4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
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>.
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)
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.
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.
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 <linux/idr.h>
45 #include <linux/lru_cache.h>
46 #include <linux/prefetch.h>
47 #include <linux/drbd.h>
48 #include "drbd_state.h"
51 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
52 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
53 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
54 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
56 # define __protected_by(x)
57 # define __protected_read_by(x)
58 # define __protected_write_by(x)
59 # define __must_hold(x)
62 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
64 /* module parameter, defined in drbd_main.c */
65 extern unsigned int minor_count;
66 extern int disable_sendpage;
68 extern unsigned int cn_idx;
70 #ifdef CONFIG_DRBD_FAULT_INJECTION
71 extern int enable_faults;
72 extern int fault_rate;
73 extern int fault_devs;
76 extern char usermode_helper[];
79 /* I don't remember why XCPU ...
80 * This is used to wake the asender,
81 * and to interrupt sending the sending task
84 #define DRBD_SIG SIGXCPU
86 /* This is used to stop/restart our threads.
87 * Cannot use SIGTERM nor SIGKILL, since these
88 * are sent out by init on runlevel changes
89 * I choose SIGHUP for now.
91 #define DRBD_SIGKILL SIGHUP
93 #define ID_IN_SYNC (4711ULL)
94 #define ID_OUT_OF_SYNC (4712ULL)
95 #define ID_SYNCER (-1ULL)
97 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
103 /* to shorten dev_warn(DEV, "msg"); and relatives statements */
104 #define DEV (disk_to_dev(mdev->vdisk))
106 #define conn_printk(LEVEL, TCONN, FMT, ARGS...) \
107 printk(LEVEL "d-con %s: " FMT, TCONN->name , ## ARGS)
108 #define conn_alert(TCONN, FMT, ARGS...) conn_printk(KERN_ALERT, TCONN, FMT, ## ARGS)
109 #define conn_crit(TCONN, FMT, ARGS...) conn_printk(KERN_CRIT, TCONN, FMT, ## ARGS)
110 #define conn_err(TCONN, FMT, ARGS...) conn_printk(KERN_ERR, TCONN, FMT, ## ARGS)
111 #define conn_warn(TCONN, FMT, ARGS...) conn_printk(KERN_WARNING, TCONN, FMT, ## ARGS)
112 #define conn_notice(TCONN, FMT, ARGS...) conn_printk(KERN_NOTICE, TCONN, FMT, ## ARGS)
113 #define conn_info(TCONN, FMT, ARGS...) conn_printk(KERN_INFO, TCONN, FMT, ## ARGS)
114 #define conn_dbg(TCONN, FMT, ARGS...) conn_printk(KERN_DEBUG, TCONN, FMT, ## ARGS)
116 #define D_ASSERT(exp) if (!(exp)) \
117 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
120 * expect - Make an assertion
122 * Unlike the assert macro, this macro returns a boolean result.
124 #define expect(exp) ({ \
125 bool _bool = (exp); \
127 dev_err(DEV, "ASSERTION %s FAILED in %s\n", \
132 /* Defines to control fault insertion */
134 DRBD_FAULT_MD_WR = 0, /* meta data write */
135 DRBD_FAULT_MD_RD = 1, /* read */
136 DRBD_FAULT_RS_WR = 2, /* resync */
137 DRBD_FAULT_RS_RD = 3,
138 DRBD_FAULT_DT_WR = 4, /* data */
139 DRBD_FAULT_DT_RD = 5,
140 DRBD_FAULT_DT_RA = 6, /* data read ahead */
141 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */
142 DRBD_FAULT_AL_EE = 8, /* alloc ee */
143 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
149 _drbd_insert_fault(struct drbd_conf *mdev, unsigned int type);
152 drbd_insert_fault(struct drbd_conf *mdev, unsigned int type) {
153 #ifdef CONFIG_DRBD_FAULT_INJECTION
155 (enable_faults & (1<<type)) &&
156 _drbd_insert_fault(mdev, type);
162 /* integer division, round _UP_ to the next integer */
163 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
164 /* usual integer division */
165 #define div_floor(A, B) ((A)/(B))
167 /* drbd_meta-data.c (still in drbd_main.c) */
168 /* 4th incarnation of the disk layout. */
169 #define DRBD_MD_MAGIC (DRBD_MAGIC+4)
171 extern struct drbd_conf **minor_table;
172 extern struct ratelimit_state drbd_ratelimit_state;
176 /* receiver (data socket) */
178 P_DATA_REPLY = 0x01, /* Response to P_DATA_REQUEST */
179 P_RS_DATA_REPLY = 0x02, /* Response to P_RS_DATA_REQUEST */
182 P_BECOME_SYNC_TARGET = 0x05,
183 P_BECOME_SYNC_SOURCE = 0x06,
184 P_UNPLUG_REMOTE = 0x07, /* Used at various times to hint the peer */
185 P_DATA_REQUEST = 0x08, /* Used to ask for a data block */
186 P_RS_DATA_REQUEST = 0x09, /* Used to ask for a data block for resync */
193 P_AUTH_CHALLENGE = 0x10,
194 P_AUTH_RESPONSE = 0x11,
195 P_STATE_CHG_REQ = 0x12,
197 /* asender (meta socket */
200 P_RECV_ACK = 0x15, /* Used in protocol B */
201 P_WRITE_ACK = 0x16, /* Used in protocol C */
202 P_RS_WRITE_ACK = 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
203 P_DISCARD_ACK = 0x18, /* Used in proto C, two-primaries conflict detection */
204 P_NEG_ACK = 0x19, /* Sent if local disk is unusable */
205 P_NEG_DREPLY = 0x1a, /* Local disk is broken... */
206 P_NEG_RS_DREPLY = 0x1b, /* Local disk is broken... */
207 P_BARRIER_ACK = 0x1c,
208 P_STATE_CHG_REPLY = 0x1d,
210 /* "new" commands, no longer fitting into the ordering scheme above */
212 P_OV_REQUEST = 0x1e, /* data socket */
214 P_OV_RESULT = 0x20, /* meta socket */
215 P_CSUM_RS_REQUEST = 0x21, /* data socket */
216 P_RS_IS_IN_SYNC = 0x22, /* meta socket */
217 P_SYNC_PARAM89 = 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
218 P_COMPRESSED_BITMAP = 0x24, /* compressed or otherwise encoded bitmap transfer */
219 /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */
220 /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */
221 P_DELAY_PROBE = 0x27, /* is used on BOTH sockets */
222 P_OUT_OF_SYNC = 0x28, /* Mark as out of sync (Outrunning), data socket */
223 P_RS_CANCEL = 0x29, /* meta: Used to cancel RS_DATA_REQUEST packet by SyncSource */
224 P_CONN_ST_CHG_REQ = 0x2a, /* data sock: Connection wide state request */
225 P_CONN_ST_CHG_REPLY = 0x2b, /* meta sock: Connection side state req reply */
228 P_MAY_IGNORE = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
229 P_MAX_OPT_CMD = 0x101,
231 /* special command ids for handshake */
233 P_HAND_SHAKE_M = 0xfff1, /* First Packet on the MetaSock */
234 P_HAND_SHAKE_S = 0xfff2, /* First Packet on the Socket */
236 P_HAND_SHAKE = 0xfffe /* FIXED for the next century! */
239 extern const char *cmdname(enum drbd_packet cmd);
241 /* for sending/receiving the bitmap,
242 * possibly in some encoding scheme */
245 * stores total bits and long words
246 * of the bitmap, so we don't need to
247 * call the accessor functions over and again. */
248 unsigned long bm_bits;
249 unsigned long bm_words;
250 /* during xfer, current position within the bitmap */
251 unsigned long bit_offset;
252 unsigned long word_offset;
254 /* statistics; index: (h->command == P_BITMAP) */
259 extern void INFO_bm_xfer_stats(struct drbd_conf *mdev,
260 const char *direction, struct bm_xfer_ctx *c);
262 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
264 /* word_offset counts "native long words" (32 or 64 bit),
266 * Encoded packet may end at an unaligned bit offset.
267 * In case a fallback clear text packet is transmitted in
268 * between, we adjust this offset back to the last 64bit
269 * aligned "native long word", which makes coding and decoding
270 * the plain text bitmap much more convenient. */
271 #if BITS_PER_LONG == 64
272 c->word_offset = c->bit_offset >> 6;
273 #elif BITS_PER_LONG == 32
274 c->word_offset = c->bit_offset >> 5;
275 c->word_offset &= ~(1UL);
277 # error "unsupported BITS_PER_LONG"
282 #define __packed __attribute__((packed))
285 /* This is the layout for a packet on the wire.
