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_genl_api.h>
48 #include <linux/drbd.h>
49 #include "drbd_state.h"
52 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
53 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
54 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
55 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
57 # define __protected_by(x)
58 # define __protected_read_by(x)
59 # define __protected_write_by(x)
60 # define __must_hold(x)
63 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
65 /* module parameter, defined in drbd_main.c */
66 extern unsigned int minor_count;
67 extern int disable_sendpage;
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 extern struct ratelimit_state drbd_ratelimit_state;
168 extern struct idr minors; /* RCU, updates: genl_lock() */
169 extern struct list_head drbd_tconns; /* RCU, updates: genl_lock() */
173 /* receiver (data socket) */
175 P_DATA_REPLY = 0x01, /* Response to P_DATA_REQUEST */
176 P_RS_DATA_REPLY = 0x02, /* Response to P_RS_DATA_REQUEST */
179 P_BECOME_SYNC_TARGET = 0x05,
180 P_BECOME_SYNC_SOURCE = 0x06,
181 P_UNPLUG_REMOTE = 0x07, /* Used at various times to hint the peer */
182 P_DATA_REQUEST = 0x08, /* Used to ask for a data block */
183 P_RS_DATA_REQUEST = 0x09, /* Used to ask for a data block for resync */
190 P_AUTH_CHALLENGE = 0x10,
191 P_AUTH_RESPONSE = 0x11,
192 P_STATE_CHG_REQ = 0x12,
194 /* asender (meta socket */
197 P_RECV_ACK = 0x15, /* Used in protocol B */
198 P_WRITE_ACK = 0x16, /* Used in protocol C */
199 P_RS_WRITE_ACK = 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
200 P_SUPERSEDED = 0x18, /* Used in proto C, two-primaries conflict detection */
201 P_NEG_ACK = 0x19, /* Sent if local disk is unusable */
202 P_NEG_DREPLY = 0x1a, /* Local disk is broken... */
203 P_NEG_RS_DREPLY = 0x1b, /* Local disk is broken... */
204 P_BARRIER_ACK = 0x1c,
205 P_STATE_CHG_REPLY = 0x1d,
207 /* "new" commands, no longer fitting into the ordering scheme above */
209 P_OV_REQUEST = 0x1e, /* data socket */
211 P_OV_RESULT = 0x20, /* meta socket */
212 P_CSUM_RS_REQUEST = 0x21, /* data socket */
213 P_RS_IS_IN_SYNC = 0x22, /* meta socket */
214 P_SYNC_PARAM89 = 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
215 P_COMPRESSED_BITMAP = 0x24, /* compressed or otherwise encoded bitmap transfer */
216 /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */
217 /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */
218 P_DELAY_PROBE = 0x27, /* is used on BOTH sockets */
219 P_OUT_OF_SYNC = 0x28, /* Mark as out of sync (Outrunning), data socket */
220 P_RS_CANCEL = 0x29, /* meta: Used to cancel RS_DATA_REQUEST packet by SyncSource */
221 P_CONN_ST_CHG_REQ = 0x2a, /* data sock: Connection wide state request */
222 P_CONN_ST_CHG_REPLY = 0x2b, /* meta sock: Connection side state req reply */
223 P_RETRY_WRITE = 0x2c, /* Protocol C: retry conflicting write request */
224 P_PROTOCOL_UPDATE = 0x2d, /* data sock: is used in established connections */
226 P_MAY_IGNORE = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
227 P_MAX_OPT_CMD = 0x101,
229 /* special command ids for handshake */
231 P_INITIAL_META = 0xfff1, /* First Packet on the MetaSock */
232 P_INITIAL_DATA = 0xfff2, /* First Packet on the Socket */
234 P_CONNECTION_FEATURES = 0xfffe /* FIXED for the next century! */
237 extern const char *cmdname(enum drbd_packet cmd);
239 /* for sending/receiving the bitmap,
240 * possibly in some encoding scheme */
243 * stores total bits and long words
244 * of the bitmap, so we don't need to
245 * call the accessor functions over and again. */
246 unsigned long bm_bits;
247 unsigned long bm_words;
248 /* during xfer, current position within the bitmap */
249 unsigned long bit_offset;
250 unsigned long word_offset;
252 /* statistics; index: (h->command == P_BITMAP) */
257 extern void INFO_bm_xfer_stats(struct drbd_conf *mdev,
258 const char *direction, struct bm_xfer_ctx *c);
260 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
262 /* word_offset counts "native long words" (32 or 64 bit),
264 * Encoded packet may end at an unaligned bit offset.
265 * In case a fallback clear text packet is transmitted in
266 * between, we adjust this offset back to the last 64bit
267 * aligned "native long word", which makes coding and decoding
268 * the plain text bitmap much more convenient. */
269 #if BITS_PER_LONG == 64
270 c->word_offset = c->bit_offset >> 6;
271 #elif BITS_PER_LONG == 32
272 c->word_offset = c->bit_offset >> 5;
273 c->word_offset &= ~(1UL);
275 # error "unsupported BITS_PER_LONG"
280 #define __packed __attribute__((packed))
283 /* This is the layout for a packet on the wire.
284 * The byteorder is the network byte order.
285 * (except block_id and barrier fields.
286 * these are pointers to local structs
287 * and have no relevance for the partner,
288 * which just echoes them as received.)
290 * NOTE that the payload starts at a long aligned offset,
291 * regardless of 32 or 64 bit arch!
296 u16 length; /* bytes of data after this header */
299 /* Header for big packets, Used for data packets exceeding 64kB */
301 u16 magic; /* use DRBD_MAGIC_BIG here */
314 extern unsigned int drbd_header_size(struct drbd_tconn *tconn);
316 /* these defines must not be changed without changing the protocol version */
317 #define DP_HARDBARRIER 1 /* depricated */
318 #define DP_RW_SYNC 2 /* equals REQ_SYNC */
319 #define DP_MAY_SET_IN_SYNC 4
320 #define DP_UNPLUG 8 /* not used anymore */
321 #define DP_FUA 16 /* equals REQ_FUA */
322 #define DP_FLUSH 32 /* equals REQ_FLUSH */
323 #define DP_DISCARD 64 /* equals REQ_DISCARD */
324 #define DP_SEND_RECEIVE_ACK 128 /* This is a proto B write request */
325 #define DP_SEND_WRITE_ACK 256 /* This is a proto C write request */
328 u64 sector; /* 64 bits sector number */
329 u64 block_id; /* to identify the request in protocol B&C */
335 * commands which share a struct:
337 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
338 * P_SUPERSEDED (proto C, two-primaries conflict detection)
340 * P_DATA_REQUEST, P_RS_DATA_REQUEST
353 u32 pad; /* to multiple of 8 Byte */
357 * commands with their own struct for additional fields:
358 * P_CONNECTION_FEATURES
365 struct p_connection_features {
370 /* should be more than enough for future enhancements
371 * for now, feature_flags and the reserved array shall be zero.
379 u32 barrier; /* barrier number _handle_ only */
380 u32 pad; /* to multiple of 8 Byte */
383 struct p_barrier_ack {
391 /* Since protocol version 88 and higher. */
395 struct p_rs_param_89 {
397 /* protocol version 89: */
398 char verify_alg[SHARED_SECRET_MAX];
399 char csums_alg[SHARED_SECRET_MAX];
402 struct p_rs_param_95 {
404 char verify_alg[SHARED_SECRET_MAX];
405 char csums_alg[SHARED_SECRET_MAX];
412 enum drbd_conn_flags {
413 CF_DISCARD_MY_DATA = 1,
425 /* Since protocol version 87 and higher. */
426 char integrity_alg[0];
431 u64 uuid[UI_EXTENDED_SIZE];
439 u64 d_size; /* size of disk */
440 u64 u_size; /* user requested size */
441 u64 c_size; /* current exported size */
442 u32 max_bio_size; /* Maximal size of a BIO */
443 u16 queue_order_type; /* not yet implemented in DRBD*/
444 u16 dds_flags; /* use enum dds_flags here. */
456 struct p_req_state_reply {
460 struct p_drbd06_param {
470 struct p_block_desc {
473 u32 pad; /* to multiple of 8 Byte */
476 /* Valid values for the encoding field.
