xref: /linux/drivers/block/drbd/drbd_int.h (revision d6d224429a86a62263d0944f79c36dce010a4ebb)
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
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/genhd.h>
31 #include <linux/idr.h>
32 #include <linux/dynamic_debug.h>
33 #include <net/tcp.h>
34 #include <linux/lru_cache.h>
35 #include <linux/prefetch.h>
36 #include <linux/drbd_genl_api.h>
37 #include <linux/drbd.h>
38 #include "drbd_strings.h"
39 #include "drbd_state.h"
40 #include "drbd_protocol.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 #define __drbd_printk_device(level, device, fmt, args...) \
80 	dev_printk(level, disk_to_dev((device)->vdisk), fmt, ## args)
81 #define __drbd_printk_peer_device(level, peer_device, fmt, args...) \
82 	dev_printk(level, disk_to_dev((peer_device)->device->vdisk), fmt, ## args)
83 #define __drbd_printk_resource(level, resource, fmt, args...) \
84 	printk(level "drbd %s: " fmt, (resource)->name, ## args)
85 #define __drbd_printk_connection(level, connection, fmt, args...) \
86 	printk(level "drbd %s: " fmt, (connection)->resource->name, ## args)
87 
88 void drbd_printk_with_wrong_object_type(void);
89 
90 #define __drbd_printk_if_same_type(obj, type, func, level, fmt, args...) \
91 	(__builtin_types_compatible_p(typeof(obj), type) || \
92 	 __builtin_types_compatible_p(typeof(obj), const type)), \
93 	func(level, (const type)(obj), fmt, ## args)
94 
95 #define drbd_printk(level, obj, fmt, args...) \
96 	__builtin_choose_expr( \
97 	  __drbd_printk_if_same_type(obj, struct drbd_device *, \
98 			     __drbd_printk_device, level, fmt, ## args), \
99 	  __builtin_choose_expr( \
100 	    __drbd_printk_if_same_type(obj, struct drbd_resource *, \
101 			       __drbd_printk_resource, level, fmt, ## args), \
102 	    __builtin_choose_expr( \
103 	      __drbd_printk_if_same_type(obj, struct drbd_connection *, \
104 				 __drbd_printk_connection, level, fmt, ## args), \
105 	      __builtin_choose_expr( \
106 		__drbd_printk_if_same_type(obj, struct drbd_peer_device *, \
107 				 __drbd_printk_peer_device, level, fmt, ## args), \
108 		drbd_printk_with_wrong_object_type()))))
109 
110 #define drbd_dbg(obj, fmt, args...) \
111 	drbd_printk(KERN_DEBUG, obj, fmt, ## args)
112 #define drbd_alert(obj, fmt, args...) \
113 	drbd_printk(KERN_ALERT, obj, fmt, ## args)
114 #define drbd_err(obj, fmt, args...) \
115 	drbd_printk(KERN_ERR, obj, fmt, ## args)
116 #define drbd_warn(obj, fmt, args...) \
117 	drbd_printk(KERN_WARNING, obj, fmt, ## args)
118 #define drbd_info(obj, fmt, args...) \
119 	drbd_printk(KERN_INFO, obj, fmt, ## args)
120 #define drbd_emerg(obj, fmt, args...) \
121 	drbd_printk(KERN_EMERG, obj, fmt, ## args)
122 
123 #define dynamic_drbd_dbg(device, fmt, args...) \
124 	dynamic_dev_dbg(disk_to_dev(device->vdisk), fmt, ## args)
125 
126 #define D_ASSERT(device, exp)	do { \
127 	if (!(exp)) \
128 		drbd_err(device, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__); \
129 	} while (0)
130 
131 /**
132  * expect  -  Make an assertion
133  *
134  * Unlike the assert macro, this macro returns a boolean result.
135  */
136 #define expect(exp) ({								\
137 		bool _bool = (exp);						\
138 		if (!_bool)							\
139 			drbd_err(device, "ASSERTION %s FAILED in %s\n",		\
140 			        #exp, __func__);				\
141 		_bool;								\
142 		})
143 
144 /* Defines to control fault insertion */
145 enum {
146 	DRBD_FAULT_MD_WR = 0,	/* meta data write */
147 	DRBD_FAULT_MD_RD = 1,	/*           read  */
148 	DRBD_FAULT_RS_WR = 2,	/* resync          */
149 	DRBD_FAULT_RS_RD = 3,
150 	DRBD_FAULT_DT_WR = 4,	/* data            */
151 	DRBD_FAULT_DT_RD = 5,
152 	DRBD_FAULT_DT_RA = 6,	/* data read ahead */
153 	DRBD_FAULT_BM_ALLOC = 7,	/* bitmap allocation */
154 	DRBD_FAULT_AL_EE = 8,	/* alloc ee */
155 	DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
156 
157 	DRBD_FAULT_MAX,
158 };
159 
160 extern unsigned int
161 _drbd_insert_fault(struct drbd_device *device, unsigned int type);
162 
163 static inline int
164 drbd_insert_fault(struct drbd_device *device, unsigned int type) {
165 #ifdef CONFIG_DRBD_FAULT_INJECTION
166 	return drbd_fault_rate &&
167 		(drbd_enable_faults & (1<<type)) &&
168 		_drbd_insert_fault(device, type);
169 #else
170 	return 0;
171 #endif
172 }
173 
174 /* integer division, round _UP_ to the next integer */
175 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
176 /* usual integer division */
177 #define div_floor(A, B) ((A)/(B))
178 
179 extern struct ratelimit_state drbd_ratelimit_state;
180 extern struct idr drbd_devices; /* RCU, updates: genl_lock() */
181 extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */
182 
183 extern const char *cmdname(enum drbd_packet cmd);
184 
185 /* for sending/receiving the bitmap,
186  * possibly in some encoding scheme */
187 struct bm_xfer_ctx {
188 	/* "const"
189 	 * stores total bits and long words
190 	 * of the bitmap, so we don't need to
191 	 * call the accessor functions over and again. */
192 	unsigned long bm_bits;
193 	unsigned long bm_words;
194 	/* during xfer, current position within the bitmap */
195 	unsigned long bit_offset;
196 	unsigned long word_offset;
197 
198 	/* statistics; index: (h->command == P_BITMAP) */
199 	unsigned packets[2];
200 	unsigned bytes[2];
201 };
202 
203 extern void INFO_bm_xfer_stats(struct drbd_device *device,
204 		const char *direction, struct bm_xfer_ctx *c);
205 
206 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
207 {
208 	/* word_offset counts "native long words" (32 or 64 bit),
209 	 * aligned at 64 bit.
210 	 * Encoded packet may end at an unaligned bit offset.
211 	 * In case a fallback clear text packet is transmitted in
212 	 * between, we adjust this offset back to the last 64bit
213 	 * aligned "native long word", which makes coding and decoding
214 	 * the plain text bitmap much more convenient.  */
215 #if BITS_PER_LONG == 64
216 	c->word_offset = c->bit_offset >> 6;
217 #elif BITS_PER_LONG == 32
218 	c->word_offset = c->bit_offset >> 5;
219 	c->word_offset &= ~(1UL);
220 #else
221 # error "unsupported BITS_PER_LONG"
222 #endif
223 }
224 
225 extern unsigned int drbd_header_size(struct drbd_connection *connection);
226 
227 /**********************************************************************/
228 enum drbd_thread_state {
229 	NONE,
230 	RUNNING,
231 	EXITING,
232 	RESTARTING
233 };
234 
235 struct drbd_thread {
236 	spinlock_t t_lock;
237 	struct task_struct *task;
238 	struct completion stop;
239 	enum drbd_thread_state t_state;
240 	int (*function) (struct drbd_thread *);
241 	struct drbd_resource *resource;
242 	struct drbd_connection *connection;
243 	int reset_cpu_mask;
244 	const char *name;
245 };
246 
247 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
248 {
249 	/* THINK testing the t_state seems to be uncritical in all cases
250 	 * (but thread_{start,stop}), so we can read it *without* the lock.
251 	 *	--lge */
252 
253 	smp_rmb();
254 	return thi->t_state;
255 }
256 
257 struct drbd_work {
258 	struct list_head list;
259 	int (*cb)(struct drbd_work *, int cancel);
260 };
261 
262 struct drbd_device_work {
263 	struct drbd_work w;
264 	struct drbd_device *device;
265 };
266 
267 #include "drbd_interval.h"
268 
269 extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *);
270 
271 extern void lock_all_resources(void);
272 extern void unlock_all_resources(void);
273 
274 struct drbd_request {
275 	struct drbd_work w;
276 	struct drbd_device *device;
277 
278 	/* if local IO is not allowed, will be NULL.
279 	 * if local IO _is_ allowed, holds the locally submitted bio clone,
280 	 * or, after local IO completion, the ERR_PTR(error).
281 	 * see drbd_request_endio(). */
282 	struct bio *private_bio;
283 
284 	struct drbd_interval i;
285 
286 	/* epoch: used to check on "completion" whether this req was in
287 	 * the current epoch, and we therefore have to close it,
288 	 * causing a p_barrier packet to be send, starting a new epoch.
289 	 *
290 	 * This corresponds to "barrier" in struct p_barrier[_ack],
291 	 * and to "barrier_nr" in struct drbd_epoch (and various
292 	 * comments/function parameters/local variable names).
293 	 */
294 	unsigned int epoch;
295 
296 	struct list_head tl_requests; /* ring list in the transfer log */
297 	struct bio *master_bio;       /* master bio pointer */
298 
299 	/* see struct drbd_device */
300 	struct list_head req_pending_master_completion;
301 	struct list_head req_pending_local;
302 
303 	/* for generic IO accounting */
304 	unsigned long start_jif;
305 
306 	/* for DRBD internal statistics */
307 
308 	/* Minimal set of time stamps to determine if we wait for activity log
309 	 * transactions, local disk or peer.  32 bit "jiffies" are good enough,
310 	 * we don't expect a DRBD request to be stalled for several month.
311 	 */
312 
313 	/* before actual request processing */
314 	unsigned long in_actlog_jif;
315 
316 	/* local disk */
317 	unsigned long pre_submit_jif;
318 
319 	/* per connection */
320 	unsigned long pre_send_jif;
321 	unsigned long acked_jif;
322 	unsigned long net_done_jif;
323 
324 	/* Possibly even more detail to track each phase:
325 	 *  master_completion_jif
326 	 *      how long did it take to complete the master bio
327 	 *      (application visible latency)
328 	 *  allocated_jif
329 	 *      how long the master bio was blocked until we finally allocated
330 	 *      a tracking struct
331 	 *  in_actlog_jif
332 	 *      how long did we wait for activity log transactions
333 	 *
334 	 *  net_queued_jif
335 	 *      when did we finally queue it for sending
336 	 *  pre_send_jif
337 	 *      when did we start sending it
338 	 *  post_send_jif
339 	 *      how long did we block in the network stack trying to send it
340 	 *  acked_jif
341 	 *      when did we receive (or fake, in protocol A) a remote ACK
342 	 *  net_done_jif
343 	 *      when did we receive final acknowledgement (P_BARRIER_ACK),
344 	 *      or decide, e.g. on connection loss, that we do no longer expect
345 	 *      anything from this peer for this request.