286 * The byteorder is the network byte order.
287 * (except block_id and barrier fields.
288 * these are pointers to local structs
289 * and have no relevance for the partner,
290 * which just echoes them as received.)
292 * NOTE that the payload starts at a long aligned offset,
293 * regardless of 32 or 64 bit arch!
298 u16 length; /* bytes of data after this header */
301 /* Header for big packets, Used for data packets exceeding 64kB */
303 u16 magic; /* use DRBD_MAGIC_BIG here */
305 u32 length; /* Use only 24 bits of that. Ignore the highest 8 bit. */
310 struct p_header80 h80;
311 struct p_header95 h95;
317 * short commands, packets without payload, plain p_header:
320 * P_BECOME_SYNC_TARGET
321 * P_BECOME_SYNC_SOURCE
326 * commands with out-of-struct payload:
327 * P_BITMAP (no additional fields)
328 * P_DATA, P_DATA_REPLY (see p_data)
329 * P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
332 /* these defines must not be changed without changing the protocol version */
333 #define DP_HARDBARRIER 1 /* depricated */
334 #define DP_RW_SYNC 2 /* equals REQ_SYNC */
335 #define DP_MAY_SET_IN_SYNC 4
336 #define DP_UNPLUG 8 /* not used anymore */
337 #define DP_FUA 16 /* equals REQ_FUA */
338 #define DP_FLUSH 32 /* equals REQ_FLUSH */
339 #define DP_DISCARD 64 /* equals REQ_DISCARD */
342 struct p_header head;
343 u64 sector; /* 64 bits sector number */
344 u64 block_id; /* to identify the request in protocol B&C */
350 * commands which share a struct:
352 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
353 * P_DISCARD_ACK (proto C, two-primaries conflict detection)
355 * P_DATA_REQUEST, P_RS_DATA_REQUEST
358 struct p_header head;
367 struct p_header head;
371 u32 pad; /* to multiple of 8 Byte */
375 * commands with their own struct for additional fields:
384 struct p_header head; /* Note: vnr will be ignored */
389 /* should be more than enough for future enhancements
390 * for now, feature_flags and the reserverd array shall be zero.
396 /* 80 bytes, FIXED for the next century */
399 struct p_header head;
400 u32 barrier; /* barrier number _handle_ only */
401 u32 pad; /* to multiple of 8 Byte */
404 struct p_barrier_ack {
405 struct p_header head;
411 struct p_header head;
414 /* Since protocol version 88 and higher. */
418 struct p_rs_param_89 {
419 struct p_header head;
421 /* protocol version 89: */
422 char verify_alg[SHARED_SECRET_MAX];
423 char csums_alg[SHARED_SECRET_MAX];
426 struct p_rs_param_95 {
427 struct p_header head;
429 char verify_alg[SHARED_SECRET_MAX];
430 char csums_alg[SHARED_SECRET_MAX];
437 enum drbd_conn_flags {
443 struct p_header head;
451 /* Since protocol version 87 and higher. */
452 char integrity_alg[0];
457 struct p_header head;
458 u64 uuid[UI_EXTENDED_SIZE];
462 struct p_header head;
467 struct p_header head;
468 u64 d_size; /* size of disk */
469 u64 u_size; /* user requested size */
470 u64 c_size; /* current exported size */
471 u32 max_bio_size; /* Maximal size of a BIO */
472 u16 queue_order_type; /* not yet implemented in DRBD*/
473 u16 dds_flags; /* use enum dds_flags here. */
477 struct p_header head;
482 struct p_header head;
487 struct p_req_state_reply {
488 struct p_header head;
492 struct p_drbd06_param {
503 struct p_header head;
509 struct p_block_desc {
510 struct p_header head;
513 u32 pad; /* to multiple of 8 Byte */
516 /* Valid values for the encoding field.
517 * Bump proto version when changing this. */
518 enum drbd_bitmap_code {
519 /* RLE_VLI_Bytes = 0,
520 * and other bit variants had been defined during
521 * algorithm evaluation. */
525 struct p_compressed_bm {
526 struct p_header head;
527 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
528 * (encoding & 0x80): polarity (set/unset) of first runlength
529 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
530 * used to pad up to head.length bytes
537 struct p_delay_probe93 {
538 struct p_header head;
539 u32 seq_num; /* sequence number to match the two probe packets */
540 u32 offset; /* usecs the probe got sent after the reference time point */
543 /* DCBP: Drbd Compressed Bitmap Packet ... */
544 static inline enum drbd_bitmap_code
545 DCBP_get_code(struct p_compressed_bm *p)
547 return (enum drbd_bitmap_code)(p->encoding & 0x0f);
551 DCBP_set_code(struct p_compressed_bm *p, enum drbd_bitmap_code code)
554 p->encoding = (p->encoding & ~0xf) | code;
558 DCBP_get_start(struct p_compressed_bm *p)
560 return (p->encoding & 0x80) != 0;
564 DCBP_set_start(struct p_compressed_bm *p, int set)
566 p->encoding = (p->encoding & ~0x80) | (set ? 0x80 : 0);
570 DCBP_get_pad_bits(struct p_compressed_bm *p)
572 return (p->encoding >> 4) & 0x7;
576 DCBP_set_pad_bits(struct p_compressed_bm *p, int n)
579 p->encoding = (p->encoding & (~0x7 << 4)) | (n << 4);
582 /* one bitmap packet, including the p_header,
583 * should fit within one _architecture independend_ page.
584 * so we need to use the fixed size 4KiB page size
585 * most architectures have used for a long time.
587 #define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header))
588 #define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
589 #define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm))
590 #if (PAGE_SIZE < 4096)
591 /* drbd_send_bitmap / receive_bitmap would break horribly */
592 #error "PAGE_SIZE too small"
596 struct p_header header;
597 struct p_handshake handshake;
599 struct p_block_ack block_ack;
600 struct p_barrier barrier;
601 struct p_barrier_ack barrier_ack;
602 struct p_rs_param_89 rs_param_89;
603 struct p_rs_param_95 rs_param_95;
604 struct p_protocol protocol;
605 struct p_sizes sizes;
606 struct p_uuids uuids;
607 struct p_state state;
608 struct p_req_state req_state;
609 struct p_req_state_reply req_state_reply;
610 struct p_block_req block_req;
611 struct p_delay_probe93 delay_probe93;
612 struct p_rs_uuid rs_uuid;
613 struct p_block_desc block_desc;
616 /**********************************************************************/
617 enum drbd_thread_state {
626 struct task_struct *task;
627 struct completion stop;
628 enum drbd_thread_state t_state;
629 int (*function) (struct drbd_thread *);
630 struct drbd_tconn *tconn;
635 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
637 /* THINK testing the t_state seems to be uncritical in all cases
638 * (but thread_{start,stop}), so we can read it *without* the lock.
646 typedef int (*drbd_work_cb)(struct drbd_work *, int cancel);
648 struct list_head list;
651 struct drbd_conf *mdev;
652 struct drbd_tconn *tconn;
656 #include "drbd_interval.h"
658 struct drbd_request {
661 /* if local IO is not allowed, will be NULL.
662 * if local IO _is_ allowed, holds the locally submitted bio clone,
663 * or, after local IO completion, the ERR_PTR(error).
664 * see drbd_request_endio(). */
665 struct bio *private_bio;
667 struct drbd_interval i;
668 unsigned int epoch; /* barrier_nr */
670 /* barrier_nr: used to check on "completion" whether this req was in
671 * the current epoch, and we therefore have to close it,
672 * starting a new epoch...
675 struct list_head tl_requests; /* ring list in the transfer log */
676 struct bio *master_bio; /* master bio pointer */
677 unsigned long rq_state; /* see comments above _req_mod() */
679 unsigned long start_time;
682 struct drbd_tl_epoch {
684 struct list_head requests; /* requests before */
685 struct drbd_tl_epoch *next; /* pointer to the next barrier */
686 unsigned int br_number; /* the barriers identifier. */
687 int n_writes; /* number of requests attached before this barrier */
691 struct list_head list;
692 unsigned int barrier_nr;
693 atomic_t epoch_size; /* increased on every request added. */
694 atomic_t active; /* increased on every req. added, and dec on every finished. */
698 /* drbd_epoch flag bits */
700 DE_HAVE_BARRIER_NUMBER,
707 EV_CLEANUP = 32, /* used as flag */
710 struct drbd_wq_barrier {
712 struct completion done;
720 struct drbd_peer_request {
722 struct drbd_epoch *epoch; /* for writes */
724 atomic_t pending_bios;
725 struct drbd_interval i;
726 /* see comments on ee flag bits below */
730 struct digest_info *digest;
735 * While corresponding bios are in flight, the only modification will be
736 * set_bit WAS_ERROR, which has to be atomic.