477 * Bump proto version when changing this. */
478 enum drbd_bitmap_code {
479 /* RLE_VLI_Bytes = 0,
480 * and other bit variants had been defined during
481 * algorithm evaluation. */
485 struct p_compressed_bm {
486 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
487 * (encoding & 0x80): polarity (set/unset) of first runlength
488 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
489 * used to pad up to head.length bytes
496 struct p_delay_probe93 {
497 u32 seq_num; /* sequence number to match the two probe packets */
498 u32 offset; /* usecs the probe got sent after the reference time point */
502 * Bitmap packets need to fit within a single page on the sender and receiver,
503 * so we are limited to 4 KiB (and not to PAGE_SIZE, which can be bigger).
505 #define DRBD_SOCKET_BUFFER_SIZE 4096
507 /**********************************************************************/
508 enum drbd_thread_state {
517 struct task_struct *task;
518 struct completion stop;
519 enum drbd_thread_state t_state;
520 int (*function) (struct drbd_thread *);
521 struct drbd_tconn *tconn;
526 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
528 /* THINK testing the t_state seems to be uncritical in all cases
529 * (but thread_{start,stop}), so we can read it *without* the lock.
537 struct list_head list;
538 int (*cb)(struct drbd_work *, int cancel);
540 struct drbd_conf *mdev;
541 struct drbd_tconn *tconn;
545 #include "drbd_interval.h"
547 extern int drbd_wait_misc(struct drbd_conf *, struct drbd_interval *);
549 struct drbd_request {
552 /* if local IO is not allowed, will be NULL.
553 * if local IO _is_ allowed, holds the locally submitted bio clone,
554 * or, after local IO completion, the ERR_PTR(error).
555 * see drbd_request_endio(). */
556 struct bio *private_bio;
558 struct drbd_interval i;
560 /* epoch: used to check on "completion" whether this req was in
561 * the current epoch, and we therefore have to close it,
562 * causing a p_barrier packet to be send, starting a new epoch.
564 * This corresponds to "barrier" in struct p_barrier[_ack],
565 * and to "barrier_nr" in struct drbd_epoch (and various
566 * comments/function parameters/local variable names).
570 struct list_head tl_requests; /* ring list in the transfer log */
571 struct bio *master_bio; /* master bio pointer */
572 unsigned long start_time;
574 /* once it hits 0, we may complete the master_bio */
575 atomic_t completion_ref;
576 /* once it hits 0, we may destroy this drbd_request object */
579 unsigned rq_state; /* see comments above _req_mod() */
583 struct drbd_tconn *tconn;
584 struct list_head list;
585 unsigned int barrier_nr;
586 atomic_t epoch_size; /* increased on every request added. */
587 atomic_t active; /* increased on every req. added, and dec on every finished. */
591 /* drbd_epoch flag bits */
593 DE_HAVE_BARRIER_NUMBER,
600 EV_CLEANUP = 32, /* used as flag */
603 struct drbd_wq_barrier {
605 struct completion done;
613 struct drbd_peer_request {
615 struct drbd_epoch *epoch; /* for writes */
617 atomic_t pending_bios;
618 struct drbd_interval i;
619 /* see comments on ee flag bits below */
623 struct digest_info *digest;
628 * While corresponding bios are in flight, the only modification will be
629 * set_bit WAS_ERROR, which has to be atomic.
630 * If no bios are in flight yet, or all have been completed,
631 * non-atomic modification to ee->flags is ok.
634 __EE_CALL_AL_COMPLETE_IO,
635 __EE_MAY_SET_IN_SYNC,
637 /* In case a barrier failed,
638 * we need to resubmit without the barrier flag. */
641 /* we may have several bios per peer request.
642 * if any of those fail, we set this flag atomically
643 * from the endio callback */
646 /* This ee has a pointer to a digest instead of a block id */
649 /* Conflicting local requests need to be restarted after this request */
650 __EE_RESTART_REQUESTS,
652 /* The peer wants a write ACK for this (wire proto C) */
655 /* Is set when net_conf had two_primaries set while creating this peer_req */
656 __EE_IN_INTERVAL_TREE,
658 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
659 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
660 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
661 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
662 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
663 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
664 #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK)
665 #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE)
667 /* flag bits per mdev */
669 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
670 MD_DIRTY, /* current uuids and flags not yet on disk */
671 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
674 CRASHED_PRIMARY, /* This node was a crashed primary.
675 * Gets cleared when the state.conn
676 * goes into C_CONNECTED state. */
679 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */
680 SUSPEND_IO, /* suspend application io */
681 BITMAP_IO, /* suspend application io;
682 once no more io in flight, start bitmap io */
683 BITMAP_IO_QUEUED, /* Started bitmap IO */
684 GO_DISKLESS, /* Disk is being detached, on io-error or admin request. */
685 WAS_IO_ERROR, /* Local disk failed returned IO error */
686 FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */
687 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
688 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
689 * the peer, if it changed there as well. */
690 NEW_CUR_UUID, /* Create new current UUID when thawing IO */
691 AL_SUSPENDED, /* Activity logging is currently suspended. */
692 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */
693 B_RS_H_DONE, /* Before resync handler done (already executed) */
694 DISCARD_MY_DATA, /* discard_my_data flag per volume */
698 struct drbd_bitmap; /* opaque for drbd_conf */
700 /* definition of bits in bm_flags to be used in drbd_bm_lock
701 * and drbd_bitmap_io and friends. */
703 /* do we need to kfree, or vfree bm_pages? */
704 BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */
706 /* currently locked for bulk operation */
707 BM_LOCKED_MASK = 0xf,
709 /* in detail, that is: */
714 /* so we can mark it locked for bulk operation,
715 * and still allow all non-bulk operations */
718 /* (test bit, count bit) allowed (common case) */
719 BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED,
721 /* testing bits, as well as setting new bits allowed, but clearing bits
722 * would be unexpected. Used during bitmap receive. Setting new bits
723 * requires sending of "out-of-sync" information, though. */
724 BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED,
726 /* for drbd_bm_write_copy_pages, everything is allowed,
727 * only concurrent bulk operations are locked out. */
728 BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED,
731 struct drbd_work_queue {
733 spinlock_t q_lock; /* to protect the list. */
734 wait_queue_head_t q_wait;
739 struct socket *socket;
740 /* this way we get our
741 * send/receive buffers off the stack */
747 u64 md_offset; /* sector offset to 'super' block */
749 u64 la_size_sect; /* last agreed size, unit sectors */
750 spinlock_t uuid_lock;
756 s32 al_offset; /* signed relative sector offset to al area */
757 s32 bm_offset; /* signed relative sector offset to bitmap */
759 /* u32 al_nr_extents; important for restoring the AL
760 * is stored into ldev->dc.al_extents, which in turn
761 * gets applied to act_log->nr_elements
765 struct drbd_backing_dev {
766 struct block_device *backing_bdev;
767 struct block_device *md_bdev;
769 struct disk_conf *disk_conf; /* RCU, for updates: mdev->tconn->conf_update */
770 sector_t known_size; /* last known size of that backing device */
782 int (*io_fn)(struct drbd_conf *mdev);
783 void (*done)(struct drbd_conf *mdev, int rv);
786 enum write_ordering_e {
793 unsigned int head_index;
795 int total; /* sum of all values */
798 extern struct fifo_buffer *fifo_alloc(int fifo_size);
800 /* flag bits per tconn */
802 NET_CONGESTED, /* The data socket is congested */
803 RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */
804 SEND_PING, /* whether asender should send a ping asap */
805 SIGNAL_ASENDER, /* whether asender wants to be interrupted */
806 GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */
807 CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */
810 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
811 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */
812 STATE_SENT, /* Do not change state/UUIDs while this is set */
813 CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
814 * pending, from drbd worker context.
815 * If set, bdi_write_congested() returns true,
816 * so shrink_page_list() would not recurse into,
817 * and potentially deadlock on, this drbd worker.
