev(struct drbd_device *device)
2090 enum drbd_disk_state disk_state = device->state.disk;
2091 /* We must check the state *before* the atomic_dec becomes visible,
2092 * or we have a theoretical race where someone hitting zero,
2093 * while state still D_FAILED, will then see D_DISKLESS in the
2094 * condition below and calling into destroy, where he must not, yet. */
2095 int i = atomic_dec_return(&device->local_cnt);
2097 /* This may be called from some endio handler,
2098 * so we must not sleep here. */
2101 D_ASSERT(device, i >= 0);
2103 if (disk_state == D_DISKLESS)
2104 /* even internal references gone, safe to destroy */
2105 drbd_device_post_work(device, DESTROY_DISK);
2106 if (disk_state == D_FAILED)
2107 /* all application IO references gone. */
2108 if (!test_and_set_bit(GOING_DISKLESS, &device->flags))
2109 drbd_device_post_work(device, GO_DISKLESS);
2110 wake_up(&device->misc_wait);
2115 static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins)
2119 /* never get a reference while D_DISKLESS */
2120 if (device->state.disk == D_DISKLESS)
2123 atomic_inc(&device->local_cnt);
2124 io_allowed = (device->state.disk >= mins);
2130 extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins);
2133 /* this throttles on-the-fly application requests
2134 * according to max_buffers settings;
2135 * maybe re-implement using semaphores? */
2136 static inline int drbd_get_max_buffers(struct drbd_device *device)
2138 struct net_conf *nc;
2142 nc = rcu_dereference(first_peer_device(device)->connection->net_conf);
2143 mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */
2149 static inline int drbd_state_is_stable(struct drbd_device *device)
2151 union drbd_dev_state s = device->state;
2153 /* DO NOT add a default clause, we want the compiler to warn us
2154 * for any newly introduced state we may have forgotten to add here */
2156 switch ((enum drbd_conns)s.conn) {
2157 /* new io only accepted when there is no connection, ... */
2159 case C_WF_CONNECTION:
2160 /* ... or there is a well established connection. */
2166 case C_PAUSED_SYNC_S:
2167 case C_PAUSED_SYNC_T:
2170 /* transitional states, IO allowed */
2171 case C_DISCONNECTING:
2175 case C_NETWORK_FAILURE:
2176 case C_PROTOCOL_ERROR:
2178 case C_WF_REPORT_PARAMS:
2179 case C_STARTING_SYNC_S:
2180 case C_STARTING_SYNC_T:
2183 /* Allow IO in BM exchange states with new protocols */
2185 if (first_peer_device(device)->connection->agreed_pro_version < 96)
2189 /* no new io accepted in these states */
2191 case C_WF_SYNC_UUID:
2197 switch ((enum drbd_disk_state)s.disk) {
2199 case D_INCONSISTENT:
2204 /* disk state is stable as well. */
2207 /* no new io accepted during transitional states */
2219 static inline int drbd_suspended(struct drbd_device *device)
2221 struct drbd_resource *resource = device->resource;
2223 return resource->susp || resource->susp_fen || resource->susp_nod;
2226 static inline bool may_inc_ap_bio(struct drbd_device *device)
2228 int mxb = drbd_get_max_buffers(device);
2230 if (drbd_suspended(device))
2232 if (test_bit(SUSPEND_IO, &device->flags))
2235 /* to avoid potential deadlock or bitmap corruption,
2236 * in various places, we only allow new application io
2237 * to start during "stable" states. */
2239 /* no new io accepted when attaching or detaching the disk */
2240 if (!drbd_state_is_stable(device))
2243 /* since some older kernels don't have atomic_add_unless,
2244 * and we are within the spinlock anyways, we have this workaround. */
2245 if (atomic_read(&device->ap_bio_cnt) > mxb)
2247 if (test_bit(BITMAP_IO, &device->flags))
2252 static inline bool inc_ap_bio_cond(struct drbd_device *device)
2256 spin_lock_irq(&device->resource->req_lock);
2257 rv = may_inc_ap_bio(device);
2259 atomic_inc(&device->ap_bio_cnt);
2260 spin_unlock_irq(&device->resource->req_lock);
2265 static inline void inc_ap_bio(struct drbd_device *device)
2268 * as long as the device is suspended
2269 * until the bitmap is no longer on the fly during connection
2270 * handshake as long as we would exceed the max_buffer limit.
2272 * to avoid races with the reconnect code,
2273 * we need to atomic_inc within the spinlock. */
2275 wait_event(device->misc_wait, inc_ap_bio_cond(device));
2278 static inline void dec_ap_bio(struct drbd_device *device)
2280 int mxb = drbd_get_max_buffers(device);
2281 int ap_bio = atomic_dec_return(&device->ap_bio_cnt);
2283 D_ASSERT(device, ap_bio >= 0);
2285 if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) {
2286 if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags))
2287 drbd_queue_work(&first_peer_device(device)->
2288 connection->sender_work,
2289 &device->bm_io_work.w);
2292 /* this currently does wake_up for every dec_ap_bio!
2293 * maybe rather introduce some type of hysteresis?
2294 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2296 wake_up(&device->misc_wait);
2299 static inline bool verify_can_do_stop_sector(struct drbd_device *device)
2301 return first_peer_device(device)->connection->agreed_pro_version >= 97 &&
2302 first_peer_device(device)->connection->agreed_pro_version != 100;
2305 static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val)
2307 int changed = device->ed_uuid != val;
2308 device->ed_uuid = val;
2312 static inline int drbd_queue_order_type(struct drbd_device *device)
2314 /* sorry, we currently have no working implementation
2315 * of distributed TCQ stuff */
2316 #ifndef QUEUE_ORDERED_NONE
2317 #define QUEUE_ORDERED_NONE 0
2319 return QUEUE_ORDERED_NONE;
2322 static inline struct drbd_connection *first_connection(struct drbd_resource *resource)
2324 return list_first_entry_or_null(&resource->connections,
2325 struct drbd_connection, connections);