737 * If no bios are in flight yet, or all have been completed,
738 * non-atomic modification to ee->flags is ok.
741 __EE_CALL_AL_COMPLETE_IO,
742 __EE_MAY_SET_IN_SYNC,
744 /* In case a barrier failed,
745 * we need to resubmit without the barrier flag. */
748 /* we may have several bios per peer request.
749 * if any of those fail, we set this flag atomically
750 * from the endio callback */
753 /* This ee has a pointer to a digest instead of a block id */
756 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
757 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
758 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
759 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
760 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
762 /* flag bits per mdev */
764 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */
765 UNPLUG_QUEUED, /* only relevant with kernel 2.4 */
766 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
767 MD_DIRTY, /* current uuids and flags not yet on disk */
768 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
771 CRASHED_PRIMARY, /* This node was a crashed primary.
772 * Gets cleared when the state.conn
773 * goes into C_CONNECTED state. */
774 NO_BARRIER_SUPP, /* underlying block device doesn't implement barriers */
777 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */
778 SUSPEND_IO, /* suspend application io */
779 BITMAP_IO, /* suspend application io;
780 once no more io in flight, start bitmap io */
781 BITMAP_IO_QUEUED, /* Started bitmap IO */
782 GO_DISKLESS, /* Disk is being detached, on io-error or admin request. */
783 WAS_IO_ERROR, /* Local disk failed returned IO error */
784 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
785 CONFIG_PENDING, /* serialization of (re)configuration requests.
786 * if set, also prevents the device from dying */
787 DEVICE_DYING, /* device became unconfigured,
788 * but worker thread is still handling the cleanup.
789 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
790 * while this is set. */
791 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
792 * the peer, if it changed there as well. */
793 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
794 NEW_CUR_UUID, /* Create new current UUID when thawing IO */
795 AL_SUSPENDED, /* Activity logging is currently suspended. */
796 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */
797 B_RS_H_DONE, /* Before resync handler done (already executed) */
800 struct drbd_bitmap; /* opaque for drbd_conf */
802 /* definition of bits in bm_flags to be used in drbd_bm_lock
803 * and drbd_bitmap_io and friends. */
805 /* do we need to kfree, or vfree bm_pages? */
806 BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */
808 /* currently locked for bulk operation */
809 BM_LOCKED_MASK = 0x7,
811 /* in detail, that is: */
816 /* (test bit, count bit) allowed (common case) */
817 BM_LOCKED_TEST_ALLOWED = 0x3,
819 /* testing bits, as well as setting new bits allowed, but clearing bits
820 * would be unexpected. Used during bitmap receive. Setting new bits
821 * requires sending of "out-of-sync" information, though. */
822 BM_LOCKED_SET_ALLOWED = 0x1,
824 /* clear is not expected while bitmap is locked for bulk operation */
828 /* TODO sort members for performance
829 * MAYBE group them further */
831 /* THINK maybe we actually want to use the default "event/%s" worker threads
832 * or similar in linux 2.6, which uses per cpu data and threads.
834 struct drbd_work_queue {
836 struct semaphore s; /* producers up it, worker down()s it */
837 spinlock_t q_lock; /* to protect the list. */
841 struct drbd_work_queue work;
843 struct socket *socket;
844 /* this way we get our
845 * send/receive buffers off the stack */
846 union p_polymorph sbuf;
847 union p_polymorph rbuf;
851 u64 md_offset; /* sector offset to 'super' block */
853 u64 la_size_sect; /* last agreed size, unit sectors */
859 s32 al_offset; /* signed relative sector offset to al area */
860 s32 bm_offset; /* signed relative sector offset to bitmap */
862 /* u32 al_nr_extents; important for restoring the AL
863 * is stored into sync_conf.al_extents, which in turn
864 * gets applied to act_log->nr_elements
868 /* for sync_conf and other types... */
869 #define NL_PACKET(name, number, fields) struct name { fields };
870 #define NL_INTEGER(pn,pr,member) int member;
871 #define NL_INT64(pn,pr,member) __u64 member;
872 #define NL_BIT(pn,pr,member) unsigned member:1;
873 #define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
874 #include "linux/drbd_nl.h"
876 struct drbd_backing_dev {
877 struct block_device *backing_bdev;
878 struct block_device *md_bdev;
880 struct disk_conf dc; /* The user provided config... */
881 sector_t known_size; /* last known size of that backing device */
885 struct drbd_conf *mdev;
886 struct completion event;
894 int (*io_fn)(struct drbd_conf *mdev);
895 void (*done)(struct drbd_conf *mdev, int rv);
898 enum write_ordering_e {
906 unsigned int head_index;
910 /* flag bits per tconn */
912 NET_CONGESTED, /* The data socket is congested */
913 DISCARD_CONCURRENT, /* Set on one node, cleared on the peer! */
914 SEND_PING, /* whether asender should send a ping asap */
915 SIGNAL_ASENDER, /* whether asender wants to be interrupted */
916 GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */
921 struct drbd_tconn { /* is a resource from the config file */
922 char *name; /* Resource name */
923 struct list_head all_tconn; /* List of all drbd_tconn, prot by global_state_lock */
924 struct idr volumes; /* <tconn, vnr> to mdev mapping */
925 enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
926 struct mutex cstate_mutex; /* Protects graceful disconnects */
929 struct net_conf *net_conf; /* protected by get_net_conf() and put_net_conf() */
930 atomic_t net_cnt; /* Users of net_conf */
931 wait_queue_head_t net_cnt_wait;
932 wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */
934 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
935 struct drbd_socket meta; /* ping/ack (metadata) packets */
936 int agreed_pro_version; /* actually used protocol version */
937 unsigned long last_received; /* in jiffies, either socket */
938 unsigned int ko_count;
941 struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */
942 struct drbd_tl_epoch *newest_tle;
943 struct drbd_tl_epoch *oldest_tle;
944 struct list_head out_of_sequence_requests;
946 struct crypto_hash *cram_hmac_tfm;
947 struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */
948 struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */
953 struct drbd_thread receiver;
954 struct drbd_thread worker;
955 struct drbd_thread asender;
956 cpumask_var_t cpu_mask;
960 struct drbd_tconn *tconn;
961 int vnr; /* volume number within the connection */
963 /* things that are stored as / read from meta data on disk */
966 /* configured by drbdsetup */
967 struct syncer_conf sync_conf;
968 struct drbd_backing_dev *ldev __protected_by(local);
970 sector_t p_size; /* partner's disk size */
971 struct request_queue *rq_queue;
972 struct block_device *this_bdev;
973 struct gendisk *vdisk;
975 struct drbd_work resync_work,
980 struct timer_list resync_timer;
981 struct timer_list md_sync_timer;
982 struct timer_list start_resync_timer;
983 struct timer_list request_timer;
984 #ifdef DRBD_DEBUG_MD_SYNC
988 } last_md_mark_dirty;
991 /* Used after attach while negotiating new disk state. */
992 union drbd_state new_state_tmp;
994 union drbd_state state;
995 wait_queue_head_t misc_wait;
996 wait_queue_head_t state_wait; /* upon each state change. */
997 unsigned int send_cnt;
998 unsigned int recv_cnt;
999 unsigned int read_cnt;
1000 unsigned int writ_cnt;
1001 unsigned int al_writ_cnt;
1002 unsigned int bm_writ_cnt;
1003 atomic_t ap_bio_cnt; /* Requests we need to complete */
1004 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
1005 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
1006 atomic_t unacked_cnt; /* Need to send replys for */
1007 atomic_t local_cnt; /* Waiting for local completion */
1009 /* Interval tree of pending local requests */
1010 struct rb_root read_requests;
1011 struct rb_root write_requests;
1013 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
1014 unsigned long rs_total;
1015 /* number of resync blocks that failed in this run */
1016 unsigned long rs_failed;
1017 /* Syncer's start time [unit jiffies] */
1018 unsigned long rs_start;
1019 /* cumulated time in PausedSyncX state [unit jiffies] */
1020 unsigned long rs_paused;
1021 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
1022 unsigned long rs_same_csum;
1023 #define DRBD_SYNC_MARKS 8
1024 #define DRBD_SYNC_MARK_STEP (3*HZ)
1025 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
1026 unsigned long rs_mark_left[DRBD_SYNC_MARKS];
1027 /* marks's time [unit jiffies] */
1028 unsigned long rs_mark_time[DRBD_SYNC_MARKS];
1029 /* current index into rs_mark_{left,time} */
1032 /* where does the admin want us to start? (sector) */
1033 sector_t ov_start_sector;
1034 /* where are we now? (sector) */
1035 sector_t ov_position;
1036 /* Start sector of out of sync range (to merge printk reporting). */
1037 sector_t ov_last_oos_start;
1038 /* size of out-of-sync range in sectors. */
1039 sector_t ov_last_oos_size;
1040 unsigned long ov_left; /* in bits */
1041 struct crypto_hash *csums_tfm;
1042 struct crypto_hash *verify_tfm;
1044 struct drbd_bitmap *bitmap;
1045 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
1047 /* Used to track operations of resync... */
1048 struct lru_cache *resync;
1049 /* Number of locked elements in resync LRU */
1050 unsigned int resync_locked;
