xref: /linux/drivers/block/drbd/drbd_int.h (revision 12871a0bd67dd4db4418e1daafcd46e9d329ef10)
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 #include <linux/prefetch.h>
46 
47 #ifdef __CHECKER__
48 # define __protected_by(x)       __attribute__((require_context(x,1,999,"rdwr")))
49 # define __protected_read_by(x)  __attribute__((require_context(x,1,999,"read")))
50 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
51 # define __must_hold(x)       __attribute__((context(x,1,1), require_context(x,1,999,"call")))
52 #else
53 # define __protected_by(x)
54 # define __protected_read_by(x)
55 # define __protected_write_by(x)
56 # define __must_hold(x)
57 #endif
58 
59 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
60 
61 /* module parameter, defined in drbd_main.c */
62 extern unsigned int minor_count;
63 extern int disable_sendpage;
64 extern int allow_oos;
65 extern unsigned int cn_idx;
66 
67 #ifdef CONFIG_DRBD_FAULT_INJECTION
68 extern int enable_faults;
69 extern int fault_rate;
70 extern int fault_devs;
71 #endif
72 
73 extern char usermode_helper[];
74 
75 
76 /* I don't remember why XCPU ...
77  * This is used to wake the asender,
78  * and to interrupt sending the sending task
79  * on disconnect.
80  */
81 #define DRBD_SIG SIGXCPU
82 
83 /* This is used to stop/restart our threads.
84  * Cannot use SIGTERM nor SIGKILL, since these
85  * are sent out by init on runlevel changes
86  * I choose SIGHUP for now.
87  */
88 #define DRBD_SIGKILL SIGHUP
89 
90 /* All EEs on the free list should have ID_VACANT (== 0)
91  * freshly allocated EEs get !ID_VACANT (== 1)
92  * so if it says "cannot dereference null pointer at address 0x00000001",
93  * it is most likely one of these :( */
94 
95 #define ID_IN_SYNC      (4711ULL)
96 #define ID_OUT_OF_SYNC  (4712ULL)
97 
98 #define ID_SYNCER (-1ULL)
99 #define ID_VACANT 0
100 #define is_syncer_block_id(id) ((id) == ID_SYNCER)
101 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
102 
103 struct drbd_conf;
104 
105 
106 /* to shorten dev_warn(DEV, "msg"); and relatives statements */
107 #define DEV (disk_to_dev(mdev->vdisk))
108 
109 #define D_ASSERT(exp)	if (!(exp)) \
110 	 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
111 
112 #define ERR_IF(exp) if (({						\
113 	int _b = (exp) != 0;						\
114 	if (_b) dev_err(DEV, "ASSERT FAILED: %s: (%s) in %s:%d\n",	\
115 			__func__, #exp, __FILE__, __LINE__);		\
116 	_b;								\
117 	}))
118 
119 /* Defines to control fault insertion */
120 enum {
121 	DRBD_FAULT_MD_WR = 0,	/* meta data write */
122 	DRBD_FAULT_MD_RD = 1,	/*           read  */
123 	DRBD_FAULT_RS_WR = 2,	/* resync          */
124 	DRBD_FAULT_RS_RD = 3,
125 	DRBD_FAULT_DT_WR = 4,	/* data            */
126 	DRBD_FAULT_DT_RD = 5,
127 	DRBD_FAULT_DT_RA = 6,	/* data read ahead */
128 	DRBD_FAULT_BM_ALLOC = 7,	/* bitmap allocation */
129 	DRBD_FAULT_AL_EE = 8,	/* alloc ee */
130 	DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
131 
132 	DRBD_FAULT_MAX,
133 };
134 
135 extern unsigned int
136 _drbd_insert_fault(struct drbd_conf *mdev, unsigned int type);
137 
138 static inline int
139 drbd_insert_fault(struct drbd_conf *mdev, unsigned int type) {
140 #ifdef CONFIG_DRBD_FAULT_INJECTION
141 	return fault_rate &&
142 		(enable_faults & (1<<type)) &&
143 		_drbd_insert_fault(mdev, type);
144 #else
145 	return 0;
146 #endif
147 }
148 
149 /* integer division, round _UP_ to the next integer */
150 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
151 /* usual integer division */
152 #define div_floor(A, B) ((A)/(B))
153 
154 /* drbd_meta-data.c (still in drbd_main.c) */
155 /* 4th incarnation of the disk layout. */
156 #define DRBD_MD_MAGIC (DRBD_MAGIC+4)
157 
158 extern struct drbd_conf **minor_table;
159 extern struct ratelimit_state drbd_ratelimit_state;
160 
161 /* on the wire */
162 enum drbd_packets {
163 	/* receiver (data socket) */
164 	P_DATA		      = 0x00,
165 	P_DATA_REPLY	      = 0x01, /* Response to P_DATA_REQUEST */
166 	P_RS_DATA_REPLY	      = 0x02, /* Response to P_RS_DATA_REQUEST */
167 	P_BARRIER	      = 0x03,
168 	P_BITMAP	      = 0x04,
169 	P_BECOME_SYNC_TARGET  = 0x05,
170 	P_BECOME_SYNC_SOURCE  = 0x06,
171 	P_UNPLUG_REMOTE	      = 0x07, /* Used at various times to hint the peer */
172 	P_DATA_REQUEST	      = 0x08, /* Used to ask for a data block */
173 	P_RS_DATA_REQUEST     = 0x09, /* Used to ask for a data block for resync */
174 	P_SYNC_PARAM	      = 0x0a,
175 	P_PROTOCOL	      = 0x0b,
176 	P_UUIDS		      = 0x0c,
177 	P_SIZES		      = 0x0d,
178 	P_STATE		      = 0x0e,
179 	P_SYNC_UUID	      = 0x0f,
180 	P_AUTH_CHALLENGE      = 0x10,
181 	P_AUTH_RESPONSE	      = 0x11,
182 	P_STATE_CHG_REQ	      = 0x12,
183 
184 	/* asender (meta socket */
185 	P_PING		      = 0x13,
186 	P_PING_ACK	      = 0x14,
187 	P_RECV_ACK	      = 0x15, /* Used in protocol B */
188 	P_WRITE_ACK	      = 0x16, /* Used in protocol C */
189 	P_RS_WRITE_ACK	      = 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
190 	P_DISCARD_ACK	      = 0x18, /* Used in proto C, two-primaries conflict detection */
191 	P_NEG_ACK	      = 0x19, /* Sent if local disk is unusable */
192 	P_NEG_DREPLY	      = 0x1a, /* Local disk is broken... */
193 	P_NEG_RS_DREPLY	      = 0x1b, /* Local disk is broken... */
194 	P_BARRIER_ACK	      = 0x1c,
195 	P_STATE_CHG_REPLY     = 0x1d,
196 
197 	/* "new" commands, no longer fitting into the ordering scheme above */
198 
199 	P_OV_REQUEST	      = 0x1e, /* data socket */
200 	P_OV_REPLY	      = 0x1f,
201 	P_OV_RESULT	      = 0x20, /* meta socket */
202 	P_CSUM_RS_REQUEST     = 0x21, /* data socket */
203 	P_RS_IS_IN_SYNC	      = 0x22, /* meta socket */
204 	P_SYNC_PARAM89	      = 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
205 	P_COMPRESSED_BITMAP   = 0x24, /* compressed or otherwise encoded bitmap transfer */
206 	/* P_CKPT_FENCE_REQ      = 0x25, * currently reserved for protocol D */
207 	/* P_CKPT_DISABLE_REQ    = 0x26, * currently reserved for protocol D */
208 	P_DELAY_PROBE         = 0x27, /* is used on BOTH sockets */
209 	P_OUT_OF_SYNC         = 0x28, /* Mark as out of sync (Outrunning), data socket */
210 	P_RS_CANCEL           = 0x29, /* meta: Used to cancel RS_DATA_REQUEST packet by SyncSource */
211 
212 	P_MAX_CMD	      = 0x2A,
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_DELAY_PROBE]         = "DelayProbe",
268 		[P_OUT_OF_SYNC]		= "OutOfSync",
269 		[P_MAX_CMD]	        = NULL,
270 	};
271 
272 	if (cmd == P_HAND_SHAKE_M)
273 		return "HandShakeM";
274 	if (cmd == P_HAND_SHAKE_S)
275 		return "HandShakeS";
276 	if (cmd == P_HAND_SHAKE)
277 		return "HandShake";
278 	if (cmd >= P_MAX_CMD)
279 		return "Unknown";
280 	return cmdnames[cmd];
281 }
282 
283 /* for sending/receiving the bitmap,
284  * possibly in some encoding scheme */
285 struct bm_xfer_ctx {
286 	/* "const"
287 	 * stores total bits and long words
288 	 * of the bitmap, so we don't need to
289 	 * call the accessor functions over and again. */
290 	unsigned long bm_bits;
291 	unsigned long bm_words;
292 	/* during xfer, current position within the bitmap */
293 	unsigned long bit_offset;
294 	unsigned long word_offset;
295 
296 	/* statistics; index: (h->command == P_BITMAP) */
297 	unsigned packets[2];
298 	unsigned bytes[2];
299 };
300 
301 extern void INFO_bm_xfer_stats(struct drbd_conf *mdev,
302 		const char *direction, struct bm_xfer_ctx *c);
303 
304 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
305 {
306 	/* word_offset counts "native long words" (32 or 64 bit),
307 	 * aligned at 64 bit.
308 	 * Encoded packet may end at an unaligned bit offset.
309 	 * In case a fallback clear text packet is transmitted in
310 	 * between, we adjust this offset back to the last 64bit
311 	 * aligned "native long word", which makes coding and decoding
312 	 * the plain text bitmap much more convenient.  */
313 #if BITS_PER_LONG == 64
314 	c->word_offset = c->bit_offset >> 6;
315 #elif BITS_PER_LONG == 32
316 	c->word_offset = c->bit_offset >> 5;
317 	c->word_offset &= ~(1UL);
318 #else
319 # error "unsupported BITS_PER_LONG"
320 #endif
321 }
322 
323 #ifndef __packed
324 #define __packed __attribute__((packed))
325 #endif
326 
327 /* This is the layout for a packet on the wire.
328  * The byteorder is the network byte order.
329  *     (except block_id and barrier fields.
330  *	these are pointers to local structs
331  *	and have no relevance for the partner,
332  *	which just echoes them as received.)
333  *
334  * NOTE that the payload starts at a long aligned offset,
335  * regardless of 32 or 64 bit arch!