821 struct drbd_tconn { /* is a resource from the config file */
822 char *name; /* Resource name */
823 struct list_head all_tconn; /* linked on global drbd_tconns */
825 struct idr volumes; /* <tconn, vnr> to mdev mapping */
826 enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
827 unsigned susp:1; /* IO suspended by user */
828 unsigned susp_nod:1; /* IO suspended because no data */
829 unsigned susp_fen:1; /* IO suspended because fence peer handler runs */
830 struct mutex cstate_mutex; /* Protects graceful disconnects */
833 struct net_conf *net_conf; /* content protected by rcu */
834 struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */
835 wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */
836 struct res_opts res_opts;
838 struct sockaddr_storage my_addr;
840 struct sockaddr_storage peer_addr;
843 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
844 struct drbd_socket meta; /* ping/ack (metadata) packets */
845 int agreed_pro_version; /* actually used protocol version */
846 unsigned long last_received; /* in jiffies, either socket */
847 unsigned int ko_count;
851 struct list_head transfer_log; /* all requests not yet fully processed */
853 struct crypto_hash *cram_hmac_tfm;
854 struct crypto_hash *integrity_tfm; /* checksums we compute, updates protected by tconn->data->mutex */
855 struct crypto_hash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */
856 struct crypto_hash *csums_tfm;
857 struct crypto_hash *verify_tfm;
862 struct drbd_epoch *current_epoch;
863 spinlock_t epoch_lock;
865 enum write_ordering_e write_ordering;
866 atomic_t current_tle_nr; /* transfer log epoch number */
867 unsigned current_tle_writes; /* writes seen within this tl epoch */
869 unsigned long last_reconnect_jif;
870 struct drbd_thread receiver;
871 struct drbd_thread worker;
872 struct drbd_thread asender;
873 cpumask_var_t cpu_mask;
876 struct drbd_work_queue sender_work;
879 /* whether this sender thread
880 * has processed a single write yet. */
881 bool seen_any_write_yet;
883 /* Which barrier number to send with the next P_BARRIER */
884 int current_epoch_nr;
886 /* how many write requests have been sent
887 * with req->epoch == current_epoch_nr.
888 * If none, no P_BARRIER will be sent. */
889 unsigned current_epoch_writes;
894 struct drbd_tconn *tconn;
895 int vnr; /* volume number within the connection */
898 /* things that are stored as / read from meta data on disk */
901 /* configured by drbdsetup */
902 struct drbd_backing_dev *ldev __protected_by(local);
904 sector_t p_size; /* partner's disk size */
905 struct request_queue *rq_queue;
906 struct block_device *this_bdev;
907 struct gendisk *vdisk;
909 unsigned long last_reattach_jif;
910 struct drbd_work resync_work,
915 struct timer_list resync_timer;
916 struct timer_list md_sync_timer;
917 struct timer_list start_resync_timer;
918 struct timer_list request_timer;
919 #ifdef DRBD_DEBUG_MD_SYNC
923 } last_md_mark_dirty;
926 /* Used after attach while negotiating new disk state. */
927 union drbd_state new_state_tmp;
929 union drbd_dev_state state;
930 wait_queue_head_t misc_wait;
931 wait_queue_head_t state_wait; /* upon each state change. */
932 unsigned int send_cnt;
933 unsigned int recv_cnt;
934 unsigned int read_cnt;
935 unsigned int writ_cnt;
936 unsigned int al_writ_cnt;
937 unsigned int bm_writ_cnt;
938 atomic_t ap_bio_cnt; /* Requests we need to complete */
939 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
940 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
941 atomic_t unacked_cnt; /* Need to send replies for */
942 atomic_t local_cnt; /* Waiting for local completion */
944 /* Interval tree of pending local requests */
945 struct rb_root read_requests;
946 struct rb_root write_requests;
948 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
949 unsigned long rs_total;
950 /* number of resync blocks that failed in this run */
951 unsigned long rs_failed;
952 /* Syncer's start time [unit jiffies] */
953 unsigned long rs_start;
954 /* cumulated time in PausedSyncX state [unit jiffies] */
955 unsigned long rs_paused;
956 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
957 unsigned long rs_same_csum;
958 #define DRBD_SYNC_MARKS 8
959 #define DRBD_SYNC_MARK_STEP (3*HZ)
960 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
961 unsigned long rs_mark_left[DRBD_SYNC_MARKS];
962 /* marks's time [unit jiffies] */
963 unsigned long rs_mark_time[DRBD_SYNC_MARKS];
964 /* current index into rs_mark_{left,time} */
967 /* where does the admin want us to start? (sector) */
968 sector_t ov_start_sector;
969 sector_t ov_stop_sector;
970 /* where are we now? (sector) */
971 sector_t ov_position;
972 /* Start sector of out of sync range (to merge printk reporting). */
973 sector_t ov_last_oos_start;
974 /* size of out-of-sync range in sectors. */
975 sector_t ov_last_oos_size;
976 unsigned long ov_left; /* in bits */
978 struct drbd_bitmap *bitmap;
979 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
981 /* Used to track operations of resync... */
982 struct lru_cache *resync;
983 /* Number of locked elements in resync LRU */
984 unsigned int resync_locked;
985 /* resync extent number waiting for application requests */
986 unsigned int resync_wenr;
991 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
992 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
993 struct list_head done_ee; /* need to send P_WRITE_ACK */
994 struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */
995 struct list_head net_ee; /* zero-copy network send in progress */
998 struct list_head resync_reads;
999 atomic_t pp_in_use; /* allocated from page pool */
1000 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */
1001 wait_queue_head_t ee_wait;
1002 struct page *md_io_page; /* one page buffer for md_io */
1003 struct drbd_md_io md_io;
1004 atomic_t md_io_in_use; /* protects the md_io, md_io_page and md_io_tmpp */
1006 wait_queue_head_t al_wait;
1007 struct lru_cache *act_log; /* activity log */
1008 unsigned int al_tr_number;
1010 int al_tr_pos; /* position of the next transaction in the journal */
1011 wait_queue_head_t seq_wait;
1012 atomic_t packet_seq;
1013 unsigned int peer_seq;
1014 spinlock_t peer_seq_lock;
1016 unsigned long comm_bm_set; /* communicated number of set bits. */
1017 struct bm_io_work bm_io_work;
1018 u64 ed_uuid; /* UUID of the exposed data */
1019 struct mutex own_state_mutex;
1020 struct mutex *state_mutex; /* either own_state_mutex or mdev->tconn->cstate_mutex */
1021 char congestion_reason; /* Why we where congested... */
1022 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
1023 atomic_t rs_sect_ev; /* for submitted resync data rate, both */
1024 int rs_last_sect_ev; /* counter to compare with */
1025 int rs_last_events; /* counter of read or write "events" (unit sectors)
1026 * on the lower level device when we last looked. */
1027 int c_sync_rate; /* current resync rate after syncer throttle magic */
1028 struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, tconn->conn_update) */
1029 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
1030 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
1031 int peer_max_bio_size;
1032 int local_max_bio_size;
1035 static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
1037 return (struct drbd_conf *)idr_find(&minors, minor);
1040 static inline unsigned int mdev_to_minor(struct drbd_conf *mdev)
1045 static inline struct drbd_conf *vnr_to_mdev(struct drbd_tconn *tconn, int vnr)
1047 return (struct drbd_conf *)idr_find(&tconn->volumes, vnr);
1051 * function declarations
1052 *************************/
1058 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1061 extern void drbd_init_set_defaults(struct drbd_conf *mdev);
1062 extern int drbd_thread_start(struct drbd_thread *thi);
1063 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1064 extern char *drbd_task_to_thread_name(struct drbd_tconn *tconn, struct task_struct *task);
1066 extern void drbd_thread_current_set_cpu(struct drbd_thread *thi);
1067 extern void drbd_calc_cpu_mask(struct drbd_tconn *tconn);
1069 #define drbd_thread_current_set_cpu(A) ({})
1070 #define drbd_calc_cpu_mask(A) ({})
1072 extern void tl_release(struct drbd_tconn *, unsigned int barrier_nr,
1073 unsigned int set_size);
1074 extern void tl_clear(struct drbd_tconn *);
1075 extern void drbd_free_sock(struct drbd_tconn *tconn);
1076 extern int drbd_send(struct drbd_tconn *tconn, struct socket *sock,
1077 void *buf, size_t size, unsigned msg_flags);
1078 extern int drbd_send_all(struct drbd_tconn *, struct socket *, void *, size_t,
1081 extern int __drbd_send_protocol(struct drbd_tconn *tconn, enum drbd_packet cmd);
1082 extern int drbd_send_protocol(struct drbd_tconn *tconn);
1083 extern int drbd_send_uuids(struct drbd_conf *mdev);
1084 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
1085 extern void drbd_gen_and_send_sync_uuid(struct drbd_conf *mdev);
1086 extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags);
1087 extern int drbd_send_state(struct drbd_conf *mdev, union drbd_state s);
