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 /* 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 ratelimit_state drbd_ratelimit_state;
172 extern struct idr minors;
173 extern struct list_head drbd_tconns;
174 extern struct rw_semaphore drbd_cfg_rwsem;
175 /* drbd_cfg_rwsem protects: drbd_tconns list,
176 note: non sleeping iterations over the idrs are protoected by RCU */
180 /* receiver (data socket) */
182 P_DATA_REPLY = 0x01, /* Response to P_DATA_REQUEST */
183 P_RS_DATA_REPLY = 0x02, /* Response to P_RS_DATA_REQUEST */
186 P_BECOME_SYNC_TARGET = 0x05,
187 P_BECOME_SYNC_SOURCE = 0x06,
188 P_UNPLUG_REMOTE = 0x07, /* Used at various times to hint the peer */
189 P_DATA_REQUEST = 0x08, /* Used to ask for a data block */
190 P_RS_DATA_REQUEST = 0x09, /* Used to ask for a data block for resync */
197 P_AUTH_CHALLENGE = 0x10,
198 P_AUTH_RESPONSE = 0x11,
199 P_STATE_CHG_REQ = 0x12,
201 /* asender (meta socket */
204 P_RECV_ACK = 0x15, /* Used in protocol B */
205 P_WRITE_ACK = 0x16, /* Used in protocol C */
206 P_RS_WRITE_ACK = 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
207 P_DISCARD_WRITE = 0x18, /* Used in proto C, two-primaries conflict detection */
208 P_NEG_ACK = 0x19, /* Sent if local disk is unusable */
209 P_NEG_DREPLY = 0x1a, /* Local disk is broken... */
210 P_NEG_RS_DREPLY = 0x1b, /* Local disk is broken... */
211 P_BARRIER_ACK = 0x1c,
212 P_STATE_CHG_REPLY = 0x1d,
214 /* "new" commands, no longer fitting into the ordering scheme above */
216 P_OV_REQUEST = 0x1e, /* data socket */
218 P_OV_RESULT = 0x20, /* meta socket */
219 P_CSUM_RS_REQUEST = 0x21, /* data socket */
220 P_RS_IS_IN_SYNC = 0x22, /* meta socket */
221 P_SYNC_PARAM89 = 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
222 P_COMPRESSED_BITMAP = 0x24, /* compressed or otherwise encoded bitmap transfer */
223 /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */
224 /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */
225 P_DELAY_PROBE = 0x27, /* is used on BOTH sockets */
226 P_OUT_OF_SYNC = 0x28, /* Mark as out of sync (Outrunning), data socket */
227 P_RS_CANCEL = 0x29, /* meta: Used to cancel RS_DATA_REQUEST packet by SyncSource */
228 P_CONN_ST_CHG_REQ = 0x2a, /* data sock: Connection wide state request */
229 P_CONN_ST_CHG_REPLY = 0x2b, /* meta sock: Connection side state req reply */
230 P_RETRY_WRITE = 0x2c, /* Protocol C: retry conflicting write request */
232 P_MAY_IGNORE = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
233 P_MAX_OPT_CMD = 0x101,
235 /* special command ids for handshake */
237 P_INITIAL_META = 0xfff1, /* First Packet on the MetaSock */
238 P_INITIAL_DATA = 0xfff2, /* First Packet on the Socket */
240 P_CONNECTION_FEATURES = 0xfffe /* FIXED for the next century! */
243 extern const char *cmdname(enum drbd_packet cmd);
245 /* for sending/receiving the bitmap,
246 * possibly in some encoding scheme */
249 * stores total bits and long words
250 * of the bitmap, so we don't need to
251 * call the accessor functions over and again. */
252 unsigned long bm_bits;
253 unsigned long bm_words;
254 /* during xfer, current position within the bitmap */
255 unsigned long bit_offset;
256 unsigned long word_offset;
258 /* statistics; index: (h->command == P_BITMAP) */
263 extern void INFO_bm_xfer_stats(struct drbd_conf *mdev,
264 const char *direction, struct bm_xfer_ctx *c);
266 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
268 /* word_offset counts "native long words" (32 or 64 bit),
270 * Encoded packet may end at an unaligned bit offset.
271 * In case a fallback clear text packet is transmitted in
272 * between, we adjust this offset back to the last 64bit
273 * aligned "native long word", which makes coding and decoding
274 * the plain text bitmap much more convenient. */
275 #if BITS_PER_LONG == 64
276 c->word_offset = c->bit_offset >> 6;
277 #elif BITS_PER_LONG == 32
278 c->word_offset = c->bit_offset >> 5;
279 c->word_offset &= ~(1UL);
281 # error "unsupported BITS_PER_LONG"
286 #define __packed __attribute__((packed))
289 /* This is the layout for a packet on the wire.
290 * The byteorder is the network byte order.
291 * (except block_id and barrier fields.
292 * these are pointers to local structs
293 * and have no relevance for the partner,
294 * which just echoes them as received.)
296 * NOTE that the payload starts at a long aligned offset,
297 * regardless of 32 or 64 bit arch!
302 u16 length; /* bytes of data after this header */
305 /* Header for big packets, Used for data packets exceeding 64kB */
307 u16 magic; /* use DRBD_MAGIC_BIG here */
320 extern unsigned int drbd_header_size(struct drbd_tconn *tconn);
322 /* these defines must not be changed without changing the protocol version */
323 #define DP_HARDBARRIER 1 /* depricated */
324 #define DP_RW_SYNC 2 /* equals REQ_SYNC */
325 #define DP_MAY_SET_IN_SYNC 4
326 #define DP_UNPLUG 8 /* not used anymore */
327 #define DP_FUA 16 /* equals REQ_FUA */
328 #define DP_FLUSH 32 /* equals REQ_FLUSH */
329 #define DP_DISCARD 64 /* equals REQ_DISCARD */
332 u64 sector; /* 64 bits sector number */
333 u64 block_id; /* to identify the request in protocol B&C */
339 * commands which share a struct:
341 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
342 * P_DISCARD_WRITE (proto C, two-primaries conflict detection)
344 * P_DATA_REQUEST, P_RS_DATA_REQUEST
357 u32 pad; /* to multiple of 8 Byte */
361 * commands with their own struct for additional fields:
362 * P_CONNECTION_FEATURES
369 struct p_connection_features {
374 /* should be more than enough for future enhancements
375 * for now, feature_flags and the reserverd array shall be zero.
383 u32 barrier; /* barrier number _handle_ only */
384 u32 pad; /* to multiple of 8 Byte */
387 struct p_barrier_ack {
395 /* Since protocol version 88 and higher. */
399 struct p_rs_param_89 {
401 /* protocol version 89: */
402 char verify_alg[SHARED_SECRET_MAX];
403 char csums_alg[SHARED_SECRET_MAX];
406 struct p_rs_param_95 {
408 char verify_alg[SHARED_SECRET_MAX];
409 char csums_alg[SHARED_SECRET_MAX];
416 enum drbd_conn_flags {
429 /* Since protocol version 87 and higher. */
430 char integrity_alg[0];
435 u64 uuid[UI_EXTENDED_SIZE];
443 u64 d_size; /* size of disk */
444 u64 u_size; /* user requested size */
445 u64 c_size; /* current exported size */
446 u32 max_bio_size; /* Maximal size of a BIO */
447 u16 queue_order_type; /* not yet implemented in DRBD*/
448 u16 dds_flags; /* use enum dds_flags here. */
460 struct p_req_state_reply {
464 struct p_drbd06_param {
480 struct p_block_desc {
483 u32 pad; /* to multiple of 8 Byte */
486 /* Valid values for the encoding field.