1051 /* resync extent number waiting for application requests */
1052 unsigned int resync_wenr;
1056 struct drbd_epoch *current_epoch;
1057 spinlock_t epoch_lock;
1058 unsigned int epochs;
1059 enum write_ordering_e write_ordering;
1060 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
1061 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
1062 struct list_head done_ee; /* need to send P_WRITE_ACK */
1063 struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */
1064 struct list_head net_ee; /* zero-copy network send in progress */
1066 int next_barrier_nr;
1067 struct list_head resync_reads;
1068 atomic_t pp_in_use; /* allocated from page pool */
1069 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */
1070 wait_queue_head_t ee_wait;
1071 struct page *md_io_page; /* one page buffer for md_io */
1072 struct page *md_io_tmpp; /* for logical_block_size != 512 */
1073 struct mutex md_io_mutex; /* protects the md_io_buffer */
1075 wait_queue_head_t al_wait;
1076 struct lru_cache *act_log; /* activity log */
1077 unsigned int al_tr_number;
1079 int al_tr_pos; /* position of the next transaction in the journal */
1080 wait_queue_head_t seq_wait;
1081 atomic_t packet_seq;
1082 unsigned int peer_seq;
1083 spinlock_t peer_seq_lock;
1085 unsigned long comm_bm_set; /* communicated number of set bits. */
1086 struct bm_io_work bm_io_work;
1087 u64 ed_uuid; /* UUID of the exposed data */
1088 struct mutex own_state_mutex;
1089 struct mutex *state_mutex; /* either own_state_mutex or mdev->tconn->cstate_mutex */
1090 char congestion_reason; /* Why we where congested... */
1091 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
1092 atomic_t rs_sect_ev; /* for submitted resync data rate, both */
1093 int rs_last_sect_ev; /* counter to compare with */
1094 int rs_last_events; /* counter of read or write "events" (unit sectors)
1095 * on the lower level device when we last looked. */
1096 int c_sync_rate; /* current resync rate after syncer throttle magic */
1097 struct fifo_buffer rs_plan_s; /* correction values of resync planer */
1098 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
1099 int rs_planed; /* resync sectors already planned */
1100 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
1101 int peer_max_bio_size;
1102 int local_max_bio_size;
1105 static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
1107 struct drbd_conf *mdev;
1109 mdev = minor < minor_count ? minor_table[minor] : NULL;
1114 static inline unsigned int mdev_to_minor(struct drbd_conf *mdev)
1119 static inline struct drbd_conf *vnr_to_mdev(struct drbd_tconn *tconn, int vnr)
1121 return (struct drbd_conf *)idr_find(&tconn->volumes, vnr);
1124 /* returns 1 if it was successful,
1125 * returns 0 if there was no data socket.
1126 * so wherever you are going to use the data.socket, e.g. do
1127 * if (!drbd_get_data_sock(mdev->tconn))
1130 * drbd_get_data_sock(mdev->tconn);
1132 static inline int drbd_get_data_sock(struct drbd_tconn *tconn)
1134 mutex_lock(&tconn->data.mutex);
1135 /* drbd_disconnect() could have called drbd_free_sock()
1136 * while we were waiting in down()... */
1137 if (unlikely(tconn->data.socket == NULL)) {
1138 mutex_unlock(&tconn->data.mutex);
1144 static inline void drbd_put_data_sock(struct drbd_tconn *tconn)
1146 mutex_unlock(&tconn->data.mutex);
1150 * function declarations
1151 *************************/
1157 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1160 extern void drbd_init_set_defaults(struct drbd_conf *mdev);
1161 extern int drbd_thread_start(struct drbd_thread *thi);
1162 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1163 extern char *drbd_task_to_thread_name(struct drbd_tconn *tconn, struct task_struct *task);
1165 extern void drbd_thread_current_set_cpu(struct drbd_thread *thi);
1166 extern void drbd_calc_cpu_mask(struct drbd_tconn *tconn);
1168 #define drbd_thread_current_set_cpu(A) ({})
1169 #define drbd_calc_cpu_mask(A) ({})
1171 extern void drbd_free_resources(struct drbd_conf *mdev);
1172 extern void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr,
1173 unsigned int set_size);
1174 extern void tl_clear(struct drbd_conf *mdev);
1175 extern void _tl_add_barrier(struct drbd_conf *, struct drbd_tl_epoch *);
1176 extern void drbd_free_sock(struct drbd_tconn *tconn);
1177 extern int drbd_send(struct drbd_tconn *tconn, struct socket *sock,
1178 void *buf, size_t size, unsigned msg_flags);
1179 extern int drbd_send_protocol(struct drbd_tconn *tconn);
1180 extern int drbd_send_uuids(struct drbd_conf *mdev);
1181 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
1182 extern int drbd_gen_and_send_sync_uuid(struct drbd_conf *mdev);
1183 extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags);
1184 extern int _conn_send_state_req(struct drbd_tconn *, int vnr, enum drbd_packet cmd,
1185 union drbd_state, union drbd_state);
1186 extern int _drbd_send_state(struct drbd_conf *mdev);
1187 extern int drbd_send_state(struct drbd_conf *mdev);
1188 extern int _conn_send_cmd(struct drbd_tconn *tconn, int vnr, struct socket *sock,
1189 enum drbd_packet cmd, struct p_header *h, size_t size,
1190 unsigned msg_flags);
1191 extern int conn_send_cmd(struct drbd_tconn *tconn, int vnr, int use_data_socket,
1192 enum drbd_packet cmd, struct p_header *h, size_t size);
1193 extern int conn_send_cmd2(struct drbd_tconn *tconn, enum drbd_packet cmd,
1194 char *data, size_t size);
1195 #define USE_DATA_SOCKET 1
1196 #define USE_META_SOCKET 0
1197 extern int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc);
1198 extern int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr,
1200 extern int drbd_send_ack(struct drbd_conf *, enum drbd_packet,
1201 struct drbd_peer_request *);
1202 extern int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packet cmd,
1203 struct p_block_req *rp);
1204 extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packet cmd,
1205 struct p_data *dp, int data_size);
1206 extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packet cmd,
1207 sector_t sector, int blksize, u64 block_id);
1208 extern int drbd_send_oos(struct drbd_conf *mdev, struct drbd_request *req);
1209 extern int drbd_send_block(struct drbd_conf *, enum drbd_packet,
1210 struct drbd_peer_request *);
1211 extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
1212 extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1213 sector_t sector, int size, u64 block_id);
1214 extern int drbd_send_drequest_csum(struct drbd_conf *mdev, sector_t sector,
1215 int size, void *digest, int digest_size,
1216 enum drbd_packet cmd);
1217 extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size);
1219 extern int drbd_send_bitmap(struct drbd_conf *mdev);
1220 extern int _drbd_send_bitmap(struct drbd_conf *mdev);
1221 extern int drbd_send_sr_reply(struct drbd_conf *mdev, enum drbd_state_rv retcode);
1222 extern int conn_send_sr_reply(struct drbd_tconn *tconn, enum drbd_state_rv retcode);
1223 extern void drbd_free_bc(struct drbd_backing_dev *ldev);
1224 extern void drbd_mdev_cleanup(struct drbd_conf *mdev);
1225 void drbd_print_uuids(struct drbd_conf *mdev, const char *text);
1227 extern void drbd_md_sync(struct drbd_conf *mdev);
1228 extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
1229 extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1230 extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1231 extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1232 extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1233 extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
1234 extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
1235 extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
1236 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1237 #ifndef DRBD_DEBUG_MD_SYNC
1238 extern void drbd_md_mark_dirty(struct drbd_conf *mdev);
1240 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1241 extern void drbd_md_mark_dirty_(struct drbd_conf *mdev,
1242 unsigned int line, const char *func);
1244 extern void drbd_queue_bitmap_io(struct drbd_conf *mdev,
1245 int (*io_fn)(struct drbd_conf *),
1246 void (*done)(struct drbd_conf *, int),
1247 char *why, enum bm_flag flags);
1248 extern int drbd_bitmap_io(struct drbd_conf *mdev,
1249 int (*io_fn)(struct drbd_conf *),
1250 char *why, enum bm_flag flags);
1251 extern int drbd_bmio_set_n_write(struct drbd_conf *mdev);
1252 extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev);
1253 extern void drbd_go_diskless(struct drbd_conf *mdev);
1254 extern void drbd_ldev_destroy(struct drbd_conf *mdev);
1258 We reserve a 128MB Block (4k aligned)
1259 * either at the end of the backing device
1260 * or on a separate meta data device. */
1262 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1263 /* The following numbers are sectors */
1264 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1265 #define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */
1266 /* Allows up to about 3.8TB */
1267 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1269 /* Since the smalles IO unit is usually 512 byte */
1270 #define MD_SECTOR_SHIFT 9
1271 #define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT)
1274 #define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1275 #define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */
1276 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1278 #if BITS_PER_LONG == 32
1280 #define cpu_to_lel(A) cpu_to_le32(A)
1281 #define lel_to_cpu(A) le32_to_cpu(A)
1282 #elif BITS_PER_LONG == 64
1284 #define cpu_to_lel(A) cpu_to_le64(A)
1285 #define lel_to_cpu(A) le64_to_cpu(A)
1287 #error "LN2 of BITS_PER_LONG unknown!"