346 	 *
347 	 *  pre_submit_jif
348 	 *  post_sub_jif
349 	 *      when did we start submiting to the lower level device,
350 	 *      and how long did we block in that submit function
351 	 *  local_completion_jif
352 	 *      how long did it take the lower level device to complete this request
353 	 */
354 
355 
356 	/* once it hits 0, we may complete the master_bio */
357 	atomic_t completion_ref;
358 	/* once it hits 0, we may destroy this drbd_request object */
359 	struct kref kref;
360 
361 	unsigned rq_state; /* see comments above _req_mod() */
362 };
363 
364 struct drbd_epoch {
365 	struct drbd_connection *connection;
366 	struct list_head list;
367 	unsigned int barrier_nr;
368 	atomic_t epoch_size; /* increased on every request added. */
369 	atomic_t active;     /* increased on every req. added, and dec on every finished. */
370 	unsigned long flags;
371 };
372 
373 /* Prototype declaration of function defined in drbd_receiver.c */
374 int drbdd_init(struct drbd_thread *);
375 int drbd_asender(struct drbd_thread *);
376 
377 /* drbd_epoch flag bits */
378 enum {
379 	DE_HAVE_BARRIER_NUMBER,
380 };
381 
382 enum epoch_event {
383 	EV_PUT,
384 	EV_GOT_BARRIER_NR,
385 	EV_BECAME_LAST,
386 	EV_CLEANUP = 32, /* used as flag */
387 };
388 
389 struct digest_info {
390 	int digest_size;
391 	void *digest;
392 };
393 
394 struct drbd_peer_request {
395 	struct drbd_work w;
396 	struct drbd_peer_device *peer_device;
397 	struct drbd_epoch *epoch; /* for writes */
398 	struct page *pages;
399 	atomic_t pending_bios;
400 	struct drbd_interval i;
401 	/* see comments on ee flag bits below */
402 	unsigned long flags;
403 	unsigned long submit_jif;
404 	union {
405 		u64 block_id;
406 		struct digest_info *digest;
407 	};
408 };
409 
410 /* ee flag bits.
411  * While corresponding bios are in flight, the only modification will be
412  * set_bit WAS_ERROR, which has to be atomic.
413  * If no bios are in flight yet, or all have been completed,
414  * non-atomic modification to ee->flags is ok.
415  */
416 enum {
417 	__EE_CALL_AL_COMPLETE_IO,
418 	__EE_MAY_SET_IN_SYNC,
419 
420 	/* is this a TRIM aka REQ_OP_DISCARD? */
421 	__EE_TRIM,
422 	/* explicit zero-out requested, or
423 	 * our lower level cannot handle trim,
424 	 * and we want to fall back to zeroout instead */
425 	__EE_ZEROOUT,
426 
427 	/* In case a barrier failed,
428 	 * we need to resubmit without the barrier flag. */
429 	__EE_RESUBMITTED,
430 
431 	/* we may have several bios per peer request.
432 	 * if any of those fail, we set this flag atomically
433 	 * from the endio callback */
434 	__EE_WAS_ERROR,
435 
436 	/* This ee has a pointer to a digest instead of a block id */
437 	__EE_HAS_DIGEST,
438 
439 	/* Conflicting local requests need to be restarted after this request */
440 	__EE_RESTART_REQUESTS,
441 
442 	/* The peer wants a write ACK for this (wire proto C) */
443 	__EE_SEND_WRITE_ACK,
444 
445 	/* Is set when net_conf had two_primaries set while creating this peer_req */
446 	__EE_IN_INTERVAL_TREE,
447 
448 	/* for debugfs: */
449 	/* has this been submitted, or does it still wait for something else? */
450 	__EE_SUBMITTED,
451 
452 	/* this is/was a write request */
453 	__EE_WRITE,
454 
455 	/* this is/was a write same request */
456 	__EE_WRITE_SAME,
457 
458 	/* this originates from application on peer
459 	 * (not some resync or verify or other DRBD internal request) */
460 	__EE_APPLICATION,
461 
462 	/* If it contains only 0 bytes, send back P_RS_DEALLOCATED */
463 	__EE_RS_THIN_REQ,
464 };
465 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
466 #define EE_MAY_SET_IN_SYNC     (1<<__EE_MAY_SET_IN_SYNC)
467 #define EE_TRIM                (1<<__EE_TRIM)
468 #define EE_ZEROOUT             (1<<__EE_ZEROOUT)
469 #define EE_RESUBMITTED         (1<<__EE_RESUBMITTED)
470 #define EE_WAS_ERROR           (1<<__EE_WAS_ERROR)
471 #define EE_HAS_DIGEST          (1<<__EE_HAS_DIGEST)
472 #define EE_RESTART_REQUESTS	(1<<__EE_RESTART_REQUESTS)
473 #define EE_SEND_WRITE_ACK	(1<<__EE_SEND_WRITE_ACK)
474 #define EE_IN_INTERVAL_TREE	(1<<__EE_IN_INTERVAL_TREE)
475 #define EE_SUBMITTED		(1<<__EE_SUBMITTED)
476 #define EE_WRITE		(1<<__EE_WRITE)
477 #define EE_WRITE_SAME		(1<<__EE_WRITE_SAME)
478 #define EE_APPLICATION		(1<<__EE_APPLICATION)
479 #define EE_RS_THIN_REQ		(1<<__EE_RS_THIN_REQ)
480 
481 /* flag bits per device */
482 enum {
483 	UNPLUG_REMOTE,		/* sending a "UnplugRemote" could help */
484 	MD_DIRTY,		/* current uuids and flags not yet on disk */
485 	USE_DEGR_WFC_T,		/* degr-wfc-timeout instead of wfc-timeout. */
486 	CL_ST_CHG_SUCCESS,
487 	CL_ST_CHG_FAIL,
488 	CRASHED_PRIMARY,	/* This node was a crashed primary.
489 				 * Gets cleared when the state.conn
490 				 * goes into C_CONNECTED state. */
491 	CONSIDER_RESYNC,
492 
493 	MD_NO_FUA,		/* Users wants us to not use FUA/FLUSH on meta data dev */
494 
495 	BITMAP_IO,		/* suspend application io;
496 				   once no more io in flight, start bitmap io */
497 	BITMAP_IO_QUEUED,       /* Started bitmap IO */
498 	WAS_IO_ERROR,		/* Local disk failed, returned IO error */
499 	WAS_READ_ERROR,		/* Local disk READ failed (set additionally to the above) */
500 	FORCE_DETACH,		/* Force-detach from local disk, aborting any pending local IO */
501 	RESYNC_AFTER_NEG,       /* Resync after online grow after the attach&negotiate finished. */
502 	RESIZE_PENDING,		/* Size change detected locally, waiting for the response from
503 				 * the peer, if it changed there as well. */
504 	NEW_CUR_UUID,		/* Create new current UUID when thawing IO */
505 	AL_SUSPENDED,		/* Activity logging is currently suspended. */
506 	AHEAD_TO_SYNC_SOURCE,   /* Ahead -> SyncSource queued */
507 	B_RS_H_DONE,		/* Before resync handler done (already executed) */
508 	DISCARD_MY_DATA,	/* discard_my_data flag per volume */
509 	READ_BALANCE_RR,
510 
511 	FLUSH_PENDING,		/* if set, device->flush_jif is when we submitted that flush
512 				 * from drbd_flush_after_epoch() */
513 
514 	/* cleared only after backing device related structures have been destroyed. */
515 	GOING_DISKLESS,		/* Disk is being detached, because of io-error, or admin request. */
516 
517 	/* to be used in drbd_device_post_work() */
518 	GO_DISKLESS,		/* tell worker to schedule cleanup before detach */
519 	DESTROY_DISK,		/* tell worker to close backing devices and destroy related structures. */
520 	MD_SYNC,		/* tell worker to call drbd_md_sync() */
521 	RS_START,		/* tell worker to start resync/OV */
522 	RS_PROGRESS,		/* tell worker that resync made significant progress */
523 	RS_DONE,		/* tell worker that resync is done */
524 };
525 
526 struct drbd_bitmap; /* opaque for drbd_device */
527 
528 /* definition of bits in bm_flags to be used in drbd_bm_lock
529  * and drbd_bitmap_io and friends. */
530 enum bm_flag {
531 	/* currently locked for bulk operation */
532 	BM_LOCKED_MASK = 0xf,
533 
534 	/* in detail, that is: */
535 	BM_DONT_CLEAR = 0x1,
536 	BM_DONT_SET   = 0x2,
537 	BM_DONT_TEST  = 0x4,
538 
539 	/* so we can mark it locked for bulk operation,
540 	 * and still allow all non-bulk operations */
541 	BM_IS_LOCKED  = 0x8,
542 
543 	/* (test bit, count bit) allowed (common case) */
544 	BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED,
545 
546 	/* testing bits, as well as setting new bits allowed, but clearing bits
547 	 * would be unexpected.  Used during bitmap receive.  Setting new bits
548 	 * requires sending of "out-of-sync" information, though. */
549 	BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED,
550 
551 	/* for drbd_bm_write_copy_pages, everything is allowed,
552 	 * only concurrent bulk operations are locked out. */
553 	BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED,
554 };
555 
556 struct drbd_work_queue {
557 	struct list_head q;
558 	spinlock_t q_lock;  /* to protect the list. */
559 	wait_queue_head_t q_wait;
560 };
561 
562 struct drbd_socket {
563 	struct mutex mutex;
564 	struct socket    *socket;
565 	/* this way we get our
566 	 * send/receive buffers off the stack */
567 	void *sbuf;
568 	void *rbuf;
569 };
570 
571 struct drbd_md {
572 	u64 md_offset;		/* sector offset to 'super' block */
573 
574 	u64 la_size_sect;	/* last agreed size, unit sectors */
575 	spinlock_t uuid_lock;
576 	u64 uuid[UI_SIZE];
577 	u64 device_uuid;
578 	u32 flags;
579 	u32 md_size_sect;
580 
581 	s32 al_offset;	/* signed relative sector offset to activity log */
582 	s32 bm_offset;	/* signed relative sector offset to bitmap */
583 
584 	/* cached value of bdev->disk_conf->meta_dev_idx (see below) */
585 	s32 meta_dev_idx;
586 
587 	/* see al_tr_number_to_on_disk_sector() */
588 	u32 al_stripes;
589 	u32 al_stripe_size_4k;
590 	u32 al_size_4k; /* cached product of the above */
591 };
592 
593 struct drbd_backing_dev {
594 	struct block_device *backing_bdev;
595 	struct block_device *md_bdev;
596 	struct drbd_md md;
597 	struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */
598 	sector_t known_size; /* last known size of that backing device */
599 };
600 
601 struct drbd_md_io {
602 	struct page *page;
603 	unsigned long start_jif;	/* last call to drbd_md_get_buffer */
604 	unsigned long submit_jif;	/* last _drbd_md_sync_page_io() submit */
605 	const char *current_use;
606 	atomic_t in_use;
607 	unsigned int done;
608 	int error;
609 };
610 
611 struct bm_io_work {
612 	struct drbd_work w;
613 	char *why;
614 	enum bm_flag flags;
615 	int (*io_fn)(struct drbd_device *device);
616 	void (*done)(struct drbd_device *device, int rv);
617 };
618 
619 struct fifo_buffer {
620 	unsigned int head_index;
621 	unsigned int size;
622 	int total; /* sum of all values */
623 	int values[];
624 };
625 extern struct fifo_buffer *fifo_alloc(unsigned int fifo_size);
626 
627 /* flag bits per connection */
628 enum {
629 	NET_CONGESTED,		/* The data socket is congested */
630 	RESOLVE_CONFLICTS,	/* Set on one node, cleared on the peer! */
631 	SEND_PING,
632 	GOT_PING_ACK,		/* set when we receive a ping_ack packet, ping_wait gets woken */
633 	CONN_WD_ST_CHG_REQ,	/* A cluster wide state change on the connection is active */
634 	CONN_WD_ST_CHG_OKAY,
635 	CONN_WD_ST_CHG_FAIL,
636 	CONN_DRY_RUN,		/* Expect disconnect after resync handshake. */
637 	CREATE_BARRIER,		/* next P_DATA is preceded by a P_BARRIER */
638 	STATE_SENT,		/* Do not change state/UUIDs while this is set */
639 	CALLBACK_PENDING,	/* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
640 				 * pending, from drbd worker context.
