1 /* SPDX-License-Identifier: GPL-2.0-or-later */
5 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
7 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
8 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
9 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
17 #include <crypto/hash.h>
18 #include <linux/compiler.h>
19 #include <linux/types.h>
20 #include <linux/list.h>
21 #include <linux/sched/signal.h>
22 #include <linux/bitops.h>
23 #include <linux/slab.h>
24 #include <linux/ratelimit.h>
25 #include <linux/tcp.h>
26 #include <linux/mutex.h>
27 #include <linux/major.h>
28 #include <linux/blkdev.h>
29 #include <linux/backing-dev.h>
30 #include <linux/idr.h>
31 #include <linux/dynamic_debug.h>
33 #include <linux/lru_cache.h>
34 #include <linux/prefetch.h>
35 #include <linux/drbd_genl_api.h>
36 #include <linux/drbd.h>
37 #include "drbd_strings.h"
38 #include "drbd_state.h"
39 #include "drbd_protocol.h"
42 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
43 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
44 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
46 # define __protected_by(x)
47 # define __protected_read_by(x)
48 # define __protected_write_by(x)
51 /* shared module parameters, defined in drbd_main.c */
52 #ifdef CONFIG_DRBD_FAULT_INJECTION
53 extern int drbd_enable_faults;
54 extern int drbd_fault_rate;
57 extern unsigned int drbd_minor_count;
58 extern char drbd_usermode_helper[];
59 extern int drbd_proc_details;
62 /* This is used to stop/restart our threads.
63 * Cannot use SIGTERM nor SIGKILL, since these
64 * are sent out by init on runlevel changes
65 * I choose SIGHUP for now.
67 #define DRBD_SIGKILL SIGHUP
69 #define ID_IN_SYNC (4711ULL)
70 #define ID_OUT_OF_SYNC (4712ULL)
71 #define ID_SYNCER (-1ULL)
73 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
76 struct drbd_connection;
78 #define __drbd_printk_device(level, device, fmt, args...) \
79 dev_printk(level, disk_to_dev((device)->vdisk), fmt, ## args)
80 #define __drbd_printk_peer_device(level, peer_device, fmt, args...) \
81 dev_printk(level, disk_to_dev((peer_device)->device->vdisk), fmt, ## args)
82 #define __drbd_printk_resource(level, resource, fmt, args...) \
83 printk(level "drbd %s: " fmt, (resource)->name, ## args)
84 #define __drbd_printk_connection(level, connection, fmt, args...) \
85 printk(level "drbd %s: " fmt, (connection)->resource->name, ## args)
87 void drbd_printk_with_wrong_object_type(void);
89 #define __drbd_printk_if_same_type(obj, type, func, level, fmt, args...) \
90 (__builtin_types_compatible_p(typeof(obj), type) || \
91 __builtin_types_compatible_p(typeof(obj), const type)), \
92 func(level, (const type)(obj), fmt, ## args)
94 #define drbd_printk(level, obj, fmt, args...) \
95 __builtin_choose_expr( \
96 __drbd_printk_if_same_type(obj, struct drbd_device *, \
97 __drbd_printk_device, level, fmt, ## args), \
98 __builtin_choose_expr( \
99 __drbd_printk_if_same_type(obj, struct drbd_resource *, \
100 __drbd_printk_resource, level, fmt, ## args), \
101 __builtin_choose_expr( \
102 __drbd_printk_if_same_type(obj, struct drbd_connection *, \
103 __drbd_printk_connection, level, fmt, ## args), \
104 __builtin_choose_expr( \
105 __drbd_printk_if_same_type(obj, struct drbd_peer_device *, \
106 __drbd_printk_peer_device, level, fmt, ## args), \
107 drbd_printk_with_wrong_object_type()))))
109 #define drbd_dbg(obj, fmt, args...) \
110 drbd_printk(KERN_DEBUG, obj, fmt, ## args)
111 #define drbd_alert(obj, fmt, args...) \
112 drbd_printk(KERN_ALERT, obj, fmt, ## args)
113 #define drbd_err(obj, fmt, args...) \
114 drbd_printk(KERN_ERR, obj, fmt, ## args)
115 #define drbd_warn(obj, fmt, args...) \
116 drbd_printk(KERN_WARNING, obj, fmt, ## args)
117 #define drbd_info(obj, fmt, args...) \
118 drbd_printk(KERN_INFO, obj, fmt, ## args)
119 #define drbd_emerg(obj, fmt, args...) \
120 drbd_printk(KERN_EMERG, obj, fmt, ## args)
122 #define dynamic_drbd_dbg(device, fmt, args...) \
123 dynamic_dev_dbg(disk_to_dev(device->vdisk), fmt, ## args)
125 #define D_ASSERT(device, exp) do { \
127 drbd_err(device, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__); \
131 * expect - Make an assertion
133 * Unlike the assert macro, this macro returns a boolean result.
135 #define expect(exp) ({ \
136 bool _bool = (exp); \
138 drbd_err(device, "ASSERTION %s FAILED in %s\n", \
143 /* Defines to control fault insertion */
145 DRBD_FAULT_MD_WR = 0, /* meta data write */
146 DRBD_FAULT_MD_RD = 1, /* read */
147 DRBD_FAULT_RS_WR = 2, /* resync */
148 DRBD_FAULT_RS_RD = 3,
149 DRBD_FAULT_DT_WR = 4, /* data */
150 DRBD_FAULT_DT_RD = 5,
151 DRBD_FAULT_DT_RA = 6, /* data read ahead */
152 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */
153 DRBD_FAULT_AL_EE = 8, /* alloc ee */
154 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
160 _drbd_insert_fault(struct drbd_device *device, unsigned int type);
163 drbd_insert_fault(struct drbd_device *device, unsigned int type) {
164 #ifdef CONFIG_DRBD_FAULT_INJECTION
165 return drbd_fault_rate &&
166 (drbd_enable_faults & (1<<type)) &&
167 _drbd_insert_fault(device, type);
173 /* integer division, round _UP_ to the next integer */
174 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
175 /* usual integer division */
176 #define div_floor(A, B) ((A)/(B))
178 extern struct ratelimit_state drbd_ratelimit_state;
179 extern struct idr drbd_devices; /* RCU, updates: genl_lock() */
180 extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */
182 extern const char *cmdname(enum drbd_packet cmd);
184 /* for sending/receiving the bitmap,
185 * possibly in some encoding scheme */
188 * stores total bits and long words
189 * of the bitmap, so we don't need to
190 * call the accessor functions over and again. */
191 unsigned long bm_bits;
192 unsigned long bm_words;
193 /* during xfer, current position within the bitmap */
194 unsigned long bit_offset;
195 unsigned long word_offset;
197 /* statistics; index: (h->command == P_BITMAP) */
202 extern void INFO_bm_xfer_stats(struct drbd_device *device,
203 const char *direction, struct bm_xfer_ctx *c);
205 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
207 /* word_offset counts "native long words" (32 or 64 bit),
209 * Encoded packet may end at an unaligned bit offset.
210 * In case a fallback clear text packet is transmitted in
211 * between, we adjust this offset back to the last 64bit
212 * aligned "native long word", which makes coding and decoding
213 * the plain text bitmap much more convenient. */
214 #if BITS_PER_LONG == 64
215 c->word_offset = c->bit_offset >> 6;
216 #elif BITS_PER_LONG == 32
217 c->word_offset = c->bit_offset >> 5;
218 c->word_offset &= ~(1UL);
220 # error "unsupported BITS_PER_LONG"
224 extern unsigned int drbd_header_size(struct drbd_connection *connection);
226 /**********************************************************************/
227 enum drbd_thread_state {
236 struct task_struct *task;
237 struct completion stop;
238 enum drbd_thread_state t_state;
239 int (*function) (struct drbd_thread *);
240 struct drbd_resource *resource;
241 struct drbd_connection *connection;
246 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
248 /* THINK testing the t_state seems to be uncritical in all cases
249 * (but thread_{start,stop}), so we can read it *without* the lock.