336  */
337 struct p_header80 {
338 	u32	  magic;
339 	u16	  command;
340 	u16	  length;	/* bytes of data after this header */
341 	u8	  payload[0];
342 } __packed;
343 
344 /* Header for big packets, Used for data packets exceeding 64kB */
345 struct p_header95 {
346 	u16	  magic;	/* use DRBD_MAGIC_BIG here */
347 	u16	  command;
348 	u32	  length;	/* Use only 24 bits of that. Ignore the highest 8 bit. */
349 	u8	  payload[0];
350 } __packed;
351 
352 union p_header {
353 	struct p_header80 h80;
354 	struct p_header95 h95;
355 };
356 
357 /*
358  * short commands, packets without payload, plain p_header:
359  *   P_PING
360  *   P_PING_ACK
361  *   P_BECOME_SYNC_TARGET
362  *   P_BECOME_SYNC_SOURCE
363  *   P_UNPLUG_REMOTE
364  */
365 
366 /*
367  * commands with out-of-struct payload:
368  *   P_BITMAP    (no additional fields)
369  *   P_DATA, P_DATA_REPLY (see p_data)
370  *   P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
371  */
372 
373 /* these defines must not be changed without changing the protocol version */
374 #define DP_HARDBARRIER	      1 /* depricated */
375 #define DP_RW_SYNC	      2 /* equals REQ_SYNC    */
376 #define DP_MAY_SET_IN_SYNC    4
377 #define DP_UNPLUG             8 /* not used anymore   */
378 #define DP_FUA               16 /* equals REQ_FUA     */
379 #define DP_FLUSH             32 /* equals REQ_FLUSH   */
380 #define DP_DISCARD           64 /* equals REQ_DISCARD */
381 
382 struct p_data {
383 	union p_header head;
384 	u64	    sector;    /* 64 bits sector number */
385 	u64	    block_id;  /* to identify the request in protocol B&C */
386 	u32	    seq_num;
387 	u32	    dp_flags;
388 } __packed;
389 
390 /*
391  * commands which share a struct:
392  *  p_block_ack:
393  *   P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
394  *   P_DISCARD_ACK (proto C, two-primaries conflict detection)
395  *  p_block_req:
396  *   P_DATA_REQUEST, P_RS_DATA_REQUEST
397  */
398 struct p_block_ack {
399 	struct p_header80 head;
400 	u64	    sector;
401 	u64	    block_id;
402 	u32	    blksize;
403 	u32	    seq_num;
404 } __packed;
405 
406 
407 struct p_block_req {
408 	struct p_header80 head;
409 	u64 sector;
410 	u64 block_id;
411 	u32 blksize;
412 	u32 pad;	/* to multiple of 8 Byte */
413 } __packed;
414 
415 /*
416  * commands with their own struct for additional fields:
417  *   P_HAND_SHAKE
418  *   P_BARRIER
419  *   P_BARRIER_ACK
420  *   P_SYNC_PARAM
421  *   ReportParams
422  */
423 
424 struct p_handshake {
425 	struct p_header80 head;	/* 8 bytes */
426 	u32 protocol_min;
427 	u32 feature_flags;
428 	u32 protocol_max;
429 
430 	/* should be more than enough for future enhancements
431 	 * for now, feature_flags and the reserverd array shall be zero.
432 	 */
433 
434 	u32 _pad;
435 	u64 reserverd[7];
436 } __packed;
437 /* 80 bytes, FIXED for the next century */
438 
439 struct p_barrier {
440 	struct p_header80 head;
441 	u32 barrier;	/* barrier number _handle_ only */
442 	u32 pad;	/* to multiple of 8 Byte */
443 } __packed;
444 
445 struct p_barrier_ack {
446 	struct p_header80 head;
447 	u32 barrier;
448 	u32 set_size;
449 } __packed;
450 
451 struct p_rs_param {
452 	struct p_header80 head;
453 	u32 rate;
454 
455 	      /* Since protocol version 88 and higher. */
456 	char verify_alg[0];
457 } __packed;
458 
459 struct p_rs_param_89 {
460 	struct p_header80 head;
461 	u32 rate;
462         /* protocol version 89: */
463 	char verify_alg[SHARED_SECRET_MAX];
464 	char csums_alg[SHARED_SECRET_MAX];
465 } __packed;
466 
467 struct p_rs_param_95 {
468 	struct p_header80 head;
469 	u32 rate;
470 	char verify_alg[SHARED_SECRET_MAX];
471 	char csums_alg[SHARED_SECRET_MAX];
472 	u32 c_plan_ahead;
473 	u32 c_delay_target;
474 	u32 c_fill_target;
475 	u32 c_max_rate;
476 } __packed;
477 
478 enum drbd_conn_flags {
479 	CF_WANT_LOSE = 1,
480 	CF_DRY_RUN = 2,
481 };
482 
483 struct p_protocol {
484 	struct p_header80 head;
485 	u32 protocol;
486 	u32 after_sb_0p;
487 	u32 after_sb_1p;
488 	u32 after_sb_2p;
489 	u32 conn_flags;
490 	u32 two_primaries;
491 
492               /* Since protocol version 87 and higher. */
493 	char integrity_alg[0];
494 
495 } __packed;
496 
497 struct p_uuids {
498 	struct p_header80 head;
499 	u64 uuid[UI_EXTENDED_SIZE];
500 } __packed;
501 
502 struct p_rs_uuid {
503 	struct p_header80 head;
504 	u64	    uuid;
505 } __packed;
506 
507 struct p_sizes {
508 	struct p_header80 head;
509 	u64	    d_size;  /* size of disk */
510 	u64	    u_size;  /* user requested size */
511 	u64	    c_size;  /* current exported size */
512 	u32	    max_bio_size;  /* Maximal size of a BIO */
513 	u16	    queue_order_type;  /* not yet implemented in DRBD*/
514 	u16	    dds_flags; /* use enum dds_flags here. */
515 } __packed;
516 
517 struct p_state {
518 	struct p_header80 head;
519 	u32	    state;
520 } __packed;
521 
522 struct p_req_state {
523 	struct p_header80 head;
524 	u32	    mask;
525 	u32	    val;
526 } __packed;
527 
528 struct p_req_state_reply {
529 	struct p_header80 head;
530 	u32	    retcode;
531 } __packed;
532 
533 struct p_drbd06_param {
534 	u64	  size;
535 	u32	  state;
536 	u32	  blksize;
537 	u32	  protocol;
538 	u32	  version;
539 	u32	  gen_cnt[5];
540 	u32	  bit_map_gen[5];
541 } __packed;
542 
543 struct p_discard {
544 	struct p_header80 head;
545 	u64	    block_id;
546 	u32	    seq_num;
547 	u32	    pad;
548 } __packed;
549 
550 struct p_block_desc {
551 	struct p_header80 head;
552 	u64 sector;
553 	u32 blksize;
554 	u32 pad;	/* to multiple of 8 Byte */
555 } __packed;
556 
557 /* Valid values for the encoding field.
558  * Bump proto version when changing this. */
559 enum drbd_bitmap_code {
560 	/* RLE_VLI_Bytes = 0,
561 	 * and other bit variants had been defined during
562 	 * algorithm evaluation. */
563 	RLE_VLI_Bits = 2,
564 };
565 
566 struct p_compressed_bm {
567 	struct p_header80 head;
568 	/* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
569 	 * (encoding & 0x80): polarity (set/unset) of first runlength
570 	 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
571 	 * used to pad up to head.length bytes
572 	 */
573 	u8 encoding;
574 
575 	u8 code[0];
576 } __packed;
577 
578 struct p_delay_probe93 {
579 	struct p_header80 head;
580 	u32     seq_num; /* sequence number to match the two probe packets */
581 	u32     offset;  /* usecs the probe got sent after the reference time point */
582 } __packed;
583 
584 /* DCBP: Drbd Compressed Bitmap Packet ... */
585 static inline enum drbd_bitmap_code
586 DCBP_get_code(struct p_compressed_bm *p)
587 {
588 	return (enum drbd_bitmap_code)(p->encoding & 0x0f);
589 }
590 
591 static inline void
592 DCBP_set_code(struct p_compressed_bm *p, enum drbd_bitmap_code code)
593 {
594 	BUG_ON(code & ~0xf);
595 	p->encoding = (p->encoding & ~0xf) | code;
596 }
597 
598 static inline int
599 DCBP_get_start(struct p_compressed_bm *p)
600 {
601 	return (p->encoding & 0x80) != 0;
602 }
603 
604 static inline void
605 DCBP_set_start(struct p_compressed_bm *p, int set)
606 {
607 	p->encoding = (p->encoding & ~0x80) | (set ? 0x80 : 0);
608 }
609 
610 static inline int
611 DCBP_get_pad_bits(struct p_compressed_bm *p)
612 {
613 	return (p->encoding >> 4) & 0x7;
614 }
615 
616 static inline void
617 DCBP_set_pad_bits(struct p_compressed_bm *p, int n)
618 {
619 	BUG_ON(n & ~0x7);
620 	p->encoding = (p->encoding & (~0x7 << 4)) | (n << 4);
621 }
622 
623 /* one bitmap packet, including the p_header,
624  * should fit within one _architecture independend_ page.
625  * so we need to use the fixed size 4KiB page size
626  * most architectures have used for a long time.
627  */
628 #define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header80))
629 #define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
630 #define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm))
631 #if (PAGE_SIZE < 4096)
632 /* drbd_send_bitmap / receive_bitmap would break horribly */
633 #error "PAGE_SIZE too small"
634 #endif
635 
636 union p_polymorph {
637         union p_header           header;
638         struct p_handshake       handshake;
639         struct p_data            data;
640         struct p_block_ack       block_ack;
641         struct p_barrier         barrier;
642         struct p_barrier_ack     barrier_ack;
643         struct p_rs_param_89     rs_param_89;
644         struct p_rs_param_95     rs_param_95;
645         struct p_protocol        protocol;
646         struct p_sizes           sizes;
647         struct p_uuids           uuids;
648         struct p_state           state;
649         struct p_req_state       req_state;
650         struct p_req_state_reply req_state_reply;
651         struct p_block_req       block_req;
652 	struct p_delay_probe93   delay_probe93;
653 	struct p_rs_uuid         rs_uuid;
654 	struct p_block_desc      block_desc;
655 } __packed;
656 
657 /**********************************************************************/
658 enum drbd_thread_state {
659 	None,
660 	Running,
661 	Exiting,
662 	Restarting
663 };
664 
665 struct drbd_thread {
666 	spinlock_t t_lock;
667 	struct task_struct *task;
668 	struct completion stop;
669 	enum drbd_thread_state t_state;
670 	int (*function) (struct drbd_thread *);
671 	struct drbd_conf *mdev;
672 	int reset_cpu_mask;
673 };
674 
675 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
676 {
677 	/* THINK testing the t_state seems to be uncritical in all cases
678 	 * (but thread_{start,stop}), so we can read it *without* the lock.
679 	 *	--lge */
680 
681 	smp_rmb();
682 	return thi->t_state;
683 }
684 
685 struct drbd_work;
686 typedef int (*drbd_work_cb)(struct drbd_conf *, struct drbd_work *, int cancel);
687 struct drbd_work {
688 	struct list_head list;
689 	drbd_work_cb cb;
690 };
691 
692 struct drbd_tl_epoch;
693 struct drbd_request {
694 	struct drbd_work w;
695 	struct drbd_conf *mdev;
696 
697 	/* if local IO is not allowed, will be NULL.
698 	 * if local IO _is_ allowed, holds the locally submitted bio clone,
699 	 * or, after local IO completion, the ERR_PTR(error).
700 	 * see drbd_endio_pri(). */
701 	struct bio *private_bio;
702 
703 	struct hlist_node collision;
704 	sector_t sector;
705 	unsigned int size;
706 	unsigned int epoch; /* barrier_nr */
707 
708 	/* barrier_nr: used to check on "completion" whether this req was in
709 	 * the current epoch, and we therefore have to close it,
710 	 * starting a new epoch...
711 	 */
712 
713 	struct list_head tl_requests; /* ring list in the transfer log */
714 	struct bio *master_bio;       /* master bio pointer */
715 	unsigned long rq_state; /* see comments above _req_mod() */
716 	int seq_num;
717 	unsigned long start_time;
718 };
719 
720 struct drbd_tl_epoch {
721 	struct drbd_work w;
722 	struct list_head requests; /* requests before */
723 	struct drbd_tl_epoch *next; /* pointer to the next barrier */
724 	unsigned int br_number;  /* the barriers identifier. */
725 	int n_writes;	/* number of requests attached before this barrier */
726 };
727 
728 struct drbd_request;
729 
730 /* These Tl_epoch_entries may be in one of 6 lists:
731    active_ee .. data packet being written
732    sync_ee   .. syncer block being written
733    done_ee   .. block written, need to send P_WRITE_ACK
734    read_ee   .. [RS]P_DATA_REQUEST being read
735 */
736 
737 struct drbd_epoch {
738 	struct list_head list;
739 	unsigned int barrier_nr;
740 	atomic_t epoch_size; /* increased on every request added. */
741 	atomic_t active;     /* increased on every req. added, and dec on every finished. */
742 	unsigned long flags;
743 };
744 
745 /* drbd_epoch flag bits */
746 enum {
747 	DE_HAVE_BARRIER_NUMBER,
748 };
749 
750 enum epoch_event {
751 	EV_PUT,
752 	EV_GOT_BARRIER_NR,
753 	EV_BECAME_LAST,
754 	EV_CLEANUP = 32, /* used as flag */
755 };
756 
757 struct drbd_wq_barrier {
758 	struct drbd_work w;
759 	struct completion done;
760 };
761 
762 struct digest_info {
763 	int digest_size;
764 	void *digest;
765 };
766 
767 struct drbd_epoch_entry {
768 	struct drbd_work w;
769 	struct hlist_node collision;
770 	struct drbd_epoch *epoch; /* for writes */
771 	struct drbd_conf *mdev;
772 	struct page *pages;
773 	atomic_t pending_bios;
774 	unsigned int size;
775 	/* see comments on ee flag bits below */
776 	unsigned long flags;
777 	sector_t sector;
778 	union {
779 		u64 block_id;
780 		struct digest_info *digest;
781 	};
782 };
783 
784 /* ee flag bits.