641 				 */
642 	DISCONNECT_SENT,
643 
644 	DEVICE_WORK_PENDING,	/* tell worker that some device has pending work */
645 };
646 
647 enum which_state { NOW, OLD = NOW, NEW };
648 
649 struct drbd_resource {
650 	char *name;
651 #ifdef CONFIG_DEBUG_FS
652 	struct dentry *debugfs_res;
653 	struct dentry *debugfs_res_volumes;
654 	struct dentry *debugfs_res_connections;
655 	struct dentry *debugfs_res_in_flight_summary;
656 #endif
657 	struct kref kref;
658 	struct idr devices;		/* volume number to device mapping */
659 	struct list_head connections;
660 	struct list_head resources;
661 	struct res_opts res_opts;
662 	struct mutex conf_update;	/* mutex for ready-copy-update of net_conf and disk_conf */
663 	struct mutex adm_mutex;		/* mutex to serialize administrative requests */
664 	spinlock_t req_lock;
665 
666 	unsigned susp:1;		/* IO suspended by user */
667 	unsigned susp_nod:1;		/* IO suspended because no data */
668 	unsigned susp_fen:1;		/* IO suspended because fence peer handler runs */
669 
670 	enum write_ordering_e write_ordering;
671 
672 	cpumask_var_t cpu_mask;
673 };
674 
675 struct drbd_thread_timing_details
676 {
677 	unsigned long start_jif;
678 	void *cb_addr;
679 	const char *caller_fn;
680 	unsigned int line;
681 	unsigned int cb_nr;
682 };
683 
684 struct drbd_connection {
685 	struct list_head connections;
686 	struct drbd_resource *resource;
687 #ifdef CONFIG_DEBUG_FS
688 	struct dentry *debugfs_conn;
689 	struct dentry *debugfs_conn_callback_history;
690 	struct dentry *debugfs_conn_oldest_requests;
691 #endif
692 	struct kref kref;
693 	struct idr peer_devices;	/* volume number to peer device mapping */
694 	enum drbd_conns cstate;		/* Only C_STANDALONE to C_WF_REPORT_PARAMS */
695 	struct mutex cstate_mutex;	/* Protects graceful disconnects */
696 	unsigned int connect_cnt;	/* Inc each time a connection is established */
697 
698 	unsigned long flags;
699 	struct net_conf *net_conf;	/* content protected by rcu */
700 	wait_queue_head_t ping_wait;	/* Woken upon reception of a ping, and a state change */
701 
702 	struct sockaddr_storage my_addr;
703 	int my_addr_len;
704 	struct sockaddr_storage peer_addr;
705 	int peer_addr_len;
706 
707 	struct drbd_socket data;	/* data/barrier/cstate/parameter packets */
708 	struct drbd_socket meta;	/* ping/ack (metadata) packets */
709 	int agreed_pro_version;		/* actually used protocol version */
710 	u32 agreed_features;
711 	unsigned long last_received;	/* in jiffies, either socket */
712 	unsigned int ko_count;
713 
714 	struct list_head transfer_log;	/* all requests not yet fully processed */
715 
716 	struct crypto_shash *cram_hmac_tfm;
717 	struct crypto_shash *integrity_tfm;  /* checksums we compute, updates protected by connection->data->mutex */
718 	struct crypto_shash *peer_integrity_tfm;  /* checksums we verify, only accessed from receiver thread  */
719 	struct crypto_shash *csums_tfm;
720 	struct crypto_shash *verify_tfm;
721 	void *int_dig_in;
722 	void *int_dig_vv;
723 
724 	/* receiver side */
725 	struct drbd_epoch *current_epoch;
726 	spinlock_t epoch_lock;
727 	unsigned int epochs;
728 	atomic_t current_tle_nr;	/* transfer log epoch number */
729 	unsigned current_tle_writes;	/* writes seen within this tl epoch */
730 
731 	unsigned long last_reconnect_jif;
732 	/* empty member on older kernels without blk_start_plug() */
733 	struct blk_plug receiver_plug;
734 	struct drbd_thread receiver;
735 	struct drbd_thread worker;
736 	struct drbd_thread ack_receiver;
737 	struct workqueue_struct *ack_sender;
738 
739 	/* cached pointers,
740 	 * so we can look up the oldest pending requests more quickly.
741 	 * protected by resource->req_lock */
742 	struct drbd_request *req_next; /* DRBD 9: todo.req_next */
743 	struct drbd_request *req_ack_pending;
744 	struct drbd_request *req_not_net_done;
745 
746 	/* sender side */
747 	struct drbd_work_queue sender_work;
748 
749 #define DRBD_THREAD_DETAILS_HIST	16
750 	unsigned int w_cb_nr; /* keeps counting up */
751 	unsigned int r_cb_nr; /* keeps counting up */
752 	struct drbd_thread_timing_details w_timing_details[DRBD_THREAD_DETAILS_HIST];
753 	struct drbd_thread_timing_details r_timing_details[DRBD_THREAD_DETAILS_HIST];
754 
755 	struct {
756 		unsigned long last_sent_barrier_jif;
757 
758 		/* whether this sender thread
759 		 * has processed a single write yet. */
760 		bool seen_any_write_yet;
761 
762 		/* Which barrier number to send with the next P_BARRIER */
763 		int current_epoch_nr;
764 
765 		/* how many write requests have been sent
766 		 * with req->epoch == current_epoch_nr.
767 		 * If none, no P_BARRIER will be sent. */
768 		unsigned current_epoch_writes;
769 	} send;
770 };
771 
772 static inline bool has_net_conf(struct drbd_connection *connection)
773 {
774 	bool has_net_conf;
775 
776 	rcu_read_lock();
777 	has_net_conf = rcu_dereference(connection->net_conf);
778 	rcu_read_unlock();
779 
780 	return has_net_conf;
781 }
782 
783 void __update_timing_details(
784 		struct drbd_thread_timing_details *tdp,
785 		unsigned int *cb_nr,
786 		void *cb,
787 		const char *fn, const unsigned int line);
788 
789 #define update_worker_timing_details(c, cb) \
790 	__update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ )
791 #define update_receiver_timing_details(c, cb) \
792 	__update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ )
793 
794 struct submit_worker {
795 	struct workqueue_struct *wq;
796 	struct work_struct worker;
797 
798 	/* protected by ..->resource->req_lock */
799 	struct list_head writes;
800 };
801 
802 struct drbd_peer_device {
803 	struct list_head peer_devices;
804 	struct drbd_device *device;
805 	struct drbd_connection *connection;
806 	struct work_struct send_acks_work;
807 #ifdef CONFIG_DEBUG_FS
808 	struct dentry *debugfs_peer_dev;
809 #endif
810 };
811 
812 struct drbd_device {
813 	struct drbd_resource *resource;
814 	struct list_head peer_devices;
815 	struct list_head pending_bitmap_io;
816 
817 	unsigned long flush_jif;
818 #ifdef CONFIG_DEBUG_FS
819 	struct dentry *debugfs_minor;
820 	struct dentry *debugfs_vol;
821 	struct dentry *debugfs_vol_oldest_requests;
822 	struct dentry *debugfs_vol_act_log_extents;
823 	struct dentry *debugfs_vol_resync_extents;
824 	struct dentry *debugfs_vol_data_gen_id;
825 	struct dentry *debugfs_vol_ed_gen_id;
826 #endif
827 
828 	unsigned int vnr;	/* volume number within the connection */
829 	unsigned int minor;	/* device minor number */
830 
831 	struct kref kref;
832 
833 	/* things that are stored as / read from meta data on disk */
834 	unsigned long flags;
835 
836 	/* configured by drbdsetup */
837 	struct drbd_backing_dev *ldev __protected_by(local);
838 
839 	sector_t p_size;     /* partner's disk size */
840 	struct request_queue *rq_queue;
841 	struct gendisk	    *vdisk;
842 
843 	unsigned long last_reattach_jif;
844 	struct drbd_work resync_work;
845 	struct drbd_work unplug_work;
846 	struct timer_list resync_timer;
847 	struct timer_list md_sync_timer;
848 	struct timer_list start_resync_timer;
849 	struct timer_list request_timer;
850 
851 	/* Used after attach while negotiating new disk state. */
852 	union drbd_state new_state_tmp;
853 
854 	union drbd_dev_state state;
855 	wait_queue_head_t misc_wait;
856 	wait_queue_head_t state_wait;  /* upon each state change. */
857 	unsigned int send_cnt;
858 	unsigned int recv_cnt;
859 	unsigned int read_cnt;
860 	unsigned int writ_cnt;
861 	unsigned int al_writ_cnt;
862 	unsigned int bm_writ_cnt;
863 	atomic_t ap_bio_cnt;	 /* Requests we need to complete */
864 	atomic_t ap_actlog_cnt;  /* Requests waiting for activity log */
865 	atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
866 	atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
867 	atomic_t unacked_cnt;	 /* Need to send replies for */
868 	atomic_t local_cnt;	 /* Waiting for local completion */
869 	atomic_t suspend_cnt;
870 
871 	/* Interval tree of pending local requests */
872 	struct rb_root read_requests;
873 	struct rb_root write_requests;
874 
875 	/* for statistics and timeouts */
876 	/* [0] read, [1] write */
877 	struct list_head pending_master_completion[2];
878 	struct list_head pending_completion[2];
879 
880 	/* use checksums for *this* resync */
881 	bool use_csums;
882 	/* blocks to resync in this run [unit BM_BLOCK_SIZE] */
883 	unsigned long rs_total;
884 	/* number of resync blocks that failed in this run */
885 	unsigned long rs_failed;
886 	/* Syncer's start time [unit jiffies] */
887 	unsigned long rs_start;
888 	/* cumulated time in PausedSyncX state [unit jiffies] */
889 	unsigned long rs_paused;
890 	/* skipped because csum was equal [unit BM_BLOCK_SIZE] */
891 	unsigned long rs_same_csum;
892 #define DRBD_SYNC_MARKS 8
893 #define DRBD_SYNC_MARK_STEP (3*HZ)
894 	/* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
895 	unsigned long rs_mark_left[DRBD_SYNC_MARKS];
896 	/* marks's time [unit jiffies] */
897 	unsigned long rs_mark_time[DRBD_SYNC_MARKS];
898 	/* current index into rs_mark_{left,time} */
899 	int rs_last_mark;
900 	unsigned long rs_last_bcast; /* [unit jiffies] */
901 
902 	/* where does the admin want us to start? (sector) */
903 	sector_t ov_start_sector;
904 	sector_t ov_stop_sector;
905 	/* where are we now? (sector) */
906 	sector_t ov_position;
907 	/* Start sector of out of sync range (to merge printk reporting). */
908 	sector_t ov_last_oos_start;
909 	/* size of out-of-sync range in sectors. */
910 	sector_t ov_last_oos_size;
911 	unsigned long ov_left; /* in bits */
912 
913 	struct drbd_bitmap *bitmap;
914 	unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
915 
916 	/* Used to track operations of resync... */
917 	struct lru_cache *resync;
918 	/* Number of locked elements in resync LRU */
919 	unsigned int resync_locked;
920 	/* resync extent number waiting for application requests */
921 	unsigned int resync_wenr;
922 
923 	int open_cnt;
924 	u64 *p_uuid;
925 
926 	struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
927 	struct list_head sync_ee;   /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
928 	struct list_head done_ee;   /* need to send P_WRITE_ACK */
929 	struct list_head read_ee;   /* [RS]P_DATA_REQUEST being read */
930 	struct list_head net_ee;    /* zero-copy network send in progress */
931 
932 	int next_barrier_nr;
933 	struct list_head resync_reads;
934 	atomic_t pp_in_use;		/* allocated from page pool */
935 	atomic_t pp_in_use_by_net;	/* sendpage()d, still referenced by tcp */
936 	wait_queue_head_t ee_wait;
937 	struct drbd_md_io md_io;
938 	spinlock_t al_lock;
939 	wait_queue_head_t al_wait;
940 	struct lru_cache *act_log;	/* activity log */
941 	unsigned int al_tr_number;
942 	int al_tr_cycle;
943 	wait_queue_head_t seq_wait;
944 	atomic_t packet_seq;
945 	unsigned int peer_seq;
946 	spinlock_t peer_seq_lock;
947 	unsigned long comm_bm_set; /* communicated number of set bits. */
948 	struct bm_io_work bm_io_work;
949 	u64 ed_uuid; /* UUID of the exposed data */
950 	struct mutex own_state_mutex;
951 	struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */
952 	char congestion_reason;  /* Why we where congested... */
