xref: /linux/drivers/block/drbd/drbd_int.h (revision c145211d1f9e2ef19e7b4c2b943f68366daa97af)
1 /*
2   drbd_int.h
3 
4   This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
5 
6   Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
7   Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
8   Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
9 
10   drbd is free software; you can redistribute it and/or modify
11   it under the terms of the GNU General Public License as published by
12   the Free Software Foundation; either version 2, or (at your option)
13   any later version.
14 
15   drbd is distributed in the hope that it will be useful,
16   but WITHOUT ANY WARRANTY; without even the implied warranty of
17   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.	See the
18   GNU General Public License for more details.
19 
20   You should have received a copy of the GNU General Public License
21   along with drbd; see the file COPYING.  If not, write to
22   the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
23 
24 */
25 
26 #ifndef _DRBD_INT_H
27 #define _DRBD_INT_H
28 
29 #include <linux/compiler.h>
30 #include <linux/types.h>
31 #include <linux/version.h>
32 #include <linux/list.h>
33 #include <linux/sched.h>
34 #include <linux/bitops.h>
35 #include <linux/slab.h>
36 #include <linux/crypto.h>
37 #include <linux/ratelimit.h>
38 #include <linux/tcp.h>
39 #include <linux/mutex.h>
40 #include <linux/major.h>
41 #include <linux/blkdev.h>
42 #include <linux/genhd.h>
43 #include <net/tcp.h>
44 #include <linux/lru_cache.h>
45 
46 #ifdef __CHECKER__
47 # define __protected_by(x)       __attribute__((require_context(x,1,999,"rdwr")))
48 # define __protected_read_by(x)  __attribute__((require_context(x,1,999,"read")))
49 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
50 # define __must_hold(x)       __attribute__((context(x,1,1), require_context(x,1,999,"call")))
51 #else
52 # define __protected_by(x)
53 # define __protected_read_by(x)
54 # define __protected_write_by(x)
55 # define __must_hold(x)
56 #endif
57 
58 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
59 
60 /* module parameter, defined in drbd_main.c */
61 extern unsigned int minor_count;
62 extern int disable_sendpage;
63 extern int allow_oos;
64 extern unsigned int cn_idx;
65 
66 #ifdef CONFIG_DRBD_FAULT_INJECTION
67 extern int enable_faults;
68 extern int fault_rate;
69 extern int fault_devs;
70 #endif
71 
72 extern char usermode_helper[];
73 
74 
75 #ifndef TRUE
76 #define TRUE 1
77 #endif
78 #ifndef FALSE
79 #define FALSE 0
80 #endif
81 
82 /* I don't remember why XCPU ...
83  * This is used to wake the asender,
84  * and to interrupt sending the sending task
85  * on disconnect.
86  */
87 #define DRBD_SIG SIGXCPU
88 
89 /* This is used to stop/restart our threads.
90  * Cannot use SIGTERM nor SIGKILL, since these
91  * are sent out by init on runlevel changes
92  * I choose SIGHUP for now.
93  */
94 #define DRBD_SIGKILL SIGHUP
95 
96 /* All EEs on the free list should have ID_VACANT (== 0)
97  * freshly allocated EEs get !ID_VACANT (== 1)
98  * so if it says "cannot dereference null pointer at address 0x00000001",
99  * it is most likely one of these :( */
100 
101 #define ID_IN_SYNC      (4711ULL)
102 #define ID_OUT_OF_SYNC  (4712ULL)
103 
104 #define ID_SYNCER (-1ULL)
105 #define ID_VACANT 0
106 #define is_syncer_block_id(id) ((id) == ID_SYNCER)
107 
108 struct drbd_conf;
109 
110 
111 /* to shorten dev_warn(DEV, "msg"); and relatives statements */
112 #define DEV (disk_to_dev(mdev->vdisk))
113 
114 #define D_ASSERT(exp)	if (!(exp)) \
115 	 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
116 
117 #define ERR_IF(exp) if (({				\
118 	int _b = (exp) != 0;				\
119 	if (_b) dev_err(DEV, "%s: (%s) in %s:%d\n",	\
120 		__func__, #exp, __FILE__, __LINE__);	\
121 	 _b;						\
122 	}))
123 
124 /* Defines to control fault insertion */
125 enum {
126 	DRBD_FAULT_MD_WR = 0,	/* meta data write */
127 	DRBD_FAULT_MD_RD = 1,	/*           read  */
128 	DRBD_FAULT_RS_WR = 2,	/* resync          */
129 	DRBD_FAULT_RS_RD = 3,
130 	DRBD_FAULT_DT_WR = 4,	/* data            */
131 	DRBD_FAULT_DT_RD = 5,
132 	DRBD_FAULT_DT_RA = 6,	/* data read ahead */
133 	DRBD_FAULT_BM_ALLOC = 7,	/* bitmap allocation */
134 	DRBD_FAULT_AL_EE = 8,	/* alloc ee */
135 
136 	DRBD_FAULT_MAX,
137 };
138 
139 #ifdef CONFIG_DRBD_FAULT_INJECTION
140 extern unsigned int
141 _drbd_insert_fault(struct drbd_conf *mdev, unsigned int type);
142 static inline int
143 drbd_insert_fault(struct drbd_conf *mdev, unsigned int type) {
144 	return fault_rate &&
145 		(enable_faults & (1<<type)) &&
146 		_drbd_insert_fault(mdev, type);
147 }
148 #define FAULT_ACTIVE(_m, _t) (drbd_insert_fault((_m), (_t)))
149 
150 #else
151 #define FAULT_ACTIVE(_m, _t) (0)
152 #endif
153 
154 /* integer division, round _UP_ to the next integer */
155 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
156 /* usual integer division */
157 #define div_floor(A, B) ((A)/(B))
158 
159 /* drbd_meta-data.c (still in drbd_main.c) */
160 /* 4th incarnation of the disk layout. */
161 #define DRBD_MD_MAGIC (DRBD_MAGIC+4)
162 
163 extern struct drbd_conf **minor_table;
164 extern struct ratelimit_state drbd_ratelimit_state;
165 
166 /* on the wire */
167 enum drbd_packets {
168 	/* receiver (data socket) */
169 	P_DATA		      = 0x00,
170 	P_DATA_REPLY	      = 0x01, /* Response to P_DATA_REQUEST */
171 	P_RS_DATA_REPLY	      = 0x02, /* Response to P_RS_DATA_REQUEST */
172 	P_BARRIER	      = 0x03,
173 	P_BITMAP	      = 0x04,
174 	P_BECOME_SYNC_TARGET  = 0x05,
175 	P_BECOME_SYNC_SOURCE  = 0x06,
176 	P_UNPLUG_REMOTE	      = 0x07, /* Used at various times to hint the peer */
177 	P_DATA_REQUEST	      = 0x08, /* Used to ask for a data block */
178 	P_RS_DATA_REQUEST     = 0x09, /* Used to ask for a data block for resync */
179 	P_SYNC_PARAM	      = 0x0a,
180 	P_PROTOCOL	      = 0x0b,
181 	P_UUIDS		      = 0x0c,
182 	P_SIZES		      = 0x0d,
183 	P_STATE		      = 0x0e,
184 	P_SYNC_UUID	      = 0x0f,
185 	P_AUTH_CHALLENGE      = 0x10,
186 	P_AUTH_RESPONSE	      = 0x11,
187 	P_STATE_CHG_REQ	      = 0x12,
188 
189 	/* asender (meta socket */
190 	P_PING		      = 0x13,
191 	P_PING_ACK	      = 0x14,
192 	P_RECV_ACK	      = 0x15, /* Used in protocol B */
193 	P_WRITE_ACK	      = 0x16, /* Used in protocol C */
194 	P_RS_WRITE_ACK	      = 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
195 	P_DISCARD_ACK	      = 0x18, /* Used in proto C, two-primaries conflict detection */
196 	P_NEG_ACK	      = 0x19, /* Sent if local disk is unusable */
197 	P_NEG_DREPLY	      = 0x1a, /* Local disk is broken... */
198 	P_NEG_RS_DREPLY	      = 0x1b, /* Local disk is broken... */
199 	P_BARRIER_ACK	      = 0x1c,
200 	P_STATE_CHG_REPLY     = 0x1d,
201 
202 	/* "new" commands, no longer fitting into the ordering scheme above */
203 
204 	P_OV_REQUEST	      = 0x1e, /* data socket */
205 	P_OV_REPLY	      = 0x1f,
206 	P_OV_RESULT	      = 0x20, /* meta socket */
207 	P_CSUM_RS_REQUEST     = 0x21, /* data socket */
208 	P_RS_IS_IN_SYNC	      = 0x22, /* meta socket */
209 	P_SYNC_PARAM89	      = 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
210 	P_COMPRESSED_BITMAP   = 0x24, /* compressed or otherwise encoded bitmap transfer */
211 
212 	P_MAX_CMD	      = 0x25,
213 	P_MAY_IGNORE	      = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
214 	P_MAX_OPT_CMD	      = 0x101,
215 
216 	/* special command ids for handshake */
217 
218 	P_HAND_SHAKE_M	      = 0xfff1, /* First Packet on the MetaSock */
219 	P_HAND_SHAKE_S	      = 0xfff2, /* First Packet on the Socket */
220 
221 	P_HAND_SHAKE	      = 0xfffe	/* FIXED for the next century! */
222 };
223 
224 static inline const char *cmdname(enum drbd_packets cmd)
225 {
226 	/* THINK may need to become several global tables
227 	 * when we want to support more than
228 	 * one PRO_VERSION */
229 	static const char *cmdnames[] = {
230 		[P_DATA]	        = "Data",
231 		[P_DATA_REPLY]	        = "DataReply",
232 		[P_RS_DATA_REPLY]	= "RSDataReply",
233 		[P_BARRIER]	        = "Barrier",
234 		[P_BITMAP]	        = "ReportBitMap",
235 		[P_BECOME_SYNC_TARGET]  = "BecomeSyncTarget",
236 		[P_BECOME_SYNC_SOURCE]  = "BecomeSyncSource",
237 		[P_UNPLUG_REMOTE]	= "UnplugRemote",
238 		[P_DATA_REQUEST]	= "DataRequest",
239 		[P_RS_DATA_REQUEST]     = "RSDataRequest",
240 		[P_SYNC_PARAM]	        = "SyncParam",
241 		[P_SYNC_PARAM89]	= "SyncParam89",
242 		[P_PROTOCOL]            = "ReportProtocol",
243 		[P_UUIDS]	        = "ReportUUIDs",
244 		[P_SIZES]	        = "ReportSizes",
245 		[P_STATE]	        = "ReportState",
246 		[P_SYNC_UUID]           = "ReportSyncUUID",
247 		[P_AUTH_CHALLENGE]      = "AuthChallenge",
248 		[P_AUTH_RESPONSE]	= "AuthResponse",
249 		[P_PING]		= "Ping",
250 		[P_PING_ACK]	        = "PingAck",
251 		[P_RECV_ACK]	        = "RecvAck",
252 		[P_WRITE_ACK]	        = "WriteAck",
253 		[P_RS_WRITE_ACK]	= "RSWriteAck",
254 		[P_DISCARD_ACK]	        = "DiscardAck",
255 		[P_NEG_ACK]	        = "NegAck",
256 		[P_NEG_DREPLY]	        = "NegDReply",
257 		[P_NEG_RS_DREPLY]	= "NegRSDReply",
258 		[P_BARRIER_ACK]	        = "BarrierAck",
259 		[P_STATE_CHG_REQ]       = "StateChgRequest",
260 		[P_STATE_CHG_REPLY]     = "StateChgReply",
261 		[P_OV_REQUEST]          = "OVRequest",
262 		[P_OV_REPLY]            = "OVReply",
263 		[P_OV_RESULT]           = "OVResult",
264 		[P_CSUM_RS_REQUEST]     = "CsumRSRequest",
265 		[P_RS_IS_IN_SYNC]	= "CsumRSIsInSync",
266 		[P_COMPRESSED_BITMAP]   = "CBitmap",
267 		[P_MAX_CMD]	        = NULL,
268 	};
269 
270 	if (cmd == P_HAND_SHAKE_M)
271 		return "HandShakeM";
272 	if (cmd == P_HAND_SHAKE_S)
273 		return "HandShakeS";
274 	if (cmd == P_HAND_SHAKE)
275 		return "HandShake";
276 	if (cmd >= P_MAX_CMD)
277 		return "Unknown";
278 	return cmdnames[cmd];
279 }
280 
281 /* for sending/receiving the bitmap,
282  * possibly in some encoding scheme */
283 struct bm_xfer_ctx {
284 	/* "const"
285 	 * stores total bits and long words
286 	 * of the bitmap, so we don't need to
287 	 * call the accessor functions over and again. */
288 	unsigned long bm_bits;
289 	unsigned long bm_words;
290 	/* during xfer, current position within the bitmap */
291 	unsigned long bit_offset;
292 	unsigned long word_offset;
293 
294 	/* statistics; index: (h->command == P_BITMAP) */
295 	unsigned packets[2];
296 	unsigned bytes[2];
297 };
298 
299 extern void INFO_bm_xfer_stats(struct drbd_conf *mdev,
300 		const char *direction, struct bm_xfer_ctx *c);
301 
302 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
303 {
304 	/* word_offset counts "native long words" (32 or 64 bit),
305 	 * aligned at 64 bit.