1291 /* 16MB sized 'bitmap extent' to track syncer usage */
1293 int rs_left; /* number of bits set (out of sync) in this extent. */
1294 int rs_failed; /* number of failed resync requests in this extent. */
1295 unsigned long flags;
1296 struct lc_element lce;
1299 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1300 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1301 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1305 * We need to store one bit for a block.
1306 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1307 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1308 * Bit 1 ==> local node thinks this block needs to be synced.
1311 #define SLEEP_TIME (HZ/10)
1313 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1314 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1315 /* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1316 * per sector of on disk bitmap */
1317 #define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */
1318 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1320 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1321 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1324 /* thus many _storage_ sectors are described by one bit */
1325 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1326 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1327 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1329 /* bit to represented kilo byte conversion */
1330 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1332 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1333 * _storage_ sector is located in */
1334 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1336 /* how much _storage_ sectors we have per bitmap sector */
1337 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1338 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1340 /* in one sector of the bitmap, we have this many activity_log extents. */
1341 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1342 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1344 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1345 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1347 /* the extent in "PER_EXTENT" below is an activity log extent
1348 * we need that many (long words/bytes) to store the bitmap
1349 * of one AL_EXTENT_SIZE chunk of storage.
1350 * we can store the bitmap for that many AL_EXTENTS within
1351 * one sector of the _on_disk_ bitmap:
1352 * bit 0 bit 37 bit 38 bit (512*8)-1
1353 * ...|........|........|.. // ..|........|
1354 * sect. 0 `296 `304 ^(512*8*8)-1
1356 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1357 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1358 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1361 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1362 #define DRBD_MAX_SECTORS_BM \
1363 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1364 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1365 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1366 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1367 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1368 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1369 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1371 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1372 /* 16 TB in units of sectors */
1373 #if BITS_PER_LONG == 32
1374 /* adjust by one page worth of bitmap,
1375 * so we won't wrap around in drbd_bm_find_next_bit.
1376 * you should use 64bit OS for that much storage, anyways. */
1377 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1379 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1380 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1381 /* corresponds to (1UL << 38) bits right now. */
1386 #define DRBD_MAX_BIO_SIZE (1U<<(9+HT_SHIFT))
1387 #define DRBD_MAX_BIO_SIZE_SAFE (1 << 12) /* Works always = 4k */
1389 #define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */
1391 extern int drbd_bm_init(struct drbd_conf *mdev);
1392 extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits);
1393 extern void drbd_bm_cleanup(struct drbd_conf *mdev);
1394 extern void drbd_bm_set_all(struct drbd_conf *mdev);
1395 extern void drbd_bm_clear_all(struct drbd_conf *mdev);
1396 /* set/clear/test only a few bits at a time */
1397 extern int drbd_bm_set_bits(
1398 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1399 extern int drbd_bm_clear_bits(
1400 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1401 extern int drbd_bm_count_bits(
1402 struct drbd_conf *mdev, const unsigned long s, const unsigned long e);
1403 /* bm_set_bits variant for use while holding drbd_bm_lock,
1404 * may process the whole bitmap in one go */
1405 extern void _drbd_bm_set_bits(struct drbd_conf *mdev,
1406 const unsigned long s, const unsigned long e);
1407 extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr);
1408 extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr);
1409 extern int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local);
1410 extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
1411 extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
1412 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev,
1413 unsigned long al_enr);
1414 extern size_t drbd_bm_words(struct drbd_conf *mdev);
1415 extern unsigned long drbd_bm_bits(struct drbd_conf *mdev);
1416 extern sector_t drbd_bm_capacity(struct drbd_conf *mdev);
1418 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1419 extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1420 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1421 extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1422 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo);
1423 extern unsigned long _drbd_bm_total_weight(struct drbd_conf *mdev);
1424 extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev);
1425 extern int drbd_bm_rs_done(struct drbd_conf *mdev);
1426 /* for receive_bitmap */
1427 extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset,
1428 size_t number, unsigned long *buffer);
1429 /* for _drbd_send_bitmap */
1430 extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset,
1431 size_t number, unsigned long *buffer);
1433 extern void drbd_bm_lock(struct drbd_conf *mdev, char *why, enum bm_flag flags);
1434 extern void drbd_bm_unlock(struct drbd_conf *mdev);
1437 extern struct kmem_cache *drbd_request_cache;
1438 extern struct kmem_cache *drbd_ee_cache; /* peer requests */
1439 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1440 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1441 extern mempool_t *drbd_request_mempool;
1442 extern mempool_t *drbd_ee_mempool;
1444 extern struct page *drbd_pp_pool; /* drbd's page pool */
1445 extern spinlock_t drbd_pp_lock;
1446 extern int drbd_pp_vacant;
1447 extern wait_queue_head_t drbd_pp_wait;
1449 extern rwlock_t global_state_lock;
1451 extern struct drbd_conf *drbd_new_device(unsigned int minor);
1452 extern void drbd_free_mdev(struct drbd_conf *mdev);
1454 struct drbd_tconn *drbd_new_tconn(char *name);
1455 extern void drbd_free_tconn(struct drbd_tconn *tconn);
1457 extern int proc_details;
1460 extern int drbd_make_request(struct request_queue *q, struct bio *bio);
1461 extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req);
1462 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1463 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1467 extern void drbd_suspend_io(struct drbd_conf *mdev);
1468 extern void drbd_resume_io(struct drbd_conf *mdev);
1469 extern char *ppsize(char *buf, unsigned long long size);
1470 extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int);
1471 enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
1472 extern enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local);
1473 extern void resync_after_online_grow(struct drbd_conf *);
1474 extern void drbd_reconsider_max_bio_size(struct drbd_conf *mdev);
1475 extern enum drbd_state_rv drbd_set_role(struct drbd_conf *mdev,
1476 enum drbd_role new_role,
1478 extern enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev);
1479 extern void drbd_try_outdate_peer_async(struct drbd_conf *mdev);
1480 extern int drbd_khelper(struct drbd_conf *mdev, char *cmd);
1483 extern int drbd_worker(struct drbd_thread *thi);
1484 extern int drbd_alter_sa(struct drbd_conf *mdev, int na);
1485 extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side);
1486 extern void resume_next_sg(struct drbd_conf *mdev);
1487 extern void suspend_other_sg(struct drbd_conf *mdev);
1488 extern int drbd_resync_finished(struct drbd_conf *mdev);