953 	atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
954 	atomic_t rs_sect_ev; /* for submitted resync data rate, both */
955 	int rs_last_sect_ev; /* counter to compare with */
956 	int rs_last_events;  /* counter of read or write "events" (unit sectors)
957 			      * on the lower level device when we last looked. */
958 	int c_sync_rate; /* current resync rate after syncer throttle magic */
959 	struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */
960 	int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
961 	atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
962 	unsigned int peer_max_bio_size;
963 	unsigned int local_max_bio_size;
964 
965 	/* any requests that would block in drbd_make_request()
966 	 * are deferred to this single-threaded work queue */
967 	struct submit_worker submit;
968 };
969 
970 struct drbd_bm_aio_ctx {
971 	struct drbd_device *device;
972 	struct list_head list; /* on device->pending_bitmap_io */;
973 	unsigned long start_jif;
974 	atomic_t in_flight;
975 	unsigned int done;
976 	unsigned flags;
977 #define BM_AIO_COPY_PAGES	1
978 #define BM_AIO_WRITE_HINTED	2
979 #define BM_AIO_WRITE_ALL_PAGES	4
980 #define BM_AIO_READ		8
981 	int error;
982 	struct kref kref;
983 };
984 
985 struct drbd_config_context {
986 	/* assigned from drbd_genlmsghdr */
987 	unsigned int minor;
988 	/* assigned from request attributes, if present */
989 	unsigned int volume;
990 #define VOLUME_UNSPECIFIED		(-1U)
991 	/* pointer into the request skb,
992 	 * limited lifetime! */
993 	char *resource_name;
994 	struct nlattr *my_addr;
995 	struct nlattr *peer_addr;
996 
997 	/* reply buffer */
998 	struct sk_buff *reply_skb;
999 	/* pointer into reply buffer */
1000 	struct drbd_genlmsghdr *reply_dh;
1001 	/* resolved from attributes, if possible */
1002 	struct drbd_device *device;
1003 	struct drbd_resource *resource;
1004 	struct drbd_connection *connection;
1005 };
1006 
1007 static inline struct drbd_device *minor_to_device(unsigned int minor)
1008 {
1009 	return (struct drbd_device *)idr_find(&drbd_devices, minor);
1010 }
1011 
1012 static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device)
1013 {
1014 	return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices);
1015 }
1016 
1017 static inline struct drbd_peer_device *
1018 conn_peer_device(struct drbd_connection *connection, int volume_number)
1019 {
1020 	return idr_find(&connection->peer_devices, volume_number);
1021 }
1022 
1023 #define for_each_resource(resource, _resources) \
1024 	list_for_each_entry(resource, _resources, resources)
1025 
1026 #define for_each_resource_rcu(resource, _resources) \
1027 	list_for_each_entry_rcu(resource, _resources, resources)
1028 
1029 #define for_each_resource_safe(resource, tmp, _resources) \
1030 	list_for_each_entry_safe(resource, tmp, _resources, resources)
1031 
1032 #define for_each_connection(connection, resource) \
1033 	list_for_each_entry(connection, &resource->connections, connections)
1034 
1035 #define for_each_connection_rcu(connection, resource) \
1036 	list_for_each_entry_rcu(connection, &resource->connections, connections)
1037 
1038 #define for_each_connection_safe(connection, tmp, resource) \
1039 	list_for_each_entry_safe(connection, tmp, &resource->connections, connections)
1040 
1041 #define for_each_peer_device(peer_device, device) \
1042 	list_for_each_entry(peer_device, &device->peer_devices, peer_devices)
1043 
1044 #define for_each_peer_device_rcu(peer_device, device) \
1045 	list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices)
1046 
1047 #define for_each_peer_device_safe(peer_device, tmp, device) \
1048 	list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices)
1049 
1050 static inline unsigned int device_to_minor(struct drbd_device *device)
1051 {
1052 	return device->minor;
1053 }
1054 
1055 /*
1056  * function declarations
1057  *************************/
1058 
1059 /* drbd_main.c */
1060 
1061 enum dds_flags {
1062 	DDSF_FORCED    = 1,
1063 	DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1064 };
1065 
1066 extern void drbd_init_set_defaults(struct drbd_device *device);
1067 extern int  drbd_thread_start(struct drbd_thread *thi);
1068 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1069 #ifdef CONFIG_SMP
1070 extern void drbd_thread_current_set_cpu(struct drbd_thread *thi);
1071 #else
1072 #define drbd_thread_current_set_cpu(A) ({})
1073 #endif
1074 extern void tl_release(struct drbd_connection *, unsigned int barrier_nr,
1075 		       unsigned int set_size);
1076 extern void tl_clear(struct drbd_connection *);
1077 extern void drbd_free_sock(struct drbd_connection *connection);
1078 extern int drbd_send(struct drbd_connection *connection, struct socket *sock,
1079 		     void *buf, size_t size, unsigned msg_flags);
1080 extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t,
1081 			 unsigned);
1082 
1083 extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd);
1084 extern int drbd_send_protocol(struct drbd_connection *connection);
1085 extern int drbd_send_uuids(struct drbd_peer_device *);
1086 extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *);
1087 extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *);
1088 extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags);
1089 extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s);
1090 extern int drbd_send_current_state(struct drbd_peer_device *);
1091 extern int drbd_send_sync_param(struct drbd_peer_device *);
1092 extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr,
1093 			    u32 set_size);
1094 extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet,
1095 			 struct drbd_peer_request *);
1096 extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet,
1097 			     struct p_block_req *rp);
1098 extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet,
1099 			     struct p_data *dp, int data_size);
1100 extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet,
1101 			    sector_t sector, int blksize, u64 block_id);
1102 extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *);
1103 extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet,
1104 			   struct drbd_peer_request *);
1105 extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req);
1106 extern int drbd_send_drequest(struct drbd_peer_device *, int cmd,
1107 			      sector_t sector, int size, u64 block_id);
1108 extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector,
1109 				   int size, void *digest, int digest_size,
1110 				   enum drbd_packet cmd);
1111 extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size);
1112 
1113 extern int drbd_send_bitmap(struct drbd_device *device);
1114 extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode);
1115 extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode);
1116 extern int drbd_send_rs_deallocated(struct drbd_peer_device *, struct drbd_peer_request *);
1117 extern void drbd_backing_dev_free(struct drbd_device *device, struct drbd_backing_dev *ldev);
1118 extern void drbd_device_cleanup(struct drbd_device *device);
1119 extern void drbd_print_uuids(struct drbd_device *device, const char *text);
1120 extern void drbd_queue_unplug(struct drbd_device *device);
1121 
1122 extern void conn_md_sync(struct drbd_connection *connection);
1123 extern void drbd_md_write(struct drbd_device *device, void *buffer);
1124 extern void drbd_md_sync(struct drbd_device *device);
1125 extern int  drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev);
1126 extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1127 extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1128 extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local);
1129 extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local);
1130 extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local);
1131 extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1132 extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local);
1133 extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local);
1134 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1135 extern void drbd_md_mark_dirty(struct drbd_device *device);
1136 extern void drbd_queue_bitmap_io(struct drbd_device *device,
1137 				 int (*io_fn)(struct drbd_device *),
1138 				 void (*done)(struct drbd_device *, int),
1139 				 char *why, enum bm_flag flags);
1140 extern int drbd_bitmap_io(struct drbd_device *device,
1141 		int (*io_fn)(struct drbd_device *),
1142 		char *why, enum bm_flag flags);
1143 extern int drbd_bitmap_io_from_worker(struct drbd_device *device,
1144 		int (*io_fn)(struct drbd_device *),
1145 		char *why, enum bm_flag flags);
1146 extern int drbd_bmio_set_n_write(struct drbd_device *device) __must_hold(local);
1147 extern int drbd_bmio_clear_n_write(struct drbd_device *device) __must_hold(local);
1148 
1149 /* Meta data layout
1150  *
1151  * We currently have two possible layouts.
1152  * Offsets in (512 byte) sectors.
1153  * external:
1154  *   |----------- md_size_sect ------------------|
1155  *   [ 4k superblock ][ activity log ][  Bitmap  ]
1156  *   | al_offset == 8 |
1157  *   | bm_offset = al_offset + X      |
1158  *  ==> bitmap sectors = md_size_sect - bm_offset
1159  *
1160  *  Variants:
1161  *     old, indexed fixed size meta data:
1162  *
1163  * internal:
1164  *            |----------- md_size_sect ------------------|
1165  * [data.....][  Bitmap  ][ activity log ][ 4k superblock ][padding*]
1166  *                        | al_offset < 0 |
1167  *            | bm_offset = al_offset - Y |
1168  *  ==> bitmap sectors = Y = al_offset - bm_offset
1169  *
1170  *  [padding*] are zero or up to 7 unused 512 Byte sectors to the
1171  *  end of the device, so that the [4k superblock] will be 4k aligned.
1172  *
1173  *  The activity log consists of 4k transaction blocks,
1174  *  which are written in a ring-buffer, or striped ring-buffer like fashion,
1175  *  which are writtensize used to be fixed 32kB,
1176  *  but is about to become configurable.
1177  */
1178 
1179 /* Our old fixed size meta data layout
1180  * allows up to about 3.8TB, so if you want more,
1181  * you need to use the "flexible" meta data format. */
1182 #define MD_128MB_SECT (128LLU << 11)  /* 128 MB, unit sectors */
1183 #define MD_4kB_SECT	 8
1184 #define MD_32kB_SECT	64
1185 
1186 /* One activity log extent represents 4M of storage */
1187 #define AL_EXTENT_SHIFT 22
1188 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1189 
1190 /* We could make these currently hardcoded constants configurable
1191  * variables at create-md time (or even re-configurable at runtime?).
1192  * Which will require some more changes to the DRBD "super block"
1193  * and attach code.
1194  *
1195  * updates per transaction:
1196  *   This many changes to the active set can be logged with one transaction.
1197  *   This number is arbitrary.
1198  * context per transaction:
1199  *   This many context extent numbers are logged with each transaction.
1200  *   This number is resulting from the transaction block size (4k), the layout
1201  *   of the transaction header, and the number of updates per transaction.
