xref: /linux/drivers/block/drbd/drbd_int.h (revision f2527d8f566a45fa00ee5abd04d1c9476d4d704f)
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3   drbd_int.h
4 
5   This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
6 
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>.
10 
11 
12 */
13 
14 #ifndef _DRBD_INT_H
15 #define _DRBD_INT_H
16 
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>
32 #include <net/tcp.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"
40 #include "drbd_polymorph_printk.h"
41 
42 #ifdef __CHECKER__
43 # define __protected_by(x)       __attribute__((require_context(x,1,999,"rdwr")))
44 # define __protected_read_by(x)  __attribute__((require_context(x,1,999,"read")))
45 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
46 #else
47 # define __protected_by(x)
48 # define __protected_read_by(x)
49 # define __protected_write_by(x)
50 #endif
51 
52 /* shared module parameters, defined in drbd_main.c */
53 #ifdef CONFIG_DRBD_FAULT_INJECTION
54 extern int drbd_enable_faults;
55 extern int drbd_fault_rate;
56 #endif
57 
58 extern unsigned int drbd_minor_count;
59 extern char drbd_usermode_helper[];
60 extern int drbd_proc_details;
61 
62 
63 /* This is used to stop/restart our threads.
64  * Cannot use SIGTERM nor SIGKILL, since these
65  * are sent out by init on runlevel changes
66  * I choose SIGHUP for now.
67  */
68 #define DRBD_SIGKILL SIGHUP
69 
70 #define ID_IN_SYNC      (4711ULL)
71 #define ID_OUT_OF_SYNC  (4712ULL)
72 #define ID_SYNCER (-1ULL)
73 
74 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
75 
76 struct drbd_device;
77 struct drbd_connection;
78 
79 /* Defines to control fault insertion */
80 enum {
81 	DRBD_FAULT_MD_WR = 0,	/* meta data write */
82 	DRBD_FAULT_MD_RD = 1,	/*           read  */
83 	DRBD_FAULT_RS_WR = 2,	/* resync          */
84 	DRBD_FAULT_RS_RD = 3,
85 	DRBD_FAULT_DT_WR = 4,	/* data            */
86 	DRBD_FAULT_DT_RD = 5,
87 	DRBD_FAULT_DT_RA = 6,	/* data read ahead */
88 	DRBD_FAULT_BM_ALLOC = 7,	/* bitmap allocation */
89 	DRBD_FAULT_AL_EE = 8,	/* alloc ee */
90 	DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
91 
92 	DRBD_FAULT_MAX,
93 };
94 
95 extern unsigned int
96 _drbd_insert_fault(struct drbd_device *device, unsigned int type);
97 
98 static inline int
99 drbd_insert_fault(struct drbd_device *device, unsigned int type) {
100 #ifdef CONFIG_DRBD_FAULT_INJECTION
101 	return drbd_fault_rate &&
102 		(drbd_enable_faults & (1<<type)) &&
103 		_drbd_insert_fault(device, type);
104 #else
105 	return 0;
106 #endif
107 }
108 
109 /* integer division, round _UP_ to the next integer */
110 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
111 /* usual integer division */
112 #define div_floor(A, B) ((A)/(B))
113 
114 extern struct ratelimit_state drbd_ratelimit_state;
115 extern struct idr drbd_devices; /* RCU, updates: genl_lock() */
116 extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */
117 
118 extern const char *cmdname(enum drbd_packet cmd);
119 
120 /* for sending/receiving the bitmap,
121  * possibly in some encoding scheme */
122 struct bm_xfer_ctx {
123 	/* "const"
124 	 * stores total bits and long words
125 	 * of the bitmap, so we don't need to
126 	 * call the accessor functions over and again. */
127 	unsigned long bm_bits;
128 	unsigned long bm_words;
129 	/* during xfer, current position within the bitmap */
130 	unsigned long bit_offset;
131 	unsigned long word_offset;
132 
133 	/* statistics; index: (h->command == P_BITMAP) */
134 	unsigned packets[2];
135 	unsigned bytes[2];
136 };
137 
138 extern void INFO_bm_xfer_stats(struct drbd_device *device,
139 		const char *direction, struct bm_xfer_ctx *c);
140 
141 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
142 {
143 	/* word_offset counts "native long words" (32 or 64 bit),
144 	 * aligned at 64 bit.
145 	 * Encoded packet may end at an unaligned bit offset.
146 	 * In case a fallback clear text packet is transmitted in
147 	 * between, we adjust this offset back to the last 64bit
148 	 * aligned "native long word", which makes coding and decoding
149 	 * the plain text bitmap much more convenient.  */
150 #if BITS_PER_LONG == 64
151 	c->word_offset = c->bit_offset >> 6;
152 #elif BITS_PER_LONG == 32
153 	c->word_offset = c->bit_offset >> 5;
154 	c->word_offset &= ~(1UL);
155 #else
156 # error "unsupported BITS_PER_LONG"
157 #endif
158 }
159 
160 extern unsigned int drbd_header_size(struct drbd_connection *connection);
161 
162 /**********************************************************************/
163 enum drbd_thread_state {
164 	NONE,
165 	RUNNING,
166 	EXITING,
167 	RESTARTING
168 };
169 
170 struct drbd_thread {
171 	spinlock_t t_lock;
172 	struct task_struct *task;
173 	struct completion stop;
174 	enum drbd_thread_state t_state;
175 	int (*function) (struct drbd_thread *);
176 	struct drbd_resource *resource;
177 	struct drbd_connection *connection;
178 	int reset_cpu_mask;
179 	const char *name;
180 };
181 
182 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
183 {
184 	/* THINK testing the t_state seems to be uncritical in all cases
185 	 * (but thread_{start,stop}), so we can read it *without* the lock.
186 	 *	--lge */
187 
188 	smp_rmb();
189 	return thi->t_state;
190 }
191 
192 struct drbd_work {
193 	struct list_head list;
194 	int (*cb)(struct drbd_work *, int cancel);
195 };
196 
197 struct drbd_device_work {
198 	struct drbd_work w;
199 	struct drbd_device *device;
200 };
201 
202 #include "drbd_interval.h"
203 
204 extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *);
205 
206 extern void lock_all_resources(void);
207 extern void unlock_all_resources(void);
208 
209 struct drbd_request {
210 	struct drbd_work w;
211 	struct drbd_device *device;
212 
213 	/* if local IO is not allowed, will be NULL.
214 	 * if local IO _is_ allowed, holds the locally submitted bio clone,
215 	 * or, after local IO completion, the ERR_PTR(error).
216 	 * see drbd_request_endio(). */
217 	struct bio *private_bio;
218 
219 	struct drbd_interval i;
220 
221 	/* epoch: used to check on "completion" whether this req was in
222 	 * the current epoch, and we therefore have to close it,
223 	 * causing a p_barrier packet to be send, starting a new epoch.
224 	 *
225 	 * This corresponds to "barrier" in struct p_barrier[_ack],
226 	 * and to "barrier_nr" in struct drbd_epoch (and various
227 	 * comments/function parameters/local variable names).
228 	 */
229 	unsigned int epoch;
230 
231 	struct list_head tl_requests; /* ring list in the transfer log */
232 	struct bio *master_bio;       /* master bio pointer */
233 
234 	/* see struct drbd_device */
235 	struct list_head req_pending_master_completion;
236 	struct list_head req_pending_local;
237 
238 	/* for generic IO accounting */
239 	unsigned long start_jif;
240 
241 	/* for DRBD internal statistics */
242 
243 	/* Minimal set of time stamps to determine if we wait for activity log
244 	 * transactions, local disk or peer.  32 bit "jiffies" are good enough,
245 	 * we don't expect a DRBD request to be stalled for several month.
246 	 */
247 
248 	/* before actual request processing */
249 	unsigned long in_actlog_jif;
250 
251 	/* local disk */
252 	unsigned long pre_submit_jif;
253 
254 	/* per connection */
255 	unsigned long pre_send_jif;
256 	unsigned long acked_jif;
257 	unsigned long net_done_jif;
258 
259 	/* Possibly even more detail to track each phase:
260 	 *  master_completion_jif
261 	 *      how long did it take to complete the master bio
262 	 *      (application visible latency)
263 	 *  allocated_jif
264 	 *      how long the master bio was blocked until we finally allocated
265 	 *      a tracking struct
266 	 *  in_actlog_jif
267 	 *      how long did we wait for activity log transactions
268 	 *
269 	 *  net_queued_jif
270 	 *      when did we finally queue it for sending
271 	 *  pre_send_jif
272 	 *      when did we start sending it
273 	 *  post_send_jif
274 	 *      how long did we block in the network stack trying to send it
275 	 *  acked_jif
276 	 *      when did we receive (or fake, in protocol A) a remote ACK
277 	 *  net_done_jif
278 	 *      when did we receive final acknowledgement (P_BARRIER_ACK),
279 	 *      or decide, e.g. on connection loss, that we do no longer expect
280 	 *      anything from this peer for this request.
281 	 *
282 	 *  pre_submit_jif
283 	 *  post_sub_jif
284 	 *      when did we start submiting to the lower level device,
285 	 *      and how long did we block in that submit function
286 	 *  local_completion_jif
287 	 *      how long did it take the lower level device to complete this request
288 	 */
289 
290 
291 	/* once it hits 0, we may complete the master_bio */
292 	atomic_t completion_ref;
293 	/* once it hits 0, we may destroy this drbd_request object */
294 	struct kref kref;
295 
296 	unsigned rq_state; /* see comments above _req_mod() */
297 };
298 
299 struct drbd_epoch {
300 	struct drbd_connection *connection;
301 	struct list_head list;
302 	unsigned int barrier_nr;
303 	atomic_t epoch_size; /* increased on every request added. */
304 	atomic_t active;     /* increased on every req. added, and dec on every finished. */
305 	unsigned long flags;
306 };
307 
308 /* Prototype declaration of function defined in drbd_receiver.c */
309 int drbdd_init(struct drbd_thread *);
310 int drbd_asender(struct drbd_thread *);
311 
312 /* drbd_epoch flag bits */
313 enum {
314 	DE_HAVE_BARRIER_NUMBER,
315 };
316 
317 enum epoch_event {
318 	EV_PUT,
319 	EV_GOT_BARRIER_NR,
320 	EV_BECAME_LAST,
321 	EV_CLEANUP = 32, /* used as flag */
322 };
323 
324 struct digest_info {
325 	int digest_size;
326 	void *digest;
327 };
328 
329 struct drbd_peer_request {
330 	struct drbd_work w;
331 	struct drbd_peer_device *peer_device;
332 	struct drbd_epoch *epoch; /* for writes */
333 	struct page *pages;
334 	blk_opf_t opf;
335 	atomic_t pending_bios;
336 	struct drbd_interval i;
337 	/* see comments on ee flag bits below */
338 	unsigned long flags;
339 	unsigned long submit_jif;
340 	union {
341 		u64 block_id;
342 		struct digest_info *digest;
343 	};
344 };
345 
346 /* Equivalent to bio_op and req_op. */
347 #define peer_req_op(peer_req) \
348 	((peer_req)->opf & REQ_OP_MASK)
349 
350 /* ee flag bits.
351  * While corresponding bios are in flight, the only modification will be
352  * set_bit WAS_ERROR, which has to be atomic.
353  * If no bios are in flight yet, or all have been completed,
354  * non-atomic modification to ee->flags is ok.
