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 mutex drbd_cfg_mutex;
178 /* receiver (data socket) */
180 P_DATA_REPLY = 0x01, /* Response to P_DATA_REQUEST */
181 P_RS_DATA_REPLY = 0x02, /* Response to P_RS_DATA_REQUEST */
184 P_BECOME_SYNC_TARGET = 0x05,
185 P_BECOME_SYNC_SOURCE = 0x06,
186 P_UNPLUG_REMOTE = 0x07, /* Used at various times to hint the peer */
187 P_DATA_REQUEST = 0x08, /* Used to ask for a data block */
188 P_RS_DATA_REQUEST = 0x09, /* Used to ask for a data block for resync */
195 P_AUTH_CHALLENGE = 0x10,
196 P_AUTH_RESPONSE = 0x11,
197 P_STATE_CHG_REQ = 0x12,
199 /* asender (meta socket */
202 P_RECV_ACK = 0x15, /* Used in protocol B */
203 P_WRITE_ACK = 0x16, /* Used in protocol C */
204 P_RS_WRITE_ACK = 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
205 P_DISCARD_WRITE = 0x18, /* Used in proto C, two-primaries conflict detection */
206 P_NEG_ACK = 0x19, /* Sent if local disk is unusable */
207 P_NEG_DREPLY = 0x1a, /* Local disk is broken... */
208 P_NEG_RS_DREPLY = 0x1b, /* Local disk is broken... */
209 P_BARRIER_ACK = 0x1c,
210 P_STATE_CHG_REPLY = 0x1d,
212 /* "new" commands, no longer fitting into the ordering scheme above */
214 P_OV_REQUEST = 0x1e, /* data socket */
216 P_OV_RESULT = 0x20, /* meta socket */
217 P_CSUM_RS_REQUEST = 0x21, /* data socket */
218 P_RS_IS_IN_SYNC = 0x22, /* meta socket */
219 P_SYNC_PARAM89 = 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
220 P_COMPRESSED_BITMAP = 0x24, /* compressed or otherwise encoded bitmap transfer */
221 /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */
222 /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */
223 P_DELAY_PROBE = 0x27, /* is used on BOTH sockets */
224 P_OUT_OF_SYNC = 0x28, /* Mark as out of sync (Outrunning), data socket */
225 P_RS_CANCEL = 0x29, /* meta: Used to cancel RS_DATA_REQUEST packet by SyncSource */
226 P_CONN_ST_CHG_REQ = 0x2a, /* data sock: Connection wide state request */
227 P_CONN_ST_CHG_REPLY = 0x2b, /* meta sock: Connection side state req reply */
228 P_RETRY_WRITE = 0x2c, /* Protocol C: retry conflicting write request */
230 P_MAY_IGNORE = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
231 P_MAX_OPT_CMD = 0x101,
233 /* special command ids for handshake */
235 P_INITIAL_META = 0xfff1, /* First Packet on the MetaSock */
236 P_INITIAL_DATA = 0xfff2, /* First Packet on the Socket */
238 P_CONNECTION_FEATURES = 0xfffe /* FIXED for the next century! */
241 extern const char *cmdname(enum drbd_packet cmd);
243 /* for sending/receiving the bitmap,
244 * possibly in some encoding scheme */
247 * stores total bits and long words
248 * of the bitmap, so we don't need to
249 * call the accessor functions over and again. */
250 unsigned long bm_bits;
251 unsigned long bm_words;
252 /* during xfer, current position within the bitmap */
253 unsigned long bit_offset;
254 unsigned long word_offset;
256 /* statistics; index: (h->command == P_BITMAP) */
261 extern void INFO_bm_xfer_stats(struct drbd_conf *mdev,
262 const char *direction, struct bm_xfer_ctx *c);
264 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
266 /* word_offset counts "native long words" (32 or 64 bit),
268 * Encoded packet may end at an unaligned bit offset.