257 struct list_head list;
258 int (*cb)(struct drbd_work *, int cancel);
261 struct drbd_device_work {
263 struct drbd_device *device;
266 #include "drbd_interval.h"
268 extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *);
270 extern void lock_all_resources(void);
271 extern void unlock_all_resources(void);
273 struct drbd_request {
275 struct drbd_device *device;
277 /* if local IO is not allowed, will be NULL.
278 * if local IO _is_ allowed, holds the locally submitted bio clone,
279 * or, after local IO completion, the ERR_PTR(error).
280 * see drbd_request_endio(). */
281 struct bio *private_bio;
283 struct drbd_interval i;
285 /* epoch: used to check on "completion" whether this req was in
286 * the current epoch, and we therefore have to close it,
287 * causing a p_barrier packet to be send, starting a new epoch.
289 * This corresponds to "barrier" in struct p_barrier[_ack],
290 * and to "barrier_nr" in struct drbd_epoch (and various
291 * comments/function parameters/local variable names).
295 struct list_head tl_requests; /* ring list in the transfer log */
296 struct bio *master_bio; /* master bio pointer */
298 /* see struct drbd_device */
299 struct list_head req_pending_master_completion;
300 struct list_head req_pending_local;
302 /* for generic IO accounting */
303 unsigned long start_jif;
305 /* for DRBD internal statistics */
307 /* Minimal set of time stamps to determine if we wait for activity log
308 * transactions, local disk or peer. 32 bit "jiffies" are good enough,
309 * we don't expect a DRBD request to be stalled for several month.
312 /* before actual request processing */
313 unsigned long in_actlog_jif;
316 unsigned long pre_submit_jif;
319 unsigned long pre_send_jif;
320 unsigned long acked_jif;
321 unsigned long net_done_jif;
323 /* Possibly even more detail to track each phase:
324 * master_completion_jif
325 * how long did it take to complete the master bio
326 * (application visible latency)
328 * how long the master bio was blocked until we finally allocated
331 * how long did we wait for activity log transactions
334 * when did we finally queue it for sending
336 * when did we start sending it
338 * how long did we block in the network stack trying to send it
340 * when did we receive (or fake, in protocol A) a remote ACK
342 * when did we receive final acknowledgement (P_BARRIER_ACK),
343 * or decide, e.g. on connection loss, that we do no longer expect
344 * anything from this peer for this request.
348 * when did we start submiting to the lower level device,
349 * and how long did we block in that submit function
350 * local_completion_jif
351 * how long did it take the lower level device to complete this request
355 /* once it hits 0, we may complete the master_bio */
356 atomic_t completion_ref;
357 /* once it hits 0, we may destroy this drbd_request object */
360 unsigned rq_state; /* see comments above _req_mod() */
364 struct drbd_connection *connection;
365 struct list_head list;
366 unsigned int barrier_nr;
367 atomic_t epoch_size; /* increased on every request added. */
368 atomic_t active; /* increased on every req. added, and dec on every finished. */
372 /* Prototype declaration of function defined in drbd_receiver.c */
373 int drbdd_init(struct drbd_thread *);
374 int drbd_asender(struct drbd_thread *);
376 /* drbd_epoch flag bits */
378 DE_HAVE_BARRIER_NUMBER,
385 EV_CLEANUP = 32, /* used as flag */
393 struct drbd_peer_request {
395 struct drbd_peer_device *peer_device;
396 struct drbd_epoch *epoch; /* for writes */
398 atomic_t pending_bios;
399 struct drbd_interval i;
400 /* see comments on ee flag bits below */
402 unsigned long submit_jif;
405 struct digest_info *digest;
410 * While corresponding bios are in flight, the only modification will be
411 * set_bit WAS_ERROR, which has to be atomic.
412 * If no bios are in flight yet, or all have been completed,
413 * non-atomic modification to ee->flags is ok.
416 __EE_CALL_AL_COMPLETE_IO,
417 __EE_MAY_SET_IN_SYNC,
419 /* is this a TRIM aka REQ_OP_DISCARD? */
421 /* explicit zero-out requested, or
422 * our lower level cannot handle trim,
423 * and we want to fall back to zeroout instead */
426 /* In case a barrier failed,
427 * we need to resubmit without the barrier flag. */
430 /* we may have several bios per peer request.
431 * if any of those fail, we set this flag atomically
432 * from the endio callback */
435 /* This ee has a pointer to a digest instead of a block id */
438 /* Conflicting local requests need to be restarted after this request */
439 __EE_RESTART_REQUESTS,
441 /* The peer wants a write ACK for this (wire proto C) */
444 /* Is set when net_conf had two_primaries set while creating this peer_req */
445 __EE_IN_INTERVAL_TREE,
448 /* has this been submitted, or does it still wait for something else? */
451 /* this is/was a write request */
454 /* this is/was a write same request */
457 /* this originates from application on peer
458 * (not some resync or verify or other DRBD internal request) */
461 /* If it contains only 0 bytes, send back P_RS_DEALLOCATED */
464 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
465 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
466 #define EE_TRIM (1<<__EE_TRIM)
467 #define EE_ZEROOUT (1<<__EE_ZEROOUT)
468 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
469 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
470 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
471 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
472 #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK)
473 #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE)
474 #define EE_SUBMITTED (1<<__EE_SUBMITTED)
475 #define EE_WRITE (1<<__EE_WRITE)
476 #define EE_WRITE_SAME (1<<__EE_WRITE_SAME)
477 #define EE_APPLICATION (1<<__EE_APPLICATION)
478 #define EE_RS_THIN_REQ (1<<__EE_RS_THIN_REQ)