355  */
356 enum {
357 	__EE_CALL_AL_COMPLETE_IO,
358 	__EE_MAY_SET_IN_SYNC,
359 
360 	/* is this a TRIM aka REQ_OP_DISCARD? */
361 	__EE_TRIM,
362 	/* explicit zero-out requested, or
363 	 * our lower level cannot handle trim,
364 	 * and we want to fall back to zeroout instead */
365 	__EE_ZEROOUT,
366 
367 	/* In case a barrier failed,
368 	 * we need to resubmit without the barrier flag. */
369 	__EE_RESUBMITTED,
370 
371 	/* we may have several bios per peer request.
372 	 * if any of those fail, we set this flag atomically
373 	 * from the endio callback */
374 	__EE_WAS_ERROR,
375 
376 	/* This ee has a pointer to a digest instead of a block id */
377 	__EE_HAS_DIGEST,
378 
379 	/* Conflicting local requests need to be restarted after this request */
380 	__EE_RESTART_REQUESTS,
381 
382 	/* The peer wants a write ACK for this (wire proto C) */
383 	__EE_SEND_WRITE_ACK,
384 
385 	/* Is set when net_conf had two_primaries set while creating this peer_req */
386 	__EE_IN_INTERVAL_TREE,
387 
388 	/* for debugfs: */
389 	/* has this been submitted, or does it still wait for something else? */
390 	__EE_SUBMITTED,
391 
392 	/* this is/was a write request */
393 	__EE_WRITE,
394 
395 	/* this is/was a write same request */
396 	__EE_WRITE_SAME,
397 
398 	/* this originates from application on peer
399 	 * (not some resync or verify or other DRBD internal request) */
400 	__EE_APPLICATION,
401 
402 	/* If it contains only 0 bytes, send back P_RS_DEALLOCATED */
403 	__EE_RS_THIN_REQ,
404 };
405 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
406 #define EE_MAY_SET_IN_SYNC     (1<<__EE_MAY_SET_IN_SYNC)
407 #define EE_TRIM                (1<<__EE_TRIM)
408 #define EE_ZEROOUT             (1<<__EE_ZEROOUT)
409 #define EE_RESUBMITTED         (1<<__EE_RESUBMITTED)
410 #define EE_WAS_ERROR           (1<<__EE_WAS_ERROR)
411 #define EE_HAS_DIGEST          (1<<__EE_HAS_DIGEST)
412 #define EE_RESTART_REQUESTS	(1<<__EE_RESTART_REQUESTS)
413 #define EE_SEND_WRITE_ACK	(1<<__EE_SEND_WRITE_ACK)
414 #define EE_IN_INTERVAL_TREE	(1<<__EE_IN_INTERVAL_TREE)
415 #define EE_SUBMITTED		(1<<__EE_SUBMITTED)
416 #define EE_WRITE		(1<<__EE_WRITE)
417 #define EE_WRITE_SAME		(1<<__EE_WRITE_SAME)
418 #define EE_APPLICATION		(1<<__EE_APPLICATION)
419 #define EE_RS_THIN_REQ		(1<<__EE_RS_THIN_REQ)
420 
421 /* flag bits per device */
422 enum {
423 	UNPLUG_REMOTE,		/* sending a "UnplugRemote" could help */
424 	MD_DIRTY,		/* current uuids and flags not yet on disk */
425 	USE_DEGR_WFC_T,		/* degr-wfc-timeout instead of wfc-timeout. */
426 	CL_ST_CHG_SUCCESS,
427 	CL_ST_CHG_FAIL,
428 	CRASHED_PRIMARY,	/* This node was a crashed primary.
429 				 * Gets cleared when the state.conn
430 				 * goes into C_CONNECTED state. */
431 	CONSIDER_RESYNC,
432 
433 	MD_NO_FUA,		/* Users wants us to not use FUA/FLUSH on meta data dev */
434 
435 	BITMAP_IO,		/* suspend application io;
436 				   once no more io in flight, start bitmap io */
437 	BITMAP_IO_QUEUED,       /* Started bitmap IO */
438 	WAS_IO_ERROR,		/* Local disk failed, returned IO error */
439 	WAS_READ_ERROR,		/* Local disk READ failed (set additionally to the above) */
440 	FORCE_DETACH,		/* Force-detach from local disk, aborting any pending local IO */
441 	RESYNC_AFTER_NEG,       /* Resync after online grow after the attach&negotiate finished. */
442 	RESIZE_PENDING,		/* Size change detected locally, waiting for the response from
443 				 * the peer, if it changed there as well. */
444 	NEW_CUR_UUID,		/* Create new current UUID when thawing IO */
445 	AL_SUSPENDED,		/* Activity logging is currently suspended. */
446 	AHEAD_TO_SYNC_SOURCE,   /* Ahead -> SyncSource queued */
447 	B_RS_H_DONE,		/* Before resync handler done (already executed) */
448 	DISCARD_MY_DATA,	/* discard_my_data flag per volume */
449 	READ_BALANCE_RR,
450 
451 	FLUSH_PENDING,		/* if set, device->flush_jif is when we submitted that flush
452 				 * from drbd_flush_after_epoch() */
453 
454 	/* cleared only after backing device related structures have been destroyed. */
455 	GOING_DISKLESS,		/* Disk is being detached, because of io-error, or admin request. */
456 
457 	/* to be used in drbd_device_post_work() */
458 	GO_DISKLESS,		/* tell worker to schedule cleanup before detach */
459 	DESTROY_DISK,		/* tell worker to close backing devices and destroy related structures. */
460 	MD_SYNC,		/* tell worker to call drbd_md_sync() */
461 	RS_START,		/* tell worker to start resync/OV */
462 	RS_PROGRESS,		/* tell worker that resync made significant progress */
463 	RS_DONE,		/* tell worker that resync is done */
464 };
465 
466 struct drbd_bitmap; /* opaque for drbd_device */
467 
468 /* definition of bits in bm_flags to be used in drbd_bm_lock
469  * and drbd_bitmap_io and friends. */
470 enum bm_flag {
471 	/* currently locked for bulk operation */
472 	BM_LOCKED_MASK = 0xf,
473 
474 	/* in detail, that is: */
475 	BM_DONT_CLEAR = 0x1,
476 	BM_DONT_SET   = 0x2,
477 	BM_DONT_TEST  = 0x4,
478 
479 	/* so we can mark it locked for bulk operation,
480 	 * and still allow all non-bulk operations */
481 	BM_IS_LOCKED  = 0x8,
482 
483 	/* (test bit, count bit) allowed (common case) */
484 	BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED,
485 
486 	/* testing bits, as well as setting new bits allowed, but clearing bits
487 	 * would be unexpected.  Used during bitmap receive.  Setting new bits
488 	 * requires sending of "out-of-sync" information, though. */
489 	BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED,
490 
491 	/* for drbd_bm_write_copy_pages, everything is allowed,
492 	 * only concurrent bulk operations are locked out. */
493 	BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED,
494 };
495 
496 struct drbd_work_queue {
497 	struct list_head q;
498 	spinlock_t q_lock;  /* to protect the list. */
499 	wait_queue_head_t q_wait;
500 };
501 
502 struct drbd_socket {
503 	struct mutex mutex;
504 	struct socket    *socket;
505 	/* this way we get our
506 	 * send/receive buffers off the stack */
507 	void *sbuf;
508 	void *rbuf;
509 };
510 
511 struct drbd_md {
512 	u64 md_offset;		/* sector offset to 'super' block */
513 
514 	u64 la_size_sect;	/* last agreed size, unit sectors */
515 	spinlock_t uuid_lock;
516 	u64 uuid[UI_SIZE];
517 	u64 device_uuid;
518 	u32 flags;
519 	u32 md_size_sect;
520 
521 	s32 al_offset;	/* signed relative sector offset to activity log */
522 	s32 bm_offset;	/* signed relative sector offset to bitmap */
523 
524 	/* cached value of bdev->disk_conf->meta_dev_idx (see below) */
525 	s32 meta_dev_idx;
526 
527 	/* see al_tr_number_to_on_disk_sector() */
528 	u32 al_stripes;
529 	u32 al_stripe_size_4k;
530 	u32 al_size_4k; /* cached product of the above */
531 };
532 
533 struct drbd_backing_dev {
534 	struct block_device *backing_bdev;
535 	struct block_device *md_bdev;
536 	struct drbd_md md;
537 	struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */
538 	sector_t known_size; /* last known size of that backing device */
539 };
540 
541 struct drbd_md_io {
542 	struct page *page;
543 	unsigned long start_jif;	/* last call to drbd_md_get_buffer */
544 	unsigned long submit_jif;	/* last _drbd_md_sync_page_io() submit */
545 	const char *current_use;
546 	atomic_t in_use;
547 	unsigned int done;
548 	int error;
549 };
550 
551 struct bm_io_work {
552 	struct drbd_work w;
553 	char *why;
554 	enum bm_flag flags;
555 	int (*io_fn)(struct drbd_device *device);
556 	void (*done)(struct drbd_device *device, int rv);
557 };
558 
559 struct fifo_buffer {
560 	unsigned int head_index;
561 	unsigned int size;
562 	int total; /* sum of all values */
563 	int values[];
564 };
565 extern struct fifo_buffer *fifo_alloc(unsigned int fifo_size);
566 
567 /* flag bits per connection */
568 enum {
569 	NET_CONGESTED,		/* The data socket is congested */
570 	RESOLVE_CONFLICTS,	/* Set on one node, cleared on the peer! */
571 	SEND_PING,
572 	GOT_PING_ACK,		/* set when we receive a ping_ack packet, ping_wait gets woken */
573 	CONN_WD_ST_CHG_REQ,	/* A cluster wide state change on the connection is active */
574 	CONN_WD_ST_CHG_OKAY,
575 	CONN_WD_ST_CHG_FAIL,
576 	CONN_DRY_RUN,		/* Expect disconnect after resync handshake. */
577 	CREATE_BARRIER,		/* next P_DATA is preceded by a P_BARRIER */
578 	STATE_SENT,		/* Do not change state/UUIDs while this is set */
579 	CALLBACK_PENDING,	/* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
580 				 * pending, from drbd worker context.