1088 extern int drbd_send_current_state(struct drbd_conf *mdev);
1089 extern int drbd_send_sync_param(struct drbd_conf *mdev);
1090 extern void drbd_send_b_ack(struct drbd_tconn *tconn, u32 barrier_nr,
1092 extern int drbd_send_ack(struct drbd_conf *, enum drbd_packet,
1093 struct drbd_peer_request *);
1094 extern void drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packet cmd,
1095 struct p_block_req *rp);
1096 extern void drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packet cmd,
1097 struct p_data *dp, int data_size);
1098 extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packet cmd,
1099 sector_t sector, int blksize, u64 block_id);
1100 extern int drbd_send_out_of_sync(struct drbd_conf *, struct drbd_request *);
1101 extern int drbd_send_block(struct drbd_conf *, enum drbd_packet,
1102 struct drbd_peer_request *);
1103 extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
1104 extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1105 sector_t sector, int size, u64 block_id);
1106 extern int drbd_send_drequest_csum(struct drbd_conf *mdev, sector_t sector,
1107 int size, void *digest, int digest_size,
1108 enum drbd_packet cmd);
1109 extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size);
1111 extern int drbd_send_bitmap(struct drbd_conf *mdev);
1112 extern void drbd_send_sr_reply(struct drbd_conf *mdev, enum drbd_state_rv retcode);
1113 extern void conn_send_sr_reply(struct drbd_tconn *tconn, enum drbd_state_rv retcode);
1114 extern void drbd_free_bc(struct drbd_backing_dev *ldev);
1115 extern void drbd_mdev_cleanup(struct drbd_conf *mdev);
1116 void drbd_print_uuids(struct drbd_conf *mdev, const char *text);
1118 extern void drbd_md_sync(struct drbd_conf *mdev);
1119 extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
1120 extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1121 extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1122 extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1123 extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
1124 extern void drbd_uuid_move_history(struct drbd_conf *mdev) __must_hold(local);
1125 extern void __drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1126 extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
1127 extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
1128 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1129 #ifndef DRBD_DEBUG_MD_SYNC
1130 extern void drbd_md_mark_dirty(struct drbd_conf *mdev);
1132 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1133 extern void drbd_md_mark_dirty_(struct drbd_conf *mdev,
1134 unsigned int line, const char *func);
1136 extern void drbd_queue_bitmap_io(struct drbd_conf *mdev,
1137 int (*io_fn)(struct drbd_conf *),
1138 void (*done)(struct drbd_conf *, int),
1139 char *why, enum bm_flag flags);
1140 extern int drbd_bitmap_io(struct drbd_conf *mdev,
1141 int (*io_fn)(struct drbd_conf *),
1142 char *why, enum bm_flag flags);
1143 extern int drbd_bmio_set_n_write(struct drbd_conf *mdev);
1144 extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev);
1145 extern void drbd_go_diskless(struct drbd_conf *mdev);
1146 extern void drbd_ldev_destroy(struct drbd_conf *mdev);
1149 We reserve a 128MB Block (4k aligned)
1150 * either at the end of the backing device
1151 * or on a separate meta data device. */
1153 /* The following numbers are sectors */
1154 /* Allows up to about 3.8TB, so if you want more,
1155 * you need to use the "flexible" meta data format. */
1156 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1157 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1158 #define MD_AL_SECTORS 64 /* = 32 kB on disk activity log ring buffer */
1159 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_SECTORS)
1161 /* we do all meta data IO in 4k blocks */
1162 #define MD_BLOCK_SHIFT 12
1163 #define MD_BLOCK_SIZE (1<<MD_BLOCK_SHIFT)
1165 /* One activity log extent represents 4M of storage */
1166 #define AL_EXTENT_SHIFT 22
1167 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1169 /* We could make these currently hardcoded constants configurable
1170 * variables at create-md time (or even re-configurable at runtime?).
1171 * Which will require some more changes to the DRBD "super block"
1174 * updates per transaction:
1175 * This many changes to the active set can be logged with one transaction.
1176 * This number is arbitrary.
1177 * context per transaction:
1178 * This many context extent numbers are logged with each transaction.
1179 * This number is resulting from the transaction block size (4k), the layout
1180 * of the transaction header, and the number of updates per transaction.
1181 * See drbd_actlog.c:struct al_transaction_on_disk
1183 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1184 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1186 #if BITS_PER_LONG == 32
1188 #define cpu_to_lel(A) cpu_to_le32(A)
1189 #define lel_to_cpu(A) le32_to_cpu(A)
1190 #elif BITS_PER_LONG == 64
1192 #define cpu_to_lel(A) cpu_to_le64(A)
1193 #define lel_to_cpu(A) le64_to_cpu(A)
1195 #error "LN2 of BITS_PER_LONG unknown!"
1199 /* 16MB sized 'bitmap extent' to track syncer usage */
1201 int rs_left; /* number of bits set (out of sync) in this extent. */
1202 int rs_failed; /* number of failed resync requests in this extent. */
1203 unsigned long flags;
1204 struct lc_element lce;
1207 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1208 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1209 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1213 * We need to store one bit for a block.
1214 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1215 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1216 * Bit 1 ==> local node thinks this block needs to be synced.
1219 #define SLEEP_TIME (HZ/10)
1221 /* We do bitmap IO in units of 4k blocks.
1222 * We also still have a hardcoded 4k per bit relation. */
1223 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1224 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1225 /* mostly arbitrarily set the represented size of one bitmap extent,
1226 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1227 * at 4k per bit resolution) */
1228 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
1229 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1231 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1232 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1235 /* thus many _storage_ sectors are described by one bit */
1236 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1237 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1238 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1240 /* bit to represented kilo byte conversion */
1241 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1243 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1244 * _storage_ sector is located in */
1245 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1247 /* how much _storage_ sectors we have per bitmap sector */
1248 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1249 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1251 /* in one sector of the bitmap, we have this many activity_log extents. */
1252 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1253 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1255 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1256 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1258 /* the extent in "PER_EXTENT" below is an activity log extent
1259 * we need that many (long words/bytes) to store the bitmap
1260 * of one AL_EXTENT_SIZE chunk of storage.
1261 * we can store the bitmap for that many AL_EXTENTS within
1262 * one sector of the _on_disk_ bitmap:
1263 * bit 0 bit 37 bit 38 bit (512*8)-1
1264 * ...|........|........|.. // ..|........|
1265 * sect. 0 `296 `304 ^(512*8*8)-1
1267 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1268 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1269 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1272 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1273 #define DRBD_MAX_SECTORS_BM \
1274 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1275 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1276 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1277 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1278 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1279 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1280 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1282 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1283 /* 16 TB in units of sectors */
1284 #if BITS_PER_LONG == 32
1285 /* adjust by one page worth of bitmap,
1286 * so we won't wrap around in drbd_bm_find_next_bit.
1287 * you should use 64bit OS for that much storage, anyways. */
1288 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1290 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1291 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1292 /* corresponds to (1UL << 38) bits right now. */
1296 /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE,
1297 * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte.