306 	 * Encoded packet may end at an unaligned bit offset.
307 	 * In case a fallback clear text packet is transmitted in
308 	 * between, we adjust this offset back to the last 64bit
309 	 * aligned "native long word", which makes coding and decoding
310 	 * the plain text bitmap much more convenient.  */
311 #if BITS_PER_LONG == 64
312 	c->word_offset = c->bit_offset >> 6;
313 #elif BITS_PER_LONG == 32
314 	c->word_offset = c->bit_offset >> 5;
315 	c->word_offset &= ~(1UL);
316 #else
317 # error "unsupported BITS_PER_LONG"
318 #endif
319 }
320 
321 #ifndef __packed
322 #define __packed __attribute__((packed))
323 #endif
324 
325 /* This is the layout for a packet on the wire.
326  * The byteorder is the network byte order.
327  *     (except block_id and barrier fields.
328  *	these are pointers to local structs
329  *	and have no relevance for the partner,
330  *	which just echoes them as received.)
331  *
332  * NOTE that the payload starts at a long aligned offset,
333  * regardless of 32 or 64 bit arch!
334  */
335 struct p_header {
336 	u32	  magic;
337 	u16	  command;
338 	u16	  length;	/* bytes of data after this header */
339 	u8	  payload[0];
340 } __packed;
341 /* 8 bytes. packet FIXED for the next century! */
342 
343 /*
344  * short commands, packets without payload, plain p_header:
345  *   P_PING
346  *   P_PING_ACK
347  *   P_BECOME_SYNC_TARGET
348  *   P_BECOME_SYNC_SOURCE
349  *   P_UNPLUG_REMOTE
350  */
351 
352 /*
353  * commands with out-of-struct payload:
354  *   P_BITMAP    (no additional fields)
355  *   P_DATA, P_DATA_REPLY (see p_data)
356  *   P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
357  */
358 
359 /* these defines must not be changed without changing the protocol version */
360 #define DP_HARDBARRIER	      1
361 #define DP_RW_SYNC	      2
362 #define DP_MAY_SET_IN_SYNC    4
363 
364 struct p_data {
365 	struct p_header head;
366 	u64	    sector;    /* 64 bits sector number */
367 	u64	    block_id;  /* to identify the request in protocol B&C */
368 	u32	    seq_num;
369 	u32	    dp_flags;
370 } __packed;
371 
372 /*
373  * commands which share a struct:
374  *  p_block_ack:
375  *   P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
376  *   P_DISCARD_ACK (proto C, two-primaries conflict detection)
377  *  p_block_req:
378  *   P_DATA_REQUEST, P_RS_DATA_REQUEST
379  */
380 struct p_block_ack {
381 	struct p_header head;
382 	u64	    sector;
383 	u64	    block_id;
384 	u32	    blksize;
385 	u32	    seq_num;
386 } __packed;
387 
388 
389 struct p_block_req {
390 	struct p_header head;
391 	u64 sector;
392 	u64 block_id;
393 	u32 blksize;
394 	u32 pad;	/* to multiple of 8 Byte */
395 } __packed;
396 
397 /*
398  * commands with their own struct for additional fields:
399  *   P_HAND_SHAKE
400  *   P_BARRIER
401  *   P_BARRIER_ACK
402  *   P_SYNC_PARAM
403  *   ReportParams
404  */
405 
406 struct p_handshake {
407 	struct p_header head;	/* 8 bytes */
408 	u32 protocol_min;
409 	u32 feature_flags;
410 	u32 protocol_max;
411 
412 	/* should be more than enough for future enhancements
413 	 * for now, feature_flags and the reserverd array shall be zero.
414 	 */
415 
416 	u32 _pad;
417 	u64 reserverd[7];
418 } __packed;
419 /* 80 bytes, FIXED for the next century */
420 
421 struct p_barrier {
422 	struct p_header head;
423 	u32 barrier;	/* barrier number _handle_ only */
424 	u32 pad;	/* to multiple of 8 Byte */
425 } __packed;
426 
427 struct p_barrier_ack {
428 	struct p_header head;
429 	u32 barrier;
430 	u32 set_size;
431 } __packed;
432 
433 struct p_rs_param {
434 	struct p_header head;
435 	u32 rate;
436 
437 	      /* Since protocol version 88 and higher. */
438 	char verify_alg[0];
439 } __packed;
440 
441 struct p_rs_param_89 {
442 	struct p_header head;
443 	u32 rate;
444         /* protocol version 89: */
445 	char verify_alg[SHARED_SECRET_MAX];
446 	char csums_alg[SHARED_SECRET_MAX];
447 } __packed;
448 
449 enum drbd_conn_flags {
450 	CF_WANT_LOSE = 1,
451 	CF_DRY_RUN = 2,
452 };
453 
454 struct p_protocol {
455 	struct p_header head;
456 	u32 protocol;
457 	u32 after_sb_0p;
458 	u32 after_sb_1p;
459 	u32 after_sb_2p;
460 	u32 conn_flags;
461 	u32 two_primaries;
462 
463               /* Since protocol version 87 and higher. */
464 	char integrity_alg[0];
465 
466 } __packed;
467 
468 struct p_uuids {
469 	struct p_header head;
470 	u64 uuid[UI_EXTENDED_SIZE];
471 } __packed;
472 
473 struct p_rs_uuid {
474 	struct p_header head;
475 	u64	    uuid;
476 } __packed;
477 
478 struct p_sizes {
479 	struct p_header head;
480 	u64	    d_size;  /* size of disk */
481 	u64	    u_size;  /* user requested size */
482 	u64	    c_size;  /* current exported size */
483 	u32	    max_segment_size;  /* Maximal size of a BIO */
484 	u32	    queue_order_type;
485 } __packed;
486 
487 struct p_state {
488 	struct p_header head;
489 	u32	    state;
490 } __packed;
491 
492 struct p_req_state {
493 	struct p_header head;
494 	u32	    mask;
495 	u32	    val;
496 } __packed;
497 
498 struct p_req_state_reply {
499 	struct p_header head;
500 	u32	    retcode;
501 } __packed;
502 
503 struct p_drbd06_param {
504 	u64	  size;
505 	u32	  state;
506 	u32	  blksize;
507 	u32	  protocol;
508 	u32	  version;
509 	u32	  gen_cnt[5];
510 	u32	  bit_map_gen[5];
511 } __packed;
512 
513 struct p_discard {
514 	struct p_header head;
515 	u64	    block_id;
516 	u32	    seq_num;
517 	u32	    pad;
518 } __packed;
519 
520 /* Valid values for the encoding field.
521  * Bump proto version when changing this. */
522 enum drbd_bitmap_code {
523 	/* RLE_VLI_Bytes = 0,
524 	 * and other bit variants had been defined during
525 	 * algorithm evaluation. */
526 	RLE_VLI_Bits = 2,
527 };
528 
529 struct p_compressed_bm {
530 	struct p_header head;
531 	/* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
532 	 * (encoding & 0x80): polarity (set/unset) of first runlength
533 	 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
534 	 * used to pad up to head.length bytes
535 	 */
536 	u8 encoding;
537 
538 	u8 code[0];
539 } __packed;
540 
541 /* DCBP: Drbd Compressed Bitmap Packet ... */
542 static inline enum drbd_bitmap_code
543 DCBP_get_code(struct p_compressed_bm *p)
544 {
545 	return (enum drbd_bitmap_code)(p->encoding & 0x0f);
546 }
547 
548 static inline void
549 DCBP_set_code(struct p_compressed_bm *p, enum drbd_bitmap_code code)
550 {
551 	BUG_ON(code & ~0xf);
552 	p->encoding = (p->encoding & ~0xf) | code;
553 }
554 
555 static inline int
556 DCBP_get_start(struct p_compressed_bm *p)
557 {
558 	return (p->encoding & 0x80) != 0;
559 }
560 
561 static inline void
562 DCBP_set_start(struct p_compressed_bm *p, int set)
563 {
564 	p->encoding = (p->encoding & ~0x80) | (set ? 0x80 : 0);
565 }
566 
567 static inline int
568 DCBP_get_pad_bits(struct p_compressed_bm *p)
569 {
570 	return (p->encoding >> 4) & 0x7;
571 }
572 
573 static inline void
574 DCBP_set_pad_bits(struct p_compressed_bm *p, int n)
575 {
576 	BUG_ON(n & ~0x7);
577 	p->encoding = (p->encoding & (~0x7 << 4)) | (n << 4);
578 }
579 
580 /* one bitmap packet, including the p_header,
581  * should fit within one _architecture independend_ page.
582  * so we need to use the fixed size 4KiB page size
583  * most architechtures have used for a long time.