581 				 */
582 	DISCONNECT_SENT,
583 
584 	DEVICE_WORK_PENDING,	/* tell worker that some device has pending work */
585 };
586 
587 enum which_state { NOW, OLD = NOW, NEW };
588 
589 struct drbd_resource {
590 	char *name;
591 #ifdef CONFIG_DEBUG_FS
592 	struct dentry *debugfs_res;
593 	struct dentry *debugfs_res_volumes;
594 	struct dentry *debugfs_res_connections;
595 	struct dentry *debugfs_res_in_flight_summary;
596 #endif
597 	struct kref kref;
598 	struct idr devices;		/* volume number to device mapping */
599 	struct list_head connections;
600 	struct list_head resources;
601 	struct res_opts res_opts;
602 	struct mutex conf_update;	/* mutex for ready-copy-update of net_conf and disk_conf */
603 	struct mutex adm_mutex;		/* mutex to serialize administrative requests */
604 	spinlock_t req_lock;
605 
606 	unsigned susp:1;		/* IO suspended by user */
607 	unsigned susp_nod:1;		/* IO suspended because no data */
608 	unsigned susp_fen:1;		/* IO suspended because fence peer handler runs */
609 
610 	enum write_ordering_e write_ordering;
611 
612 	cpumask_var_t cpu_mask;
613 };
614 
615 struct drbd_thread_timing_details
616 {
617 	unsigned long start_jif;
618 	void *cb_addr;
619 	const char *caller_fn;
620 	unsigned int line;
621 	unsigned int cb_nr;
622 };
623 
624 struct drbd_connection {
625 	struct list_head connections;
626 	struct drbd_resource *resource;
627 #ifdef CONFIG_DEBUG_FS
628 	struct dentry *debugfs_conn;
629 	struct dentry *debugfs_conn_callback_history;
630 	struct dentry *debugfs_conn_oldest_requests;
631 #endif
632 	struct kref kref;
633 	struct idr peer_devices;	/* volume number to peer device mapping */
634 	enum drbd_conns cstate;		/* Only C_STANDALONE to C_WF_REPORT_PARAMS */
635 	struct mutex cstate_mutex;	/* Protects graceful disconnects */
636 	unsigned int connect_cnt;	/* Inc each time a connection is established */
637 
638 	unsigned long flags;
639 	struct net_conf *net_conf;	/* content protected by rcu */
640 	wait_queue_head_t ping_wait;	/* Woken upon reception of a ping, and a state change */
641 
642 	struct sockaddr_storage my_addr;
643 	int my_addr_len;
644 	struct sockaddr_storage peer_addr;
645 	int peer_addr_len;
646 
647 	struct drbd_socket data;	/* data/barrier/cstate/parameter packets */
648 	struct drbd_socket meta;	/* ping/ack (metadata) packets */
649 	int agreed_pro_version;		/* actually used protocol version */
650 	u32 agreed_features;
651 	unsigned long last_received;	/* in jiffies, either socket */
652 	unsigned int ko_count;
653 
654 	struct list_head transfer_log;	/* all requests not yet fully processed */
655 
656 	struct crypto_shash *cram_hmac_tfm;
657 	struct crypto_shash *integrity_tfm;  /* checksums we compute, updates protected by connection->data->mutex */
658 	struct crypto_shash *peer_integrity_tfm;  /* checksums we verify, only accessed from receiver thread  */
659 	struct crypto_shash *csums_tfm;
660 	struct crypto_shash *verify_tfm;
661 	void *int_dig_in;
662 	void *int_dig_vv;
663 
664 	/* receiver side */
665 	struct drbd_epoch *current_epoch;
666 	spinlock_t epoch_lock;
667 	unsigned int epochs;
668 	atomic_t current_tle_nr;	/* transfer log epoch number */
669 	unsigned current_tle_writes;	/* writes seen within this tl epoch */
670 
671 	unsigned long last_reconnect_jif;
672 	/* empty member on older kernels without blk_start_plug() */
673 	struct blk_plug receiver_plug;
674 	struct drbd_thread receiver;
675 	struct drbd_thread worker;
676 	struct drbd_thread ack_receiver;
677 	struct workqueue_struct *ack_sender;
678 
679 	/* cached pointers,
680 	 * so we can look up the oldest pending requests more quickly.
681 	 * protected by resource->req_lock */
682 	struct drbd_request *req_next; /* DRBD 9: todo.req_next */
683 	struct drbd_request *req_ack_pending;
684 	struct drbd_request *req_not_net_done;
685 
686 	/* sender side */
687 	struct drbd_work_queue sender_work;
688 
689 #define DRBD_THREAD_DETAILS_HIST	16
690 	unsigned int w_cb_nr; /* keeps counting up */
691 	unsigned int r_cb_nr; /* keeps counting up */
692 	struct drbd_thread_timing_details w_timing_details[DRBD_THREAD_DETAILS_HIST];
693 	struct drbd_thread_timing_details r_timing_details[DRBD_THREAD_DETAILS_HIST];
694 
695 	struct {
696 		unsigned long last_sent_barrier_jif;
697 
698 		/* whether this sender thread
699 		 * has processed a single write yet. */
700 		bool seen_any_write_yet;
701 
702 		/* Which barrier number to send with the next P_BARRIER */
703 		int current_epoch_nr;
704 
705 		/* how many write requests have been sent
706 		 * with req->epoch == current_epoch_nr.
707 		 * If none, no P_BARRIER will be sent. */
708 		unsigned current_epoch_writes;
709 	} send;
710 };
711 
712 static inline bool has_net_conf(struct drbd_connection *connection)
713 {
714 	bool has_net_conf;
715 
716 	rcu_read_lock();
717 	has_net_conf = rcu_dereference(connection->net_conf);
718 	rcu_read_unlock();
719 
720 	return has_net_conf;
721 }
722 
723 void __update_timing_details(
724 		struct drbd_thread_timing_details *tdp,
725 		unsigned int *cb_nr,
726 		void *cb,
727 		const char *fn, const unsigned int line);
728 
729 #define update_worker_timing_details(c, cb) \
730 	__update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ )
731 #define update_receiver_timing_details(c, cb) \
732 	__update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ )
733 
734 struct submit_worker {
735 	struct workqueue_struct *wq;
736 	struct work_struct worker;
737 
738 	/* protected by ..->resource->req_lock */
739 	struct list_head writes;
740 };
741 
742 struct drbd_peer_device {
743 	struct list_head peer_devices;
744 	struct drbd_device *device;
745 	struct drbd_connection *connection;
746 	struct work_struct send_acks_work;
747 #ifdef CONFIG_DEBUG_FS
748 	struct dentry *debugfs_peer_dev;
749 #endif
750 };
751 
752 struct drbd_device {
753 	struct drbd_resource *resource;
754 	struct list_head peer_devices;
755 	struct list_head pending_bitmap_io;
756 
757 	unsigned long flush_jif;
758 #ifdef CONFIG_DEBUG_FS
759 	struct dentry *debugfs_minor;
760 	struct dentry *debugfs_vol;
761 	struct dentry *debugfs_vol_oldest_requests;
762 	struct dentry *debugfs_vol_act_log_extents;
763 	struct dentry *debugfs_vol_resync_extents;
764 	struct dentry *debugfs_vol_data_gen_id;
765 	struct dentry *debugfs_vol_ed_gen_id;
766 #endif
767 
768 	unsigned int vnr;	/* volume number within the connection */
769 	unsigned int minor;	/* device minor number */
770 
771 	struct kref kref;
772 
773 	/* things that are stored as / read from meta data on disk */
774 	unsigned long flags;
775 
776 	/* configured by drbdsetup */
777 	struct drbd_backing_dev *ldev __protected_by(local);
778 
779 	sector_t p_size;     /* partner's disk size */
780 	struct request_queue *rq_queue;
781 	struct gendisk	    *vdisk;
782 
783 	unsigned long last_reattach_jif;
784 	struct drbd_work resync_work;
785 	struct drbd_work unplug_work;
786 	struct timer_list resync_timer;
787 	struct timer_list md_sync_timer;
788 	struct timer_list start_resync_timer;
789 	struct timer_list request_timer;
790 
791 	/* Used after attach while negotiating new disk state. */
792 	union drbd_state new_state_tmp;
793 
794 	union drbd_dev_state state;
795 	wait_queue_head_t misc_wait;
796 	wait_queue_head_t state_wait;  /* upon each state change. */
797 	unsigned int send_cnt;
798 	unsigned int recv_cnt;
799 	unsigned int read_cnt;
800 	unsigned int writ_cnt;
801 	unsigned int al_writ_cnt;
802 	unsigned int bm_writ_cnt;
803 	atomic_t ap_bio_cnt;	 /* Requests we need to complete */
804 	atomic_t ap_actlog_cnt;  /* Requests waiting for activity log */
805 	atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
806 	atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
807 	atomic_t unacked_cnt;	 /* Need to send replies for */
808 	atomic_t local_cnt;	 /* Waiting for local completion */
809 	atomic_t suspend_cnt;
810 
811 	/* Interval tree of pending local requests */
812 	struct rb_root read_requests;
813 	struct rb_root write_requests;
814 
815 	/* for statistics and timeouts */
816 	/* [0] read, [1] write */
817 	struct list_head pending_master_completion[2];
818 	struct list_head pending_completion[2];
819 
820 	/* use checksums for *this* resync */
821 	bool use_csums;
822 	/* blocks to resync in this run [unit BM_BLOCK_SIZE] */
823 	unsigned long rs_total;
824 	/* number of resync blocks that failed in this run */
825 	unsigned long rs_failed;
826 	/* Syncer's start time [unit jiffies] */
827 	unsigned long rs_start;
828 	/* cumulated time in PausedSyncX state [unit jiffies] */
829 	unsigned long rs_paused;
830 	/* skipped because csum was equal [unit BM_BLOCK_SIZE] */
831 	unsigned long rs_same_csum;
832 #define DRBD_SYNC_MARKS 8
833 #define DRBD_SYNC_MARK_STEP (3*HZ)
834 	/* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
835 	unsigned long rs_mark_left[DRBD_SYNC_MARKS];
836 	/* marks's time [unit jiffies] */
837 	unsigned long rs_mark_time[DRBD_SYNC_MARKS];
838 	/* current index into rs_mark_{left,time} */
839 	int rs_last_mark;
840 	unsigned long rs_last_bcast; /* [unit jiffies] */
841 
842 	/* where does the admin want us to start? (sector) */
843 	sector_t ov_start_sector;
844 	sector_t ov_stop_sector;
845 	/* where are we now? (sector) */
846 	sector_t ov_position;
847 	/* Start sector of out of sync range (to merge printk reporting). */
848 	sector_t ov_last_oos_start;
849 	/* size of out-of-sync range in sectors. */
850 	sector_t ov_last_oos_size;
851 	unsigned long ov_left; /* in bits */
852 
853 	struct drbd_bitmap *bitmap;
854 	unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
855 
856 	/* Used to track operations of resync... */
857 	struct lru_cache *resync;
858 	/* Number of locked elements in resync LRU */
859 	unsigned int resync_locked;
860 	/* resync extent number waiting for application requests */
861 	unsigned int resync_wenr;
862 
863 	int open_cnt;
864 	u64 *p_uuid;
865 
866 	struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
867 	struct list_head sync_ee;   /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
868 	struct list_head done_ee;   /* need to send P_WRITE_ACK */
869 	struct list_head read_ee;   /* [RS]P_DATA_REQUEST being read */
870 	struct list_head net_ee;    /* zero-copy network send in progress */
871 
872 	int next_barrier_nr;
873 	struct list_head resync_reads;
874 	atomic_t pp_in_use;		/* allocated from page pool */
875 	atomic_t pp_in_use_by_net;	/* sendpage()d, still referenced by tcp */
876 	wait_queue_head_t ee_wait;
877 	struct drbd_md_io md_io;
878 	spinlock_t al_lock;
879 	wait_queue_head_t al_wait;
880 	struct lru_cache *act_log;	/* activity log */
881 	unsigned int al_tr_number;
882 	int al_tr_cycle;
883 	wait_queue_head_t seq_wait;
884 	atomic_t packet_seq;
885 	unsigned int peer_seq;
886 	spinlock_t peer_seq_lock;
887 	unsigned long comm_bm_set; /* communicated number of set bits. */
888 	struct bm_io_work bm_io_work;
889 	u64 ed_uuid; /* UUID of the exposed data */
890 	struct mutex own_state_mutex;
891 	struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */
892 	char congestion_reason;  /* Why we where congested... */
893 	atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
894 	atomic_t rs_sect_ev; /* for submitted resync data rate, both */
895 	int rs_last_sect_ev; /* counter to compare with */
896 	int rs_last_events;  /* counter of read or write "events" (unit sectors)
897 			      * on the lower level device when we last looked. */
898 	int c_sync_rate; /* current resync rate after syncer throttle magic */
899 	struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */
900 	int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
901 	atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