1298 * Since we may live in a mixed-platform cluster,
1299 * we limit us to a platform agnostic constant here for now.
1300 * A followup commit may allow even bigger BIO sizes,
1301 * once we thought that through. */
1302 #define DRBD_MAX_BIO_SIZE (1 << 20)
1303 #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1304 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1306 #define DRBD_MAX_BIO_SIZE_SAFE (1 << 12) /* Works always = 4k */
1308 #define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* Header 80 only allows packets up to 32KiB data */
1309 #define DRBD_MAX_BIO_SIZE_P95 (1 << 17) /* Protocol 95 to 99 allows bios up to 128KiB */
1311 extern int drbd_bm_init(struct drbd_conf *mdev);
1312 extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits);
1313 extern void drbd_bm_cleanup(struct drbd_conf *mdev);
1314 extern void drbd_bm_set_all(struct drbd_conf *mdev);
1315 extern void drbd_bm_clear_all(struct drbd_conf *mdev);
1316 /* set/clear/test only a few bits at a time */
1317 extern int drbd_bm_set_bits(
1318 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1319 extern int drbd_bm_clear_bits(
1320 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1321 extern int drbd_bm_count_bits(
1322 struct drbd_conf *mdev, const unsigned long s, const unsigned long e);
1323 /* bm_set_bits variant for use while holding drbd_bm_lock,
1324 * may process the whole bitmap in one go */
1325 extern void _drbd_bm_set_bits(struct drbd_conf *mdev,
1326 const unsigned long s, const unsigned long e);
1327 extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr);
1328 extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr);
1329 extern int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local);
1330 extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
1331 extern void drbd_bm_mark_for_writeout(struct drbd_conf *mdev, int page_nr);
1332 extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
1333 extern int drbd_bm_write_hinted(struct drbd_conf *mdev) __must_hold(local);
1334 extern int drbd_bm_write_all(struct drbd_conf *mdev) __must_hold(local);
1335 extern int drbd_bm_write_copy_pages(struct drbd_conf *mdev) __must_hold(local);
1336 extern size_t drbd_bm_words(struct drbd_conf *mdev);
1337 extern unsigned long drbd_bm_bits(struct drbd_conf *mdev);
1338 extern sector_t drbd_bm_capacity(struct drbd_conf *mdev);
1340 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1341 extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1342 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1343 extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1344 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo);
1345 extern unsigned long _drbd_bm_total_weight(struct drbd_conf *mdev);
1346 extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev);
1347 extern int drbd_bm_rs_done(struct drbd_conf *mdev);
1348 /* for receive_bitmap */
1349 extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset,
1350 size_t number, unsigned long *buffer);
1351 /* for _drbd_send_bitmap */
1352 extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset,
1353 size_t number, unsigned long *buffer);
1355 extern void drbd_bm_lock(struct drbd_conf *mdev, char *why, enum bm_flag flags);
1356 extern void drbd_bm_unlock(struct drbd_conf *mdev);
1359 extern struct kmem_cache *drbd_request_cache;
1360 extern struct kmem_cache *drbd_ee_cache; /* peer requests */
1361 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1362 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1363 extern mempool_t *drbd_request_mempool;
1364 extern mempool_t *drbd_ee_mempool;
1366 /* drbd's page pool, used to buffer data received from the peer,
1367 * or data requested by the peer.
1369 * This does not have an emergency reserve.
1371 * When allocating from this pool, it first takes pages from the pool.
1372 * Only if the pool is depleted will try to allocate from the system.
1374 * The assumption is that pages taken from this pool will be processed,
1375 * and given back, "quickly", and then can be recycled, so we can avoid
1376 * frequent calls to alloc_page(), and still will be able to make progress even
1377 * under memory pressure.
1379 extern struct page *drbd_pp_pool;
1380 extern spinlock_t drbd_pp_lock;
1381 extern int drbd_pp_vacant;
1382 extern wait_queue_head_t drbd_pp_wait;
1384 /* We also need a standard (emergency-reserve backed) page pool
1385 * for meta data IO (activity log, bitmap).
1386 * We can keep it global, as long as it is used as "N pages at a time".
1387 * 128 should be plenty, currently we probably can get away with as few as 1.
1389 #define DRBD_MIN_POOL_PAGES 128
1390 extern mempool_t *drbd_md_io_page_pool;
1392 /* We also need to make sure we get a bio
1393 * when we need it for housekeeping purposes */
1394 extern struct bio_set *drbd_md_io_bio_set;
1395 /* to allocate from that set */
1396 extern struct bio *bio_alloc_drbd(gfp_t gfp_mask);
1398 extern rwlock_t global_state_lock;
1400 extern int conn_lowest_minor(struct drbd_tconn *tconn);
1401 enum drbd_ret_code conn_new_minor(struct drbd_tconn *tconn, unsigned int minor, int vnr);
1402 extern void drbd_minor_destroy(struct kref *kref);
1404 extern int set_resource_options(struct drbd_tconn *tconn, struct res_opts *res_opts);
1405 extern struct drbd_tconn *conn_create(const char *name, struct res_opts *res_opts);
1406 extern void conn_destroy(struct kref *kref);
1407 struct drbd_tconn *conn_get_by_name(const char *name);
1408 extern struct drbd_tconn *conn_get_by_addrs(void *my_addr, int my_addr_len,
1409 void *peer_addr, int peer_addr_len);
1410 extern void conn_free_crypto(struct drbd_tconn *tconn);
1412 extern int proc_details;
1415 extern void __drbd_make_request(struct drbd_conf *, struct bio *, unsigned long);
1416 extern int drbd_make_request(struct request_queue *q, struct bio *bio);
1417 extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req);
1418 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1419 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1423 extern int drbd_msg_put_info(const char *info);
1424 extern void drbd_suspend_io(struct drbd_conf *mdev);
1425 extern void drbd_resume_io(struct drbd_conf *mdev);
1426 extern char *ppsize(char *buf, unsigned long long size);
1427 extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, sector_t, int);
1428 enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
1429 extern enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local);
1430 extern void resync_after_online_grow(struct drbd_conf *);
1431 extern void drbd_reconsider_max_bio_size(struct drbd_conf *mdev);
1432 extern enum drbd_state_rv drbd_set_role(struct drbd_conf *mdev,
1433 enum drbd_role new_role,
1435 extern bool conn_try_outdate_peer(struct drbd_tconn *tconn);
1436 extern void conn_try_outdate_peer_async(struct drbd_tconn *tconn);
1437 extern int drbd_khelper(struct drbd_conf *mdev, char *cmd);
1440 extern int drbd_worker(struct drbd_thread *thi);
1441 enum drbd_ret_code drbd_resync_after_valid(struct drbd_conf *mdev, int o_minor);
1442 void drbd_resync_after_changed(struct drbd_conf *mdev);
1443 extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side);
1444 extern void resume_next_sg(struct drbd_conf *mdev);
1445 extern void suspend_other_sg(struct drbd_conf *mdev);
1446 extern int drbd_resync_finished(struct drbd_conf *mdev);
1447 /* maybe rather drbd_main.c ? */
1448 extern void *drbd_md_get_buffer(struct drbd_conf *mdev);
1449 extern void drbd_md_put_buffer(struct drbd_conf *mdev);
1450 extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
1451 struct drbd_backing_dev *bdev, sector_t sector, int rw);
1452 extern void drbd_ov_out_of_sync_found(struct drbd_conf *, sector_t, int);
1453 extern void wait_until_done_or_disk_failure(struct drbd_conf *mdev, struct drbd_backing_dev *bdev,
1454 unsigned int *done);
1455 extern void drbd_rs_controller_reset(struct drbd_conf *mdev);
1457 static inline void ov_out_of_sync_print(struct drbd_conf *mdev)
1459 if (mdev->ov_last_oos_size) {
1460 dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n",
1461 (unsigned long long)mdev->ov_last_oos_start,
1462 (unsigned long)mdev->ov_last_oos_size);
1464 mdev->ov_last_oos_size=0;