785  * While corresponding bios are in flight, the only modification will be
786  * set_bit WAS_ERROR, which has to be atomic.
787  * If no bios are in flight yet, or all have been completed,
788  * non-atomic modification to ee->flags is ok.
789  */
790 enum {
791 	__EE_CALL_AL_COMPLETE_IO,
792 	__EE_MAY_SET_IN_SYNC,
793 
794 	/* In case a barrier failed,
795 	 * we need to resubmit without the barrier flag. */
796 	__EE_RESUBMITTED,
797 
798 	/* we may have several bios per epoch entry.
799 	 * if any of those fail, we set this flag atomically
800 	 * from the endio callback */
801 	__EE_WAS_ERROR,
802 
803 	/* This ee has a pointer to a digest instead of a block id */
804 	__EE_HAS_DIGEST,
805 };
806 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
807 #define EE_MAY_SET_IN_SYNC     (1<<__EE_MAY_SET_IN_SYNC)
808 #define	EE_RESUBMITTED         (1<<__EE_RESUBMITTED)
809 #define EE_WAS_ERROR           (1<<__EE_WAS_ERROR)
810 #define EE_HAS_DIGEST          (1<<__EE_HAS_DIGEST)
811 
812 /* global flag bits */
813 enum {
814 	CREATE_BARRIER,		/* next P_DATA is preceded by a P_BARRIER */
815 	SIGNAL_ASENDER,		/* whether asender wants to be interrupted */
816 	SEND_PING,		/* whether asender should send a ping asap */
817 
818 	UNPLUG_QUEUED,		/* only relevant with kernel 2.4 */
819 	UNPLUG_REMOTE,		/* sending a "UnplugRemote" could help */
820 	MD_DIRTY,		/* current uuids and flags not yet on disk */
821 	DISCARD_CONCURRENT,	/* Set on one node, cleared on the peer! */
822 	USE_DEGR_WFC_T,		/* degr-wfc-timeout instead of wfc-timeout. */
823 	CLUSTER_ST_CHANGE,	/* Cluster wide state change going on... */
824 	CL_ST_CHG_SUCCESS,
825 	CL_ST_CHG_FAIL,
826 	CRASHED_PRIMARY,	/* This node was a crashed primary.
827 				 * Gets cleared when the state.conn
828 				 * goes into C_CONNECTED state. */
829 	NO_BARRIER_SUPP,	/* underlying block device doesn't implement barriers */
830 	CONSIDER_RESYNC,
831 
832 	MD_NO_FUA,		/* Users wants us to not use FUA/FLUSH on meta data dev */
833 	SUSPEND_IO,		/* suspend application io */
834 	BITMAP_IO,		/* suspend application io;
835 				   once no more io in flight, start bitmap io */
836 	BITMAP_IO_QUEUED,       /* Started bitmap IO */
837 	GO_DISKLESS,		/* Disk is being detached, on io-error or admin request. */
838 	WAS_IO_ERROR,		/* Local disk failed returned IO error */
839 	RESYNC_AFTER_NEG,       /* Resync after online grow after the attach&negotiate finished. */
840 	NET_CONGESTED,		/* The data socket is congested */
841 
842 	CONFIG_PENDING,		/* serialization of (re)configuration requests.
843 				 * if set, also prevents the device from dying */
844 	DEVICE_DYING,		/* device became unconfigured,
845 				 * but worker thread is still handling the cleanup.
846 				 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
847 				 * while this is set. */
848 	RESIZE_PENDING,		/* Size change detected locally, waiting for the response from
849 				 * the peer, if it changed there as well. */
850 	CONN_DRY_RUN,		/* Expect disconnect after resync handshake. */
851 	GOT_PING_ACK,		/* set when we receive a ping_ack packet, misc wait gets woken */
852 	NEW_CUR_UUID,		/* Create new current UUID when thawing IO */
853 	AL_SUSPENDED,		/* Activity logging is currently suspended. */
854 	AHEAD_TO_SYNC_SOURCE,   /* Ahead -> SyncSource queued */
855 };
856 
857 struct drbd_bitmap; /* opaque for drbd_conf */
858 
859 /* definition of bits in bm_flags to be used in drbd_bm_lock
860  * and drbd_bitmap_io and friends. */
861 enum bm_flag {
862 	/* do we need to kfree, or vfree bm_pages? */
863 	BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */
864 
865 	/* currently locked for bulk operation */
866 	BM_LOCKED_MASK = 0x7,
867 
868 	/* in detail, that is: */
869 	BM_DONT_CLEAR = 0x1,
870 	BM_DONT_SET   = 0x2,
871 	BM_DONT_TEST  = 0x4,
872 
873 	/* (test bit, count bit) allowed (common case) */
874 	BM_LOCKED_TEST_ALLOWED = 0x3,
875 
876 	/* testing bits, as well as setting new bits allowed, but clearing bits
877 	 * would be unexpected.  Used during bitmap receive.  Setting new bits
878 	 * requires sending of "out-of-sync" information, though. */
879 	BM_LOCKED_SET_ALLOWED = 0x1,
880 
881 	/* clear is not expected while bitmap is locked for bulk operation */
882 };
883 
884 
885 /* TODO sort members for performance
886  * MAYBE group them further */
887 
888 /* THINK maybe we actually want to use the default "event/%s" worker threads
889  * or similar in linux 2.6, which uses per cpu data and threads.
890  */
891 struct drbd_work_queue {
892 	struct list_head q;
893 	struct semaphore s; /* producers up it, worker down()s it */
894 	spinlock_t q_lock;  /* to protect the list. */
895 };
896 
897 struct drbd_socket {
898 	struct drbd_work_queue work;
899 	struct mutex mutex;
900 	struct socket    *socket;
901 	/* this way we get our
902 	 * send/receive buffers off the stack */
903 	union p_polymorph sbuf;
904 	union p_polymorph rbuf;
905 };
906 
907 struct drbd_md {
908 	u64 md_offset;		/* sector offset to 'super' block */
909 
910 	u64 la_size_sect;	/* last agreed size, unit sectors */
911 	u64 uuid[UI_SIZE];
912 	u64 device_uuid;
913 	u32 flags;
914 	u32 md_size_sect;
915 
916 	s32 al_offset;	/* signed relative sector offset to al area */
917 	s32 bm_offset;	/* signed relative sector offset to bitmap */
918 
919 	/* u32 al_nr_extents;	   important for restoring the AL
920 	 * is stored into  sync_conf.al_extents, which in turn
921 	 * gets applied to act_log->nr_elements
922 	 */
923 };
924 
925 /* for sync_conf and other types... */
926 #define NL_PACKET(name, number, fields) struct name { fields };
927 #define NL_INTEGER(pn,pr,member) int member;
928 #define NL_INT64(pn,pr,member) __u64 member;
929 #define NL_BIT(pn,pr,member)   unsigned member:1;
930 #define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
931 #include "linux/drbd_nl.h"
932 
933 struct drbd_backing_dev {
934 	struct block_device *backing_bdev;
935 	struct block_device *md_bdev;
936 	struct drbd_md md;
937 	struct disk_conf dc; /* The user provided config... */
938 	sector_t known_size; /* last known size of that backing device */
939 };
940 
941 struct drbd_md_io {
942 	struct drbd_conf *mdev;
943 	struct completion event;
944 	int error;
945 };
946 
947 struct bm_io_work {
948 	struct drbd_work w;
949 	char *why;
950 	enum bm_flag flags;
951 	int (*io_fn)(struct drbd_conf *mdev);
952 	void (*done)(struct drbd_conf *mdev, int rv);
953 };
954 
955 enum write_ordering_e {
956 	WO_none,
957 	WO_drain_io,
958 	WO_bdev_flush,
959 };
960 
961 struct fifo_buffer {
962 	int *values;
963 	unsigned int head_index;
964 	unsigned int size;
965 };
966 
967 struct drbd_conf {
968 	/* things that are stored as / read from meta data on disk */
969 	unsigned long flags;
970 
971 	/* configured by drbdsetup */
972 	struct net_conf *net_conf; /* protected by get_net_conf() and put_net_conf() */
973 	struct syncer_conf sync_conf;
974 	struct drbd_backing_dev *ldev __protected_by(local);
975 
976 	sector_t p_size;     /* partner's disk size */
977 	struct request_queue *rq_queue;
978 	struct block_device *this_bdev;
979 	struct gendisk	    *vdisk;
980 
981 	struct drbd_socket data; /* data/barrier/cstate/parameter packets */
982 	struct drbd_socket meta; /* ping/ack (metadata) packets */
983 	int agreed_pro_version;  /* actually used protocol version */
984 	unsigned long last_received; /* in jiffies, either socket */
985 	unsigned int ko_count;
986 	struct drbd_work  resync_work,
987 			  unplug_work,
988 			  go_diskless,
989 			  md_sync_work,
990 			  start_resync_work;
991 	struct timer_list resync_timer;
992 	struct timer_list md_sync_timer;
993 	struct timer_list start_resync_timer;
994 	struct timer_list request_timer;
995 #ifdef DRBD_DEBUG_MD_SYNC
996 	struct {
997 		unsigned int line;
998 		const char* func;
999 	} last_md_mark_dirty;
1000 #endif
1001 
1002 	/* Used after attach while negotiating new disk state. */
1003 	union drbd_state new_state_tmp;
1004 
1005 	union drbd_state state;
1006 	wait_queue_head_t misc_wait;
1007 	wait_queue_head_t state_wait;  /* upon each state change. */
1008 	wait_queue_head_t net_cnt_wait;
1009 	unsigned int send_cnt;
1010 	unsigned int recv_cnt;
1011 	unsigned int read_cnt;
1012 	unsigned int writ_cnt;
1013 	unsigned int al_writ_cnt;
1014 	unsigned int bm_writ_cnt;
1015 	atomic_t ap_bio_cnt;	 /* Requests we need to complete */
1016 	atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
1017 	atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
1018 	atomic_t unacked_cnt;	 /* Need to send replys for */
1019 	atomic_t local_cnt;	 /* Waiting for local completion */
1020 	atomic_t net_cnt;	 /* Users of net_conf */
1021 	spinlock_t req_lock;
1022 	struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */
1023 	struct drbd_tl_epoch *newest_tle;
1024 	struct drbd_tl_epoch *oldest_tle;
1025 	struct list_head out_of_sequence_requests;
1026 	struct hlist_head *tl_hash;
1027 	unsigned int tl_hash_s;
1028 
1029 	/* blocks to resync in this run [unit BM_BLOCK_SIZE] */
1030 	unsigned long rs_total;
1031 	/* number of resync blocks that failed in this run */
1032 	unsigned long rs_failed;
1033 	/* Syncer's start time [unit jiffies] */
1034 	unsigned long rs_start;
1035 	/* cumulated time in PausedSyncX state [unit jiffies] */
1036 	unsigned long rs_paused;
1037 	/* skipped because csum was equal [unit BM_BLOCK_SIZE] */
1038 	unsigned long rs_same_csum;
1039 #define DRBD_SYNC_MARKS 8
1040 #define DRBD_SYNC_MARK_STEP (3*HZ)
1041 	/* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
1042 	unsigned long rs_mark_left[DRBD_SYNC_MARKS];
1043 	/* marks's time [unit jiffies] */
1044 	unsigned long rs_mark_time[DRBD_SYNC_MARKS];
1045 	/* current index into rs_mark_{left,time} */
1046 	int rs_last_mark;
1047 
1048 	/* where does the admin want us to start? (sector) */
1049 	sector_t ov_start_sector;
1050 	/* where are we now? (sector) */
1051 	sector_t ov_position;
1052 	/* Start sector of out of sync range (to merge printk reporting). */
1053 	sector_t ov_last_oos_start;
1054 	/* size of out-of-sync range in sectors. */
1055 	sector_t ov_last_oos_size;
1056 	unsigned long ov_left; /* in bits */
1057 	struct crypto_hash *csums_tfm;
1058 	struct crypto_hash *verify_tfm;
1059 
1060 	struct drbd_thread receiver;
1061 	struct drbd_thread worker;
1062 	struct drbd_thread asender;
1063 	struct drbd_bitmap *bitmap;
1064 	unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
1065 
1066 	/* Used to track operations of resync... */
1067 	struct lru_cache *resync;
1068 	/* Number of locked elements in resync LRU */
1069 	unsigned int resync_locked;
1070 	/* resync extent number waiting for application requests */
1071 	unsigned int resync_wenr;
1072 
1073 	int open_cnt;
1074 	u64 *p_uuid;
1075 	struct drbd_epoch *current_epoch;
1076 	spinlock_t epoch_lock;
1077 	unsigned int epochs;
1078 	enum write_ordering_e write_ordering;
1079 	struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
1080 	struct list_head sync_ee;   /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