269 * In case a fallback clear text packet is transmitted in
270 * between, we adjust this offset back to the last 64bit
271 * aligned "native long word", which makes coding and decoding
272 * the plain text bitmap much more convenient. */
273 #if BITS_PER_LONG == 64
274 c->word_offset = c->bit_offset >> 6;
275 #elif BITS_PER_LONG == 32
276 c->word_offset = c->bit_offset >> 5;
277 c->word_offset &= ~(1UL);
279 # error "unsupported BITS_PER_LONG"
284 #define __packed __attribute__((packed))
287 /* This is the layout for a packet on the wire.
288 * The byteorder is the network byte order.
289 * (except block_id and barrier fields.
290 * these are pointers to local structs
291 * and have no relevance for the partner,
292 * which just echoes them as received.)
294 * NOTE that the payload starts at a long aligned offset,
295 * regardless of 32 or 64 bit arch!
300 u16 length; /* bytes of data after this header */
303 /* Header for big packets, Used for data packets exceeding 64kB */
305 u16 magic; /* use DRBD_MAGIC_BIG here */
307 u32 length; /* Use only 24 bits of that. Ignore the highest 8 bit. */
312 struct p_header80 h80;
313 struct p_header95 h95;
319 * short commands, packets without payload, plain p_header:
322 * P_BECOME_SYNC_TARGET
323 * P_BECOME_SYNC_SOURCE
328 * commands with out-of-struct payload:
329 * P_BITMAP (no additional fields)
330 * P_DATA, P_DATA_REPLY (see p_data)
331 * P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
334 /* these defines must not be changed without changing the protocol version */
335 #define DP_HARDBARRIER 1 /* depricated */
336 #define DP_RW_SYNC 2 /* equals REQ_SYNC */
337 #define DP_MAY_SET_IN_SYNC 4
338 #define DP_UNPLUG 8 /* not used anymore */
339 #define DP_FUA 16 /* equals REQ_FUA */
340 #define DP_FLUSH 32 /* equals REQ_FLUSH */
341 #define DP_DISCARD 64 /* equals REQ_DISCARD */
344 struct p_header head;
345 u64 sector; /* 64 bits sector number */
346 u64 block_id; /* to identify the request in protocol B&C */
352 * commands which share a struct:
354 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
355 * P_DISCARD_WRITE (proto C, two-primaries conflict detection)
357 * P_DATA_REQUEST, P_RS_DATA_REQUEST
360 struct p_header head;
368 struct p_header head;
372 u32 pad; /* to multiple of 8 Byte */
376 * commands with their own struct for additional fields:
377 * P_CONNECTION_FEATURES
384 struct p_connection_features {
385 struct p_header head; /* Note: vnr will be ignored */
390 /* should be more than enough for future enhancements
391 * for now, feature_flags and the reserverd array shall be zero.
397 /* 80 bytes, FIXED for the next century */
400 struct p_header head;
401 u32 barrier; /* barrier number _handle_ only */
402 u32 pad; /* to multiple of 8 Byte */
405 struct p_barrier_ack {
406 struct p_header head;
412 struct p_header head;
415 /* Since protocol version 88 and higher. */
419 struct p_rs_param_89 {
420 struct p_header head;
422 /* protocol version 89: */
423 char verify_alg[SHARED_SECRET_MAX];
424 char csums_alg[SHARED_SECRET_MAX];
427 struct p_rs_param_95 {
428 struct p_header head;
430 char verify_alg[SHARED_SECRET_MAX];
431 char csums_alg[SHARED_SECRET_MAX];
438 enum drbd_conn_flags {
444 struct p_header head;
452 /* Since protocol version 87 and higher. */
453 char integrity_alg[0];
458 struct p_header head;
459 u64 uuid[UI_EXTENDED_SIZE];
463 struct p_header head;
468 struct p_header head;
469 u64 d_size; /* size of disk */
470 u64 u_size; /* user requested size */
471 u64 c_size; /* current exported size */
472 u32 max_bio_size; /* Maximal size of a BIO */
473 u16 queue_order_type; /* not yet implemented in DRBD*/
474 u16 dds_flags; /* use enum dds_flags here. */
478 struct p_header head;
483 struct p_header head;
488 struct p_req_state_reply {
489 struct p_header head;
493 struct p_drbd06_param {
504 struct p_header head;
510 struct p_block_desc {
511 struct p_header head;
514 u32 pad; /* to multiple of 8 Byte */
517 /* Valid values for the encoding field.