487 * Bump proto version when changing this. */
488 enum drbd_bitmap_code {
489 /* RLE_VLI_Bytes = 0,
490 * and other bit variants had been defined during
491 * algorithm evaluation. */
495 struct p_compressed_bm {
496 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
497 * (encoding & 0x80): polarity (set/unset) of first runlength
498 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
499 * used to pad up to head.length bytes
506 struct p_delay_probe93 {
507 u32 seq_num; /* sequence number to match the two probe packets */
508 u32 offset; /* usecs the probe got sent after the reference time point */
512 * Bitmap packets need to fit within a single page on the sender and receiver,
513 * so we are limited to 4 KiB (and not to PAGE_SIZE, which can be bigger).
515 #define DRBD_SOCKET_BUFFER_SIZE 4096
517 /**********************************************************************/
518 enum drbd_thread_state {
527 struct task_struct *task;
528 struct completion stop;
529 enum drbd_thread_state t_state;
530 int (*function) (struct drbd_thread *);
531 struct drbd_tconn *tconn;
536 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
538 /* THINK testing the t_state seems to be uncritical in all cases
539 * (but thread_{start,stop}), so we can read it *without* the lock.
547 struct list_head list;
548 int (*cb)(struct drbd_work *, int cancel);
550 struct drbd_conf *mdev;
551 struct drbd_tconn *tconn;
555 #include "drbd_interval.h"
557 extern int drbd_wait_misc(struct drbd_conf *, struct drbd_interval *);
559 struct drbd_request {
562 /* if local IO is not allowed, will be NULL.
563 * if local IO _is_ allowed, holds the locally submitted bio clone,
564 * or, after local IO completion, the ERR_PTR(error).
565 * see drbd_request_endio(). */
566 struct bio *private_bio;
568 struct drbd_interval i;
569 unsigned int epoch; /* barrier_nr */
571 /* barrier_nr: used to check on "completion" whether this req was in
572 * the current epoch, and we therefore have to close it,
573 * starting a new epoch...
576 struct list_head tl_requests; /* ring list in the transfer log */
577 struct bio *master_bio; /* master bio pointer */
578 unsigned long rq_state; /* see comments above _req_mod() */
580 unsigned long start_time;
583 struct drbd_tl_epoch {
585 struct list_head requests; /* requests before */
586 struct drbd_tl_epoch *next; /* pointer to the next barrier */
587 unsigned int br_number; /* the barriers identifier. */
588 int n_writes; /* number of requests attached before this barrier */
592 struct list_head list;
593 unsigned int barrier_nr;
594 atomic_t epoch_size; /* increased on every request added. */
595 atomic_t active; /* increased on every req. added, and dec on every finished. */
599 /* drbd_epoch flag bits */
601 DE_HAVE_BARRIER_NUMBER,
608 EV_CLEANUP = 32, /* used as flag */
611 struct drbd_wq_barrier {
613 struct completion done;
621 struct drbd_peer_request {
623 struct drbd_epoch *epoch; /* for writes */
625 atomic_t pending_bios;
626 struct drbd_interval i;
627 /* see comments on ee flag bits below */
631 struct digest_info *digest;
636 * While corresponding bios are in flight, the only modification will be
637 * set_bit WAS_ERROR, which has to be atomic.
638 * If no bios are in flight yet, or all have been completed,
639 * non-atomic modification to ee->flags is ok.
642 __EE_CALL_AL_COMPLETE_IO,
643 __EE_MAY_SET_IN_SYNC,
645 /* In case a barrier failed,
646 * we need to resubmit without the barrier flag. */
649 /* we may have several bios per peer request.
650 * if any of those fail, we set this flag atomically
651 * from the endio callback */
654 /* This ee has a pointer to a digest instead of a block id */
657 /* Conflicting local requests need to be restarted after this request */
658 __EE_RESTART_REQUESTS,
660 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
661 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
662 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
663 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
664 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
665 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
667 /* flag bits per mdev */
669 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */
670 UNPLUG_QUEUED, /* only relevant with kernel 2.4 */
671 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
672 MD_DIRTY, /* current uuids and flags not yet on disk */
673 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
676 CRASHED_PRIMARY, /* This node was a crashed primary.
677 * Gets cleared when the state.conn
678 * goes into C_CONNECTED state. */
679 NO_BARRIER_SUPP, /* underlying block device doesn't implement barriers */
682 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */
683 SUSPEND_IO, /* suspend application io */
684 BITMAP_IO, /* suspend application io;
685 once no more io in flight, start bitmap io */
686 BITMAP_IO_QUEUED, /* Started bitmap IO */
687 GO_DISKLESS, /* Disk is being detached, on io-error or admin request. */
688 WAS_IO_ERROR, /* Local disk failed returned IO error */
689 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
690 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
691 * the peer, if it changed there as well. */
692 NEW_CUR_UUID, /* Create new current UUID when thawing IO */
693 AL_SUSPENDED, /* Activity logging is currently suspended. */
694 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */
695 B_RS_H_DONE, /* Before resync handler done (already executed) */
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 = 0x7,
709 /* in detail, that is: */
714 /* (test bit, count bit) allowed (common case) */
715 BM_LOCKED_TEST_ALLOWED = 0x3,
717 /* testing bits, as well as setting new bits allowed, but clearing bits
718 * would be unexpected. Used during bitmap receive. Setting new bits
719 * requires sending of "out-of-sync" information, though. */
720 BM_LOCKED_SET_ALLOWED = 0x1,
722 /* clear is not expected while bitmap is locked for bulk operation */
726 /* TODO sort members for performance
727 * MAYBE group them further */
729 /* THINK maybe we actually want to use the default "event/%s" worker threads
730 * or similar in linux 2.6, which uses per cpu data and threads.
732 struct drbd_work_queue {
734 struct semaphore s; /* producers up it, worker down()s it */
735 spinlock_t q_lock; /* to protect the list. */
739 struct drbd_work_queue work;
741 struct socket *socket;
742 /* this way we get our
743 * send/receive buffers off the stack */
749 u64 md_offset; /* sector offset to 'super' block */
751 u64 la_size_sect; /* last agreed size, unit sectors */
757 s32 al_offset; /* signed relative sector offset to al area */
758 s32 bm_offset; /* signed relative sector offset to bitmap */
760 /* u32 al_nr_extents; important for restoring the AL
761 * is stored into ldev->dc.al_extents, which in turn
762 * gets applied to act_log->nr_elements
766 struct drbd_backing_dev {
767 struct block_device *backing_bdev;
768 struct block_device *md_bdev;
770 struct disk_conf dc; /* The user provided config... */
771 sector_t known_size; /* last known size of that backing device */
775 struct drbd_conf *mdev;
776 struct completion event;
784 int (*io_fn)(struct drbd_conf *mdev);
785 void (*done)(struct drbd_conf *mdev, int rv);
788 enum write_ordering_e {
796 unsigned int head_index;
800 /* flag bits per tconn */
802 NET_CONGESTED, /* The data socket is congested */
803 DISCARD_CONCURRENT, /* 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 */
809 CONFIG_PENDING, /* serialization of (re)configuration requests.
810 * if set, also prevents the device from dying */
811 OBJECT_DYING, /* device became unconfigured,
812 * but worker thread is still handling the cleanup.