1489 /* maybe rather drbd_main.c ? */
1490 extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
1491 struct drbd_backing_dev *bdev, sector_t sector, int rw);
1492 extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int);
1493 extern void drbd_rs_controller_reset(struct drbd_conf *mdev);
1495 static inline void ov_oos_print(struct drbd_conf *mdev)
1497 if (mdev->ov_last_oos_size) {
1498 dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n",
1499 (unsigned long long)mdev->ov_last_oos_start,
1500 (unsigned long)mdev->ov_last_oos_size);
1502 mdev->ov_last_oos_size=0;
1506 extern void drbd_csum_bio(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
1507 extern void drbd_csum_ee(struct drbd_conf *, struct crypto_hash *,
1508 struct drbd_peer_request *, void *);
1509 /* worker callbacks */
1510 extern int w_req_cancel_conflict(struct drbd_work *, int);
1511 extern int w_read_retry_remote(struct drbd_work *, int);
1512 extern int w_e_end_data_req(struct drbd_work *, int);
1513 extern int w_e_end_rsdata_req(struct drbd_work *, int);
1514 extern int w_e_end_csum_rs_req(struct drbd_work *, int);
1515 extern int w_e_end_ov_reply(struct drbd_work *, int);
1516 extern int w_e_end_ov_req(struct drbd_work *, int);
1517 extern int w_ov_finished(struct drbd_work *, int);
1518 extern int w_resync_timer(struct drbd_work *, int);
1519 extern int w_resume_next_sg(struct drbd_work *, int);
1520 extern int w_send_write_hint(struct drbd_work *, int);
1521 extern int w_send_dblock(struct drbd_work *, int);
1522 extern int w_send_barrier(struct drbd_work *, int);
1523 extern int w_send_read_req(struct drbd_work *, int);
1524 extern int w_prev_work_done(struct drbd_work *, int);
1525 extern int w_e_reissue(struct drbd_work *, int);
1526 extern int w_restart_disk_io(struct drbd_work *, int);
1527 extern int w_send_oos(struct drbd_work *, int);
1528 extern int w_start_resync(struct drbd_work *, int);
1530 extern void resync_timer_fn(unsigned long data);
1531 extern void start_resync_timer_fn(unsigned long data);
1533 /* drbd_receiver.c */
1534 extern int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector);
1535 extern int drbd_submit_peer_request(struct drbd_conf *,
1536 struct drbd_peer_request *, const unsigned,
1538 extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list);
1539 extern struct drbd_peer_request *drbd_alloc_ee(struct drbd_conf *,
1540 u64, sector_t, unsigned int,
1541 gfp_t) __must_hold(local);
1542 extern void drbd_free_some_ee(struct drbd_conf *, struct drbd_peer_request *,
1544 #define drbd_free_ee(m,e) drbd_free_some_ee(m, e, 0)
1545 #define drbd_free_net_ee(m,e) drbd_free_some_ee(m, e, 1)
1546 extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1547 struct list_head *head);
1548 extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1549 struct list_head *head);
1550 extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled);
1551 extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed);
1552 extern void drbd_flush_workqueue(struct drbd_conf *mdev);
1554 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1555 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1556 static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1557 char __user *optval, int optlen)
1560 if (level == SOL_SOCKET)
1561 err = sock_setsockopt(sock, level, optname, optval, optlen);
1563 err = sock->ops->setsockopt(sock, level, optname, optval,
1568 static inline void drbd_tcp_cork(struct socket *sock)
1571 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1572 (char __user *)&val, sizeof(val));
1575 static inline void drbd_tcp_uncork(struct socket *sock)
1578 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1579 (char __user *)&val, sizeof(val));
1582 static inline void drbd_tcp_nodelay(struct socket *sock)
1585 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1586 (char __user *)&val, sizeof(val));
1589 static inline void drbd_tcp_quickack(struct socket *sock)
1592 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1593 (char __user *)&val, sizeof(val));
1596 void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo);
1599 extern struct proc_dir_entry *drbd_proc;
1600 extern const struct file_operations drbd_proc_fops;
1601 extern const char *drbd_conn_str(enum drbd_conns s);
1602 extern const char *drbd_role_str(enum drbd_role s);
1605 extern void drbd_al_begin_io(struct drbd_conf *mdev, sector_t sector);
1606 extern void drbd_al_complete_io(struct drbd_conf *mdev, sector_t sector);
1607 extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector);
1608 extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1609 extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1610 extern void drbd_rs_cancel_all(struct drbd_conf *mdev);
1611 extern int drbd_rs_del_all(struct drbd_conf *mdev);
1612 extern void drbd_rs_failed_io(struct drbd_conf *mdev,
1613 sector_t sector, int size);
1614 extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *);
1615 extern void drbd_advance_rs_marks(struct drbd_conf *mdev, unsigned long still_to_go);
1616 extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
1617 int size, const char *file, const unsigned int line);
1618 #define drbd_set_in_sync(mdev, sector, size) \
1619 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1620 extern int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
1621 int size, const char *file, const unsigned int line);
1622 #define drbd_set_out_of_sync(mdev, sector, size) \
1623 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1624 extern void drbd_al_apply_to_bm(struct drbd_conf *mdev);
1625 extern void drbd_al_shrink(struct drbd_conf *mdev);
1630 void drbd_nl_cleanup(void);
1631 int __init drbd_nl_init(void);
1632 void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state);
1633 void drbd_bcast_sync_progress(struct drbd_conf *mdev);
1634 void drbd_bcast_ee(struct drbd_conf *, const char *, const int, const char *,
1635 const char *, const struct drbd_peer_request *);
1639 * inline helper functions
1640 *************************/
1642 /* see also page_chain_add and friends in drbd_receiver.c */
1643 static inline struct page *page_chain_next(struct page *page)
1645 return (struct page *)page_private(page);
1647 #define page_chain_for_each(page) \
1648 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1649 page = page_chain_next(page))
1650 #define page_chain_for_each_safe(page, n) \
1651 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1653 static inline int drbd_bio_has_active_page(struct bio *bio)
1655 struct bio_vec *bvec;
1658 __bio_for_each_segment(bvec, bio, i, 0) {
1659 if (page_count(bvec->bv_page) > 1)
1666 static inline int drbd_ee_has_active_page(struct drbd_peer_request *peer_req)
1668 struct page *page = peer_req->pages;
1669 page_chain_for_each(page) {
1670 if (page_count(page) > 1)
1676 static inline enum drbd_state_rv
1677 _drbd_set_state(struct drbd_conf *mdev, union drbd_state ns,
1678 enum chg_state_flags flags, struct completion *done)
1680 enum drbd_state_rv rv;
1682 read_lock(&global_state_lock);
1683 rv = __drbd_set_state(mdev, ns, flags, done);
1684 read_unlock(&global_state_lock);
1689 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1690 static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach, const char *where)
1692 switch (mdev->ldev->dc.on_io_error) {
1695 if (__ratelimit(&drbd_ratelimit_state))
1696 dev_err(DEV, "Local IO failed in %s.\n", where);
1697 if (mdev->state.disk > D_INCONSISTENT)
1698 _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_HARD, NULL);
1701 /* NOTE fall through to detach case if forcedetach set */
1703 case EP_CALL_HELPER:
1704 set_bit(WAS_IO_ERROR, &mdev->flags);
1705 if (mdev->state.disk > D_FAILED) {
1706 _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL);
1708 "Local IO failed in %s. Detaching...\n", where);
1715 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1716 * @mdev: DRBD device.
1717 * @error: Error code passed to the IO completion callback
1718 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1720 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1722 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1723 static inline void drbd_chk_io_error_(struct drbd_conf *mdev,
1724 int error, int forcedetach, const char *where)
1727 unsigned long flags;
1728 spin_lock_irqsave(&mdev->tconn->req_lock, flags);
1729 __drbd_chk_io_error_(mdev, forcedetach, where);
1730 spin_unlock_irqrestore(&mdev->tconn->req_lock, flags);
1736 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1737 * @bdev: Meta data block device.