902 	unsigned int peer_max_bio_size;
903 	unsigned int local_max_bio_size;
904 
905 	/* any requests that would block in drbd_make_request()
906 	 * are deferred to this single-threaded work queue */
907 	struct submit_worker submit;
908 };
909 
910 struct drbd_bm_aio_ctx {
911 	struct drbd_device *device;
912 	struct list_head list; /* on device->pending_bitmap_io */;
913 	unsigned long start_jif;
914 	atomic_t in_flight;
915 	unsigned int done;
916 	unsigned flags;
917 #define BM_AIO_COPY_PAGES	1
918 #define BM_AIO_WRITE_HINTED	2
919 #define BM_AIO_WRITE_ALL_PAGES	4
920 #define BM_AIO_READ		8
921 	int error;
922 	struct kref kref;
923 };
924 
925 struct drbd_config_context {
926 	/* assigned from drbd_genlmsghdr */
927 	unsigned int minor;
928 	/* assigned from request attributes, if present */
929 	unsigned int volume;
930 #define VOLUME_UNSPECIFIED		(-1U)
931 	/* pointer into the request skb,
932 	 * limited lifetime! */
933 	char *resource_name;
934 	struct nlattr *my_addr;
935 	struct nlattr *peer_addr;
936 
937 	/* reply buffer */
938 	struct sk_buff *reply_skb;
939 	/* pointer into reply buffer */
940 	struct drbd_genlmsghdr *reply_dh;
941 	/* resolved from attributes, if possible */
942 	struct drbd_device *device;
943 	struct drbd_resource *resource;
944 	struct drbd_connection *connection;
945 };
946 
947 static inline struct drbd_device *minor_to_device(unsigned int minor)
948 {
949 	return (struct drbd_device *)idr_find(&drbd_devices, minor);
950 }
951 
952 static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device)
953 {
954 	return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices);
955 }
956 
957 static inline struct drbd_peer_device *
958 conn_peer_device(struct drbd_connection *connection, int volume_number)
959 {
960 	return idr_find(&connection->peer_devices, volume_number);
961 }
962 
963 #define for_each_resource(resource, _resources) \
964 	list_for_each_entry(resource, _resources, resources)
965 
966 #define for_each_resource_rcu(resource, _resources) \
967 	list_for_each_entry_rcu(resource, _resources, resources)
968 
969 #define for_each_resource_safe(resource, tmp, _resources) \
970 	list_for_each_entry_safe(resource, tmp, _resources, resources)
971 
972 #define for_each_connection(connection, resource) \
973 	list_for_each_entry(connection, &resource->connections, connections)
974 
975 #define for_each_connection_rcu(connection, resource) \
976 	list_for_each_entry_rcu(connection, &resource->connections, connections)
977 
978 #define for_each_connection_safe(connection, tmp, resource) \
979 	list_for_each_entry_safe(connection, tmp, &resource->connections, connections)
980 
981 #define for_each_peer_device(peer_device, device) \
982 	list_for_each_entry(peer_device, &device->peer_devices, peer_devices)
983 
984 #define for_each_peer_device_rcu(peer_device, device) \
985 	list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices)
986 
987 #define for_each_peer_device_safe(peer_device, tmp, device) \
988 	list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices)
989 
990 static inline unsigned int device_to_minor(struct drbd_device *device)
991 {
992 	return device->minor;
993 }
994 
995 /*
996  * function declarations
997  *************************/
998 
999 /* drbd_main.c */
1000 
1001 enum dds_flags {
1002 	DDSF_FORCED    = 1,
1003 	DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1004 };
1005 
1006 extern void drbd_init_set_defaults(struct drbd_device *device);
1007 extern int  drbd_thread_start(struct drbd_thread *thi);
1008 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1009 #ifdef CONFIG_SMP
1010 extern void drbd_thread_current_set_cpu(struct drbd_thread *thi);
1011 #else
1012 #define drbd_thread_current_set_cpu(A) ({})
1013 #endif
1014 extern void tl_release(struct drbd_connection *, unsigned int barrier_nr,
1015 		       unsigned int set_size);
1016 extern void tl_clear(struct drbd_connection *);
1017 extern void drbd_free_sock(struct drbd_connection *connection);
1018 extern int drbd_send(struct drbd_connection *connection, struct socket *sock,
1019 		     void *buf, size_t size, unsigned msg_flags);
1020 extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t,
1021 			 unsigned);
1022 
1023 extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd);
1024 extern int drbd_send_protocol(struct drbd_connection *connection);
1025 extern int drbd_send_uuids(struct drbd_peer_device *);
1026 extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *);
1027 extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *);
1028 extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags);
1029 extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s);
1030 extern int drbd_send_current_state(struct drbd_peer_device *);
1031 extern int drbd_send_sync_param(struct drbd_peer_device *);
1032 extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr,
1033 			    u32 set_size);
1034 extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet,
1035 			 struct drbd_peer_request *);
1036 extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet,
1037 			     struct p_block_req *rp);
1038 extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet,
1039 			     struct p_data *dp, int data_size);
1040 extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet,
1041 			    sector_t sector, int blksize, u64 block_id);
1042 extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *);
1043 extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet,
1044 			   struct drbd_peer_request *);
1045 extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req);
1046 extern int drbd_send_drequest(struct drbd_peer_device *, int cmd,
1047 			      sector_t sector, int size, u64 block_id);
1048 extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector,
1049 				   int size, void *digest, int digest_size,
1050 				   enum drbd_packet cmd);
1051 extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size);
1052 
1053 extern int drbd_send_bitmap(struct drbd_device *device);
1054 extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode);
1055 extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode);
1056 extern int drbd_send_rs_deallocated(struct drbd_peer_device *, struct drbd_peer_request *);
1057 extern void drbd_backing_dev_free(struct drbd_device *device, struct drbd_backing_dev *ldev);
1058 extern void drbd_device_cleanup(struct drbd_device *device);
1059 extern void drbd_print_uuids(struct drbd_device *device, const char *text);
1060 extern void drbd_queue_unplug(struct drbd_device *device);
1061 
1062 extern void conn_md_sync(struct drbd_connection *connection);
1063 extern void drbd_md_write(struct drbd_device *device, void *buffer);
1064 extern void drbd_md_sync(struct drbd_device *device);
1065 extern int  drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev);
1066 extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1067 extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1068 extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local);
1069 extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local);
1070 extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local);
1071 extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1072 extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local);
1073 extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local);
1074 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1075 extern void drbd_md_mark_dirty(struct drbd_device *device);
1076 extern void drbd_queue_bitmap_io(struct drbd_device *device,
1077 				 int (*io_fn)(struct drbd_device *),
1078 				 void (*done)(struct drbd_device *, int),
1079 				 char *why, enum bm_flag flags);
1080 extern int drbd_bitmap_io(struct drbd_device *device,
1081 		int (*io_fn)(struct drbd_device *),
1082 		char *why, enum bm_flag flags);
1083 extern int drbd_bitmap_io_from_worker(struct drbd_device *device,
1084 		int (*io_fn)(struct drbd_device *),
1085 		char *why, enum bm_flag flags);
1086 extern int drbd_bmio_set_n_write(struct drbd_device *device) __must_hold(local);
1087 extern int drbd_bmio_clear_n_write(struct drbd_device *device) __must_hold(local);
1088 
1089 /* Meta data layout
1090  *
1091  * We currently have two possible layouts.
1092  * Offsets in (512 byte) sectors.
1093  * external:
1094  *   |----------- md_size_sect ------------------|
1095  *   [ 4k superblock ][ activity log ][  Bitmap  ]
1096  *   | al_offset == 8 |
1097  *   | bm_offset = al_offset + X      |
1098  *  ==> bitmap sectors = md_size_sect - bm_offset
1099  *
1100  *  Variants:
1101  *     old, indexed fixed size meta data:
1102  *
1103  * internal:
1104  *            |----------- md_size_sect ------------------|
1105  * [data.....][  Bitmap  ][ activity log ][ 4k superblock ][padding*]
1106  *                        | al_offset < 0 |
1107  *            | bm_offset = al_offset - Y |
1108  *  ==> bitmap sectors = Y = al_offset - bm_offset
1109  *
1110  *  [padding*] are zero or up to 7 unused 512 Byte sectors to the
1111  *  end of the device, so that the [4k superblock] will be 4k aligned.
1112  *
1113  *  The activity log consists of 4k transaction blocks,
1114  *  which are written in a ring-buffer, or striped ring-buffer like fashion,
1115  *  which are writtensize used to be fixed 32kB,
1116  *  but is about to become configurable.
1117  */
1118 
1119 /* Our old fixed size meta data layout
1120  * allows up to about 3.8TB, so if you want more,
1121  * you need to use the "flexible" meta data format. */
1122 #define MD_128MB_SECT (128LLU << 11)  /* 128 MB, unit sectors */
1123 #define MD_4kB_SECT	 8
1124 #define MD_32kB_SECT	64
1125 
1126 /* One activity log extent represents 4M of storage */
1127 #define AL_EXTENT_SHIFT 22
1128 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1129 
1130 /* We could make these currently hardcoded constants configurable
1131  * variables at create-md time (or even re-configurable at runtime?).
1132  * Which will require some more changes to the DRBD "super block"
1133  * and attach code.
1134  *
1135  * updates per transaction:
1136  *   This many changes to the active set can be logged with one transaction.
1137  *   This number is arbitrary.
1138  * context per transaction:
1139  *   This many context extent numbers are logged with each transaction.
1140  *   This number is resulting from the transaction block size (4k), the layout
1141  *   of the transaction header, and the number of updates per transaction.
1142  *   See drbd_actlog.c:struct al_transaction_on_disk
1143  * */
1144 #define AL_UPDATES_PER_TRANSACTION	 64	// arbitrary
1145 #define AL_CONTEXT_PER_TRANSACTION	919	// (4096 - 36 - 6*64)/4
1146 
1147 #if BITS_PER_LONG == 32
1148 #define LN2_BPL 5
1149 #define cpu_to_lel(A) cpu_to_le32(A)
1150 #define lel_to_cpu(A) le32_to_cpu(A)
1151 #elif BITS_PER_LONG == 64
1152 #define LN2_BPL 6
1153 #define cpu_to_lel(A) cpu_to_le64(A)
1154 #define lel_to_cpu(A) le64_to_cpu(A)
1155 #else
1156 #error "LN2 of BITS_PER_LONG unknown!"
1157 #endif
1158 
1159 /* resync bitmap */
1160 /* 16MB sized 'bitmap extent' to track syncer usage */
1161 struct bm_extent {
1162 	int rs_left; /* number of bits set (out of sync) in this extent. */
1163 	int rs_failed; /* number of failed resync requests in this extent. */
1164 	unsigned long flags;
1165 	struct lc_element lce;
1166 };
1167 
1168 #define BME_NO_WRITES  0  /* bm_extent.flags: no more requests on this one! */
1169 #define BME_LOCKED     1  /* bm_extent.flags: syncer active on this one. */
1170 #define BME_PRIORITY   2  /* finish resync IO on this extent ASAP! App IO waiting! */
1171 
1172 /* drbd_bitmap.c */
1173 /*
1174  * We need to store one bit for a block.