1081 	struct list_head done_ee;   /* send ack */
1082 	struct list_head read_ee;   /* IO in progress (any read) */
1083 	struct list_head net_ee;    /* zero-copy network send in progress */
1084 	struct hlist_head *ee_hash; /* is proteced by req_lock! */
1085 	unsigned int ee_hash_s;
1086 
1087 	/* this one is protected by ee_lock, single thread */
1088 	struct drbd_epoch_entry *last_write_w_barrier;
1089 
1090 	int next_barrier_nr;
1091 	struct hlist_head *app_reads_hash; /* is proteced by req_lock */
1092 	struct list_head resync_reads;
1093 	atomic_t pp_in_use;		/* allocated from page pool */
1094 	atomic_t pp_in_use_by_net;	/* sendpage()d, still referenced by tcp */
1095 	wait_queue_head_t ee_wait;
1096 	struct page *md_io_page;	/* one page buffer for md_io */
1097 	struct page *md_io_tmpp;	/* for logical_block_size != 512 */
1098 	struct mutex md_io_mutex;	/* protects the md_io_buffer */
1099 	spinlock_t al_lock;
1100 	wait_queue_head_t al_wait;
1101 	struct lru_cache *act_log;	/* activity log */
1102 	unsigned int al_tr_number;
1103 	int al_tr_cycle;
1104 	int al_tr_pos;   /* position of the next transaction in the journal */
1105 	struct crypto_hash *cram_hmac_tfm;
1106 	struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */
1107 	struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */
1108 	void *int_dig_out;
1109 	void *int_dig_in;
1110 	void *int_dig_vv;
1111 	wait_queue_head_t seq_wait;
1112 	atomic_t packet_seq;
1113 	unsigned int peer_seq;
1114 	spinlock_t peer_seq_lock;
1115 	unsigned int minor;
1116 	unsigned long comm_bm_set; /* communicated number of set bits. */
1117 	cpumask_var_t cpu_mask;
1118 	struct bm_io_work bm_io_work;
1119 	u64 ed_uuid; /* UUID of the exposed data */
1120 	struct mutex state_mutex;
1121 	char congestion_reason;  /* Why we where congested... */
1122 	atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
1123 	atomic_t rs_sect_ev; /* for submitted resync data rate, both */
1124 	int rs_last_sect_ev; /* counter to compare with */
1125 	int rs_last_events;  /* counter of read or write "events" (unit sectors)
1126 			      * on the lower level device when we last looked. */
1127 	int c_sync_rate; /* current resync rate after syncer throttle magic */
1128 	struct fifo_buffer rs_plan_s; /* correction values of resync planer */
1129 	int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
1130 	int rs_planed;    /* resync sectors already planned */
1131 	atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
1132 	int peer_max_bio_size;
1133 	int local_max_bio_size;
1134 };
1135 
1136 static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
1137 {
1138 	struct drbd_conf *mdev;
1139 
1140 	mdev = minor < minor_count ? minor_table[minor] : NULL;
1141 
1142 	return mdev;
1143 }
1144 
1145 static inline unsigned int mdev_to_minor(struct drbd_conf *mdev)
1146 {
1147 	return mdev->minor;
1148 }
1149 
1150 /* returns 1 if it was successful,
1151  * returns 0 if there was no data socket.
1152  * so wherever you are going to use the data.socket, e.g. do
1153  * if (!drbd_get_data_sock(mdev))
1154  *	return 0;
1155  *	CODE();
1156  * drbd_put_data_sock(mdev);
1157  */
1158 static inline int drbd_get_data_sock(struct drbd_conf *mdev)
1159 {
1160 	mutex_lock(&mdev->data.mutex);
1161 	/* drbd_disconnect() could have called drbd_free_sock()
1162 	 * while we were waiting in down()... */
1163 	if (unlikely(mdev->data.socket == NULL)) {
1164 		mutex_unlock(&mdev->data.mutex);
1165 		return 0;
1166 	}
1167 	return 1;
1168 }
1169 
1170 static inline void drbd_put_data_sock(struct drbd_conf *mdev)
1171 {
1172 	mutex_unlock(&mdev->data.mutex);
1173 }
1174 
1175 /*
1176  * function declarations
1177  *************************/
1178 
1179 /* drbd_main.c */
1180 
1181 enum chg_state_flags {
1182 	CS_HARD	= 1,
1183 	CS_VERBOSE = 2,
1184 	CS_WAIT_COMPLETE = 4,
1185 	CS_SERIALIZE    = 8,
1186 	CS_ORDERED      = CS_WAIT_COMPLETE + CS_SERIALIZE,
1187 };
1188 
1189 enum dds_flags {
1190 	DDSF_FORCED    = 1,
1191 	DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1192 };
1193 
1194 extern void drbd_init_set_defaults(struct drbd_conf *mdev);
1195 extern enum drbd_state_rv drbd_change_state(struct drbd_conf *mdev,
1196 					    enum chg_state_flags f,
1197 					    union drbd_state mask,
1198 					    union drbd_state val);
1199 extern void drbd_force_state(struct drbd_conf *, union drbd_state,
1200 			union drbd_state);
1201 extern enum drbd_state_rv _drbd_request_state(struct drbd_conf *,
1202 					      union drbd_state,
1203 					      union drbd_state,
1204 					      enum chg_state_flags);
1205 extern enum drbd_state_rv __drbd_set_state(struct drbd_conf *, union drbd_state,
1206 					   enum chg_state_flags,
1207 					   struct completion *done);
1208 extern void print_st_err(struct drbd_conf *, union drbd_state,
1209 			union drbd_state, int);
1210 extern int  drbd_thread_start(struct drbd_thread *thi);
1211 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1212 #ifdef CONFIG_SMP
1213 extern void drbd_thread_current_set_cpu(struct drbd_conf *mdev);
1214 extern void drbd_calc_cpu_mask(struct drbd_conf *mdev);
1215 #else
1216 #define drbd_thread_current_set_cpu(A) ({})
1217 #define drbd_calc_cpu_mask(A) ({})
1218 #endif
1219 extern void drbd_free_resources(struct drbd_conf *mdev);
1220 extern void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr,
1221 		       unsigned int set_size);
1222 extern void tl_clear(struct drbd_conf *mdev);
1223 extern void _tl_add_barrier(struct drbd_conf *, struct drbd_tl_epoch *);
1224 extern void drbd_free_sock(struct drbd_conf *mdev);
1225 extern int drbd_send(struct drbd_conf *mdev, struct socket *sock,
1226 			void *buf, size_t size, unsigned msg_flags);
1227 extern int drbd_send_protocol(struct drbd_conf *mdev);
1228 extern int drbd_send_uuids(struct drbd_conf *mdev);
1229 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
1230 extern int drbd_gen_and_send_sync_uuid(struct drbd_conf *mdev);
1231 extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags);
1232 extern int _drbd_send_state(struct drbd_conf *mdev);
1233 extern int drbd_send_state(struct drbd_conf *mdev);
1234 extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock,
1235 			enum drbd_packets cmd, struct p_header80 *h,
1236 			size_t size, unsigned msg_flags);
1237 #define USE_DATA_SOCKET 1
1238 #define USE_META_SOCKET 0
1239 extern int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket,
1240 			enum drbd_packets cmd, struct p_header80 *h,
1241 			size_t size);
1242 extern int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd,
1243 			char *data, size_t size);
1244 extern int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc);
1245 extern int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr,
1246 			u32 set_size);
1247 extern int drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd,
1248 			struct drbd_epoch_entry *e);
1249 extern int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packets cmd,
1250 			struct p_block_req *rp);
1251 extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd,
1252 			struct p_data *dp, int data_size);
1253 extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd,
1254 			    sector_t sector, int blksize, u64 block_id);
1255 extern int drbd_send_oos(struct drbd_conf *mdev, struct drbd_request *req);
1256 extern int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd,
1257 			   struct drbd_epoch_entry *e);
1258 extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
1259 extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1260 			      sector_t sector, int size, u64 block_id);
1261 extern int drbd_send_drequest_csum(struct drbd_conf *mdev,
1262 				   sector_t sector,int size,
1263 				   void *digest, int digest_size,
1264 				   enum drbd_packets cmd);
1265 extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size);
1266 
1267 extern int drbd_send_bitmap(struct drbd_conf *mdev);
1268 extern int _drbd_send_bitmap(struct drbd_conf *mdev);
1269 extern int drbd_send_sr_reply(struct drbd_conf *mdev, enum drbd_state_rv retcode);
1270 extern void drbd_free_bc(struct drbd_backing_dev *ldev);
1271 extern void drbd_mdev_cleanup(struct drbd_conf *mdev);
1272 void drbd_print_uuids(struct drbd_conf *mdev, const char *text);
1273 
1274 extern void drbd_md_sync(struct drbd_conf *mdev);
1275 extern int  drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
1276 extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1277 extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1278 extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1279 extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1280 extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
1281 extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
1282 extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
1283 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1284 #ifndef DRBD_DEBUG_MD_SYNC
1285 extern void drbd_md_mark_dirty(struct drbd_conf *mdev);
1286 #else
1287 #define drbd_md_mark_dirty(m)	drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1288 extern void drbd_md_mark_dirty_(struct drbd_conf *mdev,
1289 		unsigned int line, const char *func);
1290 #endif
1291 extern void drbd_queue_bitmap_io(struct drbd_conf *mdev,
1292 				 int (*io_fn)(struct drbd_conf *),
1293 				 void (*done)(struct drbd_conf *, int),
1294 				 char *why, enum bm_flag flags);
1295 extern int drbd_bitmap_io(struct drbd_conf *mdev,
1296 		int (*io_fn)(struct drbd_conf *),
1297 		char *why, enum bm_flag flags);
1298 extern int drbd_bmio_set_n_write(struct drbd_conf *mdev);
1299 extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev);
1300 extern void drbd_go_diskless(struct drbd_conf *mdev);
1301 extern void drbd_ldev_destroy(struct drbd_conf *mdev);
1302 
1303 
1304 /* Meta data layout
1305    We reserve a 128MB Block (4k aligned)
1306    * either at the end of the backing device
1307    * or on a separate meta data device. */
1308 
1309 #define MD_RESERVED_SECT (128LU << 11)  /* 128 MB, unit sectors */
1310 /* The following numbers are sectors */
1311 #define MD_AL_OFFSET 8	    /* 8 Sectors after start of meta area */
1312 #define MD_AL_MAX_SIZE 64   /* = 32 kb LOG  ~ 3776 extents ~ 14 GB Storage */
1313 /* Allows up to about 3.8TB */
1314 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1315 
1316 /* Since the smalles IO unit is usually 512 byte */
1317 #define MD_SECTOR_SHIFT	 9
1318 #define MD_SECTOR_SIZE	 (1<<MD_SECTOR_SHIFT)
1319 
1320 /* activity log */
1321 #define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1322 #define AL_EXTENT_SHIFT 22		 /* One extent represents 4M Storage */
1323 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1324 
1325 #if BITS_PER_LONG == 32
1326 #define LN2_BPL 5
1327 #define cpu_to_lel(A) cpu_to_le32(A)
1328 #define lel_to_cpu(A) le32_to_cpu(A)
1329 #elif BITS_PER_LONG == 64
1330 #define LN2_BPL 6
1331 #define cpu_to_lel(A) cpu_to_le64(A)
1332 #define lel_to_cpu(A) le64_to_cpu(A)
1333 #else
1334 #error "LN2 of BITS_PER_LONG unknown!"