1739 * BTW, for internal meta data, this happens to be the maximum capacity
1740 * we could agree upon with our peer node.
1742 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1744 switch (bdev->dc.meta_dev_idx) {
1745 case DRBD_MD_INDEX_INTERNAL:
1746 case DRBD_MD_INDEX_FLEX_INT:
1747 return bdev->md.md_offset + bdev->md.bm_offset;
1748 case DRBD_MD_INDEX_FLEX_EXT:
1750 return bdev->md.md_offset;
1755 * drbd_md_last_sector() - Return the last sector number of the meta data area
1756 * @bdev: Meta data block device.
1758 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1760 switch (bdev->dc.meta_dev_idx) {
1761 case DRBD_MD_INDEX_INTERNAL:
1762 case DRBD_MD_INDEX_FLEX_INT:
1763 return bdev->md.md_offset + MD_AL_OFFSET - 1;
1764 case DRBD_MD_INDEX_FLEX_EXT:
1766 return bdev->md.md_offset + bdev->md.md_size_sect;
1770 /* Returns the number of 512 byte sectors of the device */
1771 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1773 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1774 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0;
1778 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1779 * @bdev: Meta data block device.
1781 * returns the capacity we announce to out peer. we clip ourselves at the
1782 * various MAX_SECTORS, because if we don't, current implementation will
1783 * oops sooner or later
1785 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1788 switch (bdev->dc.meta_dev_idx) {
1789 case DRBD_MD_INDEX_INTERNAL:
1790 case DRBD_MD_INDEX_FLEX_INT:
1791 s = drbd_get_capacity(bdev->backing_bdev)
1792 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1793 drbd_md_first_sector(bdev))
1796 case DRBD_MD_INDEX_FLEX_EXT:
1797 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1798 drbd_get_capacity(bdev->backing_bdev));
1799 /* clip at maximum size the meta device can support */
1800 s = min_t(sector_t, s,
1801 BM_EXT_TO_SECT(bdev->md.md_size_sect
1802 - bdev->md.bm_offset));
1805 s = min_t(sector_t, DRBD_MAX_SECTORS,
1806 drbd_get_capacity(bdev->backing_bdev));
1812 * drbd_md_ss__() - Return the sector number of our meta data super block
1813 * @mdev: DRBD device.
1814 * @bdev: Meta data block device.
1816 static inline sector_t drbd_md_ss__(struct drbd_conf *mdev,
1817 struct drbd_backing_dev *bdev)
1819 switch (bdev->dc.meta_dev_idx) {
1820 default: /* external, some index */
1821 return MD_RESERVED_SECT * bdev->dc.meta_dev_idx;
1822 case DRBD_MD_INDEX_INTERNAL:
1823 /* with drbd08, internal meta data is always "flexible" */
1824 case DRBD_MD_INDEX_FLEX_INT:
1825 /* sizeof(struct md_on_disk_07) == 4k
1826 * position: last 4k aligned block of 4k size */
1827 if (!bdev->backing_bdev) {
1828 if (__ratelimit(&drbd_ratelimit_state)) {
1829 dev_err(DEV, "bdev->backing_bdev==NULL\n");
1834 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL)
1836 case DRBD_MD_INDEX_FLEX_EXT:
1842 drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w)
1844 unsigned long flags;
1845 spin_lock_irqsave(&q->q_lock, flags);
1846 list_add(&w->list, &q->q);
1847 up(&q->s); /* within the spinlock,
1848 see comment near end of drbd_worker() */
1849 spin_unlock_irqrestore(&q->q_lock, flags);
1853 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1855 unsigned long flags;
1856 spin_lock_irqsave(&q->q_lock, flags);
1857 list_add_tail(&w->list, &q->q);
1858 up(&q->s); /* within the spinlock,
1859 see comment near end of drbd_worker() */
1860 spin_unlock_irqrestore(&q->q_lock, flags);
1863 static inline void wake_asender(struct drbd_tconn *tconn)
1865 if (test_bit(SIGNAL_ASENDER, &tconn->flags))
1866 force_sig(DRBD_SIG, tconn->asender.task);
1869 static inline void request_ping(struct drbd_tconn *tconn)
1871 set_bit(SEND_PING, &tconn->flags);
1872 wake_asender(tconn);
1875 static inline int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock,
1876 enum drbd_packet cmd, struct p_header *h, size_t size,
1879 return _conn_send_cmd(mdev->tconn, mdev->vnr, sock, cmd, h, size, msg_flags);
1882 static inline int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket,
1883 enum drbd_packet cmd, struct p_header *h, size_t size)
1885 return conn_send_cmd(mdev->tconn, mdev->vnr, use_data_socket, cmd, h, size);
1888 static inline int drbd_send_short_cmd(struct drbd_conf *mdev,
1889 enum drbd_packet cmd)
1892 return drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, &h, sizeof(h));
1895 static inline int drbd_send_ping(struct drbd_tconn *tconn)
1898 return conn_send_cmd(tconn, 0, USE_META_SOCKET, P_PING, &h, sizeof(h));
1901 static inline int drbd_send_ping_ack(struct drbd_tconn *tconn)
1904 return conn_send_cmd(tconn, 0, USE_META_SOCKET, P_PING_ACK, &h, sizeof(h));
1907 static inline int drbd_send_state_req(struct drbd_conf *mdev,
1908 union drbd_state mask, union drbd_state val)
1910 return _conn_send_state_req(mdev->tconn, mdev->vnr, P_STATE_CHG_REQ, mask, val);
1913 static inline int conn_send_state_req(struct drbd_tconn *tconn,
1914 union drbd_state mask, union drbd_state val)
1916 enum drbd_packet cmd = tconn->agreed_pro_version < 100 ? P_STATE_CHG_REQ : P_CONN_ST_CHG_REQ;
1917 return _conn_send_state_req(tconn, 0, cmd, mask, val);
1920 static inline void drbd_thread_stop(struct drbd_thread *thi)
1922 _drbd_thread_stop(thi, false, true);
1925 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
1927 _drbd_thread_stop(thi, false, false);
1930 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
1932 _drbd_thread_stop(thi, true, false);
1935 /* counts how many answer packets packets we expect from our peer,
1936 * for either explicit application requests,
1937 * or implicit barrier packets as necessary.
1940 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1941 * it is much easier and equally valid to count what we queue for the
1942 * worker, even before it actually was queued or send.
1943 * (drbd_make_request_common; recovery path on read io-error)
1945 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1946 * _req_mod(req, DATA_RECEIVED)
1947 * [from receive_DataReply]
1948 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1949 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1950 * for some reason it is NOT decreased in got_NegAck,
1951 * but in the resulting cleanup code from report_params.
1952 * we should try to remember the reason for that...
1953 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
1954 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
1955 * [from tl_clear_barrier]
1957 static inline void inc_ap_pending(struct drbd_conf *mdev)
1959 atomic_inc(&mdev->ap_pending_cnt);
1962 #define ERR_IF_CNT_IS_NEGATIVE(which) \
1963 if (atomic_read(&mdev->which) < 0) \
1964 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
1965 __func__ , __LINE__ , \
1966 atomic_read(&mdev->which))
1968 #define dec_ap_pending(mdev) do { \
1969 typecheck(struct drbd_conf *, mdev); \
1970 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \
1971 wake_up(&mdev->misc_wait); \
1972 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
1974 /* counts how many resync-related answers we still expect from the peer
1976 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1977 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
1978 * (or P_NEG_ACK with ID_SYNCER)
1980 static inline void inc_rs_pending(struct drbd_conf *mdev)
1982 atomic_inc(&mdev->rs_pending_cnt);
1985 #define dec_rs_pending(mdev) do { \
1986 typecheck(struct drbd_conf *, mdev); \
1987 atomic_dec(&mdev->rs_pending_cnt); \
1988 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
1990 /* counts how many answers we still need to send to the peer.