1175  * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1176  * Bit 0 ==> local node thinks this block is binary identical on both nodes
1177  * Bit 1 ==> local node thinks this block needs to be synced.
1178  */
1179 
1180 #define SLEEP_TIME (HZ/10)
1181 
1182 /* We do bitmap IO in units of 4k blocks.
1183  * We also still have a hardcoded 4k per bit relation. */
1184 #define BM_BLOCK_SHIFT	12			 /* 4k per bit */
1185 #define BM_BLOCK_SIZE	 (1<<BM_BLOCK_SHIFT)
1186 /* mostly arbitrarily set the represented size of one bitmap extent,
1187  * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1188  * at 4k per bit resolution) */
1189 #define BM_EXT_SHIFT	 24	/* 16 MiB per resync extent */
1190 #define BM_EXT_SIZE	 (1<<BM_EXT_SHIFT)
1191 
1192 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1193 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1194 #endif
1195 
1196 /* thus many _storage_ sectors are described by one bit */
1197 #define BM_SECT_TO_BIT(x)   ((x)>>(BM_BLOCK_SHIFT-9))
1198 #define BM_BIT_TO_SECT(x)   ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1199 #define BM_SECT_PER_BIT     BM_BIT_TO_SECT(1)
1200 
1201 /* bit to represented kilo byte conversion */
1202 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1203 
1204 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1205  * _storage_ sector is located in */
1206 #define BM_SECT_TO_EXT(x)   ((x)>>(BM_EXT_SHIFT-9))
1207 #define BM_BIT_TO_EXT(x)    ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1208 
1209 /* first storage sector a bitmap extent corresponds to */
1210 #define BM_EXT_TO_SECT(x)   ((sector_t)(x) << (BM_EXT_SHIFT-9))
1211 /* how much _storage_ sectors we have per bitmap extent */
1212 #define BM_SECT_PER_EXT     BM_EXT_TO_SECT(1)
1213 /* how many bits are covered by one bitmap extent (resync extent) */
1214 #define BM_BITS_PER_EXT     (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1215 
1216 #define BM_BLOCKS_PER_BM_EXT_MASK  (BM_BITS_PER_EXT - 1)
1217 
1218 
1219 /* in one sector of the bitmap, we have this many activity_log extents. */
1220 #define AL_EXT_PER_BM_SECT  (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1221 
1222 /* the extent in "PER_EXTENT" below is an activity log extent
1223  * we need that many (long words/bytes) to store the bitmap
1224  *		     of one AL_EXTENT_SIZE chunk of storage.
1225  * we can store the bitmap for that many AL_EXTENTS within
1226  * one sector of the _on_disk_ bitmap:
1227  * bit	 0	  bit 37   bit 38	     bit (512*8)-1
1228  *	     ...|........|........|.. // ..|........|
1229  * sect. 0	 `296	  `304			   ^(512*8*8)-1
1230  *
1231 #define BM_WORDS_PER_EXT    ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1232 #define BM_BYTES_PER_EXT    ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 )  // 128
1233 #define BM_EXT_PER_SECT	    ( 512 / BM_BYTES_PER_EXTENT )	 //   4
1234  */
1235 
1236 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1237 /* we have a certain meta data variant that has a fixed on-disk size of 128
1238  * MiB, of which 4k are our "superblock", and 32k are the fixed size activity
1239  * log, leaving this many sectors for the bitmap.
1240  */
1241 
1242 #define DRBD_MAX_SECTORS_FIXED_BM \
1243 	  ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9)))
1244 #define DRBD_MAX_SECTORS      DRBD_MAX_SECTORS_FIXED_BM
1245 /* 16 TB in units of sectors */
1246 #if BITS_PER_LONG == 32
1247 /* adjust by one page worth of bitmap,
1248  * so we won't wrap around in drbd_bm_find_next_bit.
1249  * you should use 64bit OS for that much storage, anyways. */
1250 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1251 #else
1252 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1253 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1254 /* corresponds to (1UL << 38) bits right now. */
1255 #endif
1256 
1257 /* Estimate max bio size as 256 * PAGE_SIZE,
1258  * so for typical PAGE_SIZE of 4k, that is (1<<20) Byte.
1259  * Since we may live in a mixed-platform cluster,
1260  * we limit us to a platform agnostic constant here for now.
1261  * A followup commit may allow even bigger BIO sizes,
1262  * once we thought that through. */
1263 #define DRBD_MAX_BIO_SIZE (1U << 20)
1264 #if DRBD_MAX_BIO_SIZE > (BIO_MAX_VECS << PAGE_SHIFT)
1265 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1266 #endif
1267 #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12)       /* Works always = 4k */
1268 
1269 #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */
1270 #define DRBD_MAX_BIO_SIZE_P95    (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */
1271 
1272 /* For now, don't allow more than half of what we can "activate" in one
1273  * activity log transaction to be discarded in one go. We may need to rework
1274  * drbd_al_begin_io() to allow for even larger discard ranges */
1275 #define DRBD_MAX_BATCH_BIO_SIZE	 (AL_UPDATES_PER_TRANSACTION/2*AL_EXTENT_SIZE)
1276 #define DRBD_MAX_BBIO_SECTORS    (DRBD_MAX_BATCH_BIO_SIZE >> 9)
1277 
1278 extern int  drbd_bm_init(struct drbd_device *device);
1279 extern int  drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits);
1280 extern void drbd_bm_cleanup(struct drbd_device *device);
1281 extern void drbd_bm_set_all(struct drbd_device *device);
1282 extern void drbd_bm_clear_all(struct drbd_device *device);
1283 /* set/clear/test only a few bits at a time */
1284 extern int  drbd_bm_set_bits(
1285 		struct drbd_device *device, unsigned long s, unsigned long e);
1286 extern int  drbd_bm_clear_bits(
1287 		struct drbd_device *device, unsigned long s, unsigned long e);
1288 extern int drbd_bm_count_bits(
1289 	struct drbd_device *device, const unsigned long s, const unsigned long e);
1290 /* bm_set_bits variant for use while holding drbd_bm_lock,
1291  * may process the whole bitmap in one go */
1292 extern void _drbd_bm_set_bits(struct drbd_device *device,
1293 		const unsigned long s, const unsigned long e);
1294 extern int  drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr);
1295 extern int  drbd_bm_e_weight(struct drbd_device *device, unsigned long enr);
1296 extern int  drbd_bm_read(struct drbd_device *device) __must_hold(local);
1297 extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr);
1298 extern int  drbd_bm_write(struct drbd_device *device) __must_hold(local);
1299 extern void drbd_bm_reset_al_hints(struct drbd_device *device) __must_hold(local);
1300 extern int  drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local);
1301 extern int  drbd_bm_write_lazy(struct drbd_device *device, unsigned upper_idx) __must_hold(local);
1302 extern int drbd_bm_write_all(struct drbd_device *device) __must_hold(local);
1303 extern int  drbd_bm_write_copy_pages(struct drbd_device *device) __must_hold(local);
1304 extern size_t	     drbd_bm_words(struct drbd_device *device);
1305 extern unsigned long drbd_bm_bits(struct drbd_device *device);
1306 extern sector_t      drbd_bm_capacity(struct drbd_device *device);
1307 
1308 #define DRBD_END_OF_BITMAP	(~(unsigned long)0)
1309 extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo);
1310 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1311 extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo);
1312 extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo);
1313 extern unsigned long _drbd_bm_total_weight(struct drbd_device *device);
1314 extern unsigned long drbd_bm_total_weight(struct drbd_device *device);
1315 /* for receive_bitmap */
1316 extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset,
1317 		size_t number, unsigned long *buffer);
1318 /* for _drbd_send_bitmap */
1319 extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset,
1320 		size_t number, unsigned long *buffer);
1321 
1322 extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags);
1323 extern void drbd_bm_unlock(struct drbd_device *device);
1324 /* drbd_main.c */
1325 
1326 extern struct kmem_cache *drbd_request_cache;
1327 extern struct kmem_cache *drbd_ee_cache;	/* peer requests */
1328 extern struct kmem_cache *drbd_bm_ext_cache;	/* bitmap extents */
1329 extern struct kmem_cache *drbd_al_ext_cache;	/* activity log extents */
1330 extern mempool_t drbd_request_mempool;
1331 extern mempool_t drbd_ee_mempool;
1332 
1333 /* drbd's page pool, used to buffer data received from the peer,
1334  * or data requested by the peer.
1335  *
1336  * This does not have an emergency reserve.
1337  *
1338  * When allocating from this pool, it first takes pages from the pool.
1339  * Only if the pool is depleted will try to allocate from the system.
1340  *
1341  * The assumption is that pages taken from this pool will be processed,
1342  * and given back, "quickly", and then can be recycled, so we can avoid
1343  * frequent calls to alloc_page(), and still will be able to make progress even
1344  * under memory pressure.
1345  */
1346 extern struct page *drbd_pp_pool;
1347 extern spinlock_t   drbd_pp_lock;
1348 extern int	    drbd_pp_vacant;
1349 extern wait_queue_head_t drbd_pp_wait;
1350 
1351 /* We also need a standard (emergency-reserve backed) page pool
1352  * for meta data IO (activity log, bitmap).
1353  * We can keep it global, as long as it is used as "N pages at a time".
1354  * 128 should be plenty, currently we probably can get away with as few as 1.