480 /* flag bits per device */
482 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
483 MD_DIRTY, /* current uuids and flags not yet on disk */
484 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
487 CRASHED_PRIMARY, /* This node was a crashed primary.
488 * Gets cleared when the state.conn
489 * goes into C_CONNECTED state. */
492 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */
494 BITMAP_IO, /* suspend application io;
495 once no more io in flight, start bitmap io */
496 BITMAP_IO_QUEUED, /* Started bitmap IO */
497 WAS_IO_ERROR, /* Local disk failed, returned IO error */
498 WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */
499 FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */
500 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
501 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
502 * the peer, if it changed there as well. */
503 NEW_CUR_UUID, /* Create new current UUID when thawing IO */
504 AL_SUSPENDED, /* Activity logging is currently suspended. */
505 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */
506 B_RS_H_DONE, /* Before resync handler done (already executed) */
507 DISCARD_MY_DATA, /* discard_my_data flag per volume */
510 FLUSH_PENDING, /* if set, device->flush_jif is when we submitted that flush
511 * from drbd_flush_after_epoch() */
513 /* cleared only after backing device related structures have been destroyed. */
514 GOING_DISKLESS, /* Disk is being detached, because of io-error, or admin request. */
516 /* to be used in drbd_device_post_work() */
517 GO_DISKLESS, /* tell worker to schedule cleanup before detach */
518 DESTROY_DISK, /* tell worker to close backing devices and destroy related structures. */
519 MD_SYNC, /* tell worker to call drbd_md_sync() */
520 RS_START, /* tell worker to start resync/OV */
521 RS_PROGRESS, /* tell worker that resync made significant progress */
522 RS_DONE, /* tell worker that resync is done */
525 struct drbd_bitmap; /* opaque for drbd_device */
527 /* definition of bits in bm_flags to be used in drbd_bm_lock
528 * and drbd_bitmap_io and friends. */
530 /* currently locked for bulk operation */
531 BM_LOCKED_MASK = 0xf,
533 /* in detail, that is: */
538 /* so we can mark it locked for bulk operation,
539 * and still allow all non-bulk operations */
542 /* (test bit, count bit) allowed (common case) */
543 BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED,
545 /* testing bits, as well as setting new bits allowed, but clearing bits
546 * would be unexpected. Used during bitmap receive. Setting new bits
547 * requires sending of "out-of-sync" information, though. */
548 BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED,
550 /* for drbd_bm_write_copy_pages, everything is allowed,
551 * only concurrent bulk operations are locked out. */
552 BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED,
555 struct drbd_work_queue {
557 spinlock_t q_lock; /* to protect the list. */
558 wait_queue_head_t q_wait;
563 struct socket *socket;
564 /* this way we get our
565 * send/receive buffers off the stack */
571 u64 md_offset; /* sector offset to 'super' block */
573 u64 la_size_sect; /* last agreed size, unit sectors */
574 spinlock_t uuid_lock;
580 s32 al_offset; /* signed relative sector offset to activity log */
581 s32 bm_offset; /* signed relative sector offset to bitmap */
583 /* cached value of bdev->disk_conf->meta_dev_idx (see below) */
586 /* see al_tr_number_to_on_disk_sector() */
588 u32 al_stripe_size_4k;
589 u32 al_size_4k; /* cached product of the above */
592 struct drbd_backing_dev {
593 struct block_device *backing_bdev;
594 struct block_device *md_bdev;
596 struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */
597 sector_t known_size; /* last known size of that backing device */
602 unsigned long start_jif; /* last call to drbd_md_get_buffer */
603 unsigned long submit_jif; /* last _drbd_md_sync_page_io() submit */
604 const char *current_use;
614 int (*io_fn)(struct drbd_device *device);
615 void (*done)(struct drbd_device *device, int rv);
619 unsigned int head_index;
621 int total; /* sum of all values */
624 extern struct fifo_buffer *fifo_alloc(unsigned int fifo_size);
626 /* flag bits per connection */
628 NET_CONGESTED, /* The data socket is congested */
629 RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */
631 GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */
632 CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */
635 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
636 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */
637 STATE_SENT, /* Do not change state/UUIDs while this is set */
638 CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
639 * pending, from drbd worker context.
643 DEVICE_WORK_PENDING, /* tell worker that some device has pending work */
646 enum which_state { NOW, OLD = NOW, NEW };
648 struct drbd_resource {
650 #ifdef CONFIG_DEBUG_FS
651 struct dentry *debugfs_res;
652 struct dentry *debugfs_res_volumes;
653 struct dentry *debugfs_res_connections;
654 struct dentry *debugfs_res_in_flight_summary;
657 struct idr devices; /* volume number to device mapping */
658 struct list_head connections;
659 struct list_head resources;
660 struct res_opts res_opts;
661 struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */
662 struct mutex adm_mutex; /* mutex to serialize administrative requests */
665 unsigned susp:1; /* IO suspended by user */
666 unsigned susp_nod:1; /* IO suspended because no data */
667 unsigned susp_fen:1; /* IO suspended because fence peer handler runs */
669 enum write_ordering_e write_ordering;
671 cpumask_var_t cpu_mask;
674 struct drbd_thread_timing_details
676 unsigned long start_jif;
678 const char *caller_fn;
683 struct drbd_connection {
684 struct list_head connections;
685 struct drbd_resource *resource;
686 #ifdef CONFIG_DEBUG_FS
687 struct dentry *debugfs_conn;
688 struct dentry *debugfs_conn_callback_history;
689 struct dentry *debugfs_conn_oldest_requests;
692 struct idr peer_devices; /* volume number to peer device mapping */
693 enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
694 struct mutex cstate_mutex; /* Protects graceful disconnects */
695 unsigned int connect_cnt; /* Inc each time a connection is established */
698 struct net_conf *net_conf; /* content protected by rcu */
699 wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */
701 struct sockaddr_storage my_addr;
703 struct sockaddr_storage peer_addr;
706 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
707 struct drbd_socket meta; /* ping/ack (metadata) packets */
708 int agreed_pro_version; /* actually used protocol version */
710 unsigned long last_received; /* in jiffies, either socket */
711 unsigned int ko_count;
713 struct list_head transfer_log; /* all requests not yet fully processed */
715 struct crypto_shash *cram_hmac_tfm;
716 struct crypto_shash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */
717 struct crypto_shash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */
718 struct crypto_shash *csums_tfm;
719 struct crypto_shash *verify_tfm;
724 struct drbd_epoch *current_epoch;
725 spinlock_t epoch_lock;
727 atomic_t current_tle_nr; /* transfer log epoch number */
728 unsigned current_tle_writes; /* writes seen within this tl epoch */
730 unsigned long last_reconnect_jif;
731 /* empty member on older kernels without blk_start_plug() */
732 struct blk_plug receiver_plug;
733 struct drbd_thread receiver;
734 struct drbd_thread worker;
735 struct drbd_thread ack_receiver;
736 struct workqueue_struct *ack_sender;
739 * so we can look up the oldest pending requests more quickly.
740 * protected by resource->req_lock */
741 struct drbd_request *req_next; /* DRBD 9: todo.req_next */
742 struct drbd_request *req_ack_pending;
743 struct drbd_request *req_not_net_done;
746 struct drbd_work_queue sender_work;
748 #define DRBD_THREAD_DETAILS_HIST 16
749 unsigned int w_cb_nr; /* keeps counting up */
750 unsigned int r_cb_nr; /* keeps counting up */
751 struct drbd_thread_timing_details w_timing_details[DRBD_THREAD_DETAILS_HIST];
752 struct drbd_thread_timing_details r_timing_details[DRBD_THREAD_DETAILS_HIST];
755 unsigned long last_sent_barrier_jif;
757 /* whether this sender thread
758 * has processed a single write yet. */
759 bool seen_any_write_yet;
761 /* Which barrier number to send with the next P_BARRIER */
762 int current_epoch_nr;
764 /* how many write requests have been sent
765 * with req->epoch == current_epoch_nr.