1335 #endif
1336 
1337 /* resync bitmap */
1338 /* 16MB sized 'bitmap extent' to track syncer usage */
1339 struct bm_extent {
1340 	int rs_left; /* number of bits set (out of sync) in this extent. */
1341 	int rs_failed; /* number of failed resync requests in this extent. */
1342 	unsigned long flags;
1343 	struct lc_element lce;
1344 };
1345 
1346 #define BME_NO_WRITES  0  /* bm_extent.flags: no more requests on this one! */
1347 #define BME_LOCKED     1  /* bm_extent.flags: syncer active on this one. */
1348 #define BME_PRIORITY   2  /* finish resync IO on this extent ASAP! App IO waiting! */
1349 
1350 /* drbd_bitmap.c */
1351 /*
1352  * We need to store one bit for a block.
1353  * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1354  * Bit 0 ==> local node thinks this block is binary identical on both nodes
1355  * Bit 1 ==> local node thinks this block needs to be synced.
1356  */
1357 
1358 #define SLEEP_TIME (HZ/10)
1359 
1360 #define BM_BLOCK_SHIFT  12			 /* 4k per bit */
1361 #define BM_BLOCK_SIZE	 (1<<BM_BLOCK_SHIFT)
1362 /* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1363  * per sector of on disk bitmap */
1364 #define BM_EXT_SHIFT	 (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3)  /* = 24 */
1365 #define BM_EXT_SIZE	 (1<<BM_EXT_SHIFT)
1366 
1367 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1368 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1369 #endif
1370 
1371 /* thus many _storage_ sectors are described by one bit */
1372 #define BM_SECT_TO_BIT(x)   ((x)>>(BM_BLOCK_SHIFT-9))
1373 #define BM_BIT_TO_SECT(x)   ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1374 #define BM_SECT_PER_BIT     BM_BIT_TO_SECT(1)
1375 
1376 /* bit to represented kilo byte conversion */
1377 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1378 
1379 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1380  * _storage_ sector is located in */
1381 #define BM_SECT_TO_EXT(x)   ((x)>>(BM_EXT_SHIFT-9))
1382 
1383 /* how much _storage_ sectors we have per bitmap sector */
1384 #define BM_EXT_TO_SECT(x)   ((sector_t)(x) << (BM_EXT_SHIFT-9))
1385 #define BM_SECT_PER_EXT     BM_EXT_TO_SECT(1)
1386 
1387 /* in one sector of the bitmap, we have this many activity_log extents. */
1388 #define AL_EXT_PER_BM_SECT  (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1389 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1390 
1391 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1392 #define BM_BLOCKS_PER_BM_EXT_MASK  ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1393 
1394 /* the extent in "PER_EXTENT" below is an activity log extent
1395  * we need that many (long words/bytes) to store the bitmap
1396  *		     of one AL_EXTENT_SIZE chunk of storage.
1397  * we can store the bitmap for that many AL_EXTENTS within
1398  * one sector of the _on_disk_ bitmap:
1399  * bit	 0	  bit 37   bit 38	     bit (512*8)-1
1400  *	     ...|........|........|.. // ..|........|
1401  * sect. 0	 `296	  `304			   ^(512*8*8)-1
1402  *
1403 #define BM_WORDS_PER_EXT    ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1404 #define BM_BYTES_PER_EXT    ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 )  // 128
1405 #define BM_EXT_PER_SECT	    ( 512 / BM_BYTES_PER_EXTENT )	 //   4
1406  */
1407 
1408 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1409 #define DRBD_MAX_SECTORS_BM \
1410 	  ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1411 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1412 #define DRBD_MAX_SECTORS      DRBD_MAX_SECTORS_BM
1413 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1414 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1415 #define DRBD_MAX_SECTORS      DRBD_MAX_SECTORS_32
1416 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1417 #else
1418 #define DRBD_MAX_SECTORS      DRBD_MAX_SECTORS_BM
1419 /* 16 TB in units of sectors */
1420 #if BITS_PER_LONG == 32
1421 /* adjust by one page worth of bitmap,
1422  * so we won't wrap around in drbd_bm_find_next_bit.
1423  * you should use 64bit OS for that much storage, anyways. */
1424 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1425 #else
1426 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1427 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1428 /* corresponds to (1UL << 38) bits right now. */
1429 #endif
1430 #endif
1431 
1432 /* Sector shift value for the "hash" functions of tl_hash and ee_hash tables.
1433  * With a value of 8 all IO in one 128K block make it to the same slot of the
1434  * hash table. */
1435 #define HT_SHIFT 8
1436 #define DRBD_MAX_BIO_SIZE (1U<<(9+HT_SHIFT))
1437 #define DRBD_MAX_BIO_SIZE_SAFE (1 << 12)       /* Works always = 4k */
1438 
1439 #define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */
1440 
1441 /* Number of elements in the app_reads_hash */
1442 #define APP_R_HSIZE 15
1443 
1444 extern int  drbd_bm_init(struct drbd_conf *mdev);
1445 extern int  drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits);
1446 extern void drbd_bm_cleanup(struct drbd_conf *mdev);
1447 extern void drbd_bm_set_all(struct drbd_conf *mdev);
1448 extern void drbd_bm_clear_all(struct drbd_conf *mdev);
1449 /* set/clear/test only a few bits at a time */
1450 extern int  drbd_bm_set_bits(
1451 		struct drbd_conf *mdev, unsigned long s, unsigned long e);
1452 extern int  drbd_bm_clear_bits(
1453 		struct drbd_conf *mdev, unsigned long s, unsigned long e);
1454 extern int drbd_bm_count_bits(
1455 	struct drbd_conf *mdev, const unsigned long s, const unsigned long e);
1456 /* bm_set_bits variant for use while holding drbd_bm_lock,
1457  * may process the whole bitmap in one go */
1458 extern void _drbd_bm_set_bits(struct drbd_conf *mdev,
1459 		const unsigned long s, const unsigned long e);
1460 extern int  drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr);
1461 extern int  drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr);
1462 extern int  drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local);
1463 extern int  drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
1464 extern int  drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
1465 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev,
1466 		unsigned long al_enr);
1467 extern size_t	     drbd_bm_words(struct drbd_conf *mdev);
1468 extern unsigned long drbd_bm_bits(struct drbd_conf *mdev);
1469 extern sector_t      drbd_bm_capacity(struct drbd_conf *mdev);
1470 
1471 #define DRBD_END_OF_BITMAP	(~(unsigned long)0)
1472 extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1473 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1474 extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1475 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo);
1476 extern unsigned long _drbd_bm_total_weight(struct drbd_conf *mdev);
1477 extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev);
1478 extern int drbd_bm_rs_done(struct drbd_conf *mdev);
1479 /* for receive_bitmap */
1480 extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset,
1481 		size_t number, unsigned long *buffer);
1482 /* for _drbd_send_bitmap */
1483 extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset,
1484 		size_t number, unsigned long *buffer);
1485 
1486 extern void drbd_bm_lock(struct drbd_conf *mdev, char *why, enum bm_flag flags);
1487 extern void drbd_bm_unlock(struct drbd_conf *mdev);
1488 /* drbd_main.c */
1489 
1490 extern struct kmem_cache *drbd_request_cache;
1491 extern struct kmem_cache *drbd_ee_cache;	/* epoch entries */
1492 extern struct kmem_cache *drbd_bm_ext_cache;	/* bitmap extents */
1493 extern struct kmem_cache *drbd_al_ext_cache;	/* activity log extents */
1494 extern mempool_t *drbd_request_mempool;
1495 extern mempool_t *drbd_ee_mempool;
1496 
1497 extern struct page *drbd_pp_pool; /* drbd's page pool */
1498 extern spinlock_t   drbd_pp_lock;
1499 extern int	    drbd_pp_vacant;
1500 extern wait_queue_head_t drbd_pp_wait;
1501 
1502 extern rwlock_t global_state_lock;
1503 
1504 extern struct drbd_conf *drbd_new_device(unsigned int minor);
1505 extern void drbd_free_mdev(struct drbd_conf *mdev);
1506 
1507 extern int proc_details;
1508 
1509 /* drbd_req */
1510 extern int drbd_make_request(struct request_queue *q, struct bio *bio);
1511 extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req);
1512 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1513 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1514 
1515 
1516 /* drbd_nl.c */
1517 extern void drbd_suspend_io(struct drbd_conf *mdev);
1518 extern void drbd_resume_io(struct drbd_conf *mdev);
1519 extern char *ppsize(char *buf, unsigned long long size);
1520 extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int);
1521 enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
1522 extern enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local);
1523 extern void resync_after_online_grow(struct drbd_conf *);
1524 extern void drbd_reconsider_max_bio_size(struct drbd_conf *mdev);
1525 extern enum drbd_state_rv drbd_set_role(struct drbd_conf *mdev,
1526 					enum drbd_role new_role,
1527 					int force);
1528 extern enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev);
1529 extern void drbd_try_outdate_peer_async(struct drbd_conf *mdev);
1530 extern int drbd_khelper(struct drbd_conf *mdev, char *cmd);
1531 
1532 /* drbd_worker.c */
1533 extern int drbd_worker(struct drbd_thread *thi);
1534 extern int drbd_alter_sa(struct drbd_conf *mdev, int na);
1535 extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side);
1536 extern void resume_next_sg(struct drbd_conf *mdev);
1537 extern void suspend_other_sg(struct drbd_conf *mdev);
1538 extern int drbd_resync_finished(struct drbd_conf *mdev);
1539 /* maybe rather drbd_main.c ? */
1540 extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
1541 		struct drbd_backing_dev *bdev, sector_t sector, int rw);
1542 extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int);
1543 extern void drbd_rs_controller_reset(struct drbd_conf *mdev);
1544 
1545 static inline void ov_oos_print(struct drbd_conf *mdev)
1546 {
1547 	if (mdev->ov_last_oos_size) {
1548 		dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n",
1549 		     (unsigned long long)mdev->ov_last_oos_start,
1550 		     (unsigned long)mdev->ov_last_oos_size);
1551 	}
1552 	mdev->ov_last_oos_size=0;
1553 }
1554 
1555 
1556 extern void drbd_csum_bio(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
1557 extern void drbd_csum_ee(struct drbd_conf *, struct crypto_hash *, struct drbd_epoch_entry *, void *);
1558 /* worker callbacks */
1559 extern int w_req_cancel_conflict(struct drbd_conf *, struct drbd_work *, int);
1560 extern int w_read_retry_remote(struct drbd_conf *, struct drbd_work *, int);
1561 extern int w_e_end_data_req(struct drbd_conf *, struct drbd_work *, int);
1562 extern int w_e_end_rsdata_req(struct drbd_conf *, struct drbd_work *, int);
1563 extern int w_e_end_csum_rs_req(struct drbd_conf *, struct drbd_work *, int);
1564 extern int w_e_end_ov_reply(struct drbd_conf *, struct drbd_work *, int);
1565 extern int w_e_end_ov_req(struct drbd_conf *, struct drbd_work *, int);
1566 extern int w_ov_finished(struct drbd_conf *, struct drbd_work *, int);