584  */
585 #define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header))
586 #define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
587 #define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm))
588 #if (PAGE_SIZE < 4096)
589 /* drbd_send_bitmap / receive_bitmap would break horribly */
590 #error "PAGE_SIZE too small"
591 #endif
592 
593 union p_polymorph {
594         struct p_header          header;
595         struct p_handshake       handshake;
596         struct p_data            data;
597         struct p_block_ack       block_ack;
598         struct p_barrier         barrier;
599         struct p_barrier_ack     barrier_ack;
600         struct p_rs_param_89     rs_param_89;
601         struct p_protocol        protocol;
602         struct p_sizes           sizes;
603         struct p_uuids           uuids;
604         struct p_state           state;
605         struct p_req_state       req_state;
606         struct p_req_state_reply req_state_reply;
607         struct p_block_req       block_req;
608 } __packed;
609 
610 /**********************************************************************/
611 enum drbd_thread_state {
612 	None,
613 	Running,
614 	Exiting,
615 	Restarting
616 };
617 
618 struct drbd_thread {
619 	spinlock_t t_lock;
620 	struct task_struct *task;
621 	struct completion stop;
622 	enum drbd_thread_state t_state;
623 	int (*function) (struct drbd_thread *);
624 	struct drbd_conf *mdev;
625 	int reset_cpu_mask;
626 };
627 
628 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
629 {
630 	/* THINK testing the t_state seems to be uncritical in all cases
631 	 * (but thread_{start,stop}), so we can read it *without* the lock.
632 	 *	--lge */
633 
634 	smp_rmb();
635 	return thi->t_state;
636 }
637 
638 
639 /*
640  * Having this as the first member of a struct provides sort of "inheritance".
641  * "derived" structs can be "drbd_queue_work()"ed.
642  * The callback should know and cast back to the descendant struct.
643  * drbd_request and drbd_epoch_entry are descendants of drbd_work.
644  */
645 struct drbd_work;
646 typedef int (*drbd_work_cb)(struct drbd_conf *, struct drbd_work *, int cancel);
647 struct drbd_work {
648 	struct list_head list;
649 	drbd_work_cb cb;
650 };
651 
652 struct drbd_tl_epoch;
653 struct drbd_request {
654 	struct drbd_work w;
655 	struct drbd_conf *mdev;
656 
657 	/* if local IO is not allowed, will be NULL.
658 	 * if local IO _is_ allowed, holds the locally submitted bio clone,
659 	 * or, after local IO completion, the ERR_PTR(error).
660 	 * see drbd_endio_pri(). */
661 	struct bio *private_bio;
662 
663 	struct hlist_node colision;
664 	sector_t sector;
665 	unsigned int size;
666 	unsigned int epoch; /* barrier_nr */
667 
668 	/* barrier_nr: used to check on "completion" whether this req was in
669 	 * the current epoch, and we therefore have to close it,
670 	 * starting a new epoch...
671 	 */
672 
673 	/* up to here, the struct layout is identical to drbd_epoch_entry;
674 	 * we might be able to use that to our advantage...  */
675 
676 	struct list_head tl_requests; /* ring list in the transfer log */
677 	struct bio *master_bio;       /* master bio pointer */
678 	unsigned long rq_state; /* see comments above _req_mod() */
679 	int seq_num;
680 	unsigned long start_time;
681 };
682 
683 struct drbd_tl_epoch {
684 	struct drbd_work w;
685 	struct list_head requests; /* requests before */
686 	struct drbd_tl_epoch *next; /* pointer to the next barrier */
687 	unsigned int br_number;  /* the barriers identifier. */
688 	int n_req;	/* number of requests attached before this barrier */
689 };
690 
691 struct drbd_request;
692 
693 /* These Tl_epoch_entries may be in one of 6 lists:
694    active_ee .. data packet being written
695    sync_ee   .. syncer block being written
696    done_ee   .. block written, need to send P_WRITE_ACK
697    read_ee   .. [RS]P_DATA_REQUEST being read
698 */
699 
700 struct drbd_epoch {
701 	struct list_head list;
702 	unsigned int barrier_nr;
703 	atomic_t epoch_size; /* increased on every request added. */
704 	atomic_t active;     /* increased on every req. added, and dec on every finished. */
705 	unsigned long flags;
706 };
707 
708 /* drbd_epoch flag bits */
709 enum {
710 	DE_BARRIER_IN_NEXT_EPOCH_ISSUED,
711 	DE_BARRIER_IN_NEXT_EPOCH_DONE,
712 	DE_CONTAINS_A_BARRIER,
713 	DE_HAVE_BARRIER_NUMBER,
714 	DE_IS_FINISHING,
715 };
716 
717 enum epoch_event {
718 	EV_PUT,
719 	EV_GOT_BARRIER_NR,
720 	EV_BARRIER_DONE,
721 	EV_BECAME_LAST,
722 	EV_CLEANUP = 32, /* used as flag */
723 };
724 
725 struct drbd_epoch_entry {
726 	struct drbd_work    w;
727 	struct drbd_conf *mdev;
728 	struct bio *private_bio;
729 	struct hlist_node colision;
730 	sector_t sector;
731 	unsigned int size;
732 	struct drbd_epoch *epoch;
733 
734 	/* up to here, the struct layout is identical to drbd_request;
735 	 * we might be able to use that to our advantage...  */
736 
737 	unsigned int flags;
738 	u64    block_id;
739 };
740 
741 struct drbd_wq_barrier {
742 	struct drbd_work w;
743 	struct completion done;
744 };
745 
746 struct digest_info {
747 	int digest_size;
748 	void *digest;
749 };
750 
751 /* ee flag bits */
752 enum {
753 	__EE_CALL_AL_COMPLETE_IO,
754 	__EE_CONFLICT_PENDING,
755 	__EE_MAY_SET_IN_SYNC,
756 	__EE_IS_BARRIER,
757 };
758 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
759 #define EE_CONFLICT_PENDING    (1<<__EE_CONFLICT_PENDING)
760 #define EE_MAY_SET_IN_SYNC     (1<<__EE_MAY_SET_IN_SYNC)
761 #define EE_IS_BARRIER          (1<<__EE_IS_BARRIER)
762 
763 /* global flag bits */
764 enum {
765 	CREATE_BARRIER,		/* next P_DATA is preceeded by a P_BARRIER */
766 	SIGNAL_ASENDER,		/* whether asender wants to be interrupted */
767 	SEND_PING,		/* whether asender should send a ping asap */
768 
769 	STOP_SYNC_TIMER,	/* tell timer to cancel itself */
770 	UNPLUG_QUEUED,		/* only relevant with kernel 2.4 */
771 	UNPLUG_REMOTE,		/* sending a "UnplugRemote" could help */
772 	MD_DIRTY,		/* current uuids and flags not yet on disk */
773 	DISCARD_CONCURRENT,	/* Set on one node, cleared on the peer! */
774 	USE_DEGR_WFC_T,		/* degr-wfc-timeout instead of wfc-timeout. */
775 	CLUSTER_ST_CHANGE,	/* Cluster wide state change going on... */
776 	CL_ST_CHG_SUCCESS,
777 	CL_ST_CHG_FAIL,
778 	CRASHED_PRIMARY,	/* This node was a crashed primary.
779 				 * Gets cleared when the state.conn
780 				 * goes into C_CONNECTED state. */
781 	WRITE_BM_AFTER_RESYNC,	/* A kmalloc() during resync failed */
782 	NO_BARRIER_SUPP,	/* underlying block device doesn't implement barriers */
783 	CONSIDER_RESYNC,
784 
785 	MD_NO_BARRIER,		/* meta data device does not support barriers,
786 				   so don't even try */
787 	SUSPEND_IO,		/* suspend application io */
788 	BITMAP_IO,		/* suspend application io;
789 				   once no more io in flight, start bitmap io */
790 	BITMAP_IO_QUEUED,       /* Started bitmap IO */
791 	RESYNC_AFTER_NEG,       /* Resync after online grow after the attach&negotiate finished. */
792 	NET_CONGESTED,		/* The data socket is congested */
793 
794 	CONFIG_PENDING,		/* serialization of (re)configuration requests.
795 				 * if set, also prevents the device from dying */
796 	DEVICE_DYING,		/* device became unconfigured,
797 				 * but worker thread is still handling the cleanup.
798 				 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
799 				 * while this is set. */
800 	RESIZE_PENDING,		/* Size change detected locally, waiting for the response from
801 				 * the peer, if it changed there as well. */
802 	CONN_DRY_RUN,		/* Expect disconnect after resync handshake. */
803 	GOT_PING_ACK,		/* set when we receive a ping_ack packet, misc wait gets woken */
804 };
805 
806 struct drbd_bitmap; /* opaque for drbd_conf */
807 
808 /* TODO sort members for performance
809  * MAYBE group them further */
810 
811 /* THINK maybe we actually want to use the default "event/%s" worker threads
812  * or similar in linux 2.6, which uses per cpu data and threads.
813  *
814  * To be general, this might need a spin_lock member.
815  * For now, please use the mdev->req_lock to protect list_head,
816  * see drbd_queue_work below.