766 * If none, no P_BARRIER will be sent. */
767 unsigned current_epoch_writes;
771 static inline bool has_net_conf(struct drbd_connection *connection)
776 has_net_conf = rcu_dereference(connection->net_conf);
782 void __update_timing_details(
783 struct drbd_thread_timing_details *tdp,
786 const char *fn, const unsigned int line);
788 #define update_worker_timing_details(c, cb) \
789 __update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ )
790 #define update_receiver_timing_details(c, cb) \
791 __update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ )
793 struct submit_worker {
794 struct workqueue_struct *wq;
795 struct work_struct worker;
797 /* protected by ..->resource->req_lock */
798 struct list_head writes;
801 struct drbd_peer_device {
802 struct list_head peer_devices;
803 struct drbd_device *device;
804 struct drbd_connection *connection;
805 struct work_struct send_acks_work;
806 #ifdef CONFIG_DEBUG_FS
807 struct dentry *debugfs_peer_dev;
812 struct drbd_resource *resource;
813 struct list_head peer_devices;
814 struct list_head pending_bitmap_io;
816 unsigned long flush_jif;
817 #ifdef CONFIG_DEBUG_FS
818 struct dentry *debugfs_minor;
819 struct dentry *debugfs_vol;
820 struct dentry *debugfs_vol_oldest_requests;
821 struct dentry *debugfs_vol_act_log_extents;
822 struct dentry *debugfs_vol_resync_extents;
823 struct dentry *debugfs_vol_data_gen_id;
824 struct dentry *debugfs_vol_ed_gen_id;
827 unsigned int vnr; /* volume number within the connection */
828 unsigned int minor; /* device minor number */
832 /* things that are stored as / read from meta data on disk */
835 /* configured by drbdsetup */
836 struct drbd_backing_dev *ldev __protected_by(local);
838 sector_t p_size; /* partner's disk size */
839 struct request_queue *rq_queue;
840 struct gendisk *vdisk;
842 unsigned long last_reattach_jif;
843 struct drbd_work resync_work;
844 struct drbd_work unplug_work;
845 struct timer_list resync_timer;
846 struct timer_list md_sync_timer;
847 struct timer_list start_resync_timer;
848 struct timer_list request_timer;
850 /* Used after attach while negotiating new disk state. */
851 union drbd_state new_state_tmp;
853 union drbd_dev_state state;
854 wait_queue_head_t misc_wait;
855 wait_queue_head_t state_wait; /* upon each state change. */
856 unsigned int send_cnt;
857 unsigned int recv_cnt;
858 unsigned int read_cnt;
859 unsigned int writ_cnt;
860 unsigned int al_writ_cnt;
861 unsigned int bm_writ_cnt;
862 atomic_t ap_bio_cnt; /* Requests we need to complete */
863 atomic_t ap_actlog_cnt; /* Requests waiting for activity log */
864 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
865 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
866 atomic_t unacked_cnt; /* Need to send replies for */
867 atomic_t local_cnt; /* Waiting for local completion */
868 atomic_t suspend_cnt;
870 /* Interval tree of pending local requests */
871 struct rb_root read_requests;
872 struct rb_root write_requests;
874 /* for statistics and timeouts */
875 /* [0] read, [1] write */
876 struct list_head pending_master_completion[2];
877 struct list_head pending_completion[2];
879 /* use checksums for *this* resync */
881 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
882 unsigned long rs_total;
883 /* number of resync blocks that failed in this run */
884 unsigned long rs_failed;
885 /* Syncer's start time [unit jiffies] */
886 unsigned long rs_start;
887 /* cumulated time in PausedSyncX state [unit jiffies] */
888 unsigned long rs_paused;
889 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
890 unsigned long rs_same_csum;
891 #define DRBD_SYNC_MARKS 8
892 #define DRBD_SYNC_MARK_STEP (3*HZ)
893 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
894 unsigned long rs_mark_left[DRBD_SYNC_MARKS];
895 /* marks's time [unit jiffies] */
896 unsigned long rs_mark_time[DRBD_SYNC_MARKS];
897 /* current index into rs_mark_{left,time} */
899 unsigned long rs_last_bcast; /* [unit jiffies] */
901 /* where does the admin want us to start? (sector) */
902 sector_t ov_start_sector;
903 sector_t ov_stop_sector;
904 /* where are we now? (sector) */
905 sector_t ov_position;
906 /* Start sector of out of sync range (to merge printk reporting). */
907 sector_t ov_last_oos_start;
908 /* size of out-of-sync range in sectors. */
909 sector_t ov_last_oos_size;
910 unsigned long ov_left; /* in bits */
912 struct drbd_bitmap *bitmap;
913 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
915 /* Used to track operations of resync... */
916 struct lru_cache *resync;
917 /* Number of locked elements in resync LRU */
918 unsigned int resync_locked;
919 /* resync extent number waiting for application requests */
920 unsigned int resync_wenr;
925 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
926 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
927 struct list_head done_ee; /* need to send P_WRITE_ACK */
928 struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */
929 struct list_head net_ee; /* zero-copy network send in progress */
932 struct list_head resync_reads;
933 atomic_t pp_in_use; /* allocated from page pool */
934 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */
935 wait_queue_head_t ee_wait;
936 struct drbd_md_io md_io;
938 wait_queue_head_t al_wait;
939 struct lru_cache *act_log; /* activity log */
940 unsigned int al_tr_number;
942 wait_queue_head_t seq_wait;
944 unsigned int peer_seq;
945 spinlock_t peer_seq_lock;
946 unsigned long comm_bm_set; /* communicated number of set bits. */
947 struct bm_io_work bm_io_work;
948 u64 ed_uuid; /* UUID of the exposed data */
949 struct mutex own_state_mutex;
950 struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */
951 char congestion_reason; /* Why we where congested... */
952 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
953 atomic_t rs_sect_ev; /* for submitted resync data rate, both */
954 int rs_last_sect_ev; /* counter to compare with */
955 int rs_last_events; /* counter of read or write "events" (unit sectors)
956 * on the lower level device when we last looked. */
957 int c_sync_rate; /* current resync rate after syncer throttle magic */
958 struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */
959 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
960 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
961 unsigned int peer_max_bio_size;
962 unsigned int local_max_bio_size;
964 /* any requests that would block in drbd_make_request()
965 * are deferred to this single-threaded work queue */
966 struct submit_worker submit;
969 struct drbd_bm_aio_ctx {
970 struct drbd_device *device;
971 struct list_head list; /* on device->pending_bitmap_io */;
972 unsigned long start_jif;
976 #define BM_AIO_COPY_PAGES 1
977 #define BM_AIO_WRITE_HINTED 2
978 #define BM_AIO_WRITE_ALL_PAGES 4
979 #define BM_AIO_READ 8
984 struct drbd_config_context {
985 /* assigned from drbd_genlmsghdr */
987 /* assigned from request attributes, if present */
989 #define VOLUME_UNSPECIFIED (-1U)
990 /* pointer into the request skb,
991 * limited lifetime! */
993 struct nlattr *my_addr;
994 struct nlattr *peer_addr;
997 struct sk_buff *reply_skb;
998 /* pointer into reply buffer */
999 struct drbd_genlmsghdr *reply_dh;
1000 /* resolved from attributes, if possible */
1001 struct drbd_device *device;
1002 struct drbd_resource *resource;
1003 struct drbd_connection *connection;
1006 static inline struct drbd_device *minor_to_device(unsigned int minor)
1008 return (struct drbd_device *)idr_find(&drbd_devices, minor);
1011 static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device)
1013 return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices);
1016 static inline struct drbd_peer_device *
1017 conn_peer_device(struct drbd_connection *connection, int volume_number)
1019 return idr_find(&connection->peer_devices, volume_number);
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 int drbd_send_rs_deallocated(struct drbd_peer_device *, struct drbd_peer_request *);
1116 extern void drbd_backing_dev_free(struct drbd_device *device, struct drbd_backing_dev *ldev);
1117 extern void drbd_device_cleanup(struct drbd_device *device);
1118 extern void drbd_print_uuids(struct drbd_device *device, const char *text);
1119 extern void drbd_queue_unplug(struct drbd_device *device);
1121 extern void conn_md_sync(struct drbd_connection *connection);
1122 extern void drbd_md_write(struct drbd_device *device, void *buffer);
1123 extern void drbd_md_sync(struct drbd_device *device);
1124 extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev);
1125 extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1126 extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1127 extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local);
1128 extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local);
1129 extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local);
1130 extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1131 extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local);
1132 extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local);
1133 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1134 extern void drbd_md_mark_dirty(struct drbd_device *device);
1135 extern void drbd_queue_bitmap_io(struct drbd_device *device,
1136 int (*io_fn)(struct drbd_device *),
1137 void (*done)(struct drbd_device *, int),
1138 char *why, enum bm_flag flags);
1139 extern int drbd_bitmap_io(struct drbd_device *device,
1140 int (*io_fn)(struct drbd_device *),
1141 char *why, enum bm_flag flags);
1142 extern int drbd_bitmap_io_from_worker(struct drbd_device *device,
1143 int (*io_fn)(struct drbd_device *),
1144 char *why, enum bm_flag flags);
1145 extern int drbd_bmio_set_n_write(struct drbd_device *device) __must_hold(local);
1146 extern int drbd_bmio_clear_n_write(struct drbd_device *device) __must_hold(local);
1150 * We currently have two possible layouts.
1151 * Offsets in (512 byte) sectors.