813 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
814 * while this is set. */
815 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
818 struct drbd_tconn { /* is a resource from the config file */
819 char *name; /* Resource name */
820 struct list_head all_tconn; /* linked on global drbd_tconns */
821 struct idr volumes; /* <tconn, vnr> to mdev mapping */
822 enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
823 unsigned susp:1; /* IO suspended by user */
824 unsigned susp_nod:1; /* IO suspended because no data */
825 unsigned susp_fen:1; /* IO suspended because fence peer handler runs */
826 struct mutex cstate_mutex; /* Protects graceful disconnects */
829 struct net_conf *net_conf; /* protected by get_net_conf() and put_net_conf() */
830 atomic_t net_cnt; /* Users of net_conf */
831 wait_queue_head_t net_cnt_wait;
832 wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */
833 struct res_opts res_opts;
835 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
836 struct drbd_socket meta; /* ping/ack (metadata) packets */
837 int agreed_pro_version; /* actually used protocol version */
838 unsigned long last_received; /* in jiffies, either socket */
839 unsigned int ko_count;
842 struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */
843 struct drbd_tl_epoch *newest_tle;
844 struct drbd_tl_epoch *oldest_tle;
845 struct list_head out_of_sequence_requests;
847 struct crypto_hash *cram_hmac_tfm;
848 struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */
849 struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */
850 struct crypto_hash *csums_tfm;
851 struct crypto_hash *verify_tfm;
855 struct drbd_thread receiver;
856 struct drbd_thread worker;
857 struct drbd_thread asender;
858 cpumask_var_t cpu_mask;
862 struct drbd_tconn *tconn;
863 int vnr; /* volume number within the connection */
865 /* things that are stored as / read from meta data on disk */
868 /* configured by drbdsetup */
869 struct drbd_backing_dev *ldev __protected_by(local);
871 sector_t p_size; /* partner's disk size */
872 struct request_queue *rq_queue;
873 struct block_device *this_bdev;
874 struct gendisk *vdisk;
876 struct drbd_work resync_work,
881 struct timer_list resync_timer;
882 struct timer_list md_sync_timer;
883 struct timer_list start_resync_timer;
884 struct timer_list request_timer;
885 #ifdef DRBD_DEBUG_MD_SYNC
889 } last_md_mark_dirty;
892 /* Used after attach while negotiating new disk state. */
893 union drbd_state new_state_tmp;
895 union drbd_dev_state state;
896 wait_queue_head_t misc_wait;
897 wait_queue_head_t state_wait; /* upon each state change. */
898 unsigned int send_cnt;
899 unsigned int recv_cnt;
900 unsigned int read_cnt;
901 unsigned int writ_cnt;
902 unsigned int al_writ_cnt;
903 unsigned int bm_writ_cnt;
904 atomic_t ap_bio_cnt; /* Requests we need to complete */
905 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
906 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
907 atomic_t unacked_cnt; /* Need to send replys for */
908 atomic_t local_cnt; /* Waiting for local completion */
910 /* Interval tree of pending local requests */
911 struct rb_root read_requests;
912 struct rb_root write_requests;
914 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
915 unsigned long rs_total;
916 /* number of resync blocks that failed in this run */
917 unsigned long rs_failed;
918 /* Syncer's start time [unit jiffies] */
919 unsigned long rs_start;
920 /* cumulated time in PausedSyncX state [unit jiffies] */
921 unsigned long rs_paused;
922 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
923 unsigned long rs_same_csum;
924 #define DRBD_SYNC_MARKS 8
925 #define DRBD_SYNC_MARK_STEP (3*HZ)
926 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
927 unsigned long rs_mark_left[DRBD_SYNC_MARKS];
928 /* marks's time [unit jiffies] */
929 unsigned long rs_mark_time[DRBD_SYNC_MARKS];
930 /* current index into rs_mark_{left,time} */
933 /* where does the admin want us to start? (sector) */
934 sector_t ov_start_sector;
935 /* where are we now? (sector) */
936 sector_t ov_position;
937 /* Start sector of out of sync range (to merge printk reporting). */
938 sector_t ov_last_oos_start;
939 /* size of out-of-sync range in sectors. */
940 sector_t ov_last_oos_size;
941 unsigned long ov_left; /* in bits */
943 struct drbd_bitmap *bitmap;
944 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
946 /* Used to track operations of resync... */
947 struct lru_cache *resync;
948 /* Number of locked elements in resync LRU */
949 unsigned int resync_locked;
950 /* resync extent number waiting for application requests */
951 unsigned int resync_wenr;
955 struct drbd_epoch *current_epoch;
956 spinlock_t epoch_lock;
958 enum write_ordering_e write_ordering;
959 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
960 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
961 struct list_head done_ee; /* need to send P_WRITE_ACK */
962 struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */
963 struct list_head net_ee; /* zero-copy network send in progress */
966 struct list_head resync_reads;
967 atomic_t pp_in_use; /* allocated from page pool */
968 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */
969 wait_queue_head_t ee_wait;
970 struct page *md_io_page; /* one page buffer for md_io */
971 struct mutex md_io_mutex; /* protects the md_io_buffer */
973 wait_queue_head_t al_wait;
974 struct lru_cache *act_log; /* activity log */
975 unsigned int al_tr_number;
977 int al_tr_pos; /* position of the next transaction in the journal */
978 wait_queue_head_t seq_wait;
980 unsigned int peer_seq;
981 spinlock_t peer_seq_lock;
983 unsigned long comm_bm_set; /* communicated number of set bits. */
984 struct bm_io_work bm_io_work;
985 u64 ed_uuid; /* UUID of the exposed data */
986 struct mutex own_state_mutex;
987 struct mutex *state_mutex; /* either own_state_mutex or mdev->tconn->cstate_mutex */
988 char congestion_reason; /* Why we where congested... */
989 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
990 atomic_t rs_sect_ev; /* for submitted resync data rate, both */
991 int rs_last_sect_ev; /* counter to compare with */
992 int rs_last_events; /* counter of read or write "events" (unit sectors)
993 * on the lower level device when we last looked. */
994 int c_sync_rate; /* current resync rate after syncer throttle magic */
995 struct fifo_buffer rs_plan_s; /* correction values of resync planer */
996 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
997 int rs_planed; /* resync sectors already planned */
998 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
999 int peer_max_bio_size;
1000 int local_max_bio_size;
1003 static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
1005 return (struct drbd_conf *)idr_find(&minors, minor);
1008 static inline unsigned int mdev_to_minor(struct drbd_conf *mdev)
1013 static inline struct drbd_conf *vnr_to_mdev(struct drbd_tconn *tconn, int vnr)
1015 return (struct drbd_conf *)idr_find(&tconn->volumes, vnr);
1019 * function declarations
1020 *************************/
1026 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1029 extern void drbd_init_set_defaults(struct drbd_conf *mdev);
1030 extern int drbd_thread_start(struct drbd_thread *thi);
1031 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1032 extern char *drbd_task_to_thread_name(struct drbd_tconn *tconn, struct task_struct *task);
1034 extern void drbd_thread_current_set_cpu(struct drbd_thread *thi);
1035 extern void drbd_calc_cpu_mask(struct drbd_tconn *tconn);
1037 #define drbd_thread_current_set_cpu(A) ({})
1038 #define drbd_calc_cpu_mask(A) ({})
1040 extern void drbd_free_resources(struct drbd_conf *mdev);
1041 extern void tl_release(struct drbd_tconn *, unsigned int barrier_nr,
1042 unsigned int set_size);
1043 extern void tl_clear(struct drbd_tconn *);
1044 extern void _tl_add_barrier(struct drbd_tconn *, struct drbd_tl_epoch *);
1045 extern void drbd_free_sock(struct drbd_tconn *tconn);
1046 extern int drbd_send(struct drbd_tconn *tconn, struct socket *sock,
1047 void *buf, size_t size, unsigned msg_flags);
1048 extern int drbd_send_all(struct drbd_tconn *, struct socket *, void *, size_t,
1051 extern int drbd_send_protocol(struct drbd_tconn *tconn);
1052 extern int drbd_send_uuids(struct drbd_conf *mdev);
1053 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
1054 extern void drbd_gen_and_send_sync_uuid(struct drbd_conf *mdev);
1055 extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags);
1056 extern int _conn_send_state_req(struct drbd_tconn *, int vnr, enum drbd_packet cmd,
1057 union drbd_state, union drbd_state);
1058 extern int _drbd_send_state(struct drbd_conf *mdev);
1059 extern int drbd_send_state(struct drbd_conf *mdev);
1060 extern int drbd_send_sync_param(struct drbd_conf *mdev);
1061 extern void drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr,