1202  *   See drbd_actlog.c:struct al_transaction_on_disk
1203  * */
1204 #define AL_UPDATES_PER_TRANSACTION	 64	// arbitrary
1205 #define AL_CONTEXT_PER_TRANSACTION	919	// (4096 - 36 - 6*64)/4
1206 
1207 #if BITS_PER_LONG == 32
1208 #define LN2_BPL 5
1209 #define cpu_to_lel(A) cpu_to_le32(A)
1210 #define lel_to_cpu(A) le32_to_cpu(A)
1211 #elif BITS_PER_LONG == 64
1212 #define LN2_BPL 6
1213 #define cpu_to_lel(A) cpu_to_le64(A)
1214 #define lel_to_cpu(A) le64_to_cpu(A)
1215 #else
1216 #error "LN2 of BITS_PER_LONG unknown!"
1217 #endif
1218 
1219 /* resync bitmap */
1220 /* 16MB sized 'bitmap extent' to track syncer usage */
1221 struct bm_extent {
1222 	int rs_left; /* number of bits set (out of sync) in this extent. */
1223 	int rs_failed; /* number of failed resync requests in this extent. */
1224 	unsigned long flags;
1225 	struct lc_element lce;
1226 };
1227 
1228 #define BME_NO_WRITES  0  /* bm_extent.flags: no more requests on this one! */
1229 #define BME_LOCKED     1  /* bm_extent.flags: syncer active on this one. */
1230 #define BME_PRIORITY   2  /* finish resync IO on this extent ASAP! App IO waiting! */
1231 
1232 /* drbd_bitmap.c */
1233 /*
1234  * We need to store one bit for a block.
1235  * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1236  * Bit 0 ==> local node thinks this block is binary identical on both nodes
1237  * Bit 1 ==> local node thinks this block needs to be synced.
1238  */
1239 
1240 #define SLEEP_TIME (HZ/10)
1241 
1242 /* We do bitmap IO in units of 4k blocks.
1243  * We also still have a hardcoded 4k per bit relation. */
1244 #define BM_BLOCK_SHIFT	12			 /* 4k per bit */
1245 #define BM_BLOCK_SIZE	 (1<<BM_BLOCK_SHIFT)
1246 /* mostly arbitrarily set the represented size of one bitmap extent,
1247  * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1248  * at 4k per bit resolution) */
1249 #define BM_EXT_SHIFT	 24	/* 16 MiB per resync extent */
1250 #define BM_EXT_SIZE	 (1<<BM_EXT_SHIFT)
1251 
1252 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1253 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1254 #endif
1255 
1256 /* thus many _storage_ sectors are described by one bit */
1257 #define BM_SECT_TO_BIT(x)   ((x)>>(BM_BLOCK_SHIFT-9))
1258 #define BM_BIT_TO_SECT(x)   ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1259 #define BM_SECT_PER_BIT     BM_BIT_TO_SECT(1)
1260 
1261 /* bit to represented kilo byte conversion */
1262 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1263 
1264 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1265  * _storage_ sector is located in */
1266 #define BM_SECT_TO_EXT(x)   ((x)>>(BM_EXT_SHIFT-9))
1267 #define BM_BIT_TO_EXT(x)    ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1268 
1269 /* first storage sector a bitmap extent corresponds to */
1270 #define BM_EXT_TO_SECT(x)   ((sector_t)(x) << (BM_EXT_SHIFT-9))
1271 /* how much _storage_ sectors we have per bitmap extent */
1272 #define BM_SECT_PER_EXT     BM_EXT_TO_SECT(1)
1273 /* how many bits are covered by one bitmap extent (resync extent) */
1274 #define BM_BITS_PER_EXT     (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1275 
1276 #define BM_BLOCKS_PER_BM_EXT_MASK  (BM_BITS_PER_EXT - 1)
1277 
1278 
1279 /* in one sector of the bitmap, we have this many activity_log extents. */
1280 #define AL_EXT_PER_BM_SECT  (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1281 
1282 /* the extent in "PER_EXTENT" below is an activity log extent
1283  * we need that many (long words/bytes) to store the bitmap
1284  *		     of one AL_EXTENT_SIZE chunk of storage.
1285  * we can store the bitmap for that many AL_EXTENTS within
1286  * one sector of the _on_disk_ bitmap:
1287  * bit	 0	  bit 37   bit 38	     bit (512*8)-1
1288  *	     ...|........|........|.. // ..|........|
1289  * sect. 0	 `296	  `304			   ^(512*8*8)-1
1290  *
1291 #define BM_WORDS_PER_EXT    ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1292 #define BM_BYTES_PER_EXT    ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 )  // 128
1293 #define BM_EXT_PER_SECT	    ( 512 / BM_BYTES_PER_EXTENT )	 //   4
1294  */
1295 
1296 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1297 /* we have a certain meta data variant that has a fixed on-disk size of 128
1298  * MiB, of which 4k are our "superblock", and 32k are the fixed size activity
1299  * log, leaving this many sectors for the bitmap.
1300  */
1301 
1302 #define DRBD_MAX_SECTORS_FIXED_BM \
1303 	  ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9)))
1304 #define DRBD_MAX_SECTORS      DRBD_MAX_SECTORS_FIXED_BM
1305 /* 16 TB in units of sectors */
1306 #if BITS_PER_LONG == 32
1307 /* adjust by one page worth of bitmap,
1308  * so we won't wrap around in drbd_bm_find_next_bit.
1309  * you should use 64bit OS for that much storage, anyways. */
1310 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1311 #else
1312 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1313 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1314 /* corresponds to (1UL << 38) bits right now. */
1315 #endif
1316 
1317 /* Estimate max bio size as 256 * PAGE_SIZE,
1318  * so for typical PAGE_SIZE of 4k, that is (1<<20) Byte.
1319  * Since we may live in a mixed-platform cluster,
1320  * we limit us to a platform agnostic constant here for now.
1321  * A followup commit may allow even bigger BIO sizes,
1322  * once we thought that through. */
1323 #define DRBD_MAX_BIO_SIZE (1U << 20)
1324 #if DRBD_MAX_BIO_SIZE > (BIO_MAX_VECS << PAGE_SHIFT)
1325 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1326 #endif
1327 #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12)       /* Works always = 4k */
1328 
1329 #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */
1330 #define DRBD_MAX_BIO_SIZE_P95    (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */
1331 
1332 /* For now, don't allow more than half of what we can "activate" in one
1333  * activity log transaction to be discarded in one go. We may need to rework
1334  * drbd_al_begin_io() to allow for even larger discard ranges */
1335 #define DRBD_MAX_BATCH_BIO_SIZE	 (AL_UPDATES_PER_TRANSACTION/2*AL_EXTENT_SIZE)
1336 #define DRBD_MAX_BBIO_SECTORS    (DRBD_MAX_BATCH_BIO_SIZE >> 9)
1337 
1338 extern int  drbd_bm_init(struct drbd_device *device);
1339 extern int  drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits);
1340 extern void drbd_bm_cleanup(struct drbd_device *device);
1341 extern void drbd_bm_set_all(struct drbd_device *device);
1342 extern void drbd_bm_clear_all(struct drbd_device *device);
1343 /* set/clear/test only a few bits at a time */
1344 extern int  drbd_bm_set_bits(
1345 		struct drbd_device *device, unsigned long s, unsigned long e);
1346 extern int  drbd_bm_clear_bits(
1347 		struct drbd_device *device, unsigned long s, unsigned long e);
1348 extern int drbd_bm_count_bits(
1349 	struct drbd_device *device, const unsigned long s, const unsigned long e);
1350 /* bm_set_bits variant for use while holding drbd_bm_lock,
1351  * may process the whole bitmap in one go */
1352 extern void _drbd_bm_set_bits(struct drbd_device *device,
1353 		const unsigned long s, const unsigned long e);
1354 extern int  drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr);
1355 extern int  drbd_bm_e_weight(struct drbd_device *device, unsigned long enr);
1356 extern int  drbd_bm_read(struct drbd_device *device) __must_hold(local);
1357 extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr);
1358 extern int  drbd_bm_write(struct drbd_device *device) __must_hold(local);
1359 extern void drbd_bm_reset_al_hints(struct drbd_device *device) __must_hold(local);
1360 extern int  drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local);
1361 extern int  drbd_bm_write_lazy(struct drbd_device *device, unsigned upper_idx) __must_hold(local);
1362 extern int drbd_bm_write_all(struct drbd_device *device) __must_hold(local);
1363 extern int  drbd_bm_write_copy_pages(struct drbd_device *device) __must_hold(local);
1364 extern size_t	     drbd_bm_words(struct drbd_device *device);
1365 extern unsigned long drbd_bm_bits(struct drbd_device *device);
1366 extern sector_t      drbd_bm_capacity(struct drbd_device *device);
1367 
1368 #define DRBD_END_OF_BITMAP	(~(unsigned long)0)
1369 extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo);
1370 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1371 extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo);
1372 extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo);
1373 extern unsigned long _drbd_bm_total_weight(struct drbd_device *device);
1374 extern unsigned long drbd_bm_total_weight(struct drbd_device *device);
1375 /* for receive_bitmap */
1376 extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset,
1377 		size_t number, unsigned long *buffer);
1378 /* for _drbd_send_bitmap */
1379 extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset,
1380 		size_t number, unsigned long *buffer);
1381 
1382 extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags);
1383 extern void drbd_bm_unlock(struct drbd_device *device);
1384 /* drbd_main.c */
1385 
1386 extern struct kmem_cache *drbd_request_cache;
1387 extern struct kmem_cache *drbd_ee_cache;	/* peer requests */
1388 extern struct kmem_cache *drbd_bm_ext_cache;	/* bitmap extents */
1389 extern struct kmem_cache *drbd_al_ext_cache;	/* activity log extents */
1390 extern mempool_t drbd_request_mempool;
1391 extern mempool_t drbd_ee_mempool;
1392 
1393 /* drbd's page pool, used to buffer data received from the peer,
1394  * or data requested by the peer.
1395  *
1396  * This does not have an emergency reserve.
1397  *
1398  * When allocating from this pool, it first takes pages from the pool.
1399  * Only if the pool is depleted will try to allocate from the system.
1400  *
1401  * The assumption is that pages taken from this pool will be processed,
1402  * and given back, "quickly", and then can be recycled, so we can avoid
1403  * frequent calls to alloc_page(), and still will be able to make progress even
1404  * under memory pressure.
1405  */
1406 extern struct page *drbd_pp_pool;
1407 extern spinlock_t   drbd_pp_lock;
1408 extern int	    drbd_pp_vacant;
1409 extern wait_queue_head_t drbd_pp_wait;
1410 
1411 /* We also need a standard (emergency-reserve backed) page pool
1412  * for meta data IO (activity log, bitmap).
1413  * We can keep it global, as long as it is used as "N pages at a time".
1414  * 128 should be plenty, currently we probably can get away with as few as 1.