1567 extern int w_resync_timer(struct drbd_conf *, struct drbd_work *, int);
1568 extern int w_resume_next_sg(struct drbd_conf *, struct drbd_work *, int);
1569 extern int w_send_write_hint(struct drbd_conf *, struct drbd_work *, int);
1570 extern int w_send_dblock(struct drbd_conf *, struct drbd_work *, int);
1571 extern int w_send_barrier(struct drbd_conf *, struct drbd_work *, int);
1572 extern int w_send_read_req(struct drbd_conf *, struct drbd_work *, int);
1573 extern int w_prev_work_done(struct drbd_conf *, struct drbd_work *, int);
1574 extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int);
1575 extern int w_restart_disk_io(struct drbd_conf *, struct drbd_work *, int);
1576 extern int w_send_oos(struct drbd_conf *, struct drbd_work *, int);
1577 extern int w_start_resync(struct drbd_conf *, struct drbd_work *, int);
1578 
1579 extern void resync_timer_fn(unsigned long data);
1580 extern void start_resync_timer_fn(unsigned long data);
1581 
1582 /* drbd_receiver.c */
1583 extern int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector);
1584 extern int drbd_submit_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e,
1585 		const unsigned rw, const int fault_type);
1586 extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list);
1587 extern struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev,
1588 					    u64 id,
1589 					    sector_t sector,
1590 					    unsigned int data_size,
1591 					    gfp_t gfp_mask) __must_hold(local);
1592 extern void drbd_free_some_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e,
1593 		int is_net);
1594 #define drbd_free_ee(m,e)	drbd_free_some_ee(m, e, 0)
1595 #define drbd_free_net_ee(m,e)	drbd_free_some_ee(m, e, 1)
1596 extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1597 		struct list_head *head);
1598 extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1599 		struct list_head *head);
1600 extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled);
1601 extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed);
1602 extern void drbd_flush_workqueue(struct drbd_conf *mdev);
1603 extern void drbd_free_tl_hash(struct drbd_conf *mdev);
1604 
1605 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1606  * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1607 static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1608 			char __user *optval, int optlen)
1609 {
1610 	int err;
1611 	if (level == SOL_SOCKET)
1612 		err = sock_setsockopt(sock, level, optname, optval, optlen);
1613 	else
1614 		err = sock->ops->setsockopt(sock, level, optname, optval,
1615 					    optlen);
1616 	return err;
1617 }
1618 
1619 static inline void drbd_tcp_cork(struct socket *sock)
1620 {
1621 	int __user val = 1;
1622 	(void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1623 			(char __user *)&val, sizeof(val));
1624 }
1625 
1626 static inline void drbd_tcp_uncork(struct socket *sock)
1627 {
1628 	int __user val = 0;
1629 	(void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1630 			(char __user *)&val, sizeof(val));
1631 }
1632 
1633 static inline void drbd_tcp_nodelay(struct socket *sock)
1634 {
1635 	int __user val = 1;
1636 	(void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1637 			(char __user *)&val, sizeof(val));
1638 }
1639 
1640 static inline void drbd_tcp_quickack(struct socket *sock)
1641 {
1642 	int __user val = 2;
1643 	(void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1644 			(char __user *)&val, sizeof(val));
1645 }
1646 
1647 void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo);
1648 
1649 /* drbd_proc.c */
1650 extern struct proc_dir_entry *drbd_proc;
1651 extern const struct file_operations drbd_proc_fops;
1652 extern const char *drbd_conn_str(enum drbd_conns s);
1653 extern const char *drbd_role_str(enum drbd_role s);
1654 
1655 /* drbd_actlog.c */
1656 extern void drbd_al_begin_io(struct drbd_conf *mdev, sector_t sector);
1657 extern void drbd_al_complete_io(struct drbd_conf *mdev, sector_t sector);
1658 extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector);
1659 extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1660 extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1661 extern void drbd_rs_cancel_all(struct drbd_conf *mdev);
1662 extern int drbd_rs_del_all(struct drbd_conf *mdev);
1663 extern void drbd_rs_failed_io(struct drbd_conf *mdev,
1664 		sector_t sector, int size);
1665 extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *);
1666 extern void drbd_advance_rs_marks(struct drbd_conf *mdev, unsigned long still_to_go);
1667 extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
1668 		int size, const char *file, const unsigned int line);
1669 #define drbd_set_in_sync(mdev, sector, size) \
1670 	__drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1671 extern int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
1672 		int size, const char *file, const unsigned int line);
1673 #define drbd_set_out_of_sync(mdev, sector, size) \
1674 	__drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1675 extern void drbd_al_apply_to_bm(struct drbd_conf *mdev);
1676 extern void drbd_al_shrink(struct drbd_conf *mdev);
1677 
1678 
1679 /* drbd_nl.c */
1680 
1681 void drbd_nl_cleanup(void);
1682 int __init drbd_nl_init(void);
1683 void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state);
1684 void drbd_bcast_sync_progress(struct drbd_conf *mdev);
1685 void drbd_bcast_ee(struct drbd_conf *mdev,
1686 		const char *reason, const int dgs,
1687 		const char* seen_hash, const char* calc_hash,
1688 		const struct drbd_epoch_entry* e);
1689 
1690 
1691 /**
1692  * DOC: DRBD State macros
1693  *
1694  * These macros are used to express state changes in easily readable form.
1695  *
1696  * The NS macros expand to a mask and a value, that can be bit ored onto the
1697  * current state as soon as the spinlock (req_lock) was taken.
1698  *
1699  * The _NS macros are used for state functions that get called with the
1700  * spinlock. These macros expand directly to the new state value.
1701  *
1702  * Besides the basic forms NS() and _NS() additional _?NS[23] are defined
1703  * to express state changes that affect more than one aspect of the state.
1704  *
1705  * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY)
1706  * Means that the network connection was established and that the peer
1707  * is in secondary role.
1708  */
1709 #define role_MASK R_MASK
1710 #define peer_MASK R_MASK
1711 #define disk_MASK D_MASK
1712 #define pdsk_MASK D_MASK
1713 #define conn_MASK C_MASK
1714 #define susp_MASK 1
1715 #define user_isp_MASK 1
1716 #define aftr_isp_MASK 1
1717 #define susp_nod_MASK 1
1718 #define susp_fen_MASK 1
1719 
1720 #define NS(T, S) \
1721 	({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \
1722 	({ union drbd_state val; val.i = 0; val.T = (S); val; })
1723 #define NS2(T1, S1, T2, S2) \
1724 	({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1725 	  mask.T2 = T2##_MASK; mask; }), \
1726 	({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1727 	  val.T2 = (S2); val; })
1728 #define NS3(T1, S1, T2, S2, T3, S3) \
1729 	({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1730 	  mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \
1731 	({ union drbd_state val;  val.i = 0; val.T1 = (S1); \
1732 	  val.T2 = (S2); val.T3 = (S3); val; })
1733 
1734 #define _NS(D, T, S) \
1735 	D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; })
1736 #define _NS2(D, T1, S1, T2, S2) \
1737 	D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1738 	__ns.T2 = (S2); __ns; })
1739 #define _NS3(D, T1, S1, T2, S2, T3, S3) \
1740 	D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1741 	__ns.T2 = (S2); __ns.T3 = (S3); __ns; })
1742 
1743 /*
1744  * inline helper functions
1745  *************************/
1746 
1747 /* see also page_chain_add and friends in drbd_receiver.c */
1748 static inline struct page *page_chain_next(struct page *page)
1749 {
1750 	return (struct page *)page_private(page);
1751 }
1752 #define page_chain_for_each(page) \
1753 	for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1754 			page = page_chain_next(page))
1755 #define page_chain_for_each_safe(page, n) \
1756 	for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1757 
1758 static inline int drbd_bio_has_active_page(struct bio *bio)
1759 {
1760 	struct bio_vec *bvec;
1761 	int i;
1762 
1763 	__bio_for_each_segment(bvec, bio, i, 0) {
1764 		if (page_count(bvec->bv_page) > 1)
1765 			return 1;
1766 	}
1767 
1768 	return 0;
1769 }
1770 
1771 static inline int drbd_ee_has_active_page(struct drbd_epoch_entry *e)
1772 {
1773 	struct page *page = e->pages;
1774 	page_chain_for_each(page) {
1775 		if (page_count(page) > 1)
1776 			return 1;
1777 	}
1778 	return 0;
1779 }
1780 
1781 
1782 static inline void drbd_state_lock(struct drbd_conf *mdev)
1783 {
1784 	wait_event(mdev->misc_wait,
1785 		   !test_and_set_bit(CLUSTER_ST_CHANGE, &mdev->flags));
1786 }
1787 
1788 static inline void drbd_state_unlock(struct drbd_conf *mdev)
1789 {
1790 	clear_bit(CLUSTER_ST_CHANGE, &mdev->flags);
1791 	wake_up(&mdev->misc_wait);
1792 }
1793 
1794 static inline enum drbd_state_rv
1795 _drbd_set_state(struct drbd_conf *mdev, union drbd_state ns,
1796 		enum chg_state_flags flags, struct completion *done)
1797 {
1798 	enum drbd_state_rv rv;
1799 
1800 	read_lock(&global_state_lock);
1801 	rv = __drbd_set_state(mdev, ns, flags, done);
1802 	read_unlock(&global_state_lock);
1803 
1804 	return rv;
1805 }
1806 
1807 /**
1808  * drbd_request_state() - Reqest a state change
1809  * @mdev:	DRBD device.
1810  * @mask:	mask of state bits to change.
1811  * @val:	value of new state bits.
1812  *
1813  * This is the most graceful way of requesting a state change. It is verbose
1814  * quite verbose in case the state change is not possible, and all those
1815  * state changes are globally serialized.
1816  */
1817 static inline int drbd_request_state(struct drbd_conf *mdev,
1818 				     union drbd_state mask,
1819 				     union drbd_state val)
1820 {
1821 	return _drbd_request_state(mdev, mask, val, CS_VERBOSE + CS_ORDERED);
1822 }
1823 
1824 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1825 static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach, const char *where)
1826 {
1827 	switch (mdev->ldev->dc.on_io_error) {
1828 	case EP_PASS_ON:
1829 		if (!forcedetach) {
1830 			if (__ratelimit(&drbd_ratelimit_state))
1831 				dev_err(DEV, "Local IO failed in %s.\n", where);
1832 			if (mdev->state.disk > D_INCONSISTENT)
1833 				_drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_HARD, NULL);
1834 			break;
1835 		}
1836 		/* NOTE fall through to detach case if forcedetach set */
1837 	case EP_DETACH:
1838 	case EP_CALL_HELPER:
1839 		set_bit(WAS_IO_ERROR, &mdev->flags);
1840 		if (mdev->state.disk > D_FAILED) {
1841 			_drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL);
1842 			dev_err(DEV,
1843 				"Local IO failed in %s. Detaching...\n", where);
1844 		}
1845 		break;
1846 	}
1847 }
1848 
1849 /**
1850  * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1851  * @mdev:	 DRBD device.