1063 extern int drbd_send_ack(struct drbd_conf *, enum drbd_packet,
1064 struct drbd_peer_request *);
1065 extern void drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packet cmd,
1066 struct p_block_req *rp);
1067 extern void drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packet cmd,
1068 struct p_data *dp, int data_size);
1069 extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packet cmd,
1070 sector_t sector, int blksize, u64 block_id);
1071 extern int drbd_send_out_of_sync(struct drbd_conf *, struct drbd_request *);
1072 extern int drbd_send_block(struct drbd_conf *, enum drbd_packet,
1073 struct drbd_peer_request *);
1074 extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
1075 extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1076 sector_t sector, int size, u64 block_id);
1077 extern int drbd_send_drequest_csum(struct drbd_conf *mdev, sector_t sector,
1078 int size, void *digest, int digest_size,
1079 enum drbd_packet cmd);
1080 extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size);
1082 extern int drbd_send_bitmap(struct drbd_conf *mdev);
1083 extern void drbd_send_sr_reply(struct drbd_conf *mdev, enum drbd_state_rv retcode);
1084 extern void conn_send_sr_reply(struct drbd_tconn *tconn, enum drbd_state_rv retcode);
1085 extern void drbd_free_bc(struct drbd_backing_dev *ldev);
1086 extern void drbd_mdev_cleanup(struct drbd_conf *mdev);
1087 void drbd_print_uuids(struct drbd_conf *mdev, const char *text);
1089 extern void drbd_md_sync(struct drbd_conf *mdev);
1090 extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
1091 extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1092 extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1093 extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1094 extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1095 extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
1096 extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
1097 extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
1098 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1099 #ifndef DRBD_DEBUG_MD_SYNC
1100 extern void drbd_md_mark_dirty(struct drbd_conf *mdev);
1102 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1103 extern void drbd_md_mark_dirty_(struct drbd_conf *mdev,
1104 unsigned int line, const char *func);
1106 extern void drbd_queue_bitmap_io(struct drbd_conf *mdev,
1107 int (*io_fn)(struct drbd_conf *),
1108 void (*done)(struct drbd_conf *, int),
1109 char *why, enum bm_flag flags);
1110 extern int drbd_bitmap_io(struct drbd_conf *mdev,
1111 int (*io_fn)(struct drbd_conf *),
1112 char *why, enum bm_flag flags);
1113 extern int drbd_bmio_set_n_write(struct drbd_conf *mdev);
1114 extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev);
1115 extern void drbd_go_diskless(struct drbd_conf *mdev);
1116 extern void drbd_ldev_destroy(struct drbd_conf *mdev);
1119 We reserve a 128MB Block (4k aligned)
1120 * either at the end of the backing device
1121 * or on a separate meta data device. */
1123 /* The following numbers are sectors */
1124 /* Allows up to about 3.8TB, so if you want more,
1125 * you need to use the "flexible" meta data format. */
1126 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1127 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1128 #define MD_AL_SECTORS 64 /* = 32 kB on disk activity log ring buffer */
1129 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_SECTORS)
1131 /* we do all meta data IO in 4k blocks */
1132 #define MD_BLOCK_SHIFT 12
1133 #define MD_BLOCK_SIZE (1<<MD_BLOCK_SHIFT)
1135 /* One activity log extent represents 4M of storage */
1136 #define AL_EXTENT_SHIFT 22
1137 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1139 /* We could make these currently hardcoded constants configurable
1140 * variables at create-md time (or even re-configurable at runtime?).
1141 * Which will require some more changes to the DRBD "super block"
1144 * updates per transaction:
1145 * This many changes to the active set can be logged with one transaction.
1146 * This number is arbitrary.
1147 * context per transaction:
1148 * This many context extent numbers are logged with each transaction.
1149 * This number is resulting from the transaction block size (4k), the layout
1150 * of the transaction header, and the number of updates per transaction.
1151 * See drbd_actlog.c:struct al_transaction_on_disk
1153 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1154 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1156 #if BITS_PER_LONG == 32
1158 #define cpu_to_lel(A) cpu_to_le32(A)
1159 #define lel_to_cpu(A) le32_to_cpu(A)
1160 #elif BITS_PER_LONG == 64
1162 #define cpu_to_lel(A) cpu_to_le64(A)
1163 #define lel_to_cpu(A) le64_to_cpu(A)
1165 #error "LN2 of BITS_PER_LONG unknown!"
1169 /* 16MB sized 'bitmap extent' to track syncer usage */
1171 int rs_left; /* number of bits set (out of sync) in this extent. */
1172 int rs_failed; /* number of failed resync requests in this extent. */
1173 unsigned long flags;
1174 struct lc_element lce;
1177 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1178 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1179 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1183 * We need to store one bit for a block.
1184 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1185 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1186 * Bit 1 ==> local node thinks this block needs to be synced.
1189 #define SLEEP_TIME (HZ/10)
1191 /* We do bitmap IO in units of 4k blocks.
1192 * We also still have a hardcoded 4k per bit relation. */
1193 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1194 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1195 /* mostly arbitrarily set the represented size of one bitmap extent,
1196 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1197 * at 4k per bit resolution) */
1198 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
1199 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1201 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1202 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1205 /* thus many _storage_ sectors are described by one bit */
1206 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1207 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1208 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1210 /* bit to represented kilo byte conversion */
1211 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1213 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1214 * _storage_ sector is located in */
1215 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1217 /* how much _storage_ sectors we have per bitmap sector */
1218 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1219 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1221 /* in one sector of the bitmap, we have this many activity_log extents. */
1222 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1223 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1225 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1226 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1228 /* the extent in "PER_EXTENT" below is an activity log extent
1229 * we need that many (long words/bytes) to store the bitmap
1230 * of one AL_EXTENT_SIZE chunk of storage.
1231 * we can store the bitmap for that many AL_EXTENTS within
1232 * one sector of the _on_disk_ bitmap:
1233 * bit 0 bit 37 bit 38 bit (512*8)-1
1234 * ...|........|........|.. // ..|........|
1235 * sect. 0 `296 `304 ^(512*8*8)-1
1237 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1238 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1239 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1242 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1243 #define DRBD_MAX_SECTORS_BM \
1244 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1245 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1246 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1247 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1248 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1249 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1250 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1252 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1253 /* 16 TB in units of sectors */
1254 #if BITS_PER_LONG == 32
1255 /* adjust by one page worth of bitmap,
1256 * so we won't wrap around in drbd_bm_find_next_bit.
1257 * you should use 64bit OS for that much storage, anyways. */
1258 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1260 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1261 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1262 /* corresponds to (1UL << 38) bits right now. */
1266 /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE,
1267 * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte.
1268 * Since we may live in a mixed-platform cluster,
1269 * we limit us to a platform agnostic constant here for now.