1355  */
1356 #define DRBD_MIN_POOL_PAGES	128
1357 extern mempool_t drbd_md_io_page_pool;
1358 
1359 /* We also need to make sure we get a bio
1360  * when we need it for housekeeping purposes */
1361 extern struct bio_set drbd_md_io_bio_set;
1362 
1363 /* And a bio_set for cloning */
1364 extern struct bio_set drbd_io_bio_set;
1365 
1366 extern struct mutex resources_mutex;
1367 
1368 extern int conn_lowest_minor(struct drbd_connection *connection);
1369 extern enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor);
1370 extern void drbd_destroy_device(struct kref *kref);
1371 extern void drbd_delete_device(struct drbd_device *device);
1372 
1373 extern struct drbd_resource *drbd_create_resource(const char *name);
1374 extern void drbd_free_resource(struct drbd_resource *resource);
1375 
1376 extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts);
1377 extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts);
1378 extern void drbd_destroy_connection(struct kref *kref);
1379 extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len,
1380 					    void *peer_addr, int peer_addr_len);
1381 extern struct drbd_resource *drbd_find_resource(const char *name);
1382 extern void drbd_destroy_resource(struct kref *kref);
1383 extern void conn_free_crypto(struct drbd_connection *connection);
1384 
1385 /* drbd_req */
1386 extern void do_submit(struct work_struct *ws);
1387 extern void __drbd_make_request(struct drbd_device *, struct bio *);
1388 void drbd_submit_bio(struct bio *bio);
1389 extern int drbd_read_remote(struct drbd_device *device, struct drbd_request *req);
1390 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1391 
1392 
1393 /* drbd_nl.c */
1394 
1395 extern struct mutex notification_mutex;
1396 
1397 extern void drbd_suspend_io(struct drbd_device *device);
1398 extern void drbd_resume_io(struct drbd_device *device);
1399 extern char *ppsize(char *buf, unsigned long long size);
1400 extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int);
1401 enum determine_dev_size {
1402 	DS_ERROR_SHRINK = -3,
1403 	DS_ERROR_SPACE_MD = -2,
1404 	DS_ERROR = -1,
1405 	DS_UNCHANGED = 0,
1406 	DS_SHRUNK = 1,
1407 	DS_GREW = 2,
1408 	DS_GREW_FROM_ZERO = 3,
1409 };
1410 extern enum determine_dev_size
1411 drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local);
1412 extern void resync_after_online_grow(struct drbd_device *);
1413 extern void drbd_reconsider_queue_parameters(struct drbd_device *device,
1414 			struct drbd_backing_dev *bdev, struct o_qlim *o);
1415 extern enum drbd_state_rv drbd_set_role(struct drbd_device *device,
1416 					enum drbd_role new_role,
1417 					int force);
1418 extern bool conn_try_outdate_peer(struct drbd_connection *connection);
1419 extern void conn_try_outdate_peer_async(struct drbd_connection *connection);
1420 extern enum drbd_peer_state conn_khelper(struct drbd_connection *connection, char *cmd);
1421 extern int drbd_khelper(struct drbd_device *device, char *cmd);
1422 
1423 /* drbd_worker.c */
1424 /* bi_end_io handlers */
1425 extern void drbd_md_endio(struct bio *bio);
1426 extern void drbd_peer_request_endio(struct bio *bio);
1427 extern void drbd_request_endio(struct bio *bio);
1428 extern int drbd_worker(struct drbd_thread *thi);
1429 enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor);
1430 void drbd_resync_after_changed(struct drbd_device *device);
1431 extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side);
1432 extern void resume_next_sg(struct drbd_device *device);
1433 extern void suspend_other_sg(struct drbd_device *device);
1434 extern int drbd_resync_finished(struct drbd_device *device);
1435 /* maybe rather drbd_main.c ? */
1436 extern void *drbd_md_get_buffer(struct drbd_device *device, const char *intent);
1437 extern void drbd_md_put_buffer(struct drbd_device *device);
1438 extern int drbd_md_sync_page_io(struct drbd_device *device,
1439 		struct drbd_backing_dev *bdev, sector_t sector, enum req_op op);
1440 extern void drbd_ov_out_of_sync_found(struct drbd_device *, sector_t, int);
1441 extern void wait_until_done_or_force_detached(struct drbd_device *device,
1442 		struct drbd_backing_dev *bdev, unsigned int *done);
1443 extern void drbd_rs_controller_reset(struct drbd_device *device);
1444 
1445 static inline void ov_out_of_sync_print(struct drbd_device *device)
1446 {
1447 	if (device->ov_last_oos_size) {
1448 		drbd_err(device, "Out of sync: start=%llu, size=%lu (sectors)\n",
1449 		     (unsigned long long)device->ov_last_oos_start,
1450 		     (unsigned long)device->ov_last_oos_size);
1451 	}
1452 	device->ov_last_oos_size = 0;
1453 }
1454 
1455 
1456 extern void drbd_csum_bio(struct crypto_shash *, struct bio *, void *);
1457 extern void drbd_csum_ee(struct crypto_shash *, struct drbd_peer_request *,
1458 			 void *);
1459 /* worker callbacks */
1460 extern int w_e_end_data_req(struct drbd_work *, int);
1461 extern int w_e_end_rsdata_req(struct drbd_work *, int);
1462 extern int w_e_end_csum_rs_req(struct drbd_work *, int);
1463 extern int w_e_end_ov_reply(struct drbd_work *, int);
1464 extern int w_e_end_ov_req(struct drbd_work *, int);
1465 extern int w_ov_finished(struct drbd_work *, int);
1466 extern int w_resync_timer(struct drbd_work *, int);
1467 extern int w_send_write_hint(struct drbd_work *, int);
1468 extern int w_send_dblock(struct drbd_work *, int);
1469 extern int w_send_read_req(struct drbd_work *, int);
1470 extern int w_e_reissue(struct drbd_work *, int);
1471 extern int w_restart_disk_io(struct drbd_work *, int);
1472 extern int w_send_out_of_sync(struct drbd_work *, int);
1473 
1474 extern void resync_timer_fn(struct timer_list *t);
1475 extern void start_resync_timer_fn(struct timer_list *t);
1476 
1477 extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req);
1478 
1479 /* drbd_receiver.c */
1480 extern int drbd_issue_discard_or_zero_out(struct drbd_device *device,
1481 		sector_t start, unsigned int nr_sectors, int flags);
1482 extern int drbd_receiver(struct drbd_thread *thi);
1483 extern int drbd_ack_receiver(struct drbd_thread *thi);
1484 extern void drbd_send_ping_wf(struct work_struct *ws);
1485 extern void drbd_send_acks_wf(struct work_struct *ws);
1486 extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device);
1487 extern bool drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector,
1488 		bool throttle_if_app_is_waiting);
1489 extern int drbd_submit_peer_request(struct drbd_peer_request *peer_req);
1490 extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *);
1491 extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64,
1492 						     sector_t, unsigned int,
1493 						     unsigned int,
1494 						     gfp_t) __must_hold(local);
1495 extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *,
1496 				 int);
1497 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1498 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
1499 extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool);
1500 extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
1501 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
1502 extern int drbd_connected(struct drbd_peer_device *);
1503 
1504 /* sets the number of 512 byte sectors of our virtual device */
1505 void drbd_set_my_capacity(struct drbd_device *device, sector_t size);
1506 
1507 /*
1508  * used to submit our private bio
1509  */
1510 static inline void drbd_submit_bio_noacct(struct drbd_device *device,
1511 					     int fault_type, struct bio *bio)
1512 {
1513 	__release(local);
1514 	if (!bio->bi_bdev) {
1515 		drbd_err(device, "drbd_submit_bio_noacct: bio->bi_bdev == NULL\n");
1516 		bio->bi_status = BLK_STS_IOERR;
1517 		bio_endio(bio);
1518 		return;
1519 	}
1520 
1521 	if (drbd_insert_fault(device, fault_type))
1522 		bio_io_error(bio);
1523 	else
1524 		submit_bio_noacct(bio);
1525 }
1526 
1527 void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev,
1528 			      enum write_ordering_e wo);
1529 
1530 /* drbd_proc.c */
1531 extern struct proc_dir_entry *drbd_proc;
1532 int drbd_seq_show(struct seq_file *seq, void *v);
1533 
1534 /* drbd_actlog.c */
1535 extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i);
1536 extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i);
1537 extern void drbd_al_begin_io_commit(struct drbd_device *device);
1538 extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i);
1539 extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i);
1540 extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i);
1541 extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector);
1542 extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector);
1543 extern int drbd_try_rs_begin_io(struct drbd_device *device, sector_t sector);
1544 extern void drbd_rs_cancel_all(struct drbd_device *device);
1545 extern int drbd_rs_del_all(struct drbd_device *device);
1546 extern void drbd_rs_failed_io(struct drbd_device *device,
1547 		sector_t sector, int size);
1548 extern void drbd_advance_rs_marks(struct drbd_device *device, unsigned long still_to_go);
1549 
1550 enum update_sync_bits_mode { RECORD_RS_FAILED, SET_OUT_OF_SYNC, SET_IN_SYNC };
1551 extern int __drbd_change_sync(struct drbd_device *device, sector_t sector, int size,
1552 		enum update_sync_bits_mode mode);
1553 #define drbd_set_in_sync(device, sector, size) \
1554 	__drbd_change_sync(device, sector, size, SET_IN_SYNC)
1555 #define drbd_set_out_of_sync(device, sector, size) \
1556 	__drbd_change_sync(device, sector, size, SET_OUT_OF_SYNC)
1557 #define drbd_rs_failed_io(device, sector, size) \
1558 	__drbd_change_sync(device, sector, size, RECORD_RS_FAILED)
1559 extern void drbd_al_shrink(struct drbd_device *device);
1560 extern int drbd_al_initialize(struct drbd_device *, void *);
1561 
1562 /* drbd_nl.c */
1563 /* state info broadcast */
1564 struct sib_info {
1565 	enum drbd_state_info_bcast_reason sib_reason;
1566 	union {
1567 		struct {
1568 			char *helper_name;
1569 			unsigned helper_exit_code;
1570 		};
1571 		struct {
1572 			union drbd_state os;
1573 			union drbd_state ns;
1574 		};
1575 	};
1576 };
1577 void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib);
1578 
1579 extern int notify_resource_state(struct sk_buff *,
1580 				  unsigned int,
1581 				  struct drbd_resource *,
1582 				  struct resource_info *,
1583 				  enum drbd_notification_type);
1584 extern int notify_device_state(struct sk_buff *,
1585 				unsigned int,
1586 				struct drbd_device *,
1587 				struct device_info *,
1588 				enum drbd_notification_type);
1589 extern int notify_connection_state(struct sk_buff *,
1590 				    unsigned int,
1591 				    struct drbd_connection *,
1592 				    struct connection_info *,
1593 				    enum drbd_notification_type);
1594 extern int notify_peer_device_state(struct sk_buff *,
1595 				     unsigned int,
1596 				     struct drbd_peer_device *,
1597 				     struct peer_device_info *,
1598 				     enum drbd_notification_type);
1599 extern void notify_helper(enum drbd_notification_type, struct drbd_device *,
1600 			  struct drbd_connection *, const char *, int);
1601 
1602 /*
1603  * inline helper functions
1604  *************************/
1605 
1606 /* see also page_chain_add and friends in drbd_receiver.c */
1607 static inline struct page *page_chain_next(struct page *page)
1608 {
1609 	return (struct page *)page_private(page);
1610 }
1611 #define page_chain_for_each(page) \
1612 	for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1613 			page = page_chain_next(page))
1614 #define page_chain_for_each_safe(page, n) \
1615 	for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1616 
1617 
1618 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req)
1619 {
1620 	struct page *page = peer_req->pages;
1621 	page_chain_for_each(page) {
1622 		if (page_count(page) > 1)
1623 			return 1;
1624 	}
1625 	return 0;
1626 }
1627 
1628 static inline union drbd_state drbd_read_state(struct drbd_device *device)
1629 {
1630 	struct drbd_resource *resource = device->resource;
1631 	union drbd_state rv;
1632 
1633 	rv.i = device->state.i;
1634 	rv.susp = resource->susp;
1635 	rv.susp_nod = resource->susp_nod;
1636 	rv.susp_fen = resource->susp_fen;
1637 
1638 	return rv;
1639 }
1640 
1641 enum drbd_force_detach_flags {
1642 	DRBD_READ_ERROR,
1643 	DRBD_WRITE_ERROR,
1644 	DRBD_META_IO_ERROR,
1645 	DRBD_FORCE_DETACH,
1646 };
1647 
1648 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1649 static inline void __drbd_chk_io_error_(struct drbd_device *device,
1650 		enum drbd_force_detach_flags df,
1651 		const char *where)
1652 {
1653 	enum drbd_io_error_p ep;
1654 
1655 	rcu_read_lock();
1656 	ep = rcu_dereference(device->ldev->disk_conf)->on_io_error;
1657 	rcu_read_unlock();
1658 	switch (ep) {
1659 	case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */
1660 		if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) {
1661 			if (drbd_ratelimit())
1662 				drbd_err(device, "Local IO failed in %s.\n", where);
1663 			if (device->state.disk > D_INCONSISTENT)
1664 				_drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL);
1665 			break;
1666 		}
1667 		fallthrough;	/* for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */
1668 	case EP_DETACH:
1669 	case EP_CALL_HELPER:
1670 		/* Remember whether we saw a READ or WRITE error.