1415  */
1416 #define DRBD_MIN_POOL_PAGES	128
1417 extern mempool_t drbd_md_io_page_pool;
1418 
1419 /* We also need to make sure we get a bio
1420  * when we need it for housekeeping purposes */
1421 extern struct bio_set drbd_md_io_bio_set;
1422 
1423 /* And a bio_set for cloning */
1424 extern struct bio_set drbd_io_bio_set;
1425 
1426 extern struct mutex resources_mutex;
1427 
1428 extern int conn_lowest_minor(struct drbd_connection *connection);
1429 extern enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor);
1430 extern void drbd_destroy_device(struct kref *kref);
1431 extern void drbd_delete_device(struct drbd_device *device);
1432 
1433 extern struct drbd_resource *drbd_create_resource(const char *name);
1434 extern void drbd_free_resource(struct drbd_resource *resource);
1435 
1436 extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts);
1437 extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts);
1438 extern void drbd_destroy_connection(struct kref *kref);
1439 extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len,
1440 					    void *peer_addr, int peer_addr_len);
1441 extern struct drbd_resource *drbd_find_resource(const char *name);
1442 extern void drbd_destroy_resource(struct kref *kref);
1443 extern void conn_free_crypto(struct drbd_connection *connection);
1444 
1445 /* drbd_req */
1446 extern void do_submit(struct work_struct *ws);
1447 extern void __drbd_make_request(struct drbd_device *, struct bio *);
1448 void drbd_submit_bio(struct bio *bio);
1449 extern int drbd_read_remote(struct drbd_device *device, struct drbd_request *req);
1450 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1451 
1452 
1453 /* drbd_nl.c */
1454 
1455 extern struct mutex notification_mutex;
1456 
1457 extern void drbd_suspend_io(struct drbd_device *device);
1458 extern void drbd_resume_io(struct drbd_device *device);
1459 extern char *ppsize(char *buf, unsigned long long size);
1460 extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int);
1461 enum determine_dev_size {
1462 	DS_ERROR_SHRINK = -3,
1463 	DS_ERROR_SPACE_MD = -2,
1464 	DS_ERROR = -1,
1465 	DS_UNCHANGED = 0,
1466 	DS_SHRUNK = 1,
1467 	DS_GREW = 2,
1468 	DS_GREW_FROM_ZERO = 3,
1469 };
1470 extern enum determine_dev_size
1471 drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local);
1472 extern void resync_after_online_grow(struct drbd_device *);
1473 extern void drbd_reconsider_queue_parameters(struct drbd_device *device,
1474 			struct drbd_backing_dev *bdev, struct o_qlim *o);
1475 extern enum drbd_state_rv drbd_set_role(struct drbd_device *device,
1476 					enum drbd_role new_role,
1477 					int force);
1478 extern bool conn_try_outdate_peer(struct drbd_connection *connection);
1479 extern void conn_try_outdate_peer_async(struct drbd_connection *connection);
1480 extern enum drbd_peer_state conn_khelper(struct drbd_connection *connection, char *cmd);
1481 extern int drbd_khelper(struct drbd_device *device, char *cmd);
1482 
1483 /* drbd_worker.c */
1484 /* bi_end_io handlers */
1485 extern void drbd_md_endio(struct bio *bio);
1486 extern void drbd_peer_request_endio(struct bio *bio);
1487 extern void drbd_request_endio(struct bio *bio);
1488 extern int drbd_worker(struct drbd_thread *thi);
1489 enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor);
1490 void drbd_resync_after_changed(struct drbd_device *device);
1491 extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side);
1492 extern void resume_next_sg(struct drbd_device *device);
1493 extern void suspend_other_sg(struct drbd_device *device);
1494 extern int drbd_resync_finished(struct drbd_device *device);
1495 /* maybe rather drbd_main.c ? */
1496 extern void *drbd_md_get_buffer(struct drbd_device *device, const char *intent);
1497 extern void drbd_md_put_buffer(struct drbd_device *device);
1498 extern int drbd_md_sync_page_io(struct drbd_device *device,
1499 		struct drbd_backing_dev *bdev, sector_t sector, int op);
1500 extern void drbd_ov_out_of_sync_found(struct drbd_device *, sector_t, int);
1501 extern void wait_until_done_or_force_detached(struct drbd_device *device,
1502 		struct drbd_backing_dev *bdev, unsigned int *done);
1503 extern void drbd_rs_controller_reset(struct drbd_device *device);
1504 
1505 static inline void ov_out_of_sync_print(struct drbd_device *device)
1506 {
1507 	if (device->ov_last_oos_size) {
1508 		drbd_err(device, "Out of sync: start=%llu, size=%lu (sectors)\n",
1509 		     (unsigned long long)device->ov_last_oos_start,
1510 		     (unsigned long)device->ov_last_oos_size);
1511 	}
1512 	device->ov_last_oos_size = 0;
1513 }
1514 
1515 
1516 extern void drbd_csum_bio(struct crypto_shash *, struct bio *, void *);
1517 extern void drbd_csum_ee(struct crypto_shash *, struct drbd_peer_request *,
1518 			 void *);
1519 /* worker callbacks */
1520 extern int w_e_end_data_req(struct drbd_work *, int);
1521 extern int w_e_end_rsdata_req(struct drbd_work *, int);
1522 extern int w_e_end_csum_rs_req(struct drbd_work *, int);
1523 extern int w_e_end_ov_reply(struct drbd_work *, int);
1524 extern int w_e_end_ov_req(struct drbd_work *, int);
1525 extern int w_ov_finished(struct drbd_work *, int);
1526 extern int w_resync_timer(struct drbd_work *, int);
1527 extern int w_send_write_hint(struct drbd_work *, int);
1528 extern int w_send_dblock(struct drbd_work *, int);
1529 extern int w_send_read_req(struct drbd_work *, int);
1530 extern int w_e_reissue(struct drbd_work *, int);
1531 extern int w_restart_disk_io(struct drbd_work *, int);
1532 extern int w_send_out_of_sync(struct drbd_work *, int);
1533 extern int w_start_resync(struct drbd_work *, int);
1534 
1535 extern void resync_timer_fn(struct timer_list *t);
1536 extern void start_resync_timer_fn(struct timer_list *t);
1537 
1538 extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req);
1539 
1540 /* drbd_receiver.c */
1541 extern int drbd_issue_discard_or_zero_out(struct drbd_device *device,
1542 		sector_t start, unsigned int nr_sectors, int flags);
1543 extern int drbd_receiver(struct drbd_thread *thi);
1544 extern int drbd_ack_receiver(struct drbd_thread *thi);
1545 extern void drbd_send_ping_wf(struct work_struct *ws);
1546 extern void drbd_send_acks_wf(struct work_struct *ws);
1547 extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device);
1548 extern bool drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector,
1549 		bool throttle_if_app_is_waiting);
1550 extern int drbd_submit_peer_request(struct drbd_device *,
1551 				    struct drbd_peer_request *, const unsigned,
1552 				    const unsigned, const int);
1553 extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *);
1554 extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64,
1555 						     sector_t, unsigned int,
1556 						     unsigned int,
1557 						     gfp_t) __must_hold(local);
1558 extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *,
1559 				 int);
1560 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1561 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
1562 extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool);
1563 extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
1564 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
1565 extern int drbd_connected(struct drbd_peer_device *);
1566 
1567 /* sets the number of 512 byte sectors of our virtual device */
1568 void drbd_set_my_capacity(struct drbd_device *device, sector_t size);
1569 
1570 /*
1571  * used to submit our private bio
1572  */
1573 static inline void drbd_submit_bio_noacct(struct drbd_device *device,
1574 					     int fault_type, struct bio *bio)
1575 {
1576 	__release(local);
1577 	if (!bio->bi_bdev) {
1578 		drbd_err(device, "drbd_submit_bio_noacct: bio->bi_bdev == NULL\n");
1579 		bio->bi_status = BLK_STS_IOERR;
1580 		bio_endio(bio);
1581 		return;
1582 	}
1583 
1584 	if (drbd_insert_fault(device, fault_type))
1585 		bio_io_error(bio);
1586 	else
1587 		submit_bio_noacct(bio);
1588 }
1589 
1590 void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev,
1591 			      enum write_ordering_e wo);
1592 
1593 /* drbd_proc.c */
1594 extern struct proc_dir_entry *drbd_proc;
1595 int drbd_seq_show(struct seq_file *seq, void *v);
1596 
1597 /* drbd_actlog.c */
1598 extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i);
1599 extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i);
1600 extern void drbd_al_begin_io_commit(struct drbd_device *device);
1601 extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i);
1602 extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i);
1603 extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i);
1604 extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector);
1605 extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector);
1606 extern int drbd_try_rs_begin_io(struct drbd_device *device, sector_t sector);
1607 extern void drbd_rs_cancel_all(struct drbd_device *device);
1608 extern int drbd_rs_del_all(struct drbd_device *device);
1609 extern void drbd_rs_failed_io(struct drbd_device *device,
1610 		sector_t sector, int size);
1611 extern void drbd_advance_rs_marks(struct drbd_device *device, unsigned long still_to_go);
1612 
1613 enum update_sync_bits_mode { RECORD_RS_FAILED, SET_OUT_OF_SYNC, SET_IN_SYNC };
1614 extern int __drbd_change_sync(struct drbd_device *device, sector_t sector, int size,
1615 		enum update_sync_bits_mode mode);
1616 #define drbd_set_in_sync(device, sector, size) \
1617 	__drbd_change_sync(device, sector, size, SET_IN_SYNC)
1618 #define drbd_set_out_of_sync(device, sector, size) \
1619 	__drbd_change_sync(device, sector, size, SET_OUT_OF_SYNC)
1620 #define drbd_rs_failed_io(device, sector, size) \
1621 	__drbd_change_sync(device, sector, size, RECORD_RS_FAILED)
1622 extern void drbd_al_shrink(struct drbd_device *device);
1623 extern int drbd_al_initialize(struct drbd_device *, void *);
1624 
1625 /* drbd_nl.c */
1626 /* state info broadcast */
1627 struct sib_info {
1628 	enum drbd_state_info_bcast_reason sib_reason;
1629 	union {
1630 		struct {
1631 			char *helper_name;
1632 			unsigned helper_exit_code;
1633 		};
1634 		struct {
1635 			union drbd_state os;
1636 			union drbd_state ns;
1637 		};
1638 	};
1639 };
1640 void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib);
1641 
1642 extern void notify_resource_state(struct sk_buff *,
1643 				  unsigned int,
1644 				  struct drbd_resource *,
1645 				  struct resource_info *,
1646 				  enum drbd_notification_type);
1647 extern void notify_device_state(struct sk_buff *,
1648 				unsigned int,
1649 				struct drbd_device *,
1650 				struct device_info *,
1651 				enum drbd_notification_type);
1652 extern void notify_connection_state(struct sk_buff *,
1653 				    unsigned int,
1654 				    struct drbd_connection *,
1655 				    struct connection_info *,
1656 				    enum drbd_notification_type);
1657 extern void notify_peer_device_state(struct sk_buff *,
1658 				     unsigned int,
1659 				     struct drbd_peer_device *,
1660 				     struct peer_device_info *,
1661 				     enum drbd_notification_type);
1662 extern void notify_helper(enum drbd_notification_type, struct drbd_device *,
1663 			  struct drbd_connection *, const char *, int);
1664 
1665 /*
1666  * inline helper functions
1667  *************************/
1668 
1669 /* see also page_chain_add and friends in drbd_receiver.c */
1670 static inline struct page *page_chain_next(struct page *page)
1671 {
1672 	return (struct page *)page_private(page);
1673 }
1674 #define page_chain_for_each(page) \
1675 	for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1676 			page = page_chain_next(page))
1677 #define page_chain_for_each_safe(page, n) \
1678 	for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1679 
1680 
1681 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req)
1682 {
1683 	struct page *page = peer_req->pages;
1684 	page_chain_for_each(page) {
1685 		if (page_count(page) > 1)
1686 			return 1;
1687 	}
1688 	return 0;
1689 }
1690 
1691 static inline union drbd_state drbd_read_state(struct drbd_device *device)
1692 {
1693 	struct drbd_resource *resource = device->resource;
1694 	union drbd_state rv;
1695 
1696 	rv.i = device->state.i;
1697 	rv.susp = resource->susp;
1698 	rv.susp_nod = resource->susp_nod;
1699 	rv.susp_fen = resource->susp_fen;
1700 
1701 	return rv;
1702 }
1703 
1704 enum drbd_force_detach_flags {
1705 	DRBD_READ_ERROR,
1706 	DRBD_WRITE_ERROR,
1707 	DRBD_META_IO_ERROR,
1708 	DRBD_FORCE_DETACH,
1709 };
1710 
1711 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1712 static inline void __drbd_chk_io_error_(struct drbd_device *device,
1713 		enum drbd_force_detach_flags df,
1714 		const char *where)
1715 {
1716 	enum drbd_io_error_p ep;
1717 
1718 	rcu_read_lock();
1719 	ep = rcu_dereference(device->ldev->disk_conf)->on_io_error;
1720 	rcu_read_unlock();
1721 	switch (ep) {
1722 	case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */
1723 		if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) {
1724 			if (__ratelimit(&drbd_ratelimit_state))
1725 				drbd_err(device, "Local IO failed in %s.\n", where);
1726 			if (device->state.disk > D_INCONSISTENT)
1727 				_drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL);
1728 			break;
1729 		}
1730 		fallthrough;	/* for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */
1731 	case EP_DETACH:
1732 	case EP_CALL_HELPER:
1733 		/* Remember whether we saw a READ or WRITE error.