1852  * @error:	 Error code passed to the IO completion callback
1853  * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1854  *
1855  * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1856  */
1857 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1858 static inline void drbd_chk_io_error_(struct drbd_conf *mdev,
1859 	int error, int forcedetach, const char *where)
1860 {
1861 	if (error) {
1862 		unsigned long flags;
1863 		spin_lock_irqsave(&mdev->req_lock, flags);
1864 		__drbd_chk_io_error_(mdev, forcedetach, where);
1865 		spin_unlock_irqrestore(&mdev->req_lock, flags);
1866 	}
1867 }
1868 
1869 
1870 /**
1871  * drbd_md_first_sector() - Returns the first sector number of the meta data area
1872  * @bdev:	Meta data block device.
1873  *
1874  * BTW, for internal meta data, this happens to be the maximum capacity
1875  * we could agree upon with our peer node.
1876  */
1877 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1878 {
1879 	switch (bdev->dc.meta_dev_idx) {
1880 	case DRBD_MD_INDEX_INTERNAL:
1881 	case DRBD_MD_INDEX_FLEX_INT:
1882 		return bdev->md.md_offset + bdev->md.bm_offset;
1883 	case DRBD_MD_INDEX_FLEX_EXT:
1884 	default:
1885 		return bdev->md.md_offset;
1886 	}
1887 }
1888 
1889 /**
1890  * drbd_md_last_sector() - Return the last sector number of the meta data area
1891  * @bdev:	Meta data block device.
1892  */
1893 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1894 {
1895 	switch (bdev->dc.meta_dev_idx) {
1896 	case DRBD_MD_INDEX_INTERNAL:
1897 	case DRBD_MD_INDEX_FLEX_INT:
1898 		return bdev->md.md_offset + MD_AL_OFFSET - 1;
1899 	case DRBD_MD_INDEX_FLEX_EXT:
1900 	default:
1901 		return bdev->md.md_offset + bdev->md.md_size_sect;
1902 	}
1903 }
1904 
1905 /* Returns the number of 512 byte sectors of the device */
1906 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1907 {
1908 	/* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1909 	return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0;
1910 }
1911 
1912 /**
1913  * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1914  * @bdev:	Meta data block device.
1915  *
1916  * returns the capacity we announce to out peer.  we clip ourselves at the
1917  * various MAX_SECTORS, because if we don't, current implementation will
1918  * oops sooner or later
1919  */
1920 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1921 {
1922 	sector_t s;
1923 	switch (bdev->dc.meta_dev_idx) {
1924 	case DRBD_MD_INDEX_INTERNAL:
1925 	case DRBD_MD_INDEX_FLEX_INT:
1926 		s = drbd_get_capacity(bdev->backing_bdev)
1927 			? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1928 					drbd_md_first_sector(bdev))
1929 			: 0;
1930 		break;
1931 	case DRBD_MD_INDEX_FLEX_EXT:
1932 		s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1933 				drbd_get_capacity(bdev->backing_bdev));
1934 		/* clip at maximum size the meta device can support */
1935 		s = min_t(sector_t, s,
1936 			BM_EXT_TO_SECT(bdev->md.md_size_sect
1937 				     - bdev->md.bm_offset));
1938 		break;
1939 	default:
1940 		s = min_t(sector_t, DRBD_MAX_SECTORS,
1941 				drbd_get_capacity(bdev->backing_bdev));
1942 	}
1943 	return s;
1944 }
1945 
1946 /**
1947  * drbd_md_ss__() - Return the sector number of our meta data super block
1948  * @mdev:	DRBD device.
1949  * @bdev:	Meta data block device.
1950  */
1951 static inline sector_t drbd_md_ss__(struct drbd_conf *mdev,
1952 				    struct drbd_backing_dev *bdev)
1953 {
1954 	switch (bdev->dc.meta_dev_idx) {
1955 	default: /* external, some index */
1956 		return MD_RESERVED_SECT * bdev->dc.meta_dev_idx;
1957 	case DRBD_MD_INDEX_INTERNAL:
1958 		/* with drbd08, internal meta data is always "flexible" */
1959 	case DRBD_MD_INDEX_FLEX_INT:
1960 		/* sizeof(struct md_on_disk_07) == 4k
1961 		 * position: last 4k aligned block of 4k size */
1962 		if (!bdev->backing_bdev) {
1963 			if (__ratelimit(&drbd_ratelimit_state)) {
1964 				dev_err(DEV, "bdev->backing_bdev==NULL\n");
1965 				dump_stack();
1966 			}
1967 			return 0;
1968 		}
1969 		return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL)
1970 			- MD_AL_OFFSET;
1971 	case DRBD_MD_INDEX_FLEX_EXT:
1972 		return 0;
1973 	}
1974 }
1975 
1976 static inline void
1977 drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w)
1978 {
1979 	unsigned long flags;
1980 	spin_lock_irqsave(&q->q_lock, flags);
1981 	list_add(&w->list, &q->q);
1982 	up(&q->s); /* within the spinlock,
1983 		      see comment near end of drbd_worker() */
1984 	spin_unlock_irqrestore(&q->q_lock, flags);
1985 }
1986 
1987 static inline void
1988 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1989 {
1990 	unsigned long flags;
1991 	spin_lock_irqsave(&q->q_lock, flags);
1992 	list_add_tail(&w->list, &q->q);
1993 	up(&q->s); /* within the spinlock,
1994 		      see comment near end of drbd_worker() */
1995 	spin_unlock_irqrestore(&q->q_lock, flags);
1996 }
1997 
1998 static inline void wake_asender(struct drbd_conf *mdev)
1999 {
2000 	if (test_bit(SIGNAL_ASENDER, &mdev->flags))
2001 		force_sig(DRBD_SIG, mdev->asender.task);
2002 }
2003 
2004 static inline void request_ping(struct drbd_conf *mdev)
2005 {
2006 	set_bit(SEND_PING, &mdev->flags);
2007 	wake_asender(mdev);
2008 }
2009 
2010 static inline int drbd_send_short_cmd(struct drbd_conf *mdev,
2011 	enum drbd_packets cmd)
2012 {
2013 	struct p_header80 h;
2014 	return drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, &h, sizeof(h));
2015 }
2016 
2017 static inline int drbd_send_ping(struct drbd_conf *mdev)
2018 {
2019 	struct p_header80 h;
2020 	return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING, &h, sizeof(h));
2021 }
2022 
2023 static inline int drbd_send_ping_ack(struct drbd_conf *mdev)
2024 {
2025 	struct p_header80 h;
2026 	return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING_ACK, &h, sizeof(h));
2027 }
2028 
2029 static inline void drbd_thread_stop(struct drbd_thread *thi)
2030 {
2031 	_drbd_thread_stop(thi, false, true);
2032 }
2033 
2034 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
2035 {
2036 	_drbd_thread_stop(thi, false, false);
2037 }
2038 
2039 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
2040 {
2041 	_drbd_thread_stop(thi, true, false);
2042 }
2043 
2044 /* counts how many answer packets packets we expect from our peer,
2045  * for either explicit application requests,
2046  * or implicit barrier packets as necessary.
2047  * increased:
2048  *  w_send_barrier
2049  *  _req_mod(req, queue_for_net_write or queue_for_net_read);
2050  *    it is much easier and equally valid to count what we queue for the
2051  *    worker, even before it actually was queued or send.
2052  *    (drbd_make_request_common; recovery path on read io-error)
2053  * decreased:
2054  *  got_BarrierAck (respective tl_clear, tl_clear_barrier)
2055  *  _req_mod(req, data_received)
2056  *     [from receive_DataReply]
2057  *  _req_mod(req, write_acked_by_peer or recv_acked_by_peer or neg_acked)
2058  *     [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
2059  *     for some reason it is NOT decreased in got_NegAck,
2060  *     but in the resulting cleanup code from report_params.
2061  *     we should try to remember the reason for that...
2062  *  _req_mod(req, send_failed or send_canceled)
2063  *  _req_mod(req, connection_lost_while_pending)
2064  *     [from tl_clear_barrier]
2065  */
2066 static inline void inc_ap_pending(struct drbd_conf *mdev)
2067 {
2068 	atomic_inc(&mdev->ap_pending_cnt);
2069 }
2070 
2071 #define ERR_IF_CNT_IS_NEGATIVE(which)				\
2072 	if (atomic_read(&mdev->which) < 0)			\
2073 		dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n",	\
2074 		    __func__ , __LINE__ ,			\
2075 		    atomic_read(&mdev->which))
2076 
2077 #define dec_ap_pending(mdev)	do {				\
2078 	typecheck(struct drbd_conf *, mdev);			\
2079 	if (atomic_dec_and_test(&mdev->ap_pending_cnt))		\
2080 		wake_up(&mdev->misc_wait);			\
2081 	ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
2082 
2083 /* counts how many resync-related answers we still expect from the peer
2084  *		     increase			decrease
2085  * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
2086  * C_SYNC_SOURCE sends P_RS_DATA_REPLY   (and expects P_WRITE_ACK with ID_SYNCER)
2087  *					   (or P_NEG_ACK with ID_SYNCER)
2088  */
2089 static inline void inc_rs_pending(struct drbd_conf *mdev)
2090 {
2091 	atomic_inc(&mdev->rs_pending_cnt);
2092 }
2093 
2094 #define dec_rs_pending(mdev)	do {				\
2095 	typecheck(struct drbd_conf *, mdev);			\
2096 	atomic_dec(&mdev->rs_pending_cnt);			\
2097 	ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
2098 
2099 /* counts how many answers we still need to send to the peer.
2100  * increased on
2101  *  receive_Data	unless protocol A;
2102  *			we need to send a P_RECV_ACK (proto B)
2103  *			or P_WRITE_ACK (proto C)
2104  *  receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2105  *  receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2106  *  receive_Barrier_*	we need to send a P_BARRIER_ACK
2107  */
2108 static inline void inc_unacked(struct drbd_conf *mdev)
2109 {
2110 	atomic_inc(&mdev->unacked_cnt);
2111 }
2112 
2113 #define dec_unacked(mdev)	do {				\
2114 	typecheck(struct drbd_conf *, mdev);			\
2115 	atomic_dec(&mdev->unacked_cnt);				\
2116 	ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2117 
2118 #define sub_unacked(mdev, n)	do {				\
2119 	typecheck(struct drbd_conf *, mdev);			\
2120 	atomic_sub(n, &mdev->unacked_cnt);			\
2121 	ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2122 
2123 
2124 static inline void put_net_conf(struct drbd_conf *mdev)
2125 {
2126 	if (atomic_dec_and_test(&mdev->net_cnt))
2127 		wake_up(&mdev->net_cnt_wait);
2128 }
2129 
2130 /**
2131  * get_net_conf() - Increase ref count on mdev->net_conf; Returns 0 if nothing there
2132  * @mdev:	DRBD device.
2133  *
2134  * You have to call put_net_conf() when finished working with mdev->net_conf.
2135  */
2136 static inline int get_net_conf(struct drbd_conf *mdev)
2137 {
2138 	int have_net_conf;
2139 
2140 	atomic_inc(&mdev->net_cnt);
2141 	have_net_conf = mdev->state.conn >= C_UNCONNECTED;
2142 	if (!have_net_conf)
2143 		put_net_conf(mdev);
2144 	return have_net_conf;
2145 }
2146 
2147 /**
2148  * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2149  * @M:		DRBD device.
2150  *
2151  * You have to call put_ldev() when finished working with mdev->ldev.