1468 extern void drbd_csum_bio(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
1469 extern void drbd_csum_ee(struct drbd_conf *, struct crypto_hash *,
1470 struct drbd_peer_request *, void *);
1471 /* worker callbacks */
1472 extern int w_e_end_data_req(struct drbd_work *, int);
1473 extern int w_e_end_rsdata_req(struct drbd_work *, int);
1474 extern int w_e_end_csum_rs_req(struct drbd_work *, int);
1475 extern int w_e_end_ov_reply(struct drbd_work *, int);
1476 extern int w_e_end_ov_req(struct drbd_work *, int);
1477 extern int w_ov_finished(struct drbd_work *, int);
1478 extern int w_resync_timer(struct drbd_work *, int);
1479 extern int w_send_write_hint(struct drbd_work *, int);
1480 extern int w_make_resync_request(struct drbd_work *, int);
1481 extern int w_send_dblock(struct drbd_work *, int);
1482 extern int w_send_read_req(struct drbd_work *, int);
1483 extern int w_prev_work_done(struct drbd_work *, int);
1484 extern int w_e_reissue(struct drbd_work *, int);
1485 extern int w_restart_disk_io(struct drbd_work *, int);
1486 extern int w_send_out_of_sync(struct drbd_work *, int);
1487 extern int w_start_resync(struct drbd_work *, int);
1489 extern void resync_timer_fn(unsigned long data);
1490 extern void start_resync_timer_fn(unsigned long data);
1492 /* drbd_receiver.c */
1493 extern int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector);
1494 extern int drbd_submit_peer_request(struct drbd_conf *,
1495 struct drbd_peer_request *, const unsigned,
1497 extern int drbd_free_peer_reqs(struct drbd_conf *, struct list_head *);
1498 extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_conf *, u64,
1499 sector_t, unsigned int,
1500 gfp_t) __must_hold(local);
1501 extern void __drbd_free_peer_req(struct drbd_conf *, struct drbd_peer_request *,
1503 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1504 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
1505 extern struct page *drbd_alloc_pages(struct drbd_conf *, unsigned int, bool);
1506 extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled);
1507 extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed);
1508 extern void conn_flush_workqueue(struct drbd_tconn *tconn);
1509 extern int drbd_connected(struct drbd_conf *mdev);
1510 static inline void drbd_flush_workqueue(struct drbd_conf *mdev)
1512 conn_flush_workqueue(mdev->tconn);
1515 /* Yes, there is kernel_setsockopt, but only since 2.6.18.
1516 * So we have our own copy of it here. */
1517 static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1518 char *optval, int optlen)
1520 mm_segment_t oldfs = get_fs();
1521 char __user *uoptval;
1524 uoptval = (char __user __force *)optval;
1527 if (level == SOL_SOCKET)
1528 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
1530 err = sock->ops->setsockopt(sock, level, optname, uoptval,
1536 static inline void drbd_tcp_cork(struct socket *sock)
1539 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1540 (char*)&val, sizeof(val));
1543 static inline void drbd_tcp_uncork(struct socket *sock)
1546 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1547 (char*)&val, sizeof(val));
1550 static inline void drbd_tcp_nodelay(struct socket *sock)
1553 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1554 (char*)&val, sizeof(val));
1557 static inline void drbd_tcp_quickack(struct socket *sock)
1560 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1561 (char*)&val, sizeof(val));
1564 void drbd_bump_write_ordering(struct drbd_tconn *tconn, enum write_ordering_e wo);
1567 extern struct proc_dir_entry *drbd_proc;
1568 extern const struct file_operations drbd_proc_fops;
1569 extern const char *drbd_conn_str(enum drbd_conns s);
1570 extern const char *drbd_role_str(enum drbd_role s);
1573 extern void drbd_al_begin_io(struct drbd_conf *mdev, struct drbd_interval *i);
1574 extern void drbd_al_complete_io(struct drbd_conf *mdev, struct drbd_interval *i);
1575 extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector);
1576 extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1577 extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1578 extern void drbd_rs_cancel_all(struct drbd_conf *mdev);
1579 extern int drbd_rs_del_all(struct drbd_conf *mdev);
1580 extern void drbd_rs_failed_io(struct drbd_conf *mdev,
1581 sector_t sector, int size);
1582 extern void drbd_advance_rs_marks(struct drbd_conf *mdev, unsigned long still_to_go);
1583 extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
1584 int size, const char *file, const unsigned int line);
1585 #define drbd_set_in_sync(mdev, sector, size) \
1586 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1587 extern int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
1588 int size, const char *file, const unsigned int line);
1589 #define drbd_set_out_of_sync(mdev, sector, size) \
1590 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1591 extern void drbd_al_shrink(struct drbd_conf *mdev);
1594 /* state info broadcast */
1596 enum drbd_state_info_bcast_reason sib_reason;
1600 unsigned helper_exit_code;
1603 union drbd_state os;
1604 union drbd_state ns;
1608 void drbd_bcast_event(struct drbd_conf *mdev, const struct sib_info *sib);
1611 * inline helper functions
1612 *************************/
1614 /* see also page_chain_add and friends in drbd_receiver.c */
1615 static inline struct page *page_chain_next(struct page *page)
1617 return (struct page *)page_private(page);
1619 #define page_chain_for_each(page) \
1620 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1621 page = page_chain_next(page))
1622 #define page_chain_for_each_safe(page, n) \
1623 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1626 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req)
1628 struct page *page = peer_req->pages;
1629 page_chain_for_each(page) {
1630 if (page_count(page) > 1)
1636 static inline enum drbd_state_rv
1637 _drbd_set_state(struct drbd_conf *mdev, union drbd_state ns,
1638 enum chg_state_flags flags, struct completion *done)
1640 enum drbd_state_rv rv;
1642 read_lock(&global_state_lock);
1643 rv = __drbd_set_state(mdev, ns, flags, done);
1644 read_unlock(&global_state_lock);
1649 static inline union drbd_state drbd_read_state(struct drbd_conf *mdev)
1651 union drbd_state rv;
1653 rv.i = mdev->state.i;
1654 rv.susp = mdev->tconn->susp;
1655 rv.susp_nod = mdev->tconn->susp_nod;
1656 rv.susp_fen = mdev->tconn->susp_fen;
1661 enum drbd_force_detach_flags {
1667 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1668 static inline void __drbd_chk_io_error_(struct drbd_conf *mdev,
1669 enum drbd_force_detach_flags forcedetach,
1672 enum drbd_io_error_p ep;
1675 ep = rcu_dereference(mdev->ldev->disk_conf)->on_io_error;
1678 case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */
1679 if (forcedetach == DRBD_IO_ERROR) {
1680 if (__ratelimit(&drbd_ratelimit_state))
1681 dev_err(DEV, "Local IO failed in %s.\n", where);
1682 if (mdev->state.disk > D_INCONSISTENT)
1683 _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_HARD, NULL);
1686 /* NOTE fall through to detach case if forcedetach set */
1688 case EP_CALL_HELPER:
1689 set_bit(WAS_IO_ERROR, &mdev->flags);
1690 if (forcedetach == DRBD_FORCE_DETACH)
1691 set_bit(FORCE_DETACH, &mdev->flags);
1692 if (mdev->state.disk > D_FAILED) {
1693 _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL);
1695 "Local IO failed in %s. Detaching...\n", where);
1702 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1703 * @mdev: DRBD device.
1704 * @error: Error code passed to the IO completion callback
1705 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1707 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1709 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1710 static inline void drbd_chk_io_error_(struct drbd_conf *mdev,
1711 int error, enum drbd_force_detach_flags forcedetach, const char *where)
1714 unsigned long flags;
1715 spin_lock_irqsave(&mdev->tconn->req_lock, flags);
1716 __drbd_chk_io_error_(mdev, forcedetach, where);
1717 spin_unlock_irqrestore(&mdev->tconn->req_lock, flags);
1723 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1724 * @bdev: Meta data block device.
1726 * BTW, for internal meta data, this happens to be the maximum capacity
1727 * we could agree upon with our peer node.