817  */
818 struct drbd_work_queue {
819 	struct list_head q;
820 	struct semaphore s; /* producers up it, worker down()s it */
821 	spinlock_t q_lock;  /* to protect the list. */
822 };
823 
824 struct drbd_socket {
825 	struct drbd_work_queue work;
826 	struct mutex mutex;
827 	struct socket    *socket;
828 	/* this way we get our
829 	 * send/receive buffers off the stack */
830 	union p_polymorph sbuf;
831 	union p_polymorph rbuf;
832 };
833 
834 struct drbd_md {
835 	u64 md_offset;		/* sector offset to 'super' block */
836 
837 	u64 la_size_sect;	/* last agreed size, unit sectors */
838 	u64 uuid[UI_SIZE];
839 	u64 device_uuid;
840 	u32 flags;
841 	u32 md_size_sect;
842 
843 	s32 al_offset;	/* signed relative sector offset to al area */
844 	s32 bm_offset;	/* signed relative sector offset to bitmap */
845 
846 	/* u32 al_nr_extents;	   important for restoring the AL
847 	 * is stored into  sync_conf.al_extents, which in turn
848 	 * gets applied to act_log->nr_elements
849 	 */
850 };
851 
852 /* for sync_conf and other types... */
853 #define NL_PACKET(name, number, fields) struct name { fields };
854 #define NL_INTEGER(pn,pr,member) int member;
855 #define NL_INT64(pn,pr,member) __u64 member;
856 #define NL_BIT(pn,pr,member)   unsigned member:1;
857 #define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
858 #include "linux/drbd_nl.h"
859 
860 struct drbd_backing_dev {
861 	struct block_device *backing_bdev;
862 	struct block_device *md_bdev;
863 	struct file *lo_file;
864 	struct file *md_file;
865 	struct drbd_md md;
866 	struct disk_conf dc; /* The user provided config... */
867 	sector_t known_size; /* last known size of that backing device */
868 };
869 
870 struct drbd_md_io {
871 	struct drbd_conf *mdev;
872 	struct completion event;
873 	int error;
874 };
875 
876 struct bm_io_work {
877 	struct drbd_work w;
878 	char *why;
879 	int (*io_fn)(struct drbd_conf *mdev);
880 	void (*done)(struct drbd_conf *mdev, int rv);
881 };
882 
883 enum write_ordering_e {
884 	WO_none,
885 	WO_drain_io,
886 	WO_bdev_flush,
887 	WO_bio_barrier
888 };
889 
890 struct drbd_conf {
891 	/* things that are stored as / read from meta data on disk */
892 	unsigned long flags;
893 
894 	/* configured by drbdsetup */
895 	struct net_conf *net_conf; /* protected by get_net_conf() and put_net_conf() */
896 	struct syncer_conf sync_conf;
897 	struct drbd_backing_dev *ldev __protected_by(local);
898 
899 	sector_t p_size;     /* partner's disk size */
900 	struct request_queue *rq_queue;
901 	struct block_device *this_bdev;
902 	struct gendisk	    *vdisk;
903 
904 	struct drbd_socket data; /* data/barrier/cstate/parameter packets */
905 	struct drbd_socket meta; /* ping/ack (metadata) packets */
906 	int agreed_pro_version;  /* actually used protocol version */
907 	unsigned long last_received; /* in jiffies, either socket */
908 	unsigned int ko_count;
909 	struct drbd_work  resync_work,
910 			  unplug_work,
911 			  md_sync_work;
912 	struct timer_list resync_timer;
913 	struct timer_list md_sync_timer;
914 
915 	/* Used after attach while negotiating new disk state. */
916 	union drbd_state new_state_tmp;
917 
918 	union drbd_state state;
919 	wait_queue_head_t misc_wait;
920 	wait_queue_head_t state_wait;  /* upon each state change. */
921 	unsigned int send_cnt;
922 	unsigned int recv_cnt;
923 	unsigned int read_cnt;
924 	unsigned int writ_cnt;
925 	unsigned int al_writ_cnt;
926 	unsigned int bm_writ_cnt;
927 	atomic_t ap_bio_cnt;	 /* Requests we need to complete */
928 	atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
929 	atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
930 	atomic_t unacked_cnt;	 /* Need to send replys for */
931 	atomic_t local_cnt;	 /* Waiting for local completion */
932 	atomic_t net_cnt;	 /* Users of net_conf */
933 	spinlock_t req_lock;
934 	struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */
935 	struct drbd_tl_epoch *newest_tle;
936 	struct drbd_tl_epoch *oldest_tle;
937 	struct list_head out_of_sequence_requests;
938 	struct hlist_head *tl_hash;
939 	unsigned int tl_hash_s;
940 
941 	/* blocks to sync in this run [unit BM_BLOCK_SIZE] */
942 	unsigned long rs_total;
943 	/* number of sync IOs that failed in this run */
944 	unsigned long rs_failed;
945 	/* Syncer's start time [unit jiffies] */
946 	unsigned long rs_start;
947 	/* cumulated time in PausedSyncX state [unit jiffies] */
948 	unsigned long rs_paused;
949 	/* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
950 	unsigned long rs_mark_left;
951 	/* marks's time [unit jiffies] */
952 	unsigned long rs_mark_time;
953 	/* skipped because csum was equeal [unit BM_BLOCK_SIZE] */
954 	unsigned long rs_same_csum;
955 
956 	/* where does the admin want us to start? (sector) */
957 	sector_t ov_start_sector;
958 	/* where are we now? (sector) */
959 	sector_t ov_position;
960 	/* Start sector of out of sync range (to merge printk reporting). */
961 	sector_t ov_last_oos_start;
962 	/* size of out-of-sync range in sectors. */
963 	sector_t ov_last_oos_size;
964 	unsigned long ov_left; /* in bits */
965 	struct crypto_hash *csums_tfm;
966 	struct crypto_hash *verify_tfm;
967 
968 	struct drbd_thread receiver;
969 	struct drbd_thread worker;
970 	struct drbd_thread asender;
971 	struct drbd_bitmap *bitmap;
972 	unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
973 
974 	/* Used to track operations of resync... */
975 	struct lru_cache *resync;
976 	/* Number of locked elements in resync LRU */
977 	unsigned int resync_locked;
978 	/* resync extent number waiting for application requests */
979 	unsigned int resync_wenr;
980 
981 	int open_cnt;
982 	u64 *p_uuid;
983 	struct drbd_epoch *current_epoch;
984 	spinlock_t epoch_lock;
985 	unsigned int epochs;
986 	enum write_ordering_e write_ordering;
987 	struct list_head active_ee; /* IO in progress */
988 	struct list_head sync_ee;   /* IO in progress */
989 	struct list_head done_ee;   /* send ack */
990 	struct list_head read_ee;   /* IO in progress */
991 	struct list_head net_ee;    /* zero-copy network send in progress */
992 	struct hlist_head *ee_hash; /* is proteced by req_lock! */
993 	unsigned int ee_hash_s;
994 
995 	/* this one is protected by ee_lock, single thread */
996 	struct drbd_epoch_entry *last_write_w_barrier;
997 
998 	int next_barrier_nr;
999 	struct hlist_head *app_reads_hash; /* is proteced by req_lock */
1000 	struct list_head resync_reads;
1001 	atomic_t pp_in_use;
1002 	wait_queue_head_t ee_wait;
1003 	struct page *md_io_page;	/* one page buffer for md_io */
1004 	struct page *md_io_tmpp;	/* for logical_block_size != 512 */
1005 	struct mutex md_io_mutex;	/* protects the md_io_buffer */
1006 	spinlock_t al_lock;
1007 	wait_queue_head_t al_wait;
1008 	struct lru_cache *act_log;	/* activity log */
1009 	unsigned int al_tr_number;
1010 	int al_tr_cycle;
1011 	int al_tr_pos;   /* position of the next transaction in the journal */
1012 	struct crypto_hash *cram_hmac_tfm;
1013 	struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */
1014 	struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */
1015 	void *int_dig_out;
1016 	void *int_dig_in;
1017 	void *int_dig_vv;
1018 	wait_queue_head_t seq_wait;
1019 	atomic_t packet_seq;
1020 	unsigned int peer_seq;
1021 	spinlock_t peer_seq_lock;
1022 	unsigned int minor;
1023 	unsigned long comm_bm_set; /* communicated number of set bits. */
1024 	cpumask_var_t cpu_mask;
1025 	struct bm_io_work bm_io_work;
1026 	u64 ed_uuid; /* UUID of the exposed data */
1027 	struct mutex state_mutex;
1028 	char congestion_reason;  /* Why we where congested... */
1029 };
1030 
1031 static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
1032 {
1033 	struct drbd_conf *mdev;
1034 
1035 	mdev = minor < minor_count ? minor_table[minor] : NULL;
1036 
1037 	return mdev;
1038 }
1039 
1040 static inline unsigned int mdev_to_minor(struct drbd_conf *mdev)
1041 {
1042 	return mdev->minor;
1043 }
1044 
1045 /* returns 1 if it was successfull,
1046  * returns 0 if there was no data socket.
1047  * so wherever you are going to use the data.socket, e.g. do
1048  * if (!drbd_get_data_sock(mdev))
1049  *	return 0;
1050  *	CODE();
1051  * drbd_put_data_sock(mdev);
1052  */
1053 static inline int drbd_get_data_sock(struct drbd_conf *mdev)
1054 {
1055 	mutex_lock(&mdev->data.mutex);
1056 	/* drbd_disconnect() could have called drbd_free_sock()
1057 	 * while we were waiting in down()... */
1058 	if (unlikely(mdev->data.socket == NULL)) {
1059 		mutex_unlock(&mdev->data.mutex);
1060 		return 0;
1061 	}
1062 	return 1;
1063 }
1064 
1065 static inline void drbd_put_data_sock(struct drbd_conf *mdev)
1066 {
1067 	mutex_unlock(&mdev->data.mutex);
1068 }
1069 
1070 /*
1071  * function declarations
1072  *************************/
1073 
1074 /* drbd_main.c */
1075 
1076 enum chg_state_flags {
1077 	CS_HARD	= 1,
1078 	CS_VERBOSE = 2,
1079 	CS_WAIT_COMPLETE = 4,
1080 	CS_SERIALIZE    = 8,
1081 	CS_ORDERED      = CS_WAIT_COMPLETE + CS_SERIALIZE,
1082 };
1083 
1084 extern void drbd_init_set_defaults(struct drbd_conf *mdev);
1085 extern int drbd_change_state(struct drbd_conf *mdev, enum chg_state_flags f,
1086 			union drbd_state mask, union drbd_state val);
1087 extern void drbd_force_state(struct drbd_conf *, union drbd_state,
1088 			union drbd_state);
1089 extern int _drbd_request_state(struct drbd_conf *, union drbd_state,
1090 			union drbd_state, enum chg_state_flags);
1091 extern int __drbd_set_state(struct drbd_conf *, union drbd_state,
1092 			    enum chg_state_flags, struct completion *done);
1093 extern void print_st_err(struct drbd_conf *, union drbd_state,
1094 			union drbd_state, int);
1095 extern int  drbd_thread_start(struct drbd_thread *thi);
1096 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1097 #ifdef CONFIG_SMP
1098 extern void drbd_thread_current_set_cpu(struct drbd_conf *mdev);
1099 extern void drbd_calc_cpu_mask(struct drbd_conf *mdev);
1100 #else
1101 #define drbd_thread_current_set_cpu(A) ({})
1102 #define drbd_calc_cpu_mask(A) ({})
1103 #endif
1104 extern void drbd_free_resources(struct drbd_conf *mdev);
1105 extern void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr,
1106 		       unsigned int set_size);
1107 extern void tl_clear(struct drbd_conf *mdev);
1108 extern void _tl_add_barrier(struct drbd_conf *, struct drbd_tl_epoch *);
1109 extern void drbd_free_sock(struct drbd_conf *mdev);
1110 extern int drbd_send(struct drbd_conf *mdev, struct socket *sock,
1111 			void *buf, size_t size, unsigned msg_flags);
1112 extern int drbd_send_protocol(struct drbd_conf *mdev);
1113 extern int drbd_send_uuids(struct drbd_conf *mdev);
1114 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
1115 extern int drbd_send_sync_uuid(struct drbd_conf *mdev, u64 val);
1116 extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply);
1117 extern int _drbd_send_state(struct drbd_conf *mdev);
1118 extern int drbd_send_state(struct drbd_conf *mdev);
1119 extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock,
1120 			enum drbd_packets cmd, struct p_header *h,
1121 			size_t size, unsigned msg_flags);
1122 #define USE_DATA_SOCKET 1
1123 #define USE_META_SOCKET 0
1124 extern int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket,
1125 			enum drbd_packets cmd, struct p_header *h,
1126 			size_t size);
1127 extern int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd,
1128 			char *data, size_t size);
1129 extern int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc);
1130 extern int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr,
1131 			u32 set_size);
1132 extern int drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd,
1133 			struct drbd_epoch_entry *e);
1134 extern int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packets cmd,
1135 			struct p_block_req *rp);
1136 extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd,
1137 			struct p_data *dp);
1138 extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd,
1139 			    sector_t sector, int blksize, u64 block_id);
1140 extern int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd,
1141 			   struct drbd_epoch_entry *e);
1142 extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
1143 extern int _drbd_send_barrier(struct drbd_conf *mdev,
1144 			struct drbd_tl_epoch *barrier);
1145 extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1146 			      sector_t sector, int size, u64 block_id);
1147 extern int drbd_send_drequest_csum(struct drbd_conf *mdev,
1148 				   sector_t sector,int size,
1149 				   void *digest, int digest_size,
1150 				   enum drbd_packets cmd);
1151 extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size);
1152 
1153 extern int drbd_send_bitmap(struct drbd_conf *mdev);
1154 extern int _drbd_send_bitmap(struct drbd_conf *mdev);
1155 extern int drbd_send_sr_reply(struct drbd_conf *mdev, int retcode);
1156 extern void drbd_free_bc(struct drbd_backing_dev *ldev);
1157 extern void drbd_mdev_cleanup(struct drbd_conf *mdev);
1158 
1159 /* drbd_meta-data.c (still in drbd_main.c) */
1160 extern void drbd_md_sync(struct drbd_conf *mdev);
1161 extern int  drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
1162 /* maybe define them below as inline? */
1163 extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1164 extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1165 extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1166 extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1167 extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
1168 extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
1169 extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
1170 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1171 extern void drbd_md_mark_dirty(struct drbd_conf *mdev);
1172 extern void drbd_queue_bitmap_io(struct drbd_conf *mdev,
1173 				 int (*io_fn)(struct drbd_conf *),
1174 				 void (*done)(struct drbd_conf *, int),
1175 				 char *why);
1176 extern int drbd_bmio_set_n_write(struct drbd_conf *mdev);
1177 extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev);
1178 extern int drbd_bitmap_io(struct drbd_conf *mdev, int (*io_fn)(struct drbd_conf *), char *why);
1179 
1180 
1181 /* Meta data layout
1182    We reserve a 128MB Block (4k aligned)
1183    * either at the end of the backing device
1184    * or on a separate meta data device. */
1185 
1186 #define MD_RESERVED_SECT (128LU << 11)  /* 128 MB, unit sectors */
1187 /* The following numbers are sectors */
1188 #define MD_AL_OFFSET 8	    /* 8 Sectors after start of meta area */
1189 #define MD_AL_MAX_SIZE 64   /* = 32 kb LOG  ~ 3776 extents ~ 14 GB Storage */
1190 /* Allows up to about 3.8TB */
1191 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1192 
1193 /* Since the smalles IO unit is usually 512 byte */
1194 #define MD_SECTOR_SHIFT	 9
1195 #define MD_SECTOR_SIZE	 (1<<MD_SECTOR_SHIFT)
1196 
1197 /* activity log */
1198 #define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1199 #define AL_EXTENT_SHIFT 22		 /* One extent represents 4M Storage */
1200 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1201 
1202 #if BITS_PER_LONG == 32
1203 #define LN2_BPL 5
1204 #define cpu_to_lel(A) cpu_to_le32(A)
1205 #define lel_to_cpu(A) le32_to_cpu(A)
1206 #elif BITS_PER_LONG == 64
1207 #define LN2_BPL 6
1208 #define cpu_to_lel(A) cpu_to_le64(A)
1209 #define lel_to_cpu(A) le64_to_cpu(A)
1210 #else
1211 #error "LN2 of BITS_PER_LONG unknown!"