518 * Bump proto version when changing this. */
519 enum drbd_bitmap_code {
520 /* RLE_VLI_Bytes = 0,
521 * and other bit variants had been defined during
522 * algorithm evaluation. */
526 struct p_compressed_bm {
527 struct p_header head;
528 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
529 * (encoding & 0x80): polarity (set/unset) of first runlength
530 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
531 * used to pad up to head.length bytes
538 struct p_delay_probe93 {
539 struct p_header head;
540 u32 seq_num; /* sequence number to match the two probe packets */
541 u32 offset; /* usecs the probe got sent after the reference time point */
544 /* one bitmap packet, including the p_header,
545 * should fit within one _architecture independend_ page.
546 * so we need to use the fixed size 4KiB page size
547 * most architectures have used for a long time.
549 #define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header))
550 #define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
551 #define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm))
552 #if (PAGE_SIZE < 4096)
553 /* drbd_send_bitmap / receive_bitmap would break horribly */
554 #error "PAGE_SIZE too small"
557 #define DRBD_SOCKET_BUFFER_SIZE 4096
559 /**********************************************************************/
560 enum drbd_thread_state {
569 struct task_struct *task;
570 struct completion stop;
571 enum drbd_thread_state t_state;
572 int (*function) (struct drbd_thread *);
573 struct drbd_tconn *tconn;
578 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
580 /* THINK testing the t_state seems to be uncritical in all cases
581 * (but thread_{start,stop}), so we can read it *without* the lock.
589 struct list_head list;
590 int (*cb)(struct drbd_work *, int cancel);
592 struct drbd_conf *mdev;
593 struct drbd_tconn *tconn;
597 #include "drbd_interval.h"
599 extern int drbd_wait_misc(struct drbd_conf *, struct drbd_interval *);
601 struct drbd_request {
604 /* if local IO is not allowed, will be NULL.
605 * if local IO _is_ allowed, holds the locally submitted bio clone,
606 * or, after local IO completion, the ERR_PTR(error).
607 * see drbd_request_endio(). */
608 struct bio *private_bio;
610 struct drbd_interval i;
611 unsigned int epoch; /* barrier_nr */
613 /* barrier_nr: used to check on "completion" whether this req was in
614 * the current epoch, and we therefore have to close it,
615 * starting a new epoch...
618 struct list_head tl_requests; /* ring list in the transfer log */
619 struct bio *master_bio; /* master bio pointer */
620 unsigned long rq_state; /* see comments above _req_mod() */
622 unsigned long start_time;
625 struct drbd_tl_epoch {
627 struct list_head requests; /* requests before */
628 struct drbd_tl_epoch *next; /* pointer to the next barrier */
629 unsigned int br_number; /* the barriers identifier. */
630 int n_writes; /* number of requests attached before this barrier */
634 struct list_head list;
635 unsigned int barrier_nr;
636 atomic_t epoch_size; /* increased on every request added. */
637 atomic_t active; /* increased on every req. added, and dec on every finished. */
641 /* drbd_epoch flag bits */
643 DE_HAVE_BARRIER_NUMBER,
650 EV_CLEANUP = 32, /* used as flag */
653 struct drbd_wq_barrier {
655 struct completion done;
663 struct drbd_peer_request {
665 struct drbd_epoch *epoch; /* for writes */
667 atomic_t pending_bios;
668 struct drbd_interval i;
669 /* see comments on ee flag bits below */
673 struct digest_info *digest;
678 * While corresponding bios are in flight, the only modification will be
679 * set_bit WAS_ERROR, which has to be atomic.