2152  */
2153 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2154 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2155 
2156 static inline void put_ldev(struct drbd_conf *mdev)
2157 {
2158 	int i = atomic_dec_return(&mdev->local_cnt);
2159 
2160 	/* This may be called from some endio handler,
2161 	 * so we must not sleep here. */
2162 
2163 	__release(local);
2164 	D_ASSERT(i >= 0);
2165 	if (i == 0) {
2166 		if (mdev->state.disk == D_DISKLESS)
2167 			/* even internal references gone, safe to destroy */
2168 			drbd_ldev_destroy(mdev);
2169 		if (mdev->state.disk == D_FAILED)
2170 			/* all application IO references gone. */
2171 			drbd_go_diskless(mdev);
2172 		wake_up(&mdev->misc_wait);
2173 	}
2174 }
2175 
2176 #ifndef __CHECKER__
2177 static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
2178 {
2179 	int io_allowed;
2180 
2181 	/* never get a reference while D_DISKLESS */
2182 	if (mdev->state.disk == D_DISKLESS)
2183 		return 0;
2184 
2185 	atomic_inc(&mdev->local_cnt);
2186 	io_allowed = (mdev->state.disk >= mins);
2187 	if (!io_allowed)
2188 		put_ldev(mdev);
2189 	return io_allowed;
2190 }
2191 #else
2192 extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins);
2193 #endif
2194 
2195 /* you must have an "get_ldev" reference */
2196 static inline void drbd_get_syncer_progress(struct drbd_conf *mdev,
2197 		unsigned long *bits_left, unsigned int *per_mil_done)
2198 {
2199 	/* this is to break it at compile time when we change that, in case we
2200 	 * want to support more than (1<<32) bits on a 32bit arch. */
2201 	typecheck(unsigned long, mdev->rs_total);
2202 
2203 	/* note: both rs_total and rs_left are in bits, i.e. in
2204 	 * units of BM_BLOCK_SIZE.
2205 	 * for the percentage, we don't care. */
2206 
2207 	if (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T)
2208 		*bits_left = mdev->ov_left;
2209 	else
2210 		*bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
2211 	/* >> 10 to prevent overflow,
2212 	 * +1 to prevent division by zero */
2213 	if (*bits_left > mdev->rs_total) {
2214 		/* doh. maybe a logic bug somewhere.
2215 		 * may also be just a race condition
2216 		 * between this and a disconnect during sync.
2217 		 * for now, just prevent in-kernel buffer overflow.
2218 		 */
2219 		smp_rmb();
2220 		dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2221 				drbd_conn_str(mdev->state.conn),
2222 				*bits_left, mdev->rs_total, mdev->rs_failed);
2223 		*per_mil_done = 0;
2224 	} else {
2225 		/* Make sure the division happens in long context.
2226 		 * We allow up to one petabyte storage right now,
2227 		 * at a granularity of 4k per bit that is 2**38 bits.
2228 		 * After shift right and multiplication by 1000,
2229 		 * this should still fit easily into a 32bit long,
2230 		 * so we don't need a 64bit division on 32bit arch.
2231 		 * Note: currently we don't support such large bitmaps on 32bit
2232 		 * arch anyways, but no harm done to be prepared for it here.
2233 		 */
2234 		unsigned int shift = mdev->rs_total >= (1ULL << 32) ? 16 : 10;
2235 		unsigned long left = *bits_left >> shift;
2236 		unsigned long total = 1UL + (mdev->rs_total >> shift);
2237 		unsigned long tmp = 1000UL - left * 1000UL/total;
2238 		*per_mil_done = tmp;
2239 	}
2240 }
2241 
2242 
2243 /* this throttles on-the-fly application requests
2244  * according to max_buffers settings;
2245  * maybe re-implement using semaphores? */
2246 static inline int drbd_get_max_buffers(struct drbd_conf *mdev)
2247 {
2248 	int mxb = 1000000; /* arbitrary limit on open requests */
2249 	if (get_net_conf(mdev)) {
2250 		mxb = mdev->net_conf->max_buffers;
2251 		put_net_conf(mdev);
2252 	}
2253 	return mxb;
2254 }
2255 
2256 static inline int drbd_state_is_stable(struct drbd_conf *mdev)
2257 {
2258 	union drbd_state s = mdev->state;
2259 
2260 	/* DO NOT add a default clause, we want the compiler to warn us
2261 	 * for any newly introduced state we may have forgotten to add here */
2262 
2263 	switch ((enum drbd_conns)s.conn) {
2264 	/* new io only accepted when there is no connection, ... */
2265 	case C_STANDALONE:
2266 	case C_WF_CONNECTION:
2267 	/* ... or there is a well established connection. */
2268 	case C_CONNECTED:
2269 	case C_SYNC_SOURCE:
2270 	case C_SYNC_TARGET:
2271 	case C_VERIFY_S:
2272 	case C_VERIFY_T:
2273 	case C_PAUSED_SYNC_S:
2274 	case C_PAUSED_SYNC_T:
2275 	case C_AHEAD:
2276 	case C_BEHIND:
2277 		/* transitional states, IO allowed */
2278 	case C_DISCONNECTING:
2279 	case C_UNCONNECTED:
2280 	case C_TIMEOUT:
2281 	case C_BROKEN_PIPE:
2282 	case C_NETWORK_FAILURE:
2283 	case C_PROTOCOL_ERROR:
2284 	case C_TEAR_DOWN:
2285 	case C_WF_REPORT_PARAMS:
2286 	case C_STARTING_SYNC_S:
2287 	case C_STARTING_SYNC_T:
2288 		break;
2289 
2290 		/* Allow IO in BM exchange states with new protocols */
2291 	case C_WF_BITMAP_S:
2292 		if (mdev->agreed_pro_version < 96)
2293 			return 0;
2294 		break;
2295 
2296 		/* no new io accepted in these states */
2297 	case C_WF_BITMAP_T:
2298 	case C_WF_SYNC_UUID:
2299 	case C_MASK:
2300 		/* not "stable" */
2301 		return 0;
2302 	}
2303 
2304 	switch ((enum drbd_disk_state)s.disk) {
2305 	case D_DISKLESS:
2306 	case D_INCONSISTENT:
2307 	case D_OUTDATED:
2308 	case D_CONSISTENT:
2309 	case D_UP_TO_DATE:
2310 		/* disk state is stable as well. */
2311 		break;
2312 
2313 	/* no new io accepted during tansitional states */
2314 	case D_ATTACHING:
2315 	case D_FAILED:
2316 	case D_NEGOTIATING:
2317 	case D_UNKNOWN:
2318 	case D_MASK:
2319 		/* not "stable" */
2320 		return 0;
2321 	}
2322 
2323 	return 1;
2324 }
2325 
2326 static inline int is_susp(union drbd_state s)
2327 {
2328 	return s.susp || s.susp_nod || s.susp_fen;
2329 }
2330 
2331 static inline bool may_inc_ap_bio(struct drbd_conf *mdev)
2332 {
2333 	int mxb = drbd_get_max_buffers(mdev);
2334 
2335 	if (is_susp(mdev->state))
2336 		return false;
2337 	if (test_bit(SUSPEND_IO, &mdev->flags))
2338 		return false;
2339 
2340 	/* to avoid potential deadlock or bitmap corruption,
2341 	 * in various places, we only allow new application io
2342 	 * to start during "stable" states. */
2343 
2344 	/* no new io accepted when attaching or detaching the disk */
2345 	if (!drbd_state_is_stable(mdev))
2346 		return false;
2347 
2348 	/* since some older kernels don't have atomic_add_unless,
2349 	 * and we are within the spinlock anyways, we have this workaround.  */
2350 	if (atomic_read(&mdev->ap_bio_cnt) > mxb)
2351 		return false;
2352 	if (test_bit(BITMAP_IO, &mdev->flags))
2353 		return false;
2354 	return true;
2355 }
2356 
2357 static inline bool inc_ap_bio_cond(struct drbd_conf *mdev, int count)
2358 {
2359 	bool rv = false;
2360 
2361 	spin_lock_irq(&mdev->req_lock);
2362 	rv = may_inc_ap_bio(mdev);
2363 	if (rv)
2364 		atomic_add(count, &mdev->ap_bio_cnt);
2365 	spin_unlock_irq(&mdev->req_lock);
2366 
2367 	return rv;
2368 }
2369 
2370 static inline void inc_ap_bio(struct drbd_conf *mdev, int count)
2371 {
2372 	/* we wait here
2373 	 *    as long as the device is suspended
2374 	 *    until the bitmap is no longer on the fly during connection
2375 	 *    handshake as long as we would exeed the max_buffer limit.
2376 	 *
2377 	 * to avoid races with the reconnect code,
2378 	 * we need to atomic_inc within the spinlock. */
2379 
2380 	wait_event(mdev->misc_wait, inc_ap_bio_cond(mdev, count));
2381 }
2382 
2383 static inline void dec_ap_bio(struct drbd_conf *mdev)
2384 {
2385 	int mxb = drbd_get_max_buffers(mdev);
2386 	int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt);
2387 
2388 	D_ASSERT(ap_bio >= 0);
2389 	/* this currently does wake_up for every dec_ap_bio!
2390 	 * maybe rather introduce some type of hysteresis?
2391 	 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2392 	if (ap_bio < mxb)
2393 		wake_up(&mdev->misc_wait);
2394 	if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) {
2395 		if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags))
2396 			drbd_queue_work(&mdev->data.work, &mdev->bm_io_work.w);
2397 	}
2398 }
2399 
2400 static inline int drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val)
2401 {
2402 	int changed = mdev->ed_uuid != val;
2403 	mdev->ed_uuid = val;
2404 	return changed;
2405 }
2406 
2407 static inline int seq_cmp(u32 a, u32 b)
2408 {
2409 	/* we assume wrap around at 32bit.
2410 	 * for wrap around at 24bit (old atomic_t),
2411 	 * we'd have to
2412 	 *  a <<= 8; b <<= 8;
2413 	 */
2414 	return (s32)(a) - (s32)(b);
2415 }
2416 #define seq_lt(a, b) (seq_cmp((a), (b)) < 0)
2417 #define seq_gt(a, b) (seq_cmp((a), (b)) > 0)
2418 #define seq_ge(a, b) (seq_cmp((a), (b)) >= 0)
2419 #define seq_le(a, b) (seq_cmp((a), (b)) <= 0)
2420 /* CAUTION: please no side effects in arguments! */
2421 #define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b)))
2422 
2423 static inline void update_peer_seq(struct drbd_conf *mdev, unsigned int new_seq)
2424 {
2425 	unsigned int m;
2426 	spin_lock(&mdev->peer_seq_lock);
2427 	m = seq_max(mdev->peer_seq, new_seq);
2428 	mdev->peer_seq = m;
2429 	spin_unlock(&mdev->peer_seq_lock);
2430 	if (m == new_seq)
2431 		wake_up(&mdev->seq_wait);
2432 }
2433 
2434 static inline void drbd_update_congested(struct drbd_conf *mdev)
2435 {
2436 	struct sock *sk = mdev->data.socket->sk;
2437 	if (sk->sk_wmem_queued > sk->sk_sndbuf * 4 / 5)
2438 		set_bit(NET_CONGESTED, &mdev->flags);
2439 }
2440 
2441 static inline int drbd_queue_order_type(struct drbd_conf *mdev)
2442 {
2443 	/* sorry, we currently have no working implementation
2444 	 * of distributed TCQ stuff */
2445 #ifndef QUEUE_ORDERED_NONE
2446 #define QUEUE_ORDERED_NONE 0
2447 #endif
2448 	return QUEUE_ORDERED_NONE;
2449 }
2450 
2451 static inline void drbd_md_flush(struct drbd_conf *mdev)
2452 {
2453 	int r;
2454 
2455 	if (test_bit(MD_NO_FUA, &mdev->flags))
2456 		return;
2457 
2458 	r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL);
2459 	if (r) {
2460 		set_bit(MD_NO_FUA, &mdev->flags);
2461 		dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r);
2462 	}
2463 }
2464 
2465 #endif
2466