1992 * receive_Data unless protocol A;
1993 * we need to send a P_RECV_ACK (proto B)
1994 * or P_WRITE_ACK (proto C)
1995 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1996 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1997 * receive_Barrier_* we need to send a P_BARRIER_ACK
1999 static inline void inc_unacked(struct drbd_conf *mdev)
2001 atomic_inc(&mdev->unacked_cnt);
2004 #define dec_unacked(mdev) do { \
2005 typecheck(struct drbd_conf *, mdev); \
2006 atomic_dec(&mdev->unacked_cnt); \
2007 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2009 #define sub_unacked(mdev, n) do { \
2010 typecheck(struct drbd_conf *, mdev); \
2011 atomic_sub(n, &mdev->unacked_cnt); \
2012 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2015 static inline void put_net_conf(struct drbd_tconn *tconn)
2017 if (atomic_dec_and_test(&tconn->net_cnt))
2018 wake_up(&tconn->net_cnt_wait);
2022 * get_net_conf() - Increase ref count on mdev->tconn->net_conf; Returns 0 if nothing there
2023 * @mdev: DRBD device.
2025 * You have to call put_net_conf() when finished working with mdev->tconn->net_conf.
2027 static inline int get_net_conf(struct drbd_tconn *tconn)
2031 atomic_inc(&tconn->net_cnt);
2032 have_net_conf = tconn->cstate >= C_UNCONNECTED;
2034 put_net_conf(tconn);
2035 return have_net_conf;
2039 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2042 * You have to call put_ldev() when finished working with mdev->ldev.
2044 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2045 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2047 static inline void put_ldev(struct drbd_conf *mdev)
2049 int i = atomic_dec_return(&mdev->local_cnt);
2051 /* This may be called from some endio handler,
2052 * so we must not sleep here. */
2057 if (mdev->state.disk == D_DISKLESS)
2058 /* even internal references gone, safe to destroy */
2059 drbd_ldev_destroy(mdev);
2060 if (mdev->state.disk == D_FAILED)
2061 /* all application IO references gone. */
2062 drbd_go_diskless(mdev);
2063 wake_up(&mdev->misc_wait);
2068 static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
2072 /* never get a reference while D_DISKLESS */
2073 if (mdev->state.disk == D_DISKLESS)
2076 atomic_inc(&mdev->local_cnt);
2077 io_allowed = (mdev->state.disk >= mins);
2083 extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins);
2086 /* you must have an "get_ldev" reference */
2087 static inline void drbd_get_syncer_progress(struct drbd_conf *mdev,
2088 unsigned long *bits_left, unsigned int *per_mil_done)
2090 /* this is to break it at compile time when we change that, in case we
2091 * want to support more than (1<<32) bits on a 32bit arch. */
2092 typecheck(unsigned long, mdev->rs_total);
2094 /* note: both rs_total and rs_left are in bits, i.e. in
2095 * units of BM_BLOCK_SIZE.
2096 * for the percentage, we don't care. */
2098 if (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T)
2099 *bits_left = mdev->ov_left;
2101 *bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
2102 /* >> 10 to prevent overflow,
2103 * +1 to prevent division by zero */
2104 if (*bits_left > mdev->rs_total) {
2105 /* doh. maybe a logic bug somewhere.
2106 * may also be just a race condition
2107 * between this and a disconnect during sync.
2108 * for now, just prevent in-kernel buffer overflow.
2111 dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2112 drbd_conn_str(mdev->state.conn),
2113 *bits_left, mdev->rs_total, mdev->rs_failed);
2116 /* Make sure the division happens in long context.
2117 * We allow up to one petabyte storage right now,
2118 * at a granularity of 4k per bit that is 2**38 bits.
2119 * After shift right and multiplication by 1000,
2120 * this should still fit easily into a 32bit long,
2121 * so we don't need a 64bit division on 32bit arch.
2122 * Note: currently we don't support such large bitmaps on 32bit
2123 * arch anyways, but no harm done to be prepared for it here.
2125 unsigned int shift = mdev->rs_total >= (1ULL << 32) ? 16 : 10;
2126 unsigned long left = *bits_left >> shift;
2127 unsigned long total = 1UL + (mdev->rs_total >> shift);
2128 unsigned long tmp = 1000UL - left * 1000UL/total;
2129 *per_mil_done = tmp;
2134 /* this throttles on-the-fly application requests
2135 * according to max_buffers settings;
2136 * maybe re-implement using semaphores? */
2137 static inline int drbd_get_max_buffers(struct drbd_conf *mdev)
2139 int mxb = 1000000; /* arbitrary limit on open requests */
2140 if (get_net_conf(mdev->tconn)) {
2141 mxb = mdev->tconn->net_conf->max_buffers;
2142 put_net_conf(mdev->tconn);
2147 static inline int drbd_state_is_stable(struct drbd_conf *mdev)
2149 union drbd_state s = mdev->state;
2151 /* DO NOT add a default clause, we want the compiler to warn us
2152 * for any newly introduced state we may have forgotten to add here */
2154 switch ((enum drbd_conns)s.conn) {
2155 /* new io only accepted when there is no connection, ... */
2157 case C_WF_CONNECTION:
2158 /* ... or there is a well established connection. */
2164 case C_PAUSED_SYNC_S:
2165 case C_PAUSED_SYNC_T:
2168 /* transitional states, IO allowed */
2169 case C_DISCONNECTING:
2173 case C_NETWORK_FAILURE:
2174 case C_PROTOCOL_ERROR:
2176 case C_WF_REPORT_PARAMS:
2177 case C_STARTING_SYNC_S:
2178 case C_STARTING_SYNC_T:
2181 /* Allow IO in BM exchange states with new protocols */
2183 if (mdev->tconn->agreed_pro_version < 96)
2187 /* no new io accepted in these states */
2189 case C_WF_SYNC_UUID:
2195 switch ((enum drbd_disk_state)s.disk) {
2197 case D_INCONSISTENT:
2201 /* disk state is stable as well. */
2204 /* no new io accepted during tansitional states */
2217 static inline int is_susp(union drbd_state s)
2219 return s.susp || s.susp_nod || s.susp_fen;
2222 static inline bool may_inc_ap_bio(struct drbd_conf *mdev)
2224 int mxb = drbd_get_max_buffers(mdev);
2226 if (is_susp(mdev->state))
2228 if (test_bit(SUSPEND_IO, &mdev->flags))
2231 /* to avoid potential deadlock or bitmap corruption,
2232 * in various places, we only allow new application io
2233 * to start during "stable" states. */
2235 /* no new io accepted when attaching or detaching the disk */
2236 if (!drbd_state_is_stable(mdev))
2239 /* since some older kernels don't have atomic_add_unless,
2240 * and we are within the spinlock anyways, we have this workaround. */
2241 if (atomic_read(&mdev->ap_bio_cnt) > mxb)
2243 if (test_bit(BITMAP_IO, &mdev->flags))
2248 static inline bool inc_ap_bio_cond(struct drbd_conf *mdev, int count)
2252 spin_lock_irq(&mdev->tconn->req_lock);
2253 rv = may_inc_ap_bio(mdev);
2255 atomic_add(count, &mdev->ap_bio_cnt);
2256 spin_unlock_irq(&mdev->tconn->req_lock);
2261 static inline void inc_ap_bio(struct drbd_conf *mdev, int count)
2264 * as long as the device is suspended
2265 * until the bitmap is no longer on the fly during connection
2266 * handshake as long as we would exeed the max_buffer limit.
2268 * to avoid races with the reconnect code,
2269 * we need to atomic_inc within the spinlock. */
2271 wait_event(mdev->misc_wait, inc_ap_bio_cond(mdev, count));
2274 static inline void dec_ap_bio(struct drbd_conf *mdev)
2276 int mxb = drbd_get_max_buffers(mdev);
2277 int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt);
2279 D_ASSERT(ap_bio >= 0);
2280 /* this currently does wake_up for every dec_ap_bio!
2281 * maybe rather introduce some type of hysteresis?
2282 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2284 wake_up(&mdev->misc_wait);
2285 if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) {
2286 if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags))
2287 drbd_queue_work(&mdev->tconn->data.work, &mdev->bm_io_work.w);
2291 static inline int drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val)
2293 int changed = mdev->ed_uuid != val;
2294 mdev->ed_uuid = val;
2298 static inline int drbd_queue_order_type(struct drbd_conf *mdev)
2300 /* sorry, we currently have no working implementation
2301 * of distributed TCQ stuff */
2302 #ifndef QUEUE_ORDERED_NONE
2303 #define QUEUE_ORDERED_NONE 0
2305 return QUEUE_ORDERED_NONE;
2308 static inline void drbd_md_flush(struct drbd_conf *mdev)
2312 if (test_bit(MD_NO_FUA, &mdev->flags))
2315 r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL);
2317 set_bit(MD_NO_FUA, &mdev->flags);
2318 dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r);