1270 * A followup commit may allow even bigger BIO sizes,
1271 * once we thought that through. */
1272 #define DRBD_MAX_BIO_SIZE (1 << 20)
1273 #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1274 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1276 #define DRBD_MAX_BIO_SIZE_SAFE (1 << 12) /* Works always = 4k */
1278 #define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */
1280 extern int drbd_bm_init(struct drbd_conf *mdev);
1281 extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits);
1282 extern void drbd_bm_cleanup(struct drbd_conf *mdev);
1283 extern void drbd_bm_set_all(struct drbd_conf *mdev);
1284 extern void drbd_bm_clear_all(struct drbd_conf *mdev);
1285 /* set/clear/test only a few bits at a time */
1286 extern int drbd_bm_set_bits(
1287 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1288 extern int drbd_bm_clear_bits(
1289 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1290 extern int drbd_bm_count_bits(
1291 struct drbd_conf *mdev, const unsigned long s, const unsigned long e);
1292 /* bm_set_bits variant for use while holding drbd_bm_lock,
1293 * may process the whole bitmap in one go */
1294 extern void _drbd_bm_set_bits(struct drbd_conf *mdev,
1295 const unsigned long s, const unsigned long e);
1296 extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr);
1297 extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr);
1298 extern int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local);
1299 extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
1300 extern void drbd_bm_mark_for_writeout(struct drbd_conf *mdev, int page_nr);
1301 extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
1302 extern int drbd_bm_write_hinted(struct drbd_conf *mdev) __must_hold(local);
1303 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev,
1304 unsigned long al_enr);
1305 extern size_t drbd_bm_words(struct drbd_conf *mdev);
1306 extern unsigned long drbd_bm_bits(struct drbd_conf *mdev);
1307 extern sector_t drbd_bm_capacity(struct drbd_conf *mdev);
1309 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1310 extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1311 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1312 extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1313 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo);
1314 extern unsigned long _drbd_bm_total_weight(struct drbd_conf *mdev);
1315 extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev);
1316 extern int drbd_bm_rs_done(struct drbd_conf *mdev);
1317 /* for receive_bitmap */
1318 extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset,
1319 size_t number, unsigned long *buffer);
1320 /* for _drbd_send_bitmap */
1321 extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset,
1322 size_t number, unsigned long *buffer);
1324 extern void drbd_bm_lock(struct drbd_conf *mdev, char *why, enum bm_flag flags);
1325 extern void drbd_bm_unlock(struct drbd_conf *mdev);
1328 extern struct kmem_cache *drbd_request_cache;
1329 extern struct kmem_cache *drbd_ee_cache; /* peer requests */
1330 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1331 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1332 extern mempool_t *drbd_request_mempool;
1333 extern mempool_t *drbd_ee_mempool;
1335 /* drbd's page pool, used to buffer data received from the peer,
1336 * or data requested by the peer.
1338 * This does not have an emergency reserve.
1340 * When allocating from this pool, it first takes pages from the pool.
1341 * Only if the pool is depleted will try to allocate from the system.
1343 * The assumption is that pages taken from this pool will be processed,
1344 * and given back, "quickly", and then can be recycled, so we can avoid
1345 * frequent calls to alloc_page(), and still will be able to make progress even
1346 * under memory pressure.
1348 extern struct page *drbd_pp_pool;
1349 extern spinlock_t drbd_pp_lock;
1350 extern int drbd_pp_vacant;
1351 extern wait_queue_head_t drbd_pp_wait;
1353 /* We also need a standard (emergency-reserve backed) page pool
1354 * for meta data IO (activity log, bitmap).
1355 * We can keep it global, as long as it is used as "N pages at a time".
1356 * 128 should be plenty, currently we probably can get away with as few as 1.
1358 #define DRBD_MIN_POOL_PAGES 128
1359 extern mempool_t *drbd_md_io_page_pool;
1361 /* We also need to make sure we get a bio
1362 * when we need it for housekeeping purposes */
1363 extern struct bio_set *drbd_md_io_bio_set;
1364 /* to allocate from that set */
1365 extern struct bio *bio_alloc_drbd(gfp_t gfp_mask);
1367 extern rwlock_t global_state_lock;
1369 extern int conn_lowest_minor(struct drbd_tconn *tconn);
1370 enum drbd_ret_code conn_new_minor(struct drbd_tconn *tconn, unsigned int minor, int vnr);
1371 extern void drbd_delete_device(struct drbd_conf *mdev);
1373 struct drbd_tconn *drbd_new_tconn(const char *name);
1374 extern void drbd_free_tconn(struct drbd_tconn *tconn);
1375 struct drbd_tconn *conn_by_name(const char *name);
1377 extern int proc_details;
1380 extern int __drbd_make_request(struct drbd_conf *, struct bio *, unsigned long);
1381 extern int drbd_make_request(struct request_queue *q, struct bio *bio);
1382 extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req);
1383 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1384 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1388 extern int drbd_msg_put_info(const char *info);
1389 extern void drbd_suspend_io(struct drbd_conf *mdev);
1390 extern void drbd_resume_io(struct drbd_conf *mdev);
1391 extern char *ppsize(char *buf, unsigned long long size);
1392 extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int);
1393 enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
1394 extern enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local);
1395 extern void resync_after_online_grow(struct drbd_conf *);
1396 extern void drbd_reconsider_max_bio_size(struct drbd_conf *mdev);
1397 extern enum drbd_state_rv drbd_set_role(struct drbd_conf *mdev,
1398 enum drbd_role new_role,
1400 extern bool conn_try_outdate_peer(struct drbd_tconn *tconn);
1401 extern void conn_try_outdate_peer_async(struct drbd_tconn *tconn);
1402 extern int drbd_khelper(struct drbd_conf *mdev, char *cmd);
1405 extern int drbd_worker(struct drbd_thread *thi);
1406 extern int drbd_alter_sa(struct drbd_conf *mdev, int na);
1407 extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side);
1408 extern void resume_next_sg(struct drbd_conf *mdev);
1409 extern void suspend_other_sg(struct drbd_conf *mdev);
1410 extern int drbd_resync_finished(struct drbd_conf *mdev);
1411 /* maybe rather drbd_main.c ? */
1412 extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
1413 struct drbd_backing_dev *bdev, sector_t sector, int rw);
1414 extern void drbd_ov_out_of_sync_found(struct drbd_conf *, sector_t, int);
1415 extern void drbd_rs_controller_reset(struct drbd_conf *mdev);
1417 static inline void ov_out_of_sync_print(struct drbd_conf *mdev)
1419 if (mdev->ov_last_oos_size) {
1420 dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n",
1421 (unsigned long long)mdev->ov_last_oos_start,
1422 (unsigned long)mdev->ov_last_oos_size);
1424 mdev->ov_last_oos_size=0;
1428 extern void drbd_csum_bio(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
1429 extern void drbd_csum_ee(struct drbd_conf *, struct crypto_hash *,
1430 struct drbd_peer_request *, void *);
1431 /* worker callbacks */
1432 extern int w_read_retry_remote(struct drbd_work *, int);
1433 extern int w_e_end_data_req(struct drbd_work *, int);
1434 extern int w_e_end_rsdata_req(struct drbd_work *, int);
1435 extern int w_e_end_csum_rs_req(struct drbd_work *, int);
1436 extern int w_e_end_ov_reply(struct drbd_work *, int);
1437 extern int w_e_end_ov_req(struct drbd_work *, int);
1438 extern int w_ov_finished(struct drbd_work *, int);
1439 extern int w_resync_timer(struct drbd_work *, int);
1440 extern int w_send_write_hint(struct drbd_work *, int);
1441 extern int w_make_resync_request(struct drbd_work *, int);
1442 extern int w_send_dblock(struct drbd_work *, int);
1443 extern int w_send_barrier(struct drbd_work *, int);
1444 extern int w_send_read_req(struct drbd_work *, int);
1445 extern int w_prev_work_done(struct drbd_work *, int);