1212 #endif
1213 
1214 /* resync bitmap */
1215 /* 16MB sized 'bitmap extent' to track syncer usage */
1216 struct bm_extent {
1217 	int rs_left; /* number of bits set (out of sync) in this extent. */
1218 	int rs_failed; /* number of failed resync requests in this extent. */
1219 	unsigned long flags;
1220 	struct lc_element lce;
1221 };
1222 
1223 #define BME_NO_WRITES  0  /* bm_extent.flags: no more requests on this one! */
1224 #define BME_LOCKED     1  /* bm_extent.flags: syncer active on this one. */
1225 
1226 /* drbd_bitmap.c */
1227 /*
1228  * We need to store one bit for a block.
1229  * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1230  * Bit 0 ==> local node thinks this block is binary identical on both nodes
1231  * Bit 1 ==> local node thinks this block needs to be synced.
1232  */
1233 
1234 #define BM_BLOCK_SHIFT  12			 /* 4k per bit */
1235 #define BM_BLOCK_SIZE	 (1<<BM_BLOCK_SHIFT)
1236 /* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1237  * per sector of on disk bitmap */
1238 #define BM_EXT_SHIFT	 (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3)  /* = 24 */
1239 #define BM_EXT_SIZE	 (1<<BM_EXT_SHIFT)
1240 
1241 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1242 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1243 #endif
1244 
1245 /* thus many _storage_ sectors are described by one bit */
1246 #define BM_SECT_TO_BIT(x)   ((x)>>(BM_BLOCK_SHIFT-9))
1247 #define BM_BIT_TO_SECT(x)   ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1248 #define BM_SECT_PER_BIT     BM_BIT_TO_SECT(1)
1249 
1250 /* bit to represented kilo byte conversion */
1251 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1252 
1253 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1254  * _storage_ sector is located in */
1255 #define BM_SECT_TO_EXT(x)   ((x)>>(BM_EXT_SHIFT-9))
1256 
1257 /* how much _storage_ sectors we have per bitmap sector */
1258 #define BM_EXT_TO_SECT(x)   ((sector_t)(x) << (BM_EXT_SHIFT-9))
1259 #define BM_SECT_PER_EXT     BM_EXT_TO_SECT(1)
1260 
1261 /* in one sector of the bitmap, we have this many activity_log extents. */
1262 #define AL_EXT_PER_BM_SECT  (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1263 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1264 
1265 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1266 #define BM_BLOCKS_PER_BM_EXT_MASK  ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1267 
1268 /* the extent in "PER_EXTENT" below is an activity log extent
1269  * we need that many (long words/bytes) to store the bitmap
1270  *		     of one AL_EXTENT_SIZE chunk of storage.
1271  * we can store the bitmap for that many AL_EXTENTS within
1272  * one sector of the _on_disk_ bitmap:
1273  * bit	 0	  bit 37   bit 38	     bit (512*8)-1
1274  *	     ...|........|........|.. // ..|........|
1275  * sect. 0	 `296	  `304			   ^(512*8*8)-1
1276  *
1277 #define BM_WORDS_PER_EXT    ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1278 #define BM_BYTES_PER_EXT    ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 )  // 128
1279 #define BM_EXT_PER_SECT	    ( 512 / BM_BYTES_PER_EXTENT )	 //   4
1280  */
1281 
1282 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1283 #define DRBD_MAX_SECTORS_BM \
1284 	  ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1285 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1286 #define DRBD_MAX_SECTORS      DRBD_MAX_SECTORS_BM
1287 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1288 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1289 #define DRBD_MAX_SECTORS      DRBD_MAX_SECTORS_32
1290 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1291 #else
1292 #define DRBD_MAX_SECTORS      DRBD_MAX_SECTORS_BM
1293 /* 16 TB in units of sectors */
1294 #if BITS_PER_LONG == 32
1295 /* adjust by one page worth of bitmap,
1296  * so we won't wrap around in drbd_bm_find_next_bit.
1297  * you should use 64bit OS for that much storage, anyways. */
1298 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1299 #else
1300 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0x1LU << 32)
1301 #endif
1302 #endif
1303 
1304 /* Sector shift value for the "hash" functions of tl_hash and ee_hash tables.
1305  * With a value of 6 all IO in one 32K block make it to the same slot of the
1306  * hash table. */
1307 #define HT_SHIFT 6
1308 #define DRBD_MAX_SEGMENT_SIZE (1U<<(9+HT_SHIFT))
1309 
1310 /* Number of elements in the app_reads_hash */
1311 #define APP_R_HSIZE 15
1312 
1313 extern int  drbd_bm_init(struct drbd_conf *mdev);
1314 extern int  drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors);
1315 extern void drbd_bm_cleanup(struct drbd_conf *mdev);
1316 extern void drbd_bm_set_all(struct drbd_conf *mdev);
1317 extern void drbd_bm_clear_all(struct drbd_conf *mdev);
1318 extern int  drbd_bm_set_bits(
1319 		struct drbd_conf *mdev, unsigned long s, unsigned long e);
1320 extern int  drbd_bm_clear_bits(
1321 		struct drbd_conf *mdev, unsigned long s, unsigned long e);
1322 /* bm_set_bits variant for use while holding drbd_bm_lock */
1323 extern void _drbd_bm_set_bits(struct drbd_conf *mdev,
1324 		const unsigned long s, const unsigned long e);
1325 extern int  drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr);
1326 extern int  drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr);
1327 extern int  drbd_bm_write_sect(struct drbd_conf *mdev, unsigned long enr) __must_hold(local);
1328 extern int  drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
1329 extern int  drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
1330 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev,
1331 		unsigned long al_enr);
1332 extern size_t	     drbd_bm_words(struct drbd_conf *mdev);
1333 extern unsigned long drbd_bm_bits(struct drbd_conf *mdev);
1334 extern sector_t      drbd_bm_capacity(struct drbd_conf *mdev);
1335 extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1336 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1337 extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1338 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo);
1339 extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev);
1340 extern int drbd_bm_rs_done(struct drbd_conf *mdev);
1341 /* for receive_bitmap */
1342 extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset,
1343 		size_t number, unsigned long *buffer);
1344 /* for _drbd_send_bitmap and drbd_bm_write_sect */
1345 extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset,
1346 		size_t number, unsigned long *buffer);
1347 
1348 extern void drbd_bm_lock(struct drbd_conf *mdev, char *why);
1349 extern void drbd_bm_unlock(struct drbd_conf *mdev);
1350 
1351 extern int drbd_bm_count_bits(struct drbd_conf *mdev, const unsigned long s, const unsigned long e);
1352 /* drbd_main.c */
1353 
1354 extern struct kmem_cache *drbd_request_cache;
1355 extern struct kmem_cache *drbd_ee_cache;	/* epoch entries */
1356 extern struct kmem_cache *drbd_bm_ext_cache;	/* bitmap extents */
1357 extern struct kmem_cache *drbd_al_ext_cache;	/* activity log extents */
1358 extern mempool_t *drbd_request_mempool;
1359 extern mempool_t *drbd_ee_mempool;
1360 
1361 extern struct page *drbd_pp_pool; /* drbd's page pool */
1362 extern spinlock_t   drbd_pp_lock;
1363 extern int	    drbd_pp_vacant;
1364 extern wait_queue_head_t drbd_pp_wait;
1365 
1366 extern rwlock_t global_state_lock;
1367 
1368 extern struct drbd_conf *drbd_new_device(unsigned int minor);
1369 extern void drbd_free_mdev(struct drbd_conf *mdev);
1370 
1371 extern int proc_details;
1372 
1373 /* drbd_req */
1374 extern int drbd_make_request_26(struct request_queue *q, struct bio *bio);
1375 extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req);
1376 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1377 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1378 
1379 
1380 /* drbd_nl.c */
1381 extern void drbd_suspend_io(struct drbd_conf *mdev);
1382 extern void drbd_resume_io(struct drbd_conf *mdev);
1383 extern char *ppsize(char *buf, unsigned long long size);
1384 extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int);
1385 enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
1386 extern enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *, int force) __must_hold(local);
1387 extern void resync_after_online_grow(struct drbd_conf *);
1388 extern void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int) __must_hold(local);
1389 extern int drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role,
1390 		int force);
1391 enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev);
1392 extern int drbd_khelper(struct drbd_conf *mdev, char *cmd);
1393 
1394 /* drbd_worker.c */
1395 extern int drbd_worker(struct drbd_thread *thi);
1396 extern int drbd_alter_sa(struct drbd_conf *mdev, int na);
1397 extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side);
1398 extern void resume_next_sg(struct drbd_conf *mdev);
1399 extern void suspend_other_sg(struct drbd_conf *mdev);
1400 extern int drbd_resync_finished(struct drbd_conf *mdev);
1401 /* maybe rather drbd_main.c ? */
1402 extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
1403 		struct drbd_backing_dev *bdev, sector_t sector, int rw);
1404 extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int);
1405 
1406 static inline void ov_oos_print(struct drbd_conf *mdev)
1407 {
1408 	if (mdev->ov_last_oos_size) {
1409 		dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n",
1410 		     (unsigned long long)mdev->ov_last_oos_start,
1411 		     (unsigned long)mdev->ov_last_oos_size);
1412 	}
1413 	mdev->ov_last_oos_size=0;
1414 }
1415 
1416 
1417 extern void drbd_csum(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
1418 /* worker callbacks */
1419 extern int w_req_cancel_conflict(struct drbd_conf *, struct drbd_work *, int);
1420 extern int w_read_retry_remote(struct drbd_conf *, struct drbd_work *, int);
1421 extern int w_e_end_data_req(struct drbd_conf *, struct drbd_work *, int);
1422 extern int w_e_end_rsdata_req(struct drbd_conf *, struct drbd_work *, int);
1423 extern int w_e_end_csum_rs_req(struct drbd_conf *, struct drbd_work *, int);
1424 extern int w_e_end_ov_reply(struct drbd_conf *, struct drbd_work *, int);