680 * If no bios are in flight yet, or all have been completed,
681 * non-atomic modification to ee->flags is ok.
684 __EE_CALL_AL_COMPLETE_IO,
685 __EE_MAY_SET_IN_SYNC,
687 /* In case a barrier failed,
688 * we need to resubmit without the barrier flag. */
691 /* we may have several bios per peer request.
692 * if any of those fail, we set this flag atomically
693 * from the endio callback */
696 /* This ee has a pointer to a digest instead of a block id */
699 /* Conflicting local requests need to be restarted after this request */
700 __EE_RESTART_REQUESTS,
702 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
703 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
704 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
705 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
706 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
707 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
709 /* flag bits per mdev */
711 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */
712 UNPLUG_QUEUED, /* only relevant with kernel 2.4 */
713 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
714 MD_DIRTY, /* current uuids and flags not yet on disk */
715 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
718 CRASHED_PRIMARY, /* This node was a crashed primary.
719 * Gets cleared when the state.conn
720 * goes into C_CONNECTED state. */
721 NO_BARRIER_SUPP, /* underlying block device doesn't implement barriers */
724 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */
725 SUSPEND_IO, /* suspend application io */
726 BITMAP_IO, /* suspend application io;
727 once no more io in flight, start bitmap io */
728 BITMAP_IO_QUEUED, /* Started bitmap IO */
729 GO_DISKLESS, /* Disk is being detached, on io-error or admin request. */
730 WAS_IO_ERROR, /* Local disk failed returned IO error */
731 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
732 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
733 * the peer, if it changed there as well. */
734 NEW_CUR_UUID, /* Create new current UUID when thawing IO */
735 AL_SUSPENDED, /* Activity logging is currently suspended. */
736 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */
737 B_RS_H_DONE, /* Before resync handler done (already executed) */
740 struct drbd_bitmap; /* opaque for drbd_conf */
742 /* definition of bits in bm_flags to be used in drbd_bm_lock
743 * and drbd_bitmap_io and friends. */
745 /* do we need to kfree, or vfree bm_pages? */
746 BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */
748 /* currently locked for bulk operation */
749 BM_LOCKED_MASK = 0x7,
751 /* in detail, that is: */
756 /* (test bit, count bit) allowed (common case) */
757 BM_LOCKED_TEST_ALLOWED = 0x3,
759 /* testing bits, as well as setting new bits allowed, but clearing bits
760 * would be unexpected. Used during bitmap receive. Setting new bits
761 * requires sending of "out-of-sync" information, though. */
762 BM_LOCKED_SET_ALLOWED = 0x1,
764 /* clear is not expected while bitmap is locked for bulk operation */
768 /* TODO sort members for performance
769 * MAYBE group them further */
771 /* THINK maybe we actually want to use the default "event/%s" worker threads
772 * or similar in linux 2.6, which uses per cpu data and threads.
774 struct drbd_work_queue {
776 struct semaphore s; /* producers up it, worker down()s it */
777 spinlock_t q_lock; /* to protect the list. */
781 struct drbd_work_queue work;
783 struct socket *socket;
784 /* this way we get our
785 * send/receive buffers off the stack */
791 u64 md_offset; /* sector offset to 'super' block */
793 u64 la_size_sect; /* last agreed size, unit sectors */
799 s32 al_offset; /* signed relative sector offset to al area */
800 s32 bm_offset; /* signed relative sector offset to bitmap */
802 /* u32 al_nr_extents; important for restoring the AL
803 * is stored into ldev->dc.al_extents, which in turn
804 * gets applied to act_log->nr_elements
808 struct drbd_backing_dev {
809 struct block_device *backing_bdev;
810 struct block_device *md_bdev;
812 struct disk_conf dc; /* The user provided config... */
813 sector_t known_size; /* last known size of that backing device */
817 struct drbd_conf *mdev;
818 struct completion event;
826 int (*io_fn)(struct drbd_conf *mdev);
827 void (*done)(struct drbd_conf *mdev, int rv);
830 enum write_ordering_e {
838 unsigned int head_index;
842 /* flag bits per tconn */
844 NET_CONGESTED, /* The data socket is congested */
845 DISCARD_CONCURRENT, /* Set on one node, cleared on the peer! */
846 SEND_PING, /* whether asender should send a ping asap */
847 SIGNAL_ASENDER, /* whether asender wants to be interrupted */
848 GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */
851 CONFIG_PENDING, /* serialization of (re)configuration requests.
852 * if set, also prevents the device from dying */
853 OBJECT_DYING, /* device became unconfigured,
854 * but worker thread is still handling the cleanup.
855 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
856 * while this is set. */
857 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
860 struct drbd_tconn { /* is a resource from the config file */
861 char *name; /* Resource name */
862 struct list_head all_tconn; /* linked on global drbd_tconns */
863 struct idr volumes; /* <tconn, vnr> to mdev mapping */
864 enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
865 struct mutex cstate_mutex; /* Protects graceful disconnects */
868 struct net_conf *net_conf; /* protected by get_net_conf() and put_net_conf() */
869 atomic_t net_cnt; /* Users of net_conf */
870 wait_queue_head_t net_cnt_wait;
871 wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */
872 struct res_opts res_opts;
874 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
875 struct drbd_socket meta; /* ping/ack (metadata) packets */
876 int agreed_pro_version; /* actually used protocol version */
877 unsigned long last_received; /* in jiffies, either socket */
878 unsigned int ko_count;
881 struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */
882 struct drbd_tl_epoch *newest_tle;
883 struct drbd_tl_epoch *oldest_tle;
884 struct list_head out_of_sequence_requests;
886 struct crypto_hash *cram_hmac_tfm;
887 struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */
888 struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */
889 struct crypto_hash *csums_tfm;
890 struct crypto_hash *verify_tfm;
895 struct drbd_thread receiver;
896 struct drbd_thread worker;
897 struct drbd_thread asender;
898 cpumask_var_t cpu_mask;
902 struct drbd_tconn *tconn;
903 int vnr; /* volume number within the connection */
905 /* things that are stored as / read from meta data on disk */
908 /* configured by drbdsetup */
909 struct drbd_backing_dev *ldev __protected_by(local);
911 sector_t p_size; /* partner's disk size */
912 struct request_queue *rq_queue;
913 struct block_device *this_bdev;
914 struct gendisk *vdisk;
916 struct drbd_work resync_work,
921 struct timer_list resync_timer;
922 struct timer_list md_sync_timer;
923 struct timer_list start_resync_timer;
924 struct timer_list request_timer;
925 #ifdef DRBD_DEBUG_MD_SYNC
929 } last_md_mark_dirty;
932 /* Used after attach while negotiating new disk state. */
933 union drbd_state new_state_tmp;
935 union drbd_state state;
936 wait_queue_head_t misc_wait;
937 wait_queue_head_t state_wait; /* upon each state change. */
938 unsigned int send_cnt;
939 unsigned int recv_cnt;
940 unsigned int read_cnt;
941 unsigned int writ_cnt;
942 unsigned int al_writ_cnt;
943 unsigned int bm_writ_cnt;
944 atomic_t ap_bio_cnt; /* Requests we need to complete */
945 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
946 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
947 atomic_t unacked_cnt; /* Need to send replys for */
948 atomic_t local_cnt; /* Waiting for local completion */
950 /* Interval tree of pending local requests */
951 struct rb_root read_requests;
952 struct rb_root write_requests;