1153 * |----------- md_size_sect ------------------|
1154 * [ 4k superblock ][ activity log ][ Bitmap ]
1155 * | al_offset == 8 |
1156 * | bm_offset = al_offset + X |
1157 * ==> bitmap sectors = md_size_sect - bm_offset
1160 * old, indexed fixed size meta data:
1163 * |----------- md_size_sect ------------------|
1164 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*]
1166 * | bm_offset = al_offset - Y |
1167 * ==> bitmap sectors = Y = al_offset - bm_offset
1169 * [padding*] are zero or up to 7 unused 512 Byte sectors to the
1170 * end of the device, so that the [4k superblock] will be 4k aligned.
1172 * The activity log consists of 4k transaction blocks,
1173 * which are written in a ring-buffer, or striped ring-buffer like fashion,
1174 * which are writtensize used to be fixed 32kB,
1175 * but is about to become configurable.
1178 /* Our old fixed size meta data layout
1179 * allows up to about 3.8TB, so if you want more,
1180 * you need to use the "flexible" meta data format. */
1181 #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */
1182 #define MD_4kB_SECT 8
1183 #define MD_32kB_SECT 64
1185 /* One activity log extent represents 4M of storage */
1186 #define AL_EXTENT_SHIFT 22
1187 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1189 /* We could make these currently hardcoded constants configurable
1190 * variables at create-md time (or even re-configurable at runtime?).
1191 * Which will require some more changes to the DRBD "super block"
1194 * updates per transaction:
1195 * This many changes to the active set can be logged with one transaction.
1196 * This number is arbitrary.
1197 * context per transaction:
1198 * This many context extent numbers are logged with each transaction.
1199 * This number is resulting from the transaction block size (4k), the layout
1200 * of the transaction header, and the number of updates per transaction.
1201 * See drbd_actlog.c:struct al_transaction_on_disk
1203 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1204 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1206 #if BITS_PER_LONG == 32
1208 #define cpu_to_lel(A) cpu_to_le32(A)
1209 #define lel_to_cpu(A) le32_to_cpu(A)
1210 #elif BITS_PER_LONG == 64
1212 #define cpu_to_lel(A) cpu_to_le64(A)
1213 #define lel_to_cpu(A) le64_to_cpu(A)
1215 #error "LN2 of BITS_PER_LONG unknown!"
1219 /* 16MB sized 'bitmap extent' to track syncer usage */
1221 int rs_left; /* number of bits set (out of sync) in this extent. */
1222 int rs_failed; /* number of failed resync requests in this extent. */
1223 unsigned long flags;
1224 struct lc_element lce;
1227 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1228 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1229 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1233 * We need to store one bit for a block.
1234 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1235 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1236 * Bit 1 ==> local node thinks this block needs to be synced.
1239 #define SLEEP_TIME (HZ/10)
1241 /* We do bitmap IO in units of 4k blocks.
1242 * We also still have a hardcoded 4k per bit relation. */
1243 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1244 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1245 /* mostly arbitrarily set the represented size of one bitmap extent,
1246 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1247 * at 4k per bit resolution) */
1248 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
1249 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1251 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1252 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1255 /* thus many _storage_ sectors are described by one bit */
1256 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1257 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1258 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1260 /* bit to represented kilo byte conversion */
1261 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1263 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1264 * _storage_ sector is located in */
1265 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1266 #define BM_BIT_TO_EXT(x) ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1268 /* first storage sector a bitmap extent corresponds to */
1269 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1270 /* how much _storage_ sectors we have per bitmap extent */
1271 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1272 /* how many bits are covered by one bitmap extent (resync extent) */
1273 #define BM_BITS_PER_EXT (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1275 #define BM_BLOCKS_PER_BM_EXT_MASK (BM_BITS_PER_EXT - 1)
1278 /* in one sector of the bitmap, we have this many activity_log extents. */
1279 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1281 /* the extent in "PER_EXTENT" below is an activity log extent
1282 * we need that many (long words/bytes) to store the bitmap
1283 * of one AL_EXTENT_SIZE chunk of storage.
1284 * we can store the bitmap for that many AL_EXTENTS within
1285 * one sector of the _on_disk_ bitmap:
1286 * bit 0 bit 37 bit 38 bit (512*8)-1
1287 * ...|........|........|.. // ..|........|
1288 * sect. 0 `296 `304 ^(512*8*8)-1
1290 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1291 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1292 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1295 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1296 /* we have a certain meta data variant that has a fixed on-disk size of 128
1297 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity
1298 * log, leaving this many sectors for the bitmap.
1301 #define DRBD_MAX_SECTORS_FIXED_BM \
1302 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9)))
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. */
1316 /* Estimate max bio size as 256 * PAGE_SIZE,
1317 * so for typical PAGE_SIZE of 4k, that is (1<<20) Byte.
1318 * Since we may live in a mixed-platform cluster,
1319 * we limit us to a platform agnostic constant here for now.
1320 * A followup commit may allow even bigger BIO sizes,
1321 * once we thought that through. */
1322 #define DRBD_MAX_BIO_SIZE (1U << 20)
1323 #if DRBD_MAX_BIO_SIZE > (BIO_MAX_VECS << PAGE_SHIFT)
1324 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1326 #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */
1328 #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */
1329 #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */
1331 /* For now, don't allow more than half of what we can "activate" in one
1332 * activity log transaction to be discarded in one go. We may need to rework
1333 * drbd_al_begin_io() to allow for even larger discard ranges */
1334 #define DRBD_MAX_BATCH_BIO_SIZE (AL_UPDATES_PER_TRANSACTION/2*AL_EXTENT_SIZE)
1335 #define DRBD_MAX_BBIO_SECTORS (DRBD_MAX_BATCH_BIO_SIZE >> 9)
1337 extern int drbd_bm_init(struct drbd_device *device);
1338 extern int drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits);
1339 extern void drbd_bm_cleanup(struct drbd_device *device);
1340 extern void drbd_bm_set_all(struct drbd_device *device);
1341 extern void drbd_bm_clear_all(struct drbd_device *device);
1342 /* set/clear/test only a few bits at a time */
1343 extern int drbd_bm_set_bits(
1344 struct drbd_device *device, unsigned long s, unsigned long e);
1345 extern int drbd_bm_clear_bits(
1346 struct drbd_device *device, unsigned long s, unsigned long e);
1347 extern int drbd_bm_count_bits(
1348 struct drbd_device *device, const unsigned long s, const unsigned long e);
1349 /* bm_set_bits variant for use while holding drbd_bm_lock,
1350 * may process the whole bitmap in one go */
1351 extern void _drbd_bm_set_bits(struct drbd_device *device,
1352 const unsigned long s, const unsigned long e);
1353 extern int drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr);
1354 extern int drbd_bm_e_weight(struct drbd_device *device, unsigned long enr);
1355 extern int drbd_bm_read(struct drbd_device *device) __must_hold(local);
1356 extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr);
1357 extern int drbd_bm_write(struct drbd_device *device) __must_hold(local);
1358 extern void drbd_bm_reset_al_hints(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;
1422 /* And a bio_set for cloning */
1423 extern struct bio_set drbd_io_bio_set;
1425 extern struct mutex resources_mutex;
1427 extern int conn_lowest_minor(struct drbd_connection *connection);
1428 extern enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor);
1429 extern void drbd_destroy_device(struct kref *kref);
1430 extern void drbd_delete_device(struct drbd_device *device);
1432 extern struct drbd_resource *drbd_create_resource(const char *name);
1433 extern void drbd_free_resource(struct drbd_resource *resource);
1435 extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts);
1436 extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts);
1437 extern void drbd_destroy_connection(struct kref *kref);
1438 extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len,
1439 void *peer_addr, int peer_addr_len);
1440 extern struct drbd_resource *drbd_find_resource(const char *name);
1441 extern void drbd_destroy_resource(struct kref *kref);
1442 extern void conn_free_crypto(struct drbd_connection *connection);
1445 extern void do_submit(struct work_struct *ws);
1446 extern void __drbd_make_request(struct drbd_device *, struct bio *);
1447 void drbd_submit_bio(struct bio *bio);
1448 extern int drbd_read_remote(struct drbd_device *device, struct drbd_request *req);
1449 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1454 extern struct mutex notification_mutex;
1456 extern void drbd_suspend_io(struct drbd_device *device);
1457 extern void drbd_resume_io(struct drbd_device *device);
1458 extern char *ppsize(char *buf, unsigned long long size);
1459 extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int);
1460 enum determine_dev_size {
1461 DS_ERROR_SHRINK = -3,
1462 DS_ERROR_SPACE_MD = -2,
1467 DS_GREW_FROM_ZERO = 3,
1469 extern enum determine_dev_size
1470 drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local);
1471 extern void resync_after_online_grow(struct drbd_device *);
1472 extern void drbd_reconsider_queue_parameters(struct drbd_device *device,
1473 struct drbd_backing_dev *bdev, struct o_qlim *o);
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 enum drbd_peer_state conn_khelper(struct drbd_connection *connection, char *cmd);
1480 extern int drbd_khelper(struct drbd_device *device, char *cmd);
1483 /* bi_end_io handlers */
1484 extern void drbd_md_endio(struct bio *bio);
1485 extern void drbd_peer_request_endio(struct bio *bio);
1486 extern void drbd_request_endio(struct bio *bio);
1487 extern int drbd_worker(struct drbd_thread *thi);
1488 enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor);
1489 void drbd_resync_after_changed(struct drbd_device *device);
1490 extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side);
1491 extern void resume_next_sg(struct drbd_device *device);
1492 extern void suspend_other_sg(struct drbd_device *device);
1493 extern int drbd_resync_finished(struct drbd_device *device);
1494 /* maybe rather drbd_main.c ? */
1495 extern void *drbd_md_get_buffer(struct drbd_device *device, const char *intent);
1496 extern void drbd_md_put_buffer(struct drbd_device *device);
1497 extern int drbd_md_sync_page_io(struct drbd_device *device,
1498 struct drbd_backing_dev *bdev, sector_t sector, int op);
1499 extern void drbd_ov_out_of_sync_found(struct drbd_device *, sector_t, int);
1500 extern void wait_until_done_or_force_detached(struct drbd_device *device,
1501 struct drbd_backing_dev *bdev, unsigned int *done);
1502 extern void drbd_rs_controller_reset(struct drbd_device *device);
1504 static inline void ov_out_of_sync_print(struct drbd_device *device)
1506 if (device->ov_last_oos_size) {
1507 drbd_err(device, "Out of sync: start=%llu, size=%lu (sectors)\n",
1508 (unsigned long long)device->ov_last_oos_start,
1509 (unsigned long)device->ov_last_oos_size);
1511 device->ov_last_oos_size = 0;
1515 extern void drbd_csum_bio(struct crypto_shash *, struct bio *, void *);
1516 extern void drbd_csum_ee(struct crypto_shash *, struct drbd_peer_request *,
1518 /* worker callbacks */
1519 extern int w_e_end_data_req(struct drbd_work *, int);
1520 extern int w_e_end_rsdata_req(struct drbd_work *, int);
1521 extern int w_e_end_csum_rs_req(struct drbd_work *, int);
1522 extern int w_e_end_ov_reply(struct drbd_work *, int);
1523 extern int w_e_end_ov_req(struct drbd_work *, int);
1524 extern int w_ov_finished(struct drbd_work *, int);
1525 extern int w_resync_timer(struct drbd_work *, int);
1526 extern int w_send_write_hint(struct drbd_work *, int);
1527 extern int w_send_dblock(struct drbd_work *, int);
1528 extern int w_send_read_req(struct drbd_work *, int);
1529 extern int w_e_reissue(struct drbd_work *, int);
1530 extern int w_restart_disk_io(struct drbd_work *, int);
1531 extern int w_send_out_of_sync(struct drbd_work *, int);
1532 extern int w_start_resync(struct drbd_work *, int);
1534 extern void resync_timer_fn(struct timer_list *t);
1535 extern void start_resync_timer_fn(struct timer_list *t);
1537 extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req);
1539 /* drbd_receiver.c */
1540 extern int drbd_issue_discard_or_zero_out(struct drbd_device *device,
1541 sector_t start, unsigned int nr_sectors, int flags);
1542 extern int drbd_receiver(struct drbd_thread *thi);
1543 extern int drbd_ack_receiver(struct drbd_thread *thi);
1544 extern void drbd_send_ping_wf(struct work_struct *ws);
1545 extern void drbd_send_acks_wf(struct work_struct *ws);
1546 extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device);
1547 extern bool drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector,
1548 bool throttle_if_app_is_waiting);
1549 extern int drbd_submit_peer_request(struct drbd_device *,
1550 struct drbd_peer_request *, const unsigned,
1551 const unsigned, const int);
1552 extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *);
1553 extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64,
1554 sector_t, unsigned int,
1556 gfp_t) __must_hold(local);
1557 extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *,
1559 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1560 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
1561 extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool);
1562 extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
1563 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
1564 extern int drbd_connected(struct drbd_peer_device *);
1566 /* sets the number of 512 byte sectors of our virtual device */
1567 void drbd_set_my_capacity(struct drbd_device *device, sector_t size);
1570 * used to submit our private bio
1572 static inline void drbd_submit_bio_noacct(struct drbd_device *device,
1573 int fault_type, struct bio *bio)
1576 if (!bio->bi_bdev) {
1577 drbd_err(device, "drbd_submit_bio_noacct: bio->bi_bdev == NULL\n");
1578 bio->bi_status = BLK_STS_IOERR;
1583 if (drbd_insert_fault(device, fault_type))
1586 submit_bio_noacct(bio);
1589 void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev,
1590 enum write_ordering_e wo);
1593 extern struct proc_dir_entry *drbd_proc;
1594 int drbd_seq_show(struct seq_file *seq, void *v);
1597 extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i);
1598 extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i);
1599 extern void drbd_al_begin_io_commit(struct drbd_device *device);
1600 extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i);
1601 extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i);
1602 extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i);
1603 extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector);
1604 extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector);
1605 extern int drbd_try_rs_begin_io(struct drbd_device *device, sector_t sector);
1606 extern void drbd_rs_cancel_all(struct drbd_device *device);
1607 extern int drbd_rs_del_all(struct drbd_device *device);
1608 extern void drbd_rs_failed_io(struct drbd_device *device,
1609 sector_t sector, int size);
1610 extern void drbd_advance_rs_marks(struct drbd_device *device, unsigned long still_to_go);
1612 enum update_sync_bits_mode { RECORD_RS_FAILED, SET_OUT_OF_SYNC, SET_IN_SYNC };
1613 extern int __drbd_change_sync(struct drbd_device *device, sector_t sector, int size,
1614 enum update_sync_bits_mode mode);
1615 #define drbd_set_in_sync(device, sector, size) \
1616 __drbd_change_sync(device, sector, size, SET_IN_SYNC)
1617 #define drbd_set_out_of_sync(device, sector, size) \
1618 __drbd_change_sync(device, sector, size, SET_OUT_OF_SYNC)
1619 #define drbd_rs_failed_io(device, sector, size) \
1620 __drbd_change_sync(device, sector, size, RECORD_RS_FAILED)
1621 extern void drbd_al_shrink(struct drbd_device *device);
1622 extern int drbd_al_initialize(struct drbd_device *, void *);
1625 /* state info broadcast */
1627 enum drbd_state_info_bcast_reason sib_reason;
1631 unsigned helper_exit_code;
1634 union drbd_state os;
1635 union drbd_state ns;
1639 void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib);
1641 extern int notify_resource_state(struct sk_buff *,
1643 struct drbd_resource *,
1644 struct resource_info *,
1645 enum drbd_notification_type);
1646 extern int notify_device_state(struct sk_buff *,
1648 struct drbd_device *,
1649 struct device_info *,
1650 enum drbd_notification_type);
1651 extern int notify_connection_state(struct sk_buff *,
1653 struct drbd_connection *,
1654 struct connection_info *,
1655 enum drbd_notification_type);
1656 extern int notify_peer_device_state(struct sk_buff *,
1658 struct drbd_peer_device *,
1659 struct peer_device_info *,
1660 enum drbd_notification_type);
1661 extern void notify_helper(enum drbd_notification_type, struct drbd_device *,
1662 struct drbd_connection *, const char *, int);
1665 * inline helper functions
1666 *************************/
1668 /* see also page_chain_add and friends in drbd_receiver.c */
1669 static inline struct page *page_chain_next(struct page *page)
1671 return (struct page *)page_private(page);
1673 #define page_chain_for_each(page) \
1674 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1675 page = page_chain_next(page))
1676 #define page_chain_for_each_safe(page, n) \
1677 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1680 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req)
1682 struct page *page = peer_req->pages;
1683 page_chain_for_each(page) {
1684 if (page_count(page) > 1)
1690 static inline union drbd_state drbd_read_state(struct drbd_device *device)
1692 struct drbd_resource *resource = device->resource;
1693 union drbd_state rv;
1695 rv.i = device->state.i;
1696 rv.susp = resource->susp;
1697 rv.susp_nod = resource->susp_nod;
1698 rv.susp_fen = resource->susp_fen;
1703 enum drbd_force_detach_flags {
1710 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1711 static inline void __drbd_chk_io_error_(struct drbd_device *device,
1712 enum drbd_force_detach_flags df,
1715 enum drbd_io_error_p ep;
1718 ep = rcu_dereference(device->ldev->disk_conf)->on_io_error;
1721 case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */
1722 if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) {
1723 if (__ratelimit(&drbd_ratelimit_state))
1724 drbd_err(device, "Local IO failed in %s.\n", where);
1725 if (device->state.disk > D_INCONSISTENT)
1726 _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL);
1729 fallthrough; /* for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */
1731 case EP_CALL_HELPER:
1732 /* Remember whether we saw a READ or WRITE error.
1734 * Recovery of the affected area for WRITE failure is covered
1735 * by the activity log.
1736 * READ errors may fall outside that area though. Certain READ
1737 * errors can be "healed" by writing good data to the affected
1738 * blocks, which triggers block re-allocation in lower layers.
1740 * If we can not write the bitmap after a READ error,
1741 * we may need to trigger a full sync (see w_go_diskless()).
1743 * Force-detach is not really an IO error, but rather a
1744 * desperate measure to try to deal with a completely
1745 * unresponsive lower level IO stack.
1746 * Still it should be treated as a WRITE error.
1748 * Meta IO error is always WRITE error:
1749 * we read meta data only once during attach,
1750 * which will fail in case of errors.
1752 set_bit(WAS_IO_ERROR, &device->flags);
1753 if (df == DRBD_READ_ERROR)
1754 set_bit(WAS_READ_ERROR, &device->flags);
1755 if (df == DRBD_FORCE_DETACH)
1756 set_bit(FORCE_DETACH, &device->flags);
1757 if (device->state.disk > D_FAILED) {
1758 _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL);
1760 "Local IO failed in %s. Detaching...\n", where);
1767 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1768 * @device: DRBD device.
1769 * @error: Error code passed to the IO completion callback
1770 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1772 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1774 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1775 static inline void drbd_chk_io_error_(struct drbd_device *device,
1776 int error, enum drbd_force_detach_flags forcedetach, const char *where)
1779 unsigned long flags;
1780 spin_lock_irqsave(&device->resource->req_lock, flags);
1781 __drbd_chk_io_error_(device, forcedetach, where);
1782 spin_unlock_irqrestore(&device->resource->req_lock, flags);
1788 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1789 * @bdev: Meta data block device.
1791 * BTW, for internal meta data, this happens to be the maximum capacity
1792 * we could agree upon with our peer node.
1794 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1796 switch (bdev->md.meta_dev_idx) {
1797 case DRBD_MD_INDEX_INTERNAL:
1798 case DRBD_MD_INDEX_FLEX_INT:
1799 return bdev->md.md_offset + bdev->md.bm_offset;
1800 case DRBD_MD_INDEX_FLEX_EXT:
1802 return bdev->md.md_offset;
1807 * drbd_md_last_sector() - Return the last sector number of the meta data area
1808 * @bdev: Meta data block device.
1810 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1812 switch (bdev->md.meta_dev_idx) {
1813 case DRBD_MD_INDEX_INTERNAL:
1814 case DRBD_MD_INDEX_FLEX_INT:
1815 return bdev->md.md_offset + MD_4kB_SECT -1;
1816 case DRBD_MD_INDEX_FLEX_EXT:
1818 return bdev->md.md_offset + bdev->md.md_size_sect -1;
1822 /* Returns the number of 512 byte sectors of the device */
1823 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1825 return bdev ? bdev_nr_sectors(bdev) : 0;
1829 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1830 * @bdev: Meta data block device.
1832 * returns the capacity we announce to out peer. we clip ourselves at the
1833 * various MAX_SECTORS, because if we don't, current implementation will
1834 * oops sooner or later
1836 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1840 switch (bdev->md.meta_dev_idx) {
1841 case DRBD_MD_INDEX_INTERNAL:
1842 case DRBD_MD_INDEX_FLEX_INT:
1843 s = drbd_get_capacity(bdev->backing_bdev)
1844 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1845 drbd_md_first_sector(bdev))
1848 case DRBD_MD_INDEX_FLEX_EXT:
1849 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1850 drbd_get_capacity(bdev->backing_bdev));
1851 /* clip at maximum size the meta device can support */
1852 s = min_t(sector_t, s,
1853 BM_EXT_TO_SECT(bdev->md.md_size_sect
1854 - bdev->md.bm_offset));
1857 s = min_t(sector_t, DRBD_MAX_SECTORS,
1858 drbd_get_capacity(bdev->backing_bdev));
1864 * drbd_md_ss() - Return the sector number of our meta data super block
1865 * @bdev: Meta data block device.
1867 static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev)
1869 const int meta_dev_idx = bdev->md.meta_dev_idx;
1871 if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT)
1874 /* Since drbd08, internal meta data is always "flexible".
1875 * position: last 4k aligned block of 4k size */
1876 if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL ||
1877 meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)
1878 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8;
1880 /* external, some index; this is the old fixed size layout */
1881 return MD_128MB_SECT * bdev->md.meta_dev_idx;
1885 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1887 unsigned long flags;
1888 spin_lock_irqsave(&q->q_lock, flags);
1889 list_add_tail(&w->list, &q->q);
1890 spin_unlock_irqrestore(&q->q_lock, flags);
1891 wake_up(&q->q_wait);
1895 drbd_queue_work_if_unqueued(struct drbd_work_queue *q, struct drbd_work *w)
1897 unsigned long flags;
1898 spin_lock_irqsave(&q->q_lock, flags);
1899 if (list_empty_careful(&w->list))
1900 list_add_tail(&w->list, &q->q);
1901 spin_unlock_irqrestore(&q->q_lock, flags);
1902 wake_up(&q->q_wait);
1906 drbd_device_post_work(struct drbd_device *device, int work_bit)
1908 if (!test_and_set_bit(work_bit, &device->flags)) {
1909 struct drbd_connection *connection =
1910 first_peer_device(device)->connection;
1911 struct drbd_work_queue *q = &connection->sender_work;
1912 if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags))
1913 wake_up(&q->q_wait);
1917 extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue);
1919 /* To get the ack_receiver out of the blocking network stack,
1920 * so it can change its sk_rcvtimeo from idle- to ping-timeout,
1921 * and send a ping, we need to send a signal.
1922 * Which signal we send is irrelevant. */
1923 static inline void wake_ack_receiver(struct drbd_connection *connection)
1925 struct task_struct *task = connection->ack_receiver.task;
1926 if (task && get_t_state(&connection->ack_receiver) == RUNNING)
1927 send_sig(SIGXCPU, task, 1);
1930 static inline void request_ping(struct drbd_connection *connection)
1932 set_bit(SEND_PING, &connection->flags);
1933 wake_ack_receiver(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_target_state(enum drbd_conns connection_state)
2054 return connection_state == C_SYNC_TARGET ||
2055 connection_state == C_PAUSED_SYNC_T;
2058 static inline bool is_sync_source_state(enum drbd_conns connection_state)
2060 return connection_state == C_SYNC_SOURCE ||
2061 connection_state == C_PAUSED_SYNC_S;
2064 static inline bool is_sync_state(enum drbd_conns connection_state)
2066 return is_sync_source_state(connection_state) ||
2067 is_sync_target_state(connection_state);
2071 * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev
2072 * @_device: DRBD device.
2073 * @_min_state: Minimum device state required for success.
2075 * You have to call put_ldev() when finished working with device->ldev.
2077 #define get_ldev_if_state(_device, _min_state) \
2078 (_get_ldev_if_state((_device), (_min_state)) ? \
2079 ({ __acquire(x); true; }) : false)
2080 #define get_ldev(_device) get_ldev_if_state(_device, D_INCONSISTENT)
2082 static inline void put_ldev(struct drbd_device *device)
2084 enum drbd_disk_state disk_state = device->state.disk;
2085 /* We must check the state *before* the atomic_dec becomes visible,
2086 * or we have a theoretical race where someone hitting zero,
2087 * while state still D_FAILED, will then see D_DISKLESS in the
2088 * condition below and calling into destroy, where he must not, yet. */
2089 int i = atomic_dec_return(&device->local_cnt);
2091 /* This may be called from some endio handler,
2092 * so we must not sleep here. */
2095 D_ASSERT(device, i >= 0);
2097 if (disk_state == D_DISKLESS)
2098 /* even internal references gone, safe to destroy */
2099 drbd_device_post_work(device, DESTROY_DISK);
2100 if (disk_state == D_FAILED)
2101 /* all application IO references gone. */
2102 if (!test_and_set_bit(GOING_DISKLESS, &device->flags))
2103 drbd_device_post_work(device, GO_DISKLESS);
2104 wake_up(&device->misc_wait);
2109 static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins)
2113 /* never get a reference while D_DISKLESS */
2114 if (device->state.disk == D_DISKLESS)
2117 atomic_inc(&device->local_cnt);
2118 io_allowed = (device->state.disk >= mins);
2124 extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins);
2127 /* this throttles on-the-fly application requests
2128 * according to max_buffers settings;
2129 * maybe re-implement using semaphores? */
2130 static inline int drbd_get_max_buffers(struct drbd_device *device)
2132 struct net_conf *nc;
2136 nc = rcu_dereference(first_peer_device(device)->connection->net_conf);
2137 mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */
2143 static inline int drbd_state_is_stable(struct drbd_device *device)
2145 union drbd_dev_state s = device->state;
2147 /* DO NOT add a default clause, we want the compiler to warn us
2148 * for any newly introduced state we may have forgotten to add here */
2150 switch ((enum drbd_conns)s.conn) {
2151 /* new io only accepted when there is no connection, ... */
2153 case C_WF_CONNECTION:
2154 /* ... or there is a well established connection. */
2160 case C_PAUSED_SYNC_S:
2161 case C_PAUSED_SYNC_T:
2164 /* transitional states, IO allowed */
2165 case C_DISCONNECTING:
2169 case C_NETWORK_FAILURE:
2170 case C_PROTOCOL_ERROR:
2172 case C_WF_REPORT_PARAMS:
2173 case C_STARTING_SYNC_S:
2174 case C_STARTING_SYNC_T:
2177 /* Allow IO in BM exchange states with new protocols */
2179 if (first_peer_device(device)->connection->agreed_pro_version < 96)
2183 /* no new io accepted in these states */
2185 case C_WF_SYNC_UUID:
2191 switch ((enum drbd_disk_state)s.disk) {
2193 case D_INCONSISTENT:
2198 /* disk state is stable as well. */
2201 /* no new io accepted during transitional states */
2213 static inline int drbd_suspended(struct drbd_device *device)
2215 struct drbd_resource *resource = device->resource;
2217 return resource->susp || resource->susp_fen || resource->susp_nod;
2220 static inline bool may_inc_ap_bio(struct drbd_device *device)
2222 int mxb = drbd_get_max_buffers(device);
2224 if (drbd_suspended(device))
2226 if (atomic_read(&device->suspend_cnt))
2229 /* to avoid potential deadlock or bitmap corruption,
2230 * in various places, we only allow new application io
2231 * to start during "stable" states. */
2233 /* no new io accepted when attaching or detaching the disk */
2234 if (!drbd_state_is_stable(device))
2237 /* since some older kernels don't have atomic_add_unless,
2238 * and we are within the spinlock anyways, we have this workaround. */
2239 if (atomic_read(&device->ap_bio_cnt) > mxb)
2241 if (test_bit(BITMAP_IO, &device->flags))
2246 static inline bool inc_ap_bio_cond(struct drbd_device *device)
2250 spin_lock_irq(&device->resource->req_lock);
2251 rv = may_inc_ap_bio(device);
2253 atomic_inc(&device->ap_bio_cnt);
2254 spin_unlock_irq(&device->resource->req_lock);
2259 static inline void inc_ap_bio(struct drbd_device *device)
2262 * as long as the device is suspended
2263 * until the bitmap is no longer on the fly during connection
2264 * handshake as long as we would exceed the max_buffer limit.
2266 * to avoid races with the reconnect code,
2267 * we need to atomic_inc within the spinlock. */
2269 wait_event(device->misc_wait, inc_ap_bio_cond(device));
2272 static inline void dec_ap_bio(struct drbd_device *device)
2274 int mxb = drbd_get_max_buffers(device);
2275 int ap_bio = atomic_dec_return(&device->ap_bio_cnt);
2277 D_ASSERT(device, ap_bio >= 0);
2279 if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) {
2280 if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags))
2281 drbd_queue_work(&first_peer_device(device)->
2282 connection->sender_work,
2283 &device->bm_io_work.w);
2286 /* this currently does wake_up for every dec_ap_bio!
2287 * maybe rather introduce some type of hysteresis?
2288 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2290 wake_up(&device->misc_wait);
2293 static inline bool verify_can_do_stop_sector(struct drbd_device *device)
2295 return first_peer_device(device)->connection->agreed_pro_version >= 97 &&
2296 first_peer_device(device)->connection->agreed_pro_version != 100;
2299 static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val)
2301 int changed = device->ed_uuid != val;
2302 device->ed_uuid = val;
2306 static inline int drbd_queue_order_type(struct drbd_device *device)
2308 /* sorry, we currently have no working implementation
2309 * of distributed TCQ stuff */
2310 #ifndef QUEUE_ORDERED_NONE
2311 #define QUEUE_ORDERED_NONE 0
2313 return QUEUE_ORDERED_NONE;
2316 static inline struct drbd_connection *first_connection(struct drbd_resource *resource)
2318 return list_first_entry_or_null(&resource->connections,
2319 struct drbd_connection, connections);