1446 extern int w_e_reissue(struct drbd_work *, int);
1447 extern int w_restart_disk_io(struct drbd_work *, int);
1448 extern int w_send_out_of_sync(struct drbd_work *, int);
1449 extern int w_start_resync(struct drbd_work *, int);
1451 extern void resync_timer_fn(unsigned long data);
1452 extern void start_resync_timer_fn(unsigned long data);
1454 /* drbd_receiver.c */
1455 extern int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector);
1456 extern int drbd_submit_peer_request(struct drbd_conf *,
1457 struct drbd_peer_request *, const unsigned,
1459 extern int drbd_free_peer_reqs(struct drbd_conf *, struct list_head *);
1460 extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_conf *, u64,
1461 sector_t, unsigned int,
1462 gfp_t) __must_hold(local);
1463 extern void __drbd_free_peer_req(struct drbd_conf *, struct drbd_peer_request *,
1465 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1466 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
1467 extern struct page *drbd_alloc_pages(struct drbd_conf *, unsigned int, bool);
1468 extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled);
1469 extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed);
1470 extern void conn_flush_workqueue(struct drbd_tconn *tconn);
1471 extern int drbd_connected(int vnr, void *p, void *data);
1472 static inline void drbd_flush_workqueue(struct drbd_conf *mdev)
1474 conn_flush_workqueue(mdev->tconn);
1477 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1478 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1479 static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1480 char __user *optval, int optlen)
1483 if (level == SOL_SOCKET)
1484 err = sock_setsockopt(sock, level, optname, optval, optlen);
1486 err = sock->ops->setsockopt(sock, level, optname, optval,
1491 static inline void drbd_tcp_cork(struct socket *sock)
1494 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1495 (char __user *)&val, sizeof(val));
1498 static inline void drbd_tcp_uncork(struct socket *sock)
1501 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1502 (char __user *)&val, sizeof(val));
1505 static inline void drbd_tcp_nodelay(struct socket *sock)
1508 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1509 (char __user *)&val, sizeof(val));
1512 static inline void drbd_tcp_quickack(struct socket *sock)
1515 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1516 (char __user *)&val, sizeof(val));
1519 void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo);
1522 extern struct proc_dir_entry *drbd_proc;
1523 extern const struct file_operations drbd_proc_fops;
1524 extern const char *drbd_conn_str(enum drbd_conns s);
1525 extern const char *drbd_role_str(enum drbd_role s);
1528 extern void drbd_al_begin_io(struct drbd_conf *mdev, struct drbd_interval *i);
1529 extern void drbd_al_complete_io(struct drbd_conf *mdev, struct drbd_interval *i);
1530 extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector);
1531 extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1532 extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1533 extern void drbd_rs_cancel_all(struct drbd_conf *mdev);
1534 extern int drbd_rs_del_all(struct drbd_conf *mdev);
1535 extern void drbd_rs_failed_io(struct drbd_conf *mdev,
1536 sector_t sector, int size);
1537 extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *);
1538 extern void drbd_advance_rs_marks(struct drbd_conf *mdev, unsigned long still_to_go);
1539 extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
1540 int size, const char *file, const unsigned int line);
1541 #define drbd_set_in_sync(mdev, sector, size) \
1542 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1543 extern int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
1544 int size, const char *file, const unsigned int line);
1545 #define drbd_set_out_of_sync(mdev, sector, size) \
1546 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1547 extern void drbd_al_apply_to_bm(struct drbd_conf *mdev);
1548 extern void drbd_al_shrink(struct drbd_conf *mdev);
1551 /* state info broadcast */
1553 enum drbd_state_info_bcast_reason sib_reason;
1557 unsigned helper_exit_code;
1560 union drbd_state os;
1561 union drbd_state ns;
1565 void drbd_bcast_event(struct drbd_conf *mdev, const struct sib_info *sib);
1568 * inline helper functions
1569 *************************/
1571 /* see also page_chain_add and friends in drbd_receiver.c */
1572 static inline struct page *page_chain_next(struct page *page)
1574 return (struct page *)page_private(page);
1576 #define page_chain_for_each(page) \
1577 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1578 page = page_chain_next(page))
1579 #define page_chain_for_each_safe(page, n) \
1580 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1582 static inline int drbd_bio_has_active_page(struct bio *bio)
1584 struct bio_vec *bvec;
1587 __bio_for_each_segment(bvec, bio, i, 0) {
1588 if (page_count(bvec->bv_page) > 1)
1595 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req)
1597 struct page *page = peer_req->pages;
1598 page_chain_for_each(page) {
1599 if (page_count(page) > 1)
1605 static inline enum drbd_state_rv
1606 _drbd_set_state(struct drbd_conf *mdev, union drbd_state ns,
1607 enum chg_state_flags flags, struct completion *done)
1609 enum drbd_state_rv rv;
1611 read_lock(&global_state_lock);
1612 rv = __drbd_set_state(mdev, ns, flags, done);
1613 read_unlock(&global_state_lock);
1618 static inline union drbd_state drbd_read_state(struct drbd_conf *mdev)
1620 union drbd_state rv;
1622 rv.i = mdev->state.i;
1623 rv.susp = mdev->tconn->susp;
1624 rv.susp_nod = mdev->tconn->susp_nod;
1625 rv.susp_fen = mdev->tconn->susp_fen;
1630 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1631 static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach, const char *where)
1633 switch (mdev->ldev->dc.on_io_error) {
1636 if (__ratelimit(&drbd_ratelimit_state))
1637 dev_err(DEV, "Local IO failed in %s.\n", where);
1638 if (mdev->state.disk > D_INCONSISTENT)
1639 _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_HARD, NULL);
1642 /* NOTE fall through to detach case if forcedetach set */
1644 case EP_CALL_HELPER:
1645 set_bit(WAS_IO_ERROR, &mdev->flags);
1646 if (mdev->state.disk > D_FAILED) {
1647 _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL);
1649 "Local IO failed in %s. Detaching...\n", where);
1656 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1657 * @mdev: DRBD device.
1658 * @error: Error code passed to the IO completion callback
1659 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1661 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1663 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1664 static inline void drbd_chk_io_error_(struct drbd_conf *mdev,
1665 int error, int forcedetach, const char *where)
1668 unsigned long flags;
1669 spin_lock_irqsave(&mdev->tconn->req_lock, flags);
1670 __drbd_chk_io_error_(mdev, forcedetach, where);
1671 spin_unlock_irqrestore(&mdev->tconn->req_lock, flags);
1677 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1678 * @bdev: Meta data block device.
1680 * BTW, for internal meta data, this happens to be the maximum capacity
1681 * we could agree upon with our peer node.
1683 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1685 switch (bdev->dc.meta_dev_idx) {
1686 case DRBD_MD_INDEX_INTERNAL:
1687 case DRBD_MD_INDEX_FLEX_INT:
1688 return bdev->md.md_offset + bdev->md.bm_offset;
1689 case DRBD_MD_INDEX_FLEX_EXT:
1691 return bdev->md.md_offset;
1696 * drbd_md_last_sector() - Return the last sector number of the meta data area
1697 * @bdev: Meta data block device.
1699 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1701 switch (bdev->dc.meta_dev_idx) {
1702 case DRBD_MD_INDEX_INTERNAL:
1703 case DRBD_MD_INDEX_FLEX_INT:
1704 return bdev->md.md_offset + MD_AL_OFFSET - 1;
1705 case DRBD_MD_INDEX_FLEX_EXT:
1707 return bdev->md.md_offset + bdev->md.md_size_sect;
1711 /* Returns the number of 512 byte sectors of the device */
1712 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1714 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1715 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0;
1719 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1720 * @bdev: Meta data block device.