1425 extern int w_e_end_ov_req(struct drbd_conf *, struct drbd_work *, int);
1426 extern int w_ov_finished(struct drbd_conf *, struct drbd_work *, int);
1427 extern int w_resync_inactive(struct drbd_conf *, struct drbd_work *, int);
1428 extern int w_resume_next_sg(struct drbd_conf *, struct drbd_work *, int);
1429 extern int w_io_error(struct drbd_conf *, struct drbd_work *, int);
1430 extern int w_send_write_hint(struct drbd_conf *, struct drbd_work *, int);
1431 extern int w_make_resync_request(struct drbd_conf *, struct drbd_work *, int);
1432 extern int w_send_dblock(struct drbd_conf *, struct drbd_work *, int);
1433 extern int w_send_barrier(struct drbd_conf *, struct drbd_work *, int);
1434 extern int w_send_read_req(struct drbd_conf *, struct drbd_work *, int);
1435 extern int w_prev_work_done(struct drbd_conf *, struct drbd_work *, int);
1436 extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int);
1437 
1438 extern void resync_timer_fn(unsigned long data);
1439 
1440 /* drbd_receiver.c */
1441 extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list);
1442 extern struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev,
1443 					    u64 id,
1444 					    sector_t sector,
1445 					    unsigned int data_size,
1446 					    gfp_t gfp_mask) __must_hold(local);
1447 extern void drbd_free_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e);
1448 extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1449 		struct list_head *head);
1450 extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1451 		struct list_head *head);
1452 extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled);
1453 extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed);
1454 extern void drbd_flush_workqueue(struct drbd_conf *mdev);
1455 
1456 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1457  * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1458 static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1459 			char __user *optval, int optlen)
1460 {
1461 	int err;
1462 	if (level == SOL_SOCKET)
1463 		err = sock_setsockopt(sock, level, optname, optval, optlen);
1464 	else
1465 		err = sock->ops->setsockopt(sock, level, optname, optval,
1466 					    optlen);
1467 	return err;
1468 }
1469 
1470 static inline void drbd_tcp_cork(struct socket *sock)
1471 {
1472 	int __user val = 1;
1473 	(void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1474 			(char __user *)&val, sizeof(val));
1475 }
1476 
1477 static inline void drbd_tcp_uncork(struct socket *sock)
1478 {
1479 	int __user val = 0;
1480 	(void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1481 			(char __user *)&val, sizeof(val));
1482 }
1483 
1484 static inline void drbd_tcp_nodelay(struct socket *sock)
1485 {
1486 	int __user val = 1;
1487 	(void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1488 			(char __user *)&val, sizeof(val));
1489 }
1490 
1491 static inline void drbd_tcp_quickack(struct socket *sock)
1492 {
1493 	int __user val = 1;
1494 	(void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1495 			(char __user *)&val, sizeof(val));
1496 }
1497 
1498 void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo);
1499 
1500 /* drbd_proc.c */
1501 extern struct proc_dir_entry *drbd_proc;
1502 extern const struct file_operations drbd_proc_fops;
1503 extern const char *drbd_conn_str(enum drbd_conns s);
1504 extern const char *drbd_role_str(enum drbd_role s);
1505 
1506 /* drbd_actlog.c */
1507 extern void drbd_al_begin_io(struct drbd_conf *mdev, sector_t sector);
1508 extern void drbd_al_complete_io(struct drbd_conf *mdev, sector_t sector);
1509 extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector);
1510 extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1511 extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1512 extern void drbd_rs_cancel_all(struct drbd_conf *mdev);
1513 extern int drbd_rs_del_all(struct drbd_conf *mdev);
1514 extern void drbd_rs_failed_io(struct drbd_conf *mdev,
1515 		sector_t sector, int size);
1516 extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *);
1517 extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
1518 		int size, const char *file, const unsigned int line);
1519 #define drbd_set_in_sync(mdev, sector, size) \
1520 	__drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1521 extern void __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
1522 		int size, const char *file, const unsigned int line);
1523 #define drbd_set_out_of_sync(mdev, sector, size) \
1524 	__drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1525 extern void drbd_al_apply_to_bm(struct drbd_conf *mdev);
1526 extern void drbd_al_to_on_disk_bm(struct drbd_conf *mdev);
1527 extern void drbd_al_shrink(struct drbd_conf *mdev);
1528 
1529 
1530 /* drbd_nl.c */
1531 
1532 void drbd_nl_cleanup(void);
1533 int __init drbd_nl_init(void);
1534 void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state);
1535 void drbd_bcast_sync_progress(struct drbd_conf *mdev);
1536 void drbd_bcast_ee(struct drbd_conf *mdev,
1537 		const char *reason, const int dgs,
1538 		const char* seen_hash, const char* calc_hash,
1539 		const struct drbd_epoch_entry* e);
1540 
1541 
1542 /**
1543  * DOC: DRBD State macros
1544  *
1545  * These macros are used to express state changes in easily readable form.
1546  *
1547  * The NS macros expand to a mask and a value, that can be bit ored onto the
1548  * current state as soon as the spinlock (req_lock) was taken.
1549  *
1550  * The _NS macros are used for state functions that get called with the
1551  * spinlock. These macros expand directly to the new state value.
1552  *
1553  * Besides the basic forms NS() and _NS() additional _?NS[23] are defined
1554  * to express state changes that affect more than one aspect of the state.
1555  *
1556  * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY)
1557  * Means that the network connection was established and that the peer
1558  * is in secondary role.
1559  */
1560 #define role_MASK R_MASK
1561 #define peer_MASK R_MASK
1562 #define disk_MASK D_MASK
1563 #define pdsk_MASK D_MASK
1564 #define conn_MASK C_MASK
1565 #define susp_MASK 1
1566 #define user_isp_MASK 1
1567 #define aftr_isp_MASK 1
1568 
1569 #define NS(T, S) \
1570 	({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \
1571 	({ union drbd_state val; val.i = 0; val.T = (S); val; })
1572 #define NS2(T1, S1, T2, S2) \
1573 	({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1574 	  mask.T2 = T2##_MASK; mask; }), \
1575 	({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1576 	  val.T2 = (S2); val; })
1577 #define NS3(T1, S1, T2, S2, T3, S3) \
1578 	({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1579 	  mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \
1580 	({ union drbd_state val;  val.i = 0; val.T1 = (S1); \
1581 	  val.T2 = (S2); val.T3 = (S3); val; })
1582 
1583 #define _NS(D, T, S) \
1584 	D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; })
1585 #define _NS2(D, T1, S1, T2, S2) \
1586 	D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1587 	__ns.T2 = (S2); __ns; })
1588 #define _NS3(D, T1, S1, T2, S2, T3, S3) \
1589 	D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1590 	__ns.T2 = (S2); __ns.T3 = (S3); __ns; })
1591 
1592 /*
1593  * inline helper functions
1594  *************************/
1595 
1596 static inline void drbd_state_lock(struct drbd_conf *mdev)
1597 {
1598 	wait_event(mdev->misc_wait,
1599 		   !test_and_set_bit(CLUSTER_ST_CHANGE, &mdev->flags));
1600 }
1601 
1602 static inline void drbd_state_unlock(struct drbd_conf *mdev)
1603 {
1604 	clear_bit(CLUSTER_ST_CHANGE, &mdev->flags);
1605 	wake_up(&mdev->misc_wait);
1606 }
1607 
1608 static inline int _drbd_set_state(struct drbd_conf *mdev,
1609 				   union drbd_state ns, enum chg_state_flags flags,
1610 				   struct completion *done)
1611 {
1612 	int rv;
1613 
1614 	read_lock(&global_state_lock);
1615 	rv = __drbd_set_state(mdev, ns, flags, done);
1616 	read_unlock(&global_state_lock);
1617 
1618 	return rv;
1619 }
1620 
1621 /**
1622  * drbd_request_state() - Reqest a state change
1623  * @mdev:	DRBD device.
1624  * @mask:	mask of state bits to change.
1625  * @val:	value of new state bits.
1626  *
1627  * This is the most graceful way of requesting a state change. It is verbose
1628  * quite verbose in case the state change is not possible, and all those
1629  * state changes are globally serialized.
1630  */
1631 static inline int drbd_request_state(struct drbd_conf *mdev,
1632 				     union drbd_state mask,
1633 				     union drbd_state val)
1634 {
1635 	return _drbd_request_state(mdev, mask, val, CS_VERBOSE + CS_ORDERED);
1636 }
1637 
1638 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1639 static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach, const char *where)
1640 {
1641 	switch (mdev->ldev->dc.on_io_error) {
1642 	case EP_PASS_ON:
1643 		if (!forcedetach) {
1644 			if (printk_ratelimit())
1645 				dev_err(DEV, "Local IO failed in %s."
1646 					     "Passing error on...\n", where);
1647 			break;
1648 		}
1649 		/* NOTE fall through to detach case if forcedetach set */
1650 	case EP_DETACH:
1651 	case EP_CALL_HELPER:
1652 		if (mdev->state.disk > D_FAILED) {
1653 			_drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL);
1654 			dev_err(DEV, "Local IO failed in %s."
1655 				     "Detaching...\n", where);
1656 		}
1657 		break;
1658 	}
1659 }
1660 
1661 /**
1662  * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1663  * @mdev:	 DRBD device.
1664  * @error:	 Error code passed to the IO completion callback
1665  * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1666  *
1667  * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1668  */
1669 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1670 static inline void drbd_chk_io_error_(struct drbd_conf *mdev,
1671 	int error, int forcedetach, const char *where)
1672 {
1673 	if (error) {
1674 		unsigned long flags;
1675 		spin_lock_irqsave(&mdev->req_lock, flags);
1676 		__drbd_chk_io_error_(mdev, forcedetach, where);
1677 		spin_unlock_irqrestore(&mdev->req_lock, flags);
1678 	}
1679 }
1680 
1681 
1682 /**
1683  * drbd_md_first_sector() - Returns the first sector number of the meta data area
1684  * @bdev:	Meta data block device.