1734 		 *
1735 		 * Recovery of the affected area for WRITE failure is covered
1736 		 * by the activity log.
1737 		 * READ errors may fall outside that area though. Certain READ
1738 		 * errors can be "healed" by writing good data to the affected
1739 		 * blocks, which triggers block re-allocation in lower layers.
1740 		 *
1741 		 * If we can not write the bitmap after a READ error,
1742 		 * we may need to trigger a full sync (see w_go_diskless()).
1743 		 *
1744 		 * Force-detach is not really an IO error, but rather a
1745 		 * desperate measure to try to deal with a completely
1746 		 * unresponsive lower level IO stack.
1747 		 * Still it should be treated as a WRITE error.
1748 		 *
1749 		 * Meta IO error is always WRITE error:
1750 		 * we read meta data only once during attach,
1751 		 * which will fail in case of errors.
1752 		 */
1753 		set_bit(WAS_IO_ERROR, &device->flags);
1754 		if (df == DRBD_READ_ERROR)
1755 			set_bit(WAS_READ_ERROR, &device->flags);
1756 		if (df == DRBD_FORCE_DETACH)
1757 			set_bit(FORCE_DETACH, &device->flags);
1758 		if (device->state.disk > D_FAILED) {
1759 			_drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL);
1760 			drbd_err(device,
1761 				"Local IO failed in %s. Detaching...\n", where);
1762 		}
1763 		break;
1764 	}
1765 }
1766 
1767 /**
1768  * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1769  * @device:	 DRBD device.
1770  * @error:	 Error code passed to the IO completion callback
1771  * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1772  *
1773  * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1774  */
1775 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1776 static inline void drbd_chk_io_error_(struct drbd_device *device,
1777 	int error, enum drbd_force_detach_flags forcedetach, const char *where)
1778 {
1779 	if (error) {
1780 		unsigned long flags;
1781 		spin_lock_irqsave(&device->resource->req_lock, flags);
1782 		__drbd_chk_io_error_(device, forcedetach, where);
1783 		spin_unlock_irqrestore(&device->resource->req_lock, flags);
1784 	}
1785 }
1786 
1787 
1788 /**
1789  * drbd_md_first_sector() - Returns the first sector number of the meta data area
1790  * @bdev:	Meta data block device.
1791  *
1792  * BTW, for internal meta data, this happens to be the maximum capacity
1793  * we could agree upon with our peer node.
1794  */
1795 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1796 {
1797 	switch (bdev->md.meta_dev_idx) {
1798 	case DRBD_MD_INDEX_INTERNAL:
1799 	case DRBD_MD_INDEX_FLEX_INT:
1800 		return bdev->md.md_offset + bdev->md.bm_offset;
1801 	case DRBD_MD_INDEX_FLEX_EXT:
1802 	default:
1803 		return bdev->md.md_offset;
1804 	}
1805 }
1806 
1807 /**
1808  * drbd_md_last_sector() - Return the last sector number of the meta data area
1809  * @bdev:	Meta data block device.
1810  */
1811 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1812 {
1813 	switch (bdev->md.meta_dev_idx) {
1814 	case DRBD_MD_INDEX_INTERNAL:
1815 	case DRBD_MD_INDEX_FLEX_INT:
1816 		return bdev->md.md_offset + MD_4kB_SECT -1;
1817 	case DRBD_MD_INDEX_FLEX_EXT:
1818 	default:
1819 		return bdev->md.md_offset + bdev->md.md_size_sect -1;
1820 	}
1821 }
1822 
1823 /* Returns the number of 512 byte sectors of the device */
1824 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1825 {
1826 	return bdev ? bdev_nr_sectors(bdev) : 0;
1827 }
1828 
1829 /**
1830  * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1831  * @bdev:	Meta data block device.
1832  *
1833  * returns the capacity we announce to out peer.  we clip ourselves at the
1834  * various MAX_SECTORS, because if we don't, current implementation will
1835  * oops sooner or later
1836  */
1837 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1838 {
1839 	sector_t s;
1840 
1841 	switch (bdev->md.meta_dev_idx) {
1842 	case DRBD_MD_INDEX_INTERNAL:
1843 	case DRBD_MD_INDEX_FLEX_INT:
1844 		s = drbd_get_capacity(bdev->backing_bdev)
1845 			? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1846 				drbd_md_first_sector(bdev))
1847 			: 0;
1848 		break;
1849 	case DRBD_MD_INDEX_FLEX_EXT:
1850 		s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1851 				drbd_get_capacity(bdev->backing_bdev));
1852 		/* clip at maximum size the meta device can support */
1853 		s = min_t(sector_t, s,
1854 			BM_EXT_TO_SECT(bdev->md.md_size_sect
1855 				     - bdev->md.bm_offset));
1856 		break;
1857 	default:
1858 		s = min_t(sector_t, DRBD_MAX_SECTORS,
1859 				drbd_get_capacity(bdev->backing_bdev));
1860 	}
1861 	return s;
1862 }
1863 
1864 /**
1865  * drbd_md_ss() - Return the sector number of our meta data super block
1866  * @bdev:	Meta data block device.
1867  */
1868 static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev)
1869 {
1870 	const int meta_dev_idx = bdev->md.meta_dev_idx;
1871 
1872 	if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT)
1873 		return 0;
1874 
1875 	/* Since drbd08, internal meta data is always "flexible".
1876 	 * position: last 4k aligned block of 4k size */
1877 	if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL ||
1878 	    meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)
1879 		return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8;
1880 
1881 	/* external, some index; this is the old fixed size layout */
1882 	return MD_128MB_SECT * bdev->md.meta_dev_idx;
1883 }
1884 
1885 static inline void
1886 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1887 {
1888 	unsigned long flags;
1889 	spin_lock_irqsave(&q->q_lock, flags);
1890 	list_add_tail(&w->list, &q->q);
1891 	spin_unlock_irqrestore(&q->q_lock, flags);
1892 	wake_up(&q->q_wait);
1893 }
1894 
1895 static inline void
1896 drbd_queue_work_if_unqueued(struct drbd_work_queue *q, struct drbd_work *w)
1897 {
1898 	unsigned long flags;
1899 	spin_lock_irqsave(&q->q_lock, flags);
1900 	if (list_empty_careful(&w->list))
1901 		list_add_tail(&w->list, &q->q);
1902 	spin_unlock_irqrestore(&q->q_lock, flags);
1903 	wake_up(&q->q_wait);
1904 }
1905 
1906 static inline void
1907 drbd_device_post_work(struct drbd_device *device, int work_bit)
1908 {
1909 	if (!test_and_set_bit(work_bit, &device->flags)) {
1910 		struct drbd_connection *connection =
1911 			first_peer_device(device)->connection;
1912 		struct drbd_work_queue *q = &connection->sender_work;
1913 		if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags))
1914 			wake_up(&q->q_wait);
1915 	}
1916 }
1917 
1918 extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue);
1919 
1920 /* To get the ack_receiver out of the blocking network stack,
1921  * so it can change its sk_rcvtimeo from idle- to ping-timeout,
1922  * and send a ping, we need to send a signal.
1923  * Which signal we send is irrelevant. */
1924 static inline void wake_ack_receiver(struct drbd_connection *connection)
1925 {
1926 	struct task_struct *task = connection->ack_receiver.task;
1927 	if (task && get_t_state(&connection->ack_receiver) == RUNNING)
1928 		send_sig(SIGXCPU, task, 1);
1929 }
1930 
1931 static inline void request_ping(struct drbd_connection *connection)
1932 {
1933 	set_bit(SEND_PING, &connection->flags);
1934 	wake_ack_receiver(connection);
1935 }
1936 
1937 extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *);
1938 extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *);
1939 extern int conn_send_command(struct drbd_connection *, struct drbd_socket *,
1940 			     enum drbd_packet, unsigned int, void *,
1941 			     unsigned int);
1942 extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *,
1943 			     enum drbd_packet, unsigned int, void *,
1944 			     unsigned int);
1945 
1946 extern int drbd_send_ping(struct drbd_connection *connection);
1947 extern int drbd_send_ping_ack(struct drbd_connection *connection);
1948 extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state);
1949 extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state);
1950 
1951 static inline void drbd_thread_stop(struct drbd_thread *thi)
1952 {
1953 	_drbd_thread_stop(thi, false, true);
1954 }
1955 
1956 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
1957 {
1958 	_drbd_thread_stop(thi, false, false);
1959 }
1960 
1961 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
1962 {
1963 	_drbd_thread_stop(thi, true, false);
1964 }
1965 
1966 /* counts how many answer packets packets we expect from our peer,
1967  * for either explicit application requests,
1968  * or implicit barrier packets as necessary.
1969  * increased:
1970  *  w_send_barrier
1971  *  _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1972  *    it is much easier and equally valid to count what we queue for the
1973  *    worker, even before it actually was queued or send.
1974  *    (drbd_make_request_common; recovery path on read io-error)
1975  * decreased:
1976  *  got_BarrierAck (respective tl_clear, tl_clear_barrier)
1977  *  _req_mod(req, DATA_RECEIVED)
1978  *     [from receive_DataReply]
1979  *  _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1980  *     [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1981  *     for some reason it is NOT decreased in got_NegAck,
1982  *     but in the resulting cleanup code from report_params.
1983  *     we should try to remember the reason for that...
1984  *  _req_mod(req, SEND_FAILED or SEND_CANCELED)
1985  *  _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
1986  *     [from tl_clear_barrier]
1987  */
1988 static inline void inc_ap_pending(struct drbd_device *device)
1989 {
1990 	atomic_inc(&device->ap_pending_cnt);
1991 }
1992 
1993 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line)			\
1994 	if (atomic_read(&device->which) < 0)				\
1995 		drbd_err(device, "in %s:%d: " #which " = %d < 0 !\n",	\
1996 			func, line,					\
1997 			atomic_read(&device->which))
1998 
1999 #define dec_ap_pending(device) _dec_ap_pending(device, __func__, __LINE__)
2000 static inline void _dec_ap_pending(struct drbd_device *device, const char *func, int line)
2001 {
2002 	if (atomic_dec_and_test(&device->ap_pending_cnt))
2003 		wake_up(&device->misc_wait);
2004 	ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line);
2005 }
2006 
2007 /* counts how many resync-related answers we still expect from the peer
2008  *		     increase			decrease
2009  * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
2010  * C_SYNC_SOURCE sends P_RS_DATA_REPLY   (and expects P_WRITE_ACK with ID_SYNCER)
2011  *					   (or P_NEG_ACK with ID_SYNCER)
2012  */
2013 static inline void inc_rs_pending(struct drbd_device *device)
2014 {
2015 	atomic_inc(&device->rs_pending_cnt);
2016 }
2017 
2018 #define dec_rs_pending(device) _dec_rs_pending(device, __func__, __LINE__)
2019 static inline void _dec_rs_pending(struct drbd_device *device, const char *func, int line)
2020 {
2021 	atomic_dec(&device->rs_pending_cnt);
2022 	ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line);
2023 }
2024 
2025 /* counts how many answers we still need to send to the peer.