1729 static inline sector_t _drbd_md_first_sector(int meta_dev_idx, struct drbd_backing_dev *bdev)
1731 switch (meta_dev_idx) {
1732 case DRBD_MD_INDEX_INTERNAL:
1733 case DRBD_MD_INDEX_FLEX_INT:
1734 return bdev->md.md_offset + bdev->md.bm_offset;
1735 case DRBD_MD_INDEX_FLEX_EXT:
1737 return bdev->md.md_offset;
1741 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1746 meta_dev_idx = rcu_dereference(bdev->disk_conf)->meta_dev_idx;
1749 return _drbd_md_first_sector(meta_dev_idx, bdev);
1753 * drbd_md_last_sector() - Return the last sector number of the meta data area
1754 * @bdev: Meta data block device.
1756 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1761 meta_dev_idx = rcu_dereference(bdev->disk_conf)->meta_dev_idx;
1764 switch (meta_dev_idx) {
1765 case DRBD_MD_INDEX_INTERNAL:
1766 case DRBD_MD_INDEX_FLEX_INT:
1767 return bdev->md.md_offset + MD_AL_OFFSET - 1;
1768 case DRBD_MD_INDEX_FLEX_EXT:
1770 return bdev->md.md_offset + bdev->md.md_size_sect;
1774 /* Returns the number of 512 byte sectors of the device */
1775 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1777 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1778 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0;
1782 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1783 * @bdev: Meta data block device.
1785 * returns the capacity we announce to out peer. we clip ourselves at the
1786 * various MAX_SECTORS, because if we don't, current implementation will
1787 * oops sooner or later
1789 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1795 meta_dev_idx = rcu_dereference(bdev->disk_conf)->meta_dev_idx;
1798 switch (meta_dev_idx) {
1799 case DRBD_MD_INDEX_INTERNAL:
1800 case DRBD_MD_INDEX_FLEX_INT:
1801 s = drbd_get_capacity(bdev->backing_bdev)
1802 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1803 _drbd_md_first_sector(meta_dev_idx, bdev))
1806 case DRBD_MD_INDEX_FLEX_EXT:
1807 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1808 drbd_get_capacity(bdev->backing_bdev));
1809 /* clip at maximum size the meta device can support */
1810 s = min_t(sector_t, s,
1811 BM_EXT_TO_SECT(bdev->md.md_size_sect
1812 - bdev->md.bm_offset));
1815 s = min_t(sector_t, DRBD_MAX_SECTORS,
1816 drbd_get_capacity(bdev->backing_bdev));
1822 * drbd_md_ss__() - Return the sector number of our meta data super block
1823 * @mdev: DRBD device.
1824 * @bdev: Meta data block device.
1826 static inline sector_t drbd_md_ss__(struct drbd_conf *mdev,
1827 struct drbd_backing_dev *bdev)
1832 meta_dev_idx = rcu_dereference(bdev->disk_conf)->meta_dev_idx;
1835 switch (meta_dev_idx) {
1836 default: /* external, some index */
1837 return MD_RESERVED_SECT * meta_dev_idx;
1838 case DRBD_MD_INDEX_INTERNAL:
1839 /* with drbd08, internal meta data is always "flexible" */
1840 case DRBD_MD_INDEX_FLEX_INT:
1841 /* sizeof(struct md_on_disk_07) == 4k
1842 * position: last 4k aligned block of 4k size */
1843 if (!bdev->backing_bdev) {
1844 if (__ratelimit(&drbd_ratelimit_state)) {
1845 dev_err(DEV, "bdev->backing_bdev==NULL\n");
1850 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL)
1852 case DRBD_MD_INDEX_FLEX_EXT:
1858 drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w)
1860 unsigned long flags;
1861 spin_lock_irqsave(&q->q_lock, flags);
1862 list_add(&w->list, &q->q);
1863 spin_unlock_irqrestore(&q->q_lock, flags);
1864 wake_up(&q->q_wait);
1868 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1870 unsigned long flags;
1871 spin_lock_irqsave(&q->q_lock, flags);
1872 list_add_tail(&w->list, &q->q);
1873 spin_unlock_irqrestore(&q->q_lock, flags);
1874 wake_up(&q->q_wait);
1877 static inline void wake_asender(struct drbd_tconn *tconn)
1879 if (test_bit(SIGNAL_ASENDER, &tconn->flags))
1880 force_sig(DRBD_SIG, tconn->asender.task);
1883 static inline void request_ping(struct drbd_tconn *tconn)
1885 set_bit(SEND_PING, &tconn->flags);
1886 wake_asender(tconn);
1889 extern void *conn_prepare_command(struct drbd_tconn *, struct drbd_socket *);
1890 extern void *drbd_prepare_command(struct drbd_conf *, struct drbd_socket *);
1891 extern int conn_send_command(struct drbd_tconn *, struct drbd_socket *,
1892 enum drbd_packet, unsigned int, void *,
1894 extern int drbd_send_command(struct drbd_conf *, struct drbd_socket *,
1895 enum drbd_packet, unsigned int, void *,
1898 extern int drbd_send_ping(struct drbd_tconn *tconn);
1899 extern int drbd_send_ping_ack(struct drbd_tconn *tconn);
1900 extern int drbd_send_state_req(struct drbd_conf *, union drbd_state, union drbd_state);
1901 extern int conn_send_state_req(struct drbd_tconn *, union drbd_state, union drbd_state);
1903 static inline void drbd_thread_stop(struct drbd_thread *thi)
1905 _drbd_thread_stop(thi, false, true);
1908 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
1910 _drbd_thread_stop(thi, false, false);
1913 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
1915 _drbd_thread_stop(thi, true, false);
1918 /* counts how many answer packets packets we expect from our peer,
1919 * for either explicit application requests,
1920 * or implicit barrier packets as necessary.
1923 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1924 * it is much easier and equally valid to count what we queue for the
1925 * worker, even before it actually was queued or send.
1926 * (drbd_make_request_common; recovery path on read io-error)
1928 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1929 * _req_mod(req, DATA_RECEIVED)
1930 * [from receive_DataReply]
1931 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1932 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1933 * for some reason it is NOT decreased in got_NegAck,
1934 * but in the resulting cleanup code from report_params.
1935 * we should try to remember the reason for that...
1936 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
1937 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
1938 * [from tl_clear_barrier]
1940 static inline void inc_ap_pending(struct drbd_conf *mdev)
1942 atomic_inc(&mdev->ap_pending_cnt);
1945 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \
1946 if (atomic_read(&mdev->which) < 0) \
1947 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
1949 atomic_read(&mdev->which))
1951 #define dec_ap_pending(mdev) _dec_ap_pending(mdev, __FUNCTION__, __LINE__)
1952 static inline void _dec_ap_pending(struct drbd_conf *mdev, const char *func, int line)
1954 if (atomic_dec_and_test(&mdev->ap_pending_cnt))
1955 wake_up(&mdev->misc_wait);
1956 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line);
1959 /* counts how many resync-related answers we still expect from the peer
1961 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1962 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
1963 * (or P_NEG_ACK with ID_SYNCER)
1965 static inline void inc_rs_pending(struct drbd_conf *mdev)
1967 atomic_inc(&mdev->rs_pending_cnt);
1970 #define dec_rs_pending(mdev) _dec_rs_pending(mdev, __FUNCTION__, __LINE__)
1971 static inline void _dec_rs_pending(struct drbd_conf *mdev, const char *func, int line)
1973 atomic_dec(&mdev->rs_pending_cnt);
1974 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line);
1977 /* counts how many answers we still need to send to the peer.
1979 * receive_Data unless protocol A;
1980 * we need to send a P_RECV_ACK (proto B)
1981 * or P_WRITE_ACK (proto C)
1982 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1983 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1984 * receive_Barrier_* we need to send a P_BARRIER_ACK
1986 static inline void inc_unacked(struct drbd_conf *mdev)
1988 atomic_inc(&mdev->unacked_cnt);
1991 #define dec_unacked(mdev) _dec_unacked(mdev, __FUNCTION__, __LINE__)
1992 static inline void _dec_unacked(struct drbd_conf *mdev, const char *func, int line)
1994 atomic_dec(&mdev->unacked_cnt);
1995 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
1998 #define sub_unacked(mdev, n) _sub_unacked(mdev, n, __FUNCTION__, __LINE__)
1999 static inline void _sub_unacked(struct drbd_conf *mdev, int n, const char *func, int line)
2001 atomic_sub(n, &mdev->unacked_cnt);
2002 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
2006 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2009 * You have to call put_ldev() when finished working with mdev->ldev.