1685  *
1686  * BTW, for internal meta data, this happens to be the maximum capacity
1687  * we could agree upon with our peer node.
1688  */
1689 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1690 {
1691 	switch (bdev->dc.meta_dev_idx) {
1692 	case DRBD_MD_INDEX_INTERNAL:
1693 	case DRBD_MD_INDEX_FLEX_INT:
1694 		return bdev->md.md_offset + bdev->md.bm_offset;
1695 	case DRBD_MD_INDEX_FLEX_EXT:
1696 	default:
1697 		return bdev->md.md_offset;
1698 	}
1699 }
1700 
1701 /**
1702  * drbd_md_last_sector() - Return the last sector number of the meta data area
1703  * @bdev:	Meta data block device.
1704  */
1705 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1706 {
1707 	switch (bdev->dc.meta_dev_idx) {
1708 	case DRBD_MD_INDEX_INTERNAL:
1709 	case DRBD_MD_INDEX_FLEX_INT:
1710 		return bdev->md.md_offset + MD_AL_OFFSET - 1;
1711 	case DRBD_MD_INDEX_FLEX_EXT:
1712 	default:
1713 		return bdev->md.md_offset + bdev->md.md_size_sect;
1714 	}
1715 }
1716 
1717 /* Returns the number of 512 byte sectors of the device */
1718 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1719 {
1720 	/* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1721 	return bdev ? bdev->bd_inode->i_size >> 9 : 0;
1722 }
1723 
1724 /**
1725  * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1726  * @bdev:	Meta data block device.
1727  *
1728  * returns the capacity we announce to out peer.  we clip ourselves at the
1729  * various MAX_SECTORS, because if we don't, current implementation will
1730  * oops sooner or later
1731  */
1732 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1733 {
1734 	sector_t s;
1735 	switch (bdev->dc.meta_dev_idx) {
1736 	case DRBD_MD_INDEX_INTERNAL:
1737 	case DRBD_MD_INDEX_FLEX_INT:
1738 		s = drbd_get_capacity(bdev->backing_bdev)
1739 			? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1740 					drbd_md_first_sector(bdev))
1741 			: 0;
1742 		break;
1743 	case DRBD_MD_INDEX_FLEX_EXT:
1744 		s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1745 				drbd_get_capacity(bdev->backing_bdev));
1746 		/* clip at maximum size the meta device can support */
1747 		s = min_t(sector_t, s,
1748 			BM_EXT_TO_SECT(bdev->md.md_size_sect
1749 				     - bdev->md.bm_offset));
1750 		break;
1751 	default:
1752 		s = min_t(sector_t, DRBD_MAX_SECTORS,
1753 				drbd_get_capacity(bdev->backing_bdev));
1754 	}
1755 	return s;
1756 }
1757 
1758 /**
1759  * drbd_md_ss__() - Return the sector number of our meta data super block
1760  * @mdev:	DRBD device.
1761  * @bdev:	Meta data block device.
1762  */
1763 static inline sector_t drbd_md_ss__(struct drbd_conf *mdev,
1764 				    struct drbd_backing_dev *bdev)
1765 {
1766 	switch (bdev->dc.meta_dev_idx) {
1767 	default: /* external, some index */
1768 		return MD_RESERVED_SECT * bdev->dc.meta_dev_idx;
1769 	case DRBD_MD_INDEX_INTERNAL:
1770 		/* with drbd08, internal meta data is always "flexible" */
1771 	case DRBD_MD_INDEX_FLEX_INT:
1772 		/* sizeof(struct md_on_disk_07) == 4k
1773 		 * position: last 4k aligned block of 4k size */
1774 		if (!bdev->backing_bdev) {
1775 			if (__ratelimit(&drbd_ratelimit_state)) {
1776 				dev_err(DEV, "bdev->backing_bdev==NULL\n");
1777 				dump_stack();
1778 			}
1779 			return 0;
1780 		}
1781 		return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL)
1782 			- MD_AL_OFFSET;
1783 	case DRBD_MD_INDEX_FLEX_EXT:
1784 		return 0;
1785 	}
1786 }
1787 
1788 static inline void
1789 _drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1790 {
1791 	list_add_tail(&w->list, &q->q);
1792 	up(&q->s);
1793 }
1794 
1795 static inline void
1796 drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w)
1797 {
1798 	unsigned long flags;
1799 	spin_lock_irqsave(&q->q_lock, flags);
1800 	list_add(&w->list, &q->q);
1801 	up(&q->s); /* within the spinlock,
1802 		      see comment near end of drbd_worker() */
1803 	spin_unlock_irqrestore(&q->q_lock, flags);
1804 }
1805 
1806 static inline void
1807 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1808 {
1809 	unsigned long flags;
1810 	spin_lock_irqsave(&q->q_lock, flags);
1811 	list_add_tail(&w->list, &q->q);
1812 	up(&q->s); /* within the spinlock,
1813 		      see comment near end of drbd_worker() */
1814 	spin_unlock_irqrestore(&q->q_lock, flags);
1815 }
1816 
1817 static inline void wake_asender(struct drbd_conf *mdev)
1818 {
1819 	if (test_bit(SIGNAL_ASENDER, &mdev->flags))
1820 		force_sig(DRBD_SIG, mdev->asender.task);
1821 }
1822 
1823 static inline void request_ping(struct drbd_conf *mdev)
1824 {
1825 	set_bit(SEND_PING, &mdev->flags);
1826 	wake_asender(mdev);
1827 }
1828 
1829 static inline int drbd_send_short_cmd(struct drbd_conf *mdev,
1830 	enum drbd_packets cmd)
1831 {
1832 	struct p_header h;
1833 	return drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, &h, sizeof(h));
1834 }
1835 
1836 static inline int drbd_send_ping(struct drbd_conf *mdev)
1837 {
1838 	struct p_header h;
1839 	return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING, &h, sizeof(h));
1840 }
1841 
1842 static inline int drbd_send_ping_ack(struct drbd_conf *mdev)
1843 {
1844 	struct p_header h;
1845 	return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING_ACK, &h, sizeof(h));
1846 }
1847 
1848 static inline void drbd_thread_stop(struct drbd_thread *thi)
1849 {
1850 	_drbd_thread_stop(thi, FALSE, TRUE);
1851 }
1852 
1853 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
1854 {
1855 	_drbd_thread_stop(thi, FALSE, FALSE);
1856 }
1857 
1858 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
1859 {
1860 	_drbd_thread_stop(thi, TRUE, FALSE);
1861 }
1862 
1863 /* counts how many answer packets packets we expect from our peer,
1864  * for either explicit application requests,
1865  * or implicit barrier packets as necessary.
1866  * increased:
1867  *  w_send_barrier
1868  *  _req_mod(req, queue_for_net_write or queue_for_net_read);
1869  *    it is much easier and equally valid to count what we queue for the
1870  *    worker, even before it actually was queued or send.
1871  *    (drbd_make_request_common; recovery path on read io-error)
1872  * decreased:
1873  *  got_BarrierAck (respective tl_clear, tl_clear_barrier)
1874  *  _req_mod(req, data_received)
1875  *     [from receive_DataReply]
1876  *  _req_mod(req, write_acked_by_peer or recv_acked_by_peer or neg_acked)
1877  *     [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1878  *     for some reason it is NOT decreased in got_NegAck,
1879  *     but in the resulting cleanup code from report_params.
1880  *     we should try to remember the reason for that...
1881  *  _req_mod(req, send_failed or send_canceled)
1882  *  _req_mod(req, connection_lost_while_pending)
1883  *     [from tl_clear_barrier]
1884  */
1885 static inline void inc_ap_pending(struct drbd_conf *mdev)
1886 {
1887 	atomic_inc(&mdev->ap_pending_cnt);
1888 }
1889 
1890 #define ERR_IF_CNT_IS_NEGATIVE(which)				\
1891 	if (atomic_read(&mdev->which) < 0)			\
1892 		dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n",	\
1893 		    __func__ , __LINE__ ,			\
1894 		    atomic_read(&mdev->which))
1895 
1896 #define dec_ap_pending(mdev)	do {				\
1897 	typecheck(struct drbd_conf *, mdev);			\
1898 	if (atomic_dec_and_test(&mdev->ap_pending_cnt))		\
1899 		wake_up(&mdev->misc_wait);			\
1900 	ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
1901 
1902 /* counts how many resync-related answers we still expect from the peer
1903  *		     increase			decrease
1904  * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1905  * C_SYNC_SOURCE sends P_RS_DATA_REPLY   (and expects P_WRITE_ACK whith ID_SYNCER)
1906  *					   (or P_NEG_ACK with ID_SYNCER)
1907  */
1908 static inline void inc_rs_pending(struct drbd_conf *mdev)
1909 {
1910 	atomic_inc(&mdev->rs_pending_cnt);
1911 }
1912 
1913 #define dec_rs_pending(mdev)	do {				\
1914 	typecheck(struct drbd_conf *, mdev);			\
1915 	atomic_dec(&mdev->rs_pending_cnt);			\
1916 	ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
1917 
1918 /* counts how many answers we still need to send to the peer.
1919  * increased on
1920  *  receive_Data	unless protocol A;
1921  *			we need to send a P_RECV_ACK (proto B)
1922  *			or P_WRITE_ACK (proto C)
1923  *  receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1924  *  receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1925  *  receive_Barrier_*	we need to send a P_BARRIER_ACK
1926  */
1927 static inline void inc_unacked(struct drbd_conf *mdev)
1928 {
1929 	atomic_inc(&mdev->unacked_cnt);
1930 }
1931 
1932 #define dec_unacked(mdev)	do {				\
1933 	typecheck(struct drbd_conf *, mdev);			\
1934 	atomic_dec(&mdev->unacked_cnt);				\
1935 	ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
1936 
1937 #define sub_unacked(mdev, n)	do {				\
1938 	typecheck(struct drbd_conf *, mdev);			\
1939 	atomic_sub(n, &mdev->unacked_cnt);			\
1940 	ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
1941 
1942 
1943 static inline void put_net_conf(struct drbd_conf *mdev)
1944 {
1945 	if (atomic_dec_and_test(&mdev->net_cnt))
1946 		wake_up(&mdev->misc_wait);
1947 }
1948 
1949 /**
1950  * get_net_conf() - Increase ref count on mdev->net_conf; Returns 0 if nothing there
1951  * @mdev:	DRBD device.
1952  *
1953  * You have to call put_net_conf() when finished working with mdev->net_conf.
1954  */
1955 static inline int get_net_conf(struct drbd_conf *mdev)
1956 {
1957 	int have_net_conf;
1958 
1959 	atomic_inc(&mdev->net_cnt);
1960 	have_net_conf = mdev->state.conn >= C_UNCONNECTED;
1961 	if (!have_net_conf)
1962 		put_net_conf(mdev);
1963 	return have_net_conf;
1964 }
1965 
1966 /**
1967  * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
1968  * @M:		DRBD device.
1969  *
1970  * You have to call put_ldev() when finished working with mdev->ldev.