1671 		 *
1672 		 * Recovery of the affected area for WRITE failure is covered
1673 		 * by the activity log.
1674 		 * READ errors may fall outside that area though. Certain READ
1675 		 * errors can be "healed" by writing good data to the affected
1676 		 * blocks, which triggers block re-allocation in lower layers.
1677 		 *
1678 		 * If we can not write the bitmap after a READ error,
1679 		 * we may need to trigger a full sync (see w_go_diskless()).
1680 		 *
1681 		 * Force-detach is not really an IO error, but rather a
1682 		 * desperate measure to try to deal with a completely
1683 		 * unresponsive lower level IO stack.
1684 		 * Still it should be treated as a WRITE error.
1685 		 *
1686 		 * Meta IO error is always WRITE error:
1687 		 * we read meta data only once during attach,
1688 		 * which will fail in case of errors.
1689 		 */
1690 		set_bit(WAS_IO_ERROR, &device->flags);
1691 		if (df == DRBD_READ_ERROR)
1692 			set_bit(WAS_READ_ERROR, &device->flags);
1693 		if (df == DRBD_FORCE_DETACH)
1694 			set_bit(FORCE_DETACH, &device->flags);
1695 		if (device->state.disk > D_FAILED) {
1696 			_drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL);
1697 			drbd_err(device,
1698 				"Local IO failed in %s. Detaching...\n", where);
1699 		}
1700 		break;
1701 	}
1702 }
1703 
1704 /**
1705  * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1706  * @device:	 DRBD device.
1707  * @error:	 Error code passed to the IO completion callback
1708  * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1709  *
1710  * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1711  */
1712 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1713 static inline void drbd_chk_io_error_(struct drbd_device *device,
1714 	int error, enum drbd_force_detach_flags forcedetach, const char *where)
1715 {
1716 	if (error) {
1717 		unsigned long flags;
1718 		spin_lock_irqsave(&device->resource->req_lock, flags);
1719 		__drbd_chk_io_error_(device, forcedetach, where);
1720 		spin_unlock_irqrestore(&device->resource->req_lock, flags);
1721 	}
1722 }
1723 
1724 
1725 /**
1726  * drbd_md_first_sector() - Returns the first sector number of the meta data area
1727  * @bdev:	Meta data block device.
1728  *
1729  * BTW, for internal meta data, this happens to be the maximum capacity
1730  * we could agree upon with our peer node.
1731  */
1732 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1733 {
1734 	switch (bdev->md.meta_dev_idx) {
1735 	case DRBD_MD_INDEX_INTERNAL:
1736 	case DRBD_MD_INDEX_FLEX_INT:
1737 		return bdev->md.md_offset + bdev->md.bm_offset;
1738 	case DRBD_MD_INDEX_FLEX_EXT:
1739 	default:
1740 		return bdev->md.md_offset;
1741 	}
1742 }
1743 
1744 /**
1745  * drbd_md_last_sector() - Return the last sector number of the meta data area
1746  * @bdev:	Meta data block device.
1747  */
1748 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1749 {
1750 	switch (bdev->md.meta_dev_idx) {
1751 	case DRBD_MD_INDEX_INTERNAL:
1752 	case DRBD_MD_INDEX_FLEX_INT:
1753 		return bdev->md.md_offset + MD_4kB_SECT -1;
1754 	case DRBD_MD_INDEX_FLEX_EXT:
1755 	default:
1756 		return bdev->md.md_offset + bdev->md.md_size_sect -1;
1757 	}
1758 }
1759 
1760 /* Returns the number of 512 byte sectors of the device */
1761 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1762 {
1763 	return bdev ? bdev_nr_sectors(bdev) : 0;
1764 }
1765 
1766 /**
1767  * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1768  * @bdev:	Meta data block device.
1769  *
1770  * returns the capacity we announce to out peer.  we clip ourselves at the
1771  * various MAX_SECTORS, because if we don't, current implementation will
1772  * oops sooner or later
1773  */
1774 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1775 {
1776 	sector_t s;
1777 
1778 	switch (bdev->md.meta_dev_idx) {
1779 	case DRBD_MD_INDEX_INTERNAL:
1780 	case DRBD_MD_INDEX_FLEX_INT:
1781 		s = drbd_get_capacity(bdev->backing_bdev)
1782 			? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1783 				drbd_md_first_sector(bdev))
1784 			: 0;
1785 		break;
1786 	case DRBD_MD_INDEX_FLEX_EXT:
1787 		s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1788 				drbd_get_capacity(bdev->backing_bdev));
1789 		/* clip at maximum size the meta device can support */
1790 		s = min_t(sector_t, s,
1791 			BM_EXT_TO_SECT(bdev->md.md_size_sect
1792 				     - bdev->md.bm_offset));
1793 		break;
1794 	default:
1795 		s = min_t(sector_t, DRBD_MAX_SECTORS,
1796 				drbd_get_capacity(bdev->backing_bdev));
1797 	}
1798 	return s;
1799 }
1800 
1801 /**
1802  * drbd_md_ss() - Return the sector number of our meta data super block
1803  * @bdev:	Meta data block device.
1804  */
1805 static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev)
1806 {
1807 	const int meta_dev_idx = bdev->md.meta_dev_idx;
1808 
1809 	if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT)
1810 		return 0;
1811 
1812 	/* Since drbd08, internal meta data is always "flexible".
1813 	 * position: last 4k aligned block of 4k size */
1814 	if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL ||
1815 	    meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)
1816 		return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8;
1817 
1818 	/* external, some index; this is the old fixed size layout */
1819 	return MD_128MB_SECT * bdev->md.meta_dev_idx;
1820 }
1821 
1822 static inline void
1823 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1824 {
1825 	unsigned long flags;
1826 	spin_lock_irqsave(&q->q_lock, flags);
1827 	list_add_tail(&w->list, &q->q);
1828 	spin_unlock_irqrestore(&q->q_lock, flags);
1829 	wake_up(&q->q_wait);
1830 }
1831 
1832 static inline void
1833 drbd_queue_work_if_unqueued(struct drbd_work_queue *q, struct drbd_work *w)
1834 {
1835 	unsigned long flags;
1836 	spin_lock_irqsave(&q->q_lock, flags);
1837 	if (list_empty_careful(&w->list))
1838 		list_add_tail(&w->list, &q->q);
1839 	spin_unlock_irqrestore(&q->q_lock, flags);
1840 	wake_up(&q->q_wait);
1841 }
1842 
1843 static inline void
1844 drbd_device_post_work(struct drbd_device *device, int work_bit)
1845 {
1846 	if (!test_and_set_bit(work_bit, &device->flags)) {
1847 		struct drbd_connection *connection =
1848 			first_peer_device(device)->connection;
1849 		struct drbd_work_queue *q = &connection->sender_work;
1850 		if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags))
1851 			wake_up(&q->q_wait);
1852 	}
1853 }
1854 
1855 extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue);
1856 
1857 /* To get the ack_receiver out of the blocking network stack,
1858  * so it can change its sk_rcvtimeo from idle- to ping-timeout,
1859  * and send a ping, we need to send a signal.
1860  * Which signal we send is irrelevant. */
1861 static inline void wake_ack_receiver(struct drbd_connection *connection)
1862 {
1863 	struct task_struct *task = connection->ack_receiver.task;
1864 	if (task && get_t_state(&connection->ack_receiver) == RUNNING)
1865 		send_sig(SIGXCPU, task, 1);
1866 }
1867 
1868 static inline void request_ping(struct drbd_connection *connection)
1869 {
1870 	set_bit(SEND_PING, &connection->flags);
1871 	wake_ack_receiver(connection);
1872 }
1873 
1874 extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *);
1875 extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *);
1876 extern int conn_send_command(struct drbd_connection *, struct drbd_socket *,
1877 			     enum drbd_packet, unsigned int, void *,
1878 			     unsigned int);
1879 extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *,
1880 			     enum drbd_packet, unsigned int, void *,
1881 			     unsigned int);
1882 
1883 extern int drbd_send_ping(struct drbd_connection *connection);
1884 extern int drbd_send_ping_ack(struct drbd_connection *connection);
1885 extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state);
1886 extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state);
1887 
1888 static inline void drbd_thread_stop(struct drbd_thread *thi)
1889 {
1890 	_drbd_thread_stop(thi, false, true);
1891 }
1892 
1893 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
1894 {
1895 	_drbd_thread_stop(thi, false, false);
1896 }
1897 
1898 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
1899 {
1900 	_drbd_thread_stop(thi, true, false);
1901 }
1902 
1903 /* counts how many answer packets packets we expect from our peer,
1904  * for either explicit application requests,
1905  * or implicit barrier packets as necessary.
1906  * increased:
1907  *  w_send_barrier
1908  *  _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1909  *    it is much easier and equally valid to count what we queue for the
1910  *    worker, even before it actually was queued or send.
1911  *    (drbd_make_request_common; recovery path on read io-error)
1912  * decreased:
1913  *  got_BarrierAck (respective tl_clear, tl_clear_barrier)
1914  *  _req_mod(req, DATA_RECEIVED)
1915  *     [from receive_DataReply]
1916  *  _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1917  *     [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1918  *     for some reason it is NOT decreased in got_NegAck,
1919  *     but in the resulting cleanup code from report_params.
1920  *     we should try to remember the reason for that...
1921  *  _req_mod(req, SEND_FAILED or SEND_CANCELED)
1922  *  _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
1923  *     [from tl_clear_barrier]
1924  */
1925 static inline void inc_ap_pending(struct drbd_device *device)
1926 {
1927 	atomic_inc(&device->ap_pending_cnt);
1928 }
1929 
1930 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line)			\
1931 	if (atomic_read(&device->which) < 0)				\
1932 		drbd_err(device, "in %s:%d: " #which " = %d < 0 !\n",	\
1933 			func, line,					\
1934 			atomic_read(&device->which))
1935 
1936 #define dec_ap_pending(device) _dec_ap_pending(device, __func__, __LINE__)
1937 static inline void _dec_ap_pending(struct drbd_device *device, const char *func, int line)
1938 {
1939 	if (atomic_dec_and_test(&device->ap_pending_cnt))
1940 		wake_up(&device->misc_wait);
1941 	ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line);
1942 }
1943 
1944 /* counts how many resync-related answers we still expect from the peer
1945  *		     increase			decrease
1946  * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1947  * C_SYNC_SOURCE sends P_RS_DATA_REPLY   (and expects P_WRITE_ACK with ID_SYNCER)
1948  *					   (or P_NEG_ACK with ID_SYNCER)
1949  */
1950 static inline void inc_rs_pending(struct drbd_device *device)
1951 {
1952 	atomic_inc(&device->rs_pending_cnt);
1953 }
1954 
1955 #define dec_rs_pending(device) _dec_rs_pending(device, __func__, __LINE__)
1956 static inline void _dec_rs_pending(struct drbd_device *device, const char *func, int line)
1957 {
1958 	atomic_dec(&device->rs_pending_cnt);
1959 	ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line);
1960 }
1961 
1962 /* counts how many answers we still need to send to the peer.