1722 * returns the capacity we announce to out peer. we clip ourselves at the
1723 * various MAX_SECTORS, because if we don't, current implementation will
1724 * oops sooner or later
1726 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1729 switch (bdev->dc.meta_dev_idx) {
1730 case DRBD_MD_INDEX_INTERNAL:
1731 case DRBD_MD_INDEX_FLEX_INT:
1732 s = drbd_get_capacity(bdev->backing_bdev)
1733 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1734 drbd_md_first_sector(bdev))
1737 case DRBD_MD_INDEX_FLEX_EXT:
1738 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1739 drbd_get_capacity(bdev->backing_bdev));
1740 /* clip at maximum size the meta device can support */
1741 s = min_t(sector_t, s,
1742 BM_EXT_TO_SECT(bdev->md.md_size_sect
1743 - bdev->md.bm_offset));
1746 s = min_t(sector_t, DRBD_MAX_SECTORS,
1747 drbd_get_capacity(bdev->backing_bdev));
1753 * drbd_md_ss__() - Return the sector number of our meta data super block
1754 * @mdev: DRBD device.
1755 * @bdev: Meta data block device.
1757 static inline sector_t drbd_md_ss__(struct drbd_conf *mdev,
1758 struct drbd_backing_dev *bdev)
1760 switch (bdev->dc.meta_dev_idx) {
1761 default: /* external, some index */
1762 return MD_RESERVED_SECT * bdev->dc.meta_dev_idx;
1763 case DRBD_MD_INDEX_INTERNAL:
1764 /* with drbd08, internal meta data is always "flexible" */
1765 case DRBD_MD_INDEX_FLEX_INT:
1766 /* sizeof(struct md_on_disk_07) == 4k
1767 * position: last 4k aligned block of 4k size */
1768 if (!bdev->backing_bdev) {
1769 if (__ratelimit(&drbd_ratelimit_state)) {
1770 dev_err(DEV, "bdev->backing_bdev==NULL\n");
1775 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL)
1777 case DRBD_MD_INDEX_FLEX_EXT:
1783 drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w)
1785 unsigned long flags;
1786 spin_lock_irqsave(&q->q_lock, flags);
1787 list_add(&w->list, &q->q);
1788 up(&q->s); /* within the spinlock,
1789 see comment near end of drbd_worker() */
1790 spin_unlock_irqrestore(&q->q_lock, flags);
1794 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1796 unsigned long flags;
1797 spin_lock_irqsave(&q->q_lock, flags);
1798 list_add_tail(&w->list, &q->q);
1799 up(&q->s); /* within the spinlock,
1800 see comment near end of drbd_worker() */
1801 spin_unlock_irqrestore(&q->q_lock, flags);
1804 static inline void wake_asender(struct drbd_tconn *tconn)
1806 if (test_bit(SIGNAL_ASENDER, &tconn->flags))
1807 force_sig(DRBD_SIG, tconn->asender.task);
1810 static inline void request_ping(struct drbd_tconn *tconn)
1812 set_bit(SEND_PING, &tconn->flags);
1813 wake_asender(tconn);
1816 extern void *conn_prepare_command(struct drbd_tconn *, struct drbd_socket *);
1817 extern void *drbd_prepare_command(struct drbd_conf *, struct drbd_socket *);
1818 extern int conn_send_command(struct drbd_tconn *, struct drbd_socket *,
1819 enum drbd_packet, unsigned int, void *,
1821 extern int drbd_send_command(struct drbd_conf *, struct drbd_socket *,
1822 enum drbd_packet, unsigned int, void *,
1825 extern int drbd_send_ping(struct drbd_tconn *tconn);
1826 extern int drbd_send_ping_ack(struct drbd_tconn *tconn);
1827 extern int drbd_send_state_req(struct drbd_conf *, union drbd_state, union drbd_state);
1828 extern int conn_send_state_req(struct drbd_tconn *, union drbd_state, union drbd_state);
1830 static inline void drbd_thread_stop(struct drbd_thread *thi)
1832 _drbd_thread_stop(thi, false, true);
1835 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
1837 _drbd_thread_stop(thi, false, false);
1840 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
1842 _drbd_thread_stop(thi, true, false);
1845 /* counts how many answer packets packets we expect from our peer,
1846 * for either explicit application requests,
1847 * or implicit barrier packets as necessary.
1850 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1851 * it is much easier and equally valid to count what we queue for the
1852 * worker, even before it actually was queued or send.
1853 * (drbd_make_request_common; recovery path on read io-error)
1855 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1856 * _req_mod(req, DATA_RECEIVED)
1857 * [from receive_DataReply]
1858 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1859 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1860 * for some reason it is NOT decreased in got_NegAck,
1861 * but in the resulting cleanup code from report_params.
1862 * we should try to remember the reason for that...
1863 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
1864 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
1865 * [from tl_clear_barrier]
1867 static inline void inc_ap_pending(struct drbd_conf *mdev)
1869 atomic_inc(&mdev->ap_pending_cnt);
1872 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \
1873 if (atomic_read(&mdev->which) < 0) \
1874 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
1876 atomic_read(&mdev->which))
1878 #define dec_ap_pending(mdev) _dec_ap_pending(mdev, __FUNCTION__, __LINE__)
1879 static inline void _dec_ap_pending(struct drbd_conf *mdev, const char *func, int line)
1881 if (atomic_dec_and_test(&mdev->ap_pending_cnt))
1882 wake_up(&mdev->misc_wait);
1883 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line);
1886 /* counts how many resync-related answers we still expect from the peer
1888 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1889 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
1890 * (or P_NEG_ACK with ID_SYNCER)
1892 static inline void inc_rs_pending(struct drbd_conf *mdev)
1894 atomic_inc(&mdev->rs_pending_cnt);
1897 #define dec_rs_pending(mdev) _dec_rs_pending(mdev, __FUNCTION__, __LINE__)
1898 static inline void _dec_rs_pending(struct drbd_conf *mdev, const char *func, int line)
1900 atomic_dec(&mdev->rs_pending_cnt);
1901 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line);
1904 /* counts how many answers we still need to send to the peer.
1906 * receive_Data unless protocol A;
1907 * we need to send a P_RECV_ACK (proto B)
1908 * or P_WRITE_ACK (proto C)
1909 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1910 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1911 * receive_Barrier_* we need to send a P_BARRIER_ACK
1913 static inline void inc_unacked(struct drbd_conf *mdev)
1915 atomic_inc(&mdev->unacked_cnt);
1918 #define dec_unacked(mdev) _dec_unacked(mdev, __FUNCTION__, __LINE__)
1919 static inline void _dec_unacked(struct drbd_conf *mdev, const char *func, int line)
1921 atomic_dec(&mdev->unacked_cnt);
1922 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
1925 #define sub_unacked(mdev, n) _sub_unacked(mdev, n, __FUNCTION__, __LINE__)
1926 static inline void _sub_unacked(struct drbd_conf *mdev, int n, const char *func, int line)
1928 atomic_sub(n, &mdev->unacked_cnt);
1929 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
1932 static inline void put_net_conf(struct drbd_tconn *tconn)
1934 if (atomic_dec_and_test(&tconn->net_cnt))
1935 wake_up(&tconn->net_cnt_wait);
1939 * get_net_conf() - Increase ref count on mdev->tconn->net_conf; Returns 0 if nothing there
1940 * @mdev: DRBD device.
1942 * You have to call put_net_conf() when finished working with mdev->tconn->net_conf.