2011 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2012 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2014 static inline void put_ldev(struct drbd_conf *mdev)
2016 int i = atomic_dec_return(&mdev->local_cnt);
2018 /* This may be called from some endio handler,
2019 * so we must not sleep here. */
2024 if (mdev->state.disk == D_DISKLESS)
2025 /* even internal references gone, safe to destroy */
2026 drbd_ldev_destroy(mdev);
2027 if (mdev->state.disk == D_FAILED)
2028 /* all application IO references gone. */
2029 drbd_go_diskless(mdev);
2030 wake_up(&mdev->misc_wait);
2035 static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
2039 /* never get a reference while D_DISKLESS */
2040 if (mdev->state.disk == D_DISKLESS)
2043 atomic_inc(&mdev->local_cnt);
2044 io_allowed = (mdev->state.disk >= mins);
2050 extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins);
2053 /* you must have an "get_ldev" reference */
2054 static inline void drbd_get_syncer_progress(struct drbd_conf *mdev,
2055 unsigned long *bits_left, unsigned int *per_mil_done)
2057 /* this is to break it at compile time when we change that, in case we
2058 * want to support more than (1<<32) bits on a 32bit arch. */
2059 typecheck(unsigned long, mdev->rs_total);
2061 /* note: both rs_total and rs_left are in bits, i.e. in
2062 * units of BM_BLOCK_SIZE.
2063 * for the percentage, we don't care. */
2065 if (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T)
2066 *bits_left = mdev->ov_left;
2068 *bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
2069 /* >> 10 to prevent overflow,
2070 * +1 to prevent division by zero */
2071 if (*bits_left > mdev->rs_total) {
2072 /* doh. maybe a logic bug somewhere.
2073 * may also be just a race condition
2074 * between this and a disconnect during sync.
2075 * for now, just prevent in-kernel buffer overflow.
2078 dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2079 drbd_conn_str(mdev->state.conn),
2080 *bits_left, mdev->rs_total, mdev->rs_failed);
2083 /* Make sure the division happens in long context.
2084 * We allow up to one petabyte storage right now,
2085 * at a granularity of 4k per bit that is 2**38 bits.
2086 * After shift right and multiplication by 1000,
2087 * this should still fit easily into a 32bit long,
2088 * so we don't need a 64bit division on 32bit arch.
2089 * Note: currently we don't support such large bitmaps on 32bit
2090 * arch anyways, but no harm done to be prepared for it here.
2092 unsigned int shift = mdev->rs_total > UINT_MAX ? 16 : 10;
2093 unsigned long left = *bits_left >> shift;
2094 unsigned long total = 1UL + (mdev->rs_total >> shift);
2095 unsigned long tmp = 1000UL - left * 1000UL/total;
2096 *per_mil_done = tmp;
2101 /* this throttles on-the-fly application requests
2102 * according to max_buffers settings;
2103 * maybe re-implement using semaphores? */
2104 static inline int drbd_get_max_buffers(struct drbd_conf *mdev)
2106 struct net_conf *nc;
2110 nc = rcu_dereference(mdev->tconn->net_conf);
2111 mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */
2117 static inline int drbd_state_is_stable(struct drbd_conf *mdev)
2119 union drbd_dev_state s = mdev->state;
2121 /* DO NOT add a default clause, we want the compiler to warn us
2122 * for any newly introduced state we may have forgotten to add here */
2124 switch ((enum drbd_conns)s.conn) {
2125 /* new io only accepted when there is no connection, ... */
2127 case C_WF_CONNECTION:
2128 /* ... or there is a well established connection. */
2134 case C_PAUSED_SYNC_S:
2135 case C_PAUSED_SYNC_T:
2138 /* transitional states, IO allowed */
2139 case C_DISCONNECTING:
2143 case C_NETWORK_FAILURE:
2144 case C_PROTOCOL_ERROR:
2146 case C_WF_REPORT_PARAMS:
2147 case C_STARTING_SYNC_S:
2148 case C_STARTING_SYNC_T:
2151 /* Allow IO in BM exchange states with new protocols */
2153 if (mdev->tconn->agreed_pro_version < 96)
2157 /* no new io accepted in these states */
2159 case C_WF_SYNC_UUID:
2165 switch ((enum drbd_disk_state)s.disk) {
2167 case D_INCONSISTENT:
2172 /* disk state is stable as well. */
2175 /* no new io accepted during transitional states */
2187 static inline int drbd_suspended(struct drbd_conf *mdev)
2189 struct drbd_tconn *tconn = mdev->tconn;
2191 return tconn->susp || tconn->susp_fen || tconn->susp_nod;
2194 static inline bool may_inc_ap_bio(struct drbd_conf *mdev)
2196 int mxb = drbd_get_max_buffers(mdev);
2198 if (drbd_suspended(mdev))
2200 if (test_bit(SUSPEND_IO, &mdev->flags))
2203 /* to avoid potential deadlock or bitmap corruption,
2204 * in various places, we only allow new application io
2205 * to start during "stable" states. */
2207 /* no new io accepted when attaching or detaching the disk */
2208 if (!drbd_state_is_stable(mdev))
2211 /* since some older kernels don't have atomic_add_unless,
2212 * and we are within the spinlock anyways, we have this workaround. */
2213 if (atomic_read(&mdev->ap_bio_cnt) > mxb)
2215 if (test_bit(BITMAP_IO, &mdev->flags))
2220 static inline bool inc_ap_bio_cond(struct drbd_conf *mdev)
2224 spin_lock_irq(&mdev->tconn->req_lock);
2225 rv = may_inc_ap_bio(mdev);
2227 atomic_inc(&mdev->ap_bio_cnt);
2228 spin_unlock_irq(&mdev->tconn->req_lock);
2233 static inline void inc_ap_bio(struct drbd_conf *mdev)
2236 * as long as the device is suspended
2237 * until the bitmap is no longer on the fly during connection
2238 * handshake as long as we would exceed the max_buffer limit.
2240 * to avoid races with the reconnect code,
2241 * we need to atomic_inc within the spinlock. */
2243 wait_event(mdev->misc_wait, inc_ap_bio_cond(mdev));
2246 static inline void dec_ap_bio(struct drbd_conf *mdev)
2248 int mxb = drbd_get_max_buffers(mdev);
2249 int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt);
2251 D_ASSERT(ap_bio >= 0);
2253 if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) {
2254 if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags))
2255 drbd_queue_work(&mdev->tconn->sender_work, &mdev->bm_io_work.w);
2258 /* this currently does wake_up for every dec_ap_bio!
2259 * maybe rather introduce some type of hysteresis?
2260 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2262 wake_up(&mdev->misc_wait);
2265 static inline bool verify_can_do_stop_sector(struct drbd_conf *mdev)
2267 return mdev->tconn->agreed_pro_version >= 97 &&
2268 mdev->tconn->agreed_pro_version != 100;
2271 static inline int drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val)
2273 int changed = mdev->ed_uuid != val;
2274 mdev->ed_uuid = val;
2278 static inline int drbd_queue_order_type(struct drbd_conf *mdev)
2280 /* sorry, we currently have no working implementation
2281 * of distributed TCQ stuff */
2282 #ifndef QUEUE_ORDERED_NONE
2283 #define QUEUE_ORDERED_NONE 0
2285 return QUEUE_ORDERED_NONE;
2288 static inline void drbd_md_flush(struct drbd_conf *mdev)
2292 if (test_bit(MD_NO_FUA, &mdev->flags))
2295 r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL);
2297 set_bit(MD_NO_FUA, &mdev->flags);
2298 dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r);