2026  * increased on
2027  *  receive_Data	unless protocol A;
2028  *			we need to send a P_RECV_ACK (proto B)
2029  *			or P_WRITE_ACK (proto C)
2030  *  receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2031  *  receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2032  *  receive_Barrier_*	we need to send a P_BARRIER_ACK
2033  */
2034 static inline void inc_unacked(struct drbd_device *device)
2035 {
2036 	atomic_inc(&device->unacked_cnt);
2037 }
2038 
2039 #define dec_unacked(device) _dec_unacked(device, __func__, __LINE__)
2040 static inline void _dec_unacked(struct drbd_device *device, const char *func, int line)
2041 {
2042 	atomic_dec(&device->unacked_cnt);
2043 	ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
2044 }
2045 
2046 #define sub_unacked(device, n) _sub_unacked(device, n, __func__, __LINE__)
2047 static inline void _sub_unacked(struct drbd_device *device, int n, const char *func, int line)
2048 {
2049 	atomic_sub(n, &device->unacked_cnt);
2050 	ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
2051 }
2052 
2053 static inline bool is_sync_target_state(enum drbd_conns connection_state)
2054 {
2055 	return	connection_state == C_SYNC_TARGET ||
2056 		connection_state == C_PAUSED_SYNC_T;
2057 }
2058 
2059 static inline bool is_sync_source_state(enum drbd_conns connection_state)
2060 {
2061 	return	connection_state == C_SYNC_SOURCE ||
2062 		connection_state == C_PAUSED_SYNC_S;
2063 }
2064 
2065 static inline bool is_sync_state(enum drbd_conns connection_state)
2066 {
2067 	return	is_sync_source_state(connection_state) ||
2068 		is_sync_target_state(connection_state);
2069 }
2070 
2071 /**
2072  * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev
2073  * @_device:		DRBD device.
2074  * @_min_state:		Minimum device state required for success.
2075  *
2076  * You have to call put_ldev() when finished working with device->ldev.
2077  */
2078 #define get_ldev_if_state(_device, _min_state)				\
2079 	(_get_ldev_if_state((_device), (_min_state)) ?			\
2080 	 ({ __acquire(x); true; }) : false)
2081 #define get_ldev(_device) get_ldev_if_state(_device, D_INCONSISTENT)
2082 
2083 static inline void put_ldev(struct drbd_device *device)
2084 {
2085 	enum drbd_disk_state disk_state = device->state.disk;
2086 	/* We must check the state *before* the atomic_dec becomes visible,
2087 	 * or we have a theoretical race where someone hitting zero,
2088 	 * while state still D_FAILED, will then see D_DISKLESS in the
2089 	 * condition below and calling into destroy, where he must not, yet. */
2090 	int i = atomic_dec_return(&device->local_cnt);
2091 
2092 	/* This may be called from some endio handler,
2093 	 * so we must not sleep here. */
2094 
2095 	__release(local);
2096 	D_ASSERT(device, i >= 0);
2097 	if (i == 0) {
2098 		if (disk_state == D_DISKLESS)
2099 			/* even internal references gone, safe to destroy */
2100 			drbd_device_post_work(device, DESTROY_DISK);
2101 		if (disk_state == D_FAILED)
2102 			/* all application IO references gone. */
2103 			if (!test_and_set_bit(GOING_DISKLESS, &device->flags))
2104 				drbd_device_post_work(device, GO_DISKLESS);
2105 		wake_up(&device->misc_wait);
2106 	}
2107 }
2108 
2109 #ifndef __CHECKER__
2110 static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins)
2111 {
2112 	int io_allowed;
2113 
2114 	/* never get a reference while D_DISKLESS */
2115 	if (device->state.disk == D_DISKLESS)
2116 		return 0;
2117 
2118 	atomic_inc(&device->local_cnt);
2119 	io_allowed = (device->state.disk >= mins);
2120 	if (!io_allowed)
2121 		put_ldev(device);
2122 	return io_allowed;
2123 }
2124 #else
2125 extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins);
2126 #endif
2127 
2128 /* this throttles on-the-fly application requests
2129  * according to max_buffers settings;
2130  * maybe re-implement using semaphores? */
2131 static inline int drbd_get_max_buffers(struct drbd_device *device)
2132 {
2133 	struct net_conf *nc;
2134 	int mxb;
2135 
2136 	rcu_read_lock();
2137 	nc = rcu_dereference(first_peer_device(device)->connection->net_conf);
2138 	mxb = nc ? nc->max_buffers : 1000000;  /* arbitrary limit on open requests */
2139 	rcu_read_unlock();
2140 
2141 	return mxb;
2142 }
2143 
2144 static inline int drbd_state_is_stable(struct drbd_device *device)
2145 {
2146 	union drbd_dev_state s = device->state;
2147 
2148 	/* DO NOT add a default clause, we want the compiler to warn us
2149 	 * for any newly introduced state we may have forgotten to add here */
2150 
2151 	switch ((enum drbd_conns)s.conn) {
2152 	/* new io only accepted when there is no connection, ... */
2153 	case C_STANDALONE:
2154 	case C_WF_CONNECTION:
2155 	/* ... or there is a well established connection. */
2156 	case C_CONNECTED:
2157 	case C_SYNC_SOURCE:
2158 	case C_SYNC_TARGET:
2159 	case C_VERIFY_S:
2160 	case C_VERIFY_T:
2161 	case C_PAUSED_SYNC_S:
2162 	case C_PAUSED_SYNC_T:
2163 	case C_AHEAD:
2164 	case C_BEHIND:
2165 		/* transitional states, IO allowed */
2166 	case C_DISCONNECTING:
2167 	case C_UNCONNECTED:
2168 	case C_TIMEOUT:
2169 	case C_BROKEN_PIPE:
2170 	case C_NETWORK_FAILURE:
2171 	case C_PROTOCOL_ERROR:
2172 	case C_TEAR_DOWN:
2173 	case C_WF_REPORT_PARAMS:
2174 	case C_STARTING_SYNC_S:
2175 	case C_STARTING_SYNC_T:
2176 		break;
2177 
2178 		/* Allow IO in BM exchange states with new protocols */
2179 	case C_WF_BITMAP_S:
2180 		if (first_peer_device(device)->connection->agreed_pro_version < 96)
2181 			return 0;
2182 		break;
2183 
2184 		/* no new io accepted in these states */
2185 	case C_WF_BITMAP_T:
2186 	case C_WF_SYNC_UUID:
2187 	case C_MASK:
2188 		/* not "stable" */
2189 		return 0;
2190 	}
2191 
2192 	switch ((enum drbd_disk_state)s.disk) {
2193 	case D_DISKLESS:
2194 	case D_INCONSISTENT:
2195 	case D_OUTDATED:
2196 	case D_CONSISTENT:
2197 	case D_UP_TO_DATE:
2198 	case D_FAILED:
2199 		/* disk state is stable as well. */
2200 		break;
2201 
2202 	/* no new io accepted during transitional states */
2203 	case D_ATTACHING:
2204 	case D_NEGOTIATING:
2205 	case D_UNKNOWN:
2206 	case D_MASK:
2207 		/* not "stable" */
2208 		return 0;
2209 	}
2210 
2211 	return 1;
2212 }
2213 
2214 static inline int drbd_suspended(struct drbd_device *device)
2215 {
2216 	struct drbd_resource *resource = device->resource;
2217 
2218 	return resource->susp || resource->susp_fen || resource->susp_nod;
2219 }
2220 
2221 static inline bool may_inc_ap_bio(struct drbd_device *device)
2222 {
2223 	int mxb = drbd_get_max_buffers(device);
2224 
2225 	if (drbd_suspended(device))
2226 		return false;
2227 	if (atomic_read(&device->suspend_cnt))
2228 		return false;
2229 
2230 	/* to avoid potential deadlock or bitmap corruption,
2231 	 * in various places, we only allow new application io
2232 	 * to start during "stable" states. */
2233 
2234 	/* no new io accepted when attaching or detaching the disk */
2235 	if (!drbd_state_is_stable(device))
2236 		return false;
2237 
2238 	/* since some older kernels don't have atomic_add_unless,
2239 	 * and we are within the spinlock anyways, we have this workaround.  */
2240 	if (atomic_read(&device->ap_bio_cnt) > mxb)
2241 		return false;
2242 	if (test_bit(BITMAP_IO, &device->flags))
2243 		return false;
2244 	return true;
2245 }
2246 
2247 static inline bool inc_ap_bio_cond(struct drbd_device *device)
2248 {
2249 	bool rv = false;
2250 
2251 	spin_lock_irq(&device->resource->req_lock);
2252 	rv = may_inc_ap_bio(device);
2253 	if (rv)
2254 		atomic_inc(&device->ap_bio_cnt);
2255 	spin_unlock_irq(&device->resource->req_lock);
2256 
2257 	return rv;
2258 }
2259 
2260 static inline void inc_ap_bio(struct drbd_device *device)
2261 {
2262 	/* we wait here
2263 	 *    as long as the device is suspended
2264 	 *    until the bitmap is no longer on the fly during connection
2265 	 *    handshake as long as we would exceed the max_buffer limit.
2266 	 *
2267 	 * to avoid races with the reconnect code,
2268 	 * we need to atomic_inc within the spinlock. */
2269 
2270 	wait_event(device->misc_wait, inc_ap_bio_cond(device));
2271 }
2272 
2273 static inline void dec_ap_bio(struct drbd_device *device)
2274 {
2275 	int mxb = drbd_get_max_buffers(device);
2276 	int ap_bio = atomic_dec_return(&device->ap_bio_cnt);
2277 
2278 	D_ASSERT(device, ap_bio >= 0);
2279 
2280 	if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) {
2281 		if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags))
2282 			drbd_queue_work(&first_peer_device(device)->
2283 				connection->sender_work,
2284 				&device->bm_io_work.w);
2285 	}
2286 
2287 	/* this currently does wake_up for every dec_ap_bio!
2288 	 * maybe rather introduce some type of hysteresis?
2289 	 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2290 	if (ap_bio < mxb)
2291 		wake_up(&device->misc_wait);
2292 }
2293 
2294 static inline bool verify_can_do_stop_sector(struct drbd_device *device)
2295 {
2296 	return first_peer_device(device)->connection->agreed_pro_version >= 97 &&
2297 		first_peer_device(device)->connection->agreed_pro_version != 100;
2298 }
2299 
2300 static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val)
2301 {
2302 	int changed = device->ed_uuid != val;
2303 	device->ed_uuid = val;
2304 	return changed;
2305 }
2306 
2307 static inline int drbd_queue_order_type(struct drbd_device *device)
2308 {
2309 	/* sorry, we currently have no working implementation
2310 	 * of distributed TCQ stuff */
2311 #ifndef QUEUE_ORDERED_NONE
2312 #define QUEUE_ORDERED_NONE 0
2313 #endif
2314 	return QUEUE_ORDERED_NONE;
2315 }
2316 
2317 static inline struct drbd_connection *first_connection(struct drbd_resource *resource)
2318 {
2319 	return list_first_entry_or_null(&resource->connections,
2320 				struct drbd_connection, connections);
2321 }
2322 
2323 #endif
2324