1944 static inline int get_net_conf(struct drbd_tconn *tconn)
1948 atomic_inc(&tconn->net_cnt);
1949 have_net_conf = tconn->cstate >= C_UNCONNECTED;
1951 put_net_conf(tconn);
1952 return have_net_conf;
1956 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
1959 * You have to call put_ldev() when finished working with mdev->ldev.
1961 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
1962 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
1964 static inline void put_ldev(struct drbd_conf *mdev)
1966 int i = atomic_dec_return(&mdev->local_cnt);
1968 /* This may be called from some endio handler,
1969 * so we must not sleep here. */
1974 if (mdev->state.disk == D_DISKLESS)
1975 /* even internal references gone, safe to destroy */
1976 drbd_ldev_destroy(mdev);
1977 if (mdev->state.disk == D_FAILED)
1978 /* all application IO references gone. */
1979 drbd_go_diskless(mdev);
1980 wake_up(&mdev->misc_wait);
1985 static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
1989 /* never get a reference while D_DISKLESS */
1990 if (mdev->state.disk == D_DISKLESS)
1993 atomic_inc(&mdev->local_cnt);
1994 io_allowed = (mdev->state.disk >= mins);
2000 extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins);
2003 /* you must have an "get_ldev" reference */
2004 static inline void drbd_get_syncer_progress(struct drbd_conf *mdev,
2005 unsigned long *bits_left, unsigned int *per_mil_done)
2007 /* this is to break it at compile time when we change that, in case we
2008 * want to support more than (1<<32) bits on a 32bit arch. */
2009 typecheck(unsigned long, mdev->rs_total);
2011 /* note: both rs_total and rs_left are in bits, i.e. in
2012 * units of BM_BLOCK_SIZE.
2013 * for the percentage, we don't care. */
2015 if (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T)
2016 *bits_left = mdev->ov_left;
2018 *bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
2019 /* >> 10 to prevent overflow,
2020 * +1 to prevent division by zero */
2021 if (*bits_left > mdev->rs_total) {
2022 /* doh. maybe a logic bug somewhere.
2023 * may also be just a race condition
2024 * between this and a disconnect during sync.
2025 * for now, just prevent in-kernel buffer overflow.
2028 dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2029 drbd_conn_str(mdev->state.conn),
2030 *bits_left, mdev->rs_total, mdev->rs_failed);
2033 /* Make sure the division happens in long context.
2034 * We allow up to one petabyte storage right now,
2035 * at a granularity of 4k per bit that is 2**38 bits.
2036 * After shift right and multiplication by 1000,
2037 * this should still fit easily into a 32bit long,
2038 * so we don't need a 64bit division on 32bit arch.
2039 * Note: currently we don't support such large bitmaps on 32bit
2040 * arch anyways, but no harm done to be prepared for it here.
2042 unsigned int shift = mdev->rs_total >= (1ULL << 32) ? 16 : 10;
2043 unsigned long left = *bits_left >> shift;
2044 unsigned long total = 1UL + (mdev->rs_total >> shift);
2045 unsigned long tmp = 1000UL - left * 1000UL/total;
2046 *per_mil_done = tmp;
2051 /* this throttles on-the-fly application requests
2052 * according to max_buffers settings;
2053 * maybe re-implement using semaphores? */
2054 static inline int drbd_get_max_buffers(struct drbd_conf *mdev)
2056 int mxb = 1000000; /* arbitrary limit on open requests */
2057 if (get_net_conf(mdev->tconn)) {
2058 mxb = mdev->tconn->net_conf->max_buffers;
2059 put_net_conf(mdev->tconn);
2064 static inline int drbd_state_is_stable(struct drbd_conf *mdev)
2066 union drbd_dev_state s = mdev->state;
2068 /* DO NOT add a default clause, we want the compiler to warn us
2069 * for any newly introduced state we may have forgotten to add here */
2071 switch ((enum drbd_conns)s.conn) {
2072 /* new io only accepted when there is no connection, ... */
2074 case C_WF_CONNECTION:
2075 /* ... or there is a well established connection. */
2081 case C_PAUSED_SYNC_S:
2082 case C_PAUSED_SYNC_T:
2085 /* transitional states, IO allowed */
2086 case C_DISCONNECTING:
2090 case C_NETWORK_FAILURE:
2091 case C_PROTOCOL_ERROR:
2093 case C_WF_REPORT_PARAMS:
2094 case C_STARTING_SYNC_S:
2095 case C_STARTING_SYNC_T:
2098 /* Allow IO in BM exchange states with new protocols */
2100 if (mdev->tconn->agreed_pro_version < 96)
2104 /* no new io accepted in these states */
2106 case C_WF_SYNC_UUID:
2112 switch ((enum drbd_disk_state)s.disk) {
2114 case D_INCONSISTENT:
2118 /* disk state is stable as well. */
2121 /* no new io accepted during tansitional states */
2134 static inline int drbd_suspended(struct drbd_conf *mdev)
2136 struct drbd_tconn *tconn = mdev->tconn;
2138 return tconn->susp || tconn->susp_fen || tconn->susp_nod;
2141 static inline bool may_inc_ap_bio(struct drbd_conf *mdev)
2143 int mxb = drbd_get_max_buffers(mdev);
2145 if (drbd_suspended(mdev))
2147 if (test_bit(SUSPEND_IO, &mdev->flags))
2150 /* to avoid potential deadlock or bitmap corruption,
2151 * in various places, we only allow new application io
2152 * to start during "stable" states. */
2154 /* no new io accepted when attaching or detaching the disk */
2155 if (!drbd_state_is_stable(mdev))
2158 /* since some older kernels don't have atomic_add_unless,
2159 * and we are within the spinlock anyways, we have this workaround. */
2160 if (atomic_read(&mdev->ap_bio_cnt) > mxb)
2162 if (test_bit(BITMAP_IO, &mdev->flags))
2167 static inline bool inc_ap_bio_cond(struct drbd_conf *mdev)
2171 spin_lock_irq(&mdev->tconn->req_lock);
2172 rv = may_inc_ap_bio(mdev);
2174 atomic_inc(&mdev->ap_bio_cnt);
2175 spin_unlock_irq(&mdev->tconn->req_lock);
2180 static inline void inc_ap_bio(struct drbd_conf *mdev)
2183 * as long as the device is suspended
2184 * until the bitmap is no longer on the fly during connection
2185 * handshake as long as we would exeed the max_buffer limit.
2187 * to avoid races with the reconnect code,
2188 * we need to atomic_inc within the spinlock. */
2190 wait_event(mdev->misc_wait, inc_ap_bio_cond(mdev));
2193 static inline void dec_ap_bio(struct drbd_conf *mdev)
2195 int mxb = drbd_get_max_buffers(mdev);
2196 int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt);
2198 D_ASSERT(ap_bio >= 0);
2199 /* this currently does wake_up for every dec_ap_bio!
2200 * maybe rather introduce some type of hysteresis?
2201 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2203 wake_up(&mdev->misc_wait);
2204 if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) {
2205 if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags))
2206 drbd_queue_work(&mdev->tconn->data.work, &mdev->bm_io_work.w);
2210 static inline int drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val)
2212 int changed = mdev->ed_uuid != val;
2213 mdev->ed_uuid = val;
2217 static inline int drbd_queue_order_type(struct drbd_conf *mdev)
2219 /* sorry, we currently have no working implementation
2220 * of distributed TCQ stuff */
2221 #ifndef QUEUE_ORDERED_NONE
2222 #define QUEUE_ORDERED_NONE 0
2224 return QUEUE_ORDERED_NONE;
2227 static inline void drbd_md_flush(struct drbd_conf *mdev)
2231 if (test_bit(MD_NO_FUA, &mdev->flags))
2234 r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL);
2236 set_bit(MD_NO_FUA, &mdev->flags);
2237 dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r);