1971  */
1972 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
1973 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
1974 
1975 static inline void put_ldev(struct drbd_conf *mdev)
1976 {
1977 	__release(local);
1978 	if (atomic_dec_and_test(&mdev->local_cnt))
1979 		wake_up(&mdev->misc_wait);
1980 	D_ASSERT(atomic_read(&mdev->local_cnt) >= 0);
1981 }
1982 
1983 #ifndef __CHECKER__
1984 static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
1985 {
1986 	int io_allowed;
1987 
1988 	atomic_inc(&mdev->local_cnt);
1989 	io_allowed = (mdev->state.disk >= mins);
1990 	if (!io_allowed)
1991 		put_ldev(mdev);
1992 	return io_allowed;
1993 }
1994 #else
1995 extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins);
1996 #endif
1997 
1998 /* you must have an "get_ldev" reference */
1999 static inline void drbd_get_syncer_progress(struct drbd_conf *mdev,
2000 		unsigned long *bits_left, unsigned int *per_mil_done)
2001 {
2002 	/*
2003 	 * this is to break it at compile time when we change that
2004 	 * (we may feel 4TB maximum storage per drbd is not enough)
2005 	 */
2006 	typecheck(unsigned long, mdev->rs_total);
2007 
2008 	/* note: both rs_total and rs_left are in bits, i.e. in
2009 	 * units of BM_BLOCK_SIZE.
2010 	 * for the percentage, we don't care. */
2011 
2012 	*bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
2013 	/* >> 10 to prevent overflow,
2014 	 * +1 to prevent division by zero */
2015 	if (*bits_left > mdev->rs_total) {
2016 		/* doh. maybe a logic bug somewhere.
2017 		 * may also be just a race condition
2018 		 * between this and a disconnect during sync.
2019 		 * for now, just prevent in-kernel buffer overflow.
2020 		 */
2021 		smp_rmb();
2022 		dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2023 				drbd_conn_str(mdev->state.conn),
2024 				*bits_left, mdev->rs_total, mdev->rs_failed);
2025 		*per_mil_done = 0;
2026 	} else {
2027 		/* make sure the calculation happens in long context */
2028 		unsigned long tmp = 1000UL -
2029 				(*bits_left >> 10)*1000UL
2030 				/ ((mdev->rs_total >> 10) + 1UL);
2031 		*per_mil_done = tmp;
2032 	}
2033 }
2034 
2035 
2036 /* this throttles on-the-fly application requests
2037  * according to max_buffers settings;
2038  * maybe re-implement using semaphores? */
2039 static inline int drbd_get_max_buffers(struct drbd_conf *mdev)
2040 {
2041 	int mxb = 1000000; /* arbitrary limit on open requests */
2042 	if (get_net_conf(mdev)) {
2043 		mxb = mdev->net_conf->max_buffers;
2044 		put_net_conf(mdev);
2045 	}
2046 	return mxb;
2047 }
2048 
2049 static inline int drbd_state_is_stable(union drbd_state s)
2050 {
2051 
2052 	/* DO NOT add a default clause, we want the compiler to warn us
2053 	 * for any newly introduced state we may have forgotten to add here */
2054 
2055 	switch ((enum drbd_conns)s.conn) {
2056 	/* new io only accepted when there is no connection, ... */
2057 	case C_STANDALONE:
2058 	case C_WF_CONNECTION:
2059 	/* ... or there is a well established connection. */
2060 	case C_CONNECTED:
2061 	case C_SYNC_SOURCE:
2062 	case C_SYNC_TARGET:
2063 	case C_VERIFY_S:
2064 	case C_VERIFY_T:
2065 	case C_PAUSED_SYNC_S:
2066 	case C_PAUSED_SYNC_T:
2067 		/* maybe stable, look at the disk state */
2068 		break;
2069 
2070 	/* no new io accepted during tansitional states
2071 	 * like handshake or teardown */
2072 	case C_DISCONNECTING:
2073 	case C_UNCONNECTED:
2074 	case C_TIMEOUT:
2075 	case C_BROKEN_PIPE:
2076 	case C_NETWORK_FAILURE:
2077 	case C_PROTOCOL_ERROR:
2078 	case C_TEAR_DOWN:
2079 	case C_WF_REPORT_PARAMS:
2080 	case C_STARTING_SYNC_S:
2081 	case C_STARTING_SYNC_T:
2082 	case C_WF_BITMAP_S:
2083 	case C_WF_BITMAP_T:
2084 	case C_WF_SYNC_UUID:
2085 	case C_MASK:
2086 		/* not "stable" */
2087 		return 0;
2088 	}
2089 
2090 	switch ((enum drbd_disk_state)s.disk) {
2091 	case D_DISKLESS:
2092 	case D_INCONSISTENT:
2093 	case D_OUTDATED:
2094 	case D_CONSISTENT:
2095 	case D_UP_TO_DATE:
2096 		/* disk state is stable as well. */
2097 		break;
2098 
2099 	/* no new io accepted during tansitional states */
2100 	case D_ATTACHING:
2101 	case D_FAILED:
2102 	case D_NEGOTIATING:
2103 	case D_UNKNOWN:
2104 	case D_MASK:
2105 		/* not "stable" */
2106 		return 0;
2107 	}
2108 
2109 	return 1;
2110 }
2111 
2112 static inline int __inc_ap_bio_cond(struct drbd_conf *mdev)
2113 {
2114 	int mxb = drbd_get_max_buffers(mdev);
2115 
2116 	if (mdev->state.susp)
2117 		return 0;
2118 	if (test_bit(SUSPEND_IO, &mdev->flags))
2119 		return 0;
2120 
2121 	/* to avoid potential deadlock or bitmap corruption,
2122 	 * in various places, we only allow new application io
2123 	 * to start during "stable" states. */
2124 
2125 	/* no new io accepted when attaching or detaching the disk */
2126 	if (!drbd_state_is_stable(mdev->state))
2127 		return 0;
2128 
2129 	/* since some older kernels don't have atomic_add_unless,
2130 	 * and we are within the spinlock anyways, we have this workaround.  */
2131 	if (atomic_read(&mdev->ap_bio_cnt) > mxb)
2132 		return 0;
2133 	if (test_bit(BITMAP_IO, &mdev->flags))
2134 		return 0;
2135 	return 1;
2136 }
2137 
2138 /* I'd like to use wait_event_lock_irq,
2139  * but I'm not sure when it got introduced,
2140  * and not sure when it has 3 or 4 arguments */
2141 static inline void inc_ap_bio(struct drbd_conf *mdev, int one_or_two)
2142 {
2143 	/* compare with after_state_ch,
2144 	 * os.conn != C_WF_BITMAP_S && ns.conn == C_WF_BITMAP_S */
2145 	DEFINE_WAIT(wait);
2146 
2147 	/* we wait here
2148 	 *    as long as the device is suspended
2149 	 *    until the bitmap is no longer on the fly during connection
2150 	 *    handshake as long as we would exeed the max_buffer limit.
2151 	 *
2152 	 * to avoid races with the reconnect code,
2153 	 * we need to atomic_inc within the spinlock. */
2154 
2155 	spin_lock_irq(&mdev->req_lock);
2156 	while (!__inc_ap_bio_cond(mdev)) {
2157 		prepare_to_wait(&mdev->misc_wait, &wait, TASK_UNINTERRUPTIBLE);
2158 		spin_unlock_irq(&mdev->req_lock);
2159 		schedule();
2160 		finish_wait(&mdev->misc_wait, &wait);
2161 		spin_lock_irq(&mdev->req_lock);
2162 	}
2163 	atomic_add(one_or_two, &mdev->ap_bio_cnt);
2164 	spin_unlock_irq(&mdev->req_lock);
2165 }
2166 
2167 static inline void dec_ap_bio(struct drbd_conf *mdev)
2168 {
2169 	int mxb = drbd_get_max_buffers(mdev);
2170 	int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt);
2171 
2172 	D_ASSERT(ap_bio >= 0);
2173 	/* this currently does wake_up for every dec_ap_bio!
2174 	 * maybe rather introduce some type of hysteresis?
2175 	 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2176 	if (ap_bio < mxb)
2177 		wake_up(&mdev->misc_wait);
2178 	if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) {
2179 		if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags))
2180 			drbd_queue_work(&mdev->data.work, &mdev->bm_io_work.w);
2181 	}
2182 }
2183 
2184 static inline void drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val)
2185 {
2186 	mdev->ed_uuid = val;
2187 }
2188 
2189 static inline int seq_cmp(u32 a, u32 b)
2190 {
2191 	/* we assume wrap around at 32bit.
2192 	 * for wrap around at 24bit (old atomic_t),
2193 	 * we'd have to
2194 	 *  a <<= 8; b <<= 8;
2195 	 */
2196 	return (s32)(a) - (s32)(b);
2197 }
2198 #define seq_lt(a, b) (seq_cmp((a), (b)) < 0)
2199 #define seq_gt(a, b) (seq_cmp((a), (b)) > 0)
2200 #define seq_ge(a, b) (seq_cmp((a), (b)) >= 0)
2201 #define seq_le(a, b) (seq_cmp((a), (b)) <= 0)
2202 /* CAUTION: please no side effects in arguments! */
2203 #define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b)))
2204 
2205 static inline void update_peer_seq(struct drbd_conf *mdev, unsigned int new_seq)
2206 {
2207 	unsigned int m;
2208 	spin_lock(&mdev->peer_seq_lock);
2209 	m = seq_max(mdev->peer_seq, new_seq);
2210 	mdev->peer_seq = m;
2211 	spin_unlock(&mdev->peer_seq_lock);
2212 	if (m == new_seq)
2213 		wake_up(&mdev->seq_wait);
2214 }
2215 
2216 static inline void drbd_update_congested(struct drbd_conf *mdev)
2217 {
2218 	struct sock *sk = mdev->data.socket->sk;
2219 	if (sk->sk_wmem_queued > sk->sk_sndbuf * 4 / 5)
2220 		set_bit(NET_CONGESTED, &mdev->flags);
2221 }
2222 
2223 static inline int drbd_queue_order_type(struct drbd_conf *mdev)
2224 {
2225 	/* sorry, we currently have no working implementation
2226 	 * of distributed TCQ stuff */
2227 #ifndef QUEUE_ORDERED_NONE
2228 #define QUEUE_ORDERED_NONE 0
2229 #endif
2230 	return QUEUE_ORDERED_NONE;
2231 }
2232 
2233 static inline void drbd_blk_run_queue(struct request_queue *q)
2234 {
2235 	if (q && q->unplug_fn)
2236 		q->unplug_fn(q);
2237 }
2238 
2239 static inline void drbd_kick_lo(struct drbd_conf *mdev)
2240 {
2241 	if (get_ldev(mdev)) {
2242 		drbd_blk_run_queue(bdev_get_queue(mdev->ldev->backing_bdev));
2243 		put_ldev(mdev);
2244 	}
2245 }
2246 
2247 static inline void drbd_md_flush(struct drbd_conf *mdev)
2248 {
2249 	int r;
2250 
2251 	if (test_bit(MD_NO_BARRIER, &mdev->flags))
2252 		return;
2253 
2254 	r = blkdev_issue_flush(mdev->ldev->md_bdev, NULL);
2255 	if (r) {
2256 		set_bit(MD_NO_BARRIER, &mdev->flags);
2257 		dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r);
2258 	}
2259 }
2260 
2261 #endif
2262