1963  * increased on
1964  *  receive_Data	unless protocol A;
1965  *			we need to send a P_RECV_ACK (proto B)
1966  *			or P_WRITE_ACK (proto C)
1967  *  receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1968  *  receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1969  *  receive_Barrier_*	we need to send a P_BARRIER_ACK
1970  */
1971 static inline void inc_unacked(struct drbd_device *device)
1972 {
1973 	atomic_inc(&device->unacked_cnt);
1974 }
1975 
1976 #define dec_unacked(device) _dec_unacked(device, __func__, __LINE__)
1977 static inline void _dec_unacked(struct drbd_device *device, const char *func, int line)
1978 {
1979 	atomic_dec(&device->unacked_cnt);
1980 	ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
1981 }
1982 
1983 #define sub_unacked(device, n) _sub_unacked(device, n, __func__, __LINE__)
1984 static inline void _sub_unacked(struct drbd_device *device, int n, const char *func, int line)
1985 {
1986 	atomic_sub(n, &device->unacked_cnt);
1987 	ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
1988 }
1989 
1990 static inline bool is_sync_target_state(enum drbd_conns connection_state)
1991 {
1992 	return	connection_state == C_SYNC_TARGET ||
1993 		connection_state == C_PAUSED_SYNC_T;
1994 }
1995 
1996 static inline bool is_sync_source_state(enum drbd_conns connection_state)
1997 {
1998 	return	connection_state == C_SYNC_SOURCE ||
1999 		connection_state == C_PAUSED_SYNC_S;
2000 }
2001 
2002 static inline bool is_sync_state(enum drbd_conns connection_state)
2003 {
2004 	return	is_sync_source_state(connection_state) ||
2005 		is_sync_target_state(connection_state);
2006 }
2007 
2008 /**
2009  * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev
2010  * @_device:		DRBD device.
2011  * @_min_state:		Minimum device state required for success.
2012  *
2013  * You have to call put_ldev() when finished working with device->ldev.
2014  */
2015 #define get_ldev_if_state(_device, _min_state)				\
2016 	(_get_ldev_if_state((_device), (_min_state)) ?			\
2017 	 ({ __acquire(x); true; }) : false)
2018 #define get_ldev(_device) get_ldev_if_state(_device, D_INCONSISTENT)
2019 
2020 static inline void put_ldev(struct drbd_device *device)
2021 {
2022 	enum drbd_disk_state disk_state = device->state.disk;
2023 	/* We must check the state *before* the atomic_dec becomes visible,
2024 	 * or we have a theoretical race where someone hitting zero,
2025 	 * while state still D_FAILED, will then see D_DISKLESS in the
2026 	 * condition below and calling into destroy, where he must not, yet. */
2027 	int i = atomic_dec_return(&device->local_cnt);
2028 
2029 	/* This may be called from some endio handler,
2030 	 * so we must not sleep here. */
2031 
2032 	__release(local);
2033 	D_ASSERT(device, i >= 0);
2034 	if (i == 0) {
2035 		if (disk_state == D_DISKLESS)
2036 			/* even internal references gone, safe to destroy */
2037 			drbd_device_post_work(device, DESTROY_DISK);
2038 		if (disk_state == D_FAILED)
2039 			/* all application IO references gone. */
2040 			if (!test_and_set_bit(GOING_DISKLESS, &device->flags))
2041 				drbd_device_post_work(device, GO_DISKLESS);
2042 		wake_up(&device->misc_wait);
2043 	}
2044 }
2045 
2046 #ifndef __CHECKER__
2047 static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins)
2048 {
2049 	int io_allowed;
2050 
2051 	/* never get a reference while D_DISKLESS */
2052 	if (device->state.disk == D_DISKLESS)
2053 		return 0;
2054 
2055 	atomic_inc(&device->local_cnt);
2056 	io_allowed = (device->state.disk >= mins);
2057 	if (!io_allowed)
2058 		put_ldev(device);
2059 	return io_allowed;
2060 }
2061 #else
2062 extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins);
2063 #endif
2064 
2065 /* this throttles on-the-fly application requests
2066  * according to max_buffers settings;
2067  * maybe re-implement using semaphores? */
2068 static inline int drbd_get_max_buffers(struct drbd_device *device)
2069 {
2070 	struct net_conf *nc;
2071 	int mxb;
2072 
2073 	rcu_read_lock();
2074 	nc = rcu_dereference(first_peer_device(device)->connection->net_conf);
2075 	mxb = nc ? nc->max_buffers : 1000000;  /* arbitrary limit on open requests */
2076 	rcu_read_unlock();
2077 
2078 	return mxb;
2079 }
2080 
2081 static inline int drbd_state_is_stable(struct drbd_device *device)
2082 {
2083 	union drbd_dev_state s = device->state;
2084 
2085 	/* DO NOT add a default clause, we want the compiler to warn us
2086 	 * for any newly introduced state we may have forgotten to add here */
2087 
2088 	switch ((enum drbd_conns)s.conn) {
2089 	/* new io only accepted when there is no connection, ... */
2090 	case C_STANDALONE:
2091 	case C_WF_CONNECTION:
2092 	/* ... or there is a well established connection. */
2093 	case C_CONNECTED:
2094 	case C_SYNC_SOURCE:
2095 	case C_SYNC_TARGET:
2096 	case C_VERIFY_S:
2097 	case C_VERIFY_T:
2098 	case C_PAUSED_SYNC_S:
2099 	case C_PAUSED_SYNC_T:
2100 	case C_AHEAD:
2101 	case C_BEHIND:
2102 		/* transitional states, IO allowed */
2103 	case C_DISCONNECTING:
2104 	case C_UNCONNECTED:
2105 	case C_TIMEOUT:
2106 	case C_BROKEN_PIPE:
2107 	case C_NETWORK_FAILURE:
2108 	case C_PROTOCOL_ERROR:
2109 	case C_TEAR_DOWN:
2110 	case C_WF_REPORT_PARAMS:
2111 	case C_STARTING_SYNC_S:
2112 	case C_STARTING_SYNC_T:
2113 		break;
2114 
2115 		/* Allow IO in BM exchange states with new protocols */
2116 	case C_WF_BITMAP_S:
2117 		if (first_peer_device(device)->connection->agreed_pro_version < 96)
2118 			return 0;
2119 		break;
2120 
2121 		/* no new io accepted in these states */
2122 	case C_WF_BITMAP_T:
2123 	case C_WF_SYNC_UUID:
2124 	case C_MASK:
2125 		/* not "stable" */
2126 		return 0;
2127 	}
2128 
2129 	switch ((enum drbd_disk_state)s.disk) {
2130 	case D_DISKLESS:
2131 	case D_INCONSISTENT:
2132 	case D_OUTDATED:
2133 	case D_CONSISTENT:
2134 	case D_UP_TO_DATE:
2135 	case D_FAILED:
2136 		/* disk state is stable as well. */
2137 		break;
2138 
2139 	/* no new io accepted during transitional states */
2140 	case D_ATTACHING:
2141 	case D_NEGOTIATING:
2142 	case D_UNKNOWN:
2143 	case D_MASK:
2144 		/* not "stable" */
2145 		return 0;
2146 	}
2147 
2148 	return 1;
2149 }
2150 
2151 static inline int drbd_suspended(struct drbd_device *device)
2152 {
2153 	struct drbd_resource *resource = device->resource;
2154 
2155 	return resource->susp || resource->susp_fen || resource->susp_nod;
2156 }
2157 
2158 static inline bool may_inc_ap_bio(struct drbd_device *device)
2159 {
2160 	int mxb = drbd_get_max_buffers(device);
2161 
2162 	if (drbd_suspended(device))
2163 		return false;
2164 	if (atomic_read(&device->suspend_cnt))
2165 		return false;
2166 
2167 	/* to avoid potential deadlock or bitmap corruption,
2168 	 * in various places, we only allow new application io
2169 	 * to start during "stable" states. */
2170 
2171 	/* no new io accepted when attaching or detaching the disk */
2172 	if (!drbd_state_is_stable(device))
2173 		return false;
2174 
2175 	/* since some older kernels don't have atomic_add_unless,
2176 	 * and we are within the spinlock anyways, we have this workaround.  */
2177 	if (atomic_read(&device->ap_bio_cnt) > mxb)
2178 		return false;
2179 	if (test_bit(BITMAP_IO, &device->flags))
2180 		return false;
2181 	return true;
2182 }
2183 
2184 static inline bool inc_ap_bio_cond(struct drbd_device *device)
2185 {
2186 	bool rv = false;
2187 
2188 	spin_lock_irq(&device->resource->req_lock);
2189 	rv = may_inc_ap_bio(device);
2190 	if (rv)
2191 		atomic_inc(&device->ap_bio_cnt);
2192 	spin_unlock_irq(&device->resource->req_lock);
2193 
2194 	return rv;
2195 }
2196 
2197 static inline void inc_ap_bio(struct drbd_device *device)
2198 {
2199 	/* we wait here
2200 	 *    as long as the device is suspended
2201 	 *    until the bitmap is no longer on the fly during connection
2202 	 *    handshake as long as we would exceed the max_buffer limit.
2203 	 *
2204 	 * to avoid races with the reconnect code,
2205 	 * we need to atomic_inc within the spinlock. */
2206 
2207 	wait_event(device->misc_wait, inc_ap_bio_cond(device));
2208 }
2209 
2210 static inline void dec_ap_bio(struct drbd_device *device)
2211 {
2212 	int mxb = drbd_get_max_buffers(device);
2213 	int ap_bio = atomic_dec_return(&device->ap_bio_cnt);
2214 
2215 	D_ASSERT(device, ap_bio >= 0);
2216 
2217 	if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) {
2218 		if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags))
2219 			drbd_queue_work(&first_peer_device(device)->
2220 				connection->sender_work,
2221 				&device->bm_io_work.w);
2222 	}
2223 
2224 	/* this currently does wake_up for every dec_ap_bio!
2225 	 * maybe rather introduce some type of hysteresis?
2226 	 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2227 	if (ap_bio < mxb)
2228 		wake_up(&device->misc_wait);
2229 }
2230 
2231 static inline bool verify_can_do_stop_sector(struct drbd_device *device)
2232 {
2233 	return first_peer_device(device)->connection->agreed_pro_version >= 97 &&
2234 		first_peer_device(device)->connection->agreed_pro_version != 100;
2235 }
2236 
2237 static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val)
2238 {
2239 	int changed = device->ed_uuid != val;
2240 	device->ed_uuid = val;
2241 	return changed;
2242 }
2243 
2244 static inline int drbd_queue_order_type(struct drbd_device *device)
2245 {
2246 	/* sorry, we currently have no working implementation
2247 	 * of distributed TCQ stuff */
2248 #ifndef QUEUE_ORDERED_NONE
2249 #define QUEUE_ORDERED_NONE 0
2250 #endif
2251 	return QUEUE_ORDERED_NONE;
2252 }
2253 
2254 static inline struct drbd_connection *first_connection(struct drbd_resource *resource)
2255 {
2256 	return list_first_entry_or_null(&resource->connections,
2257 				struct drbd_connection, connections);
2258 }
2259 
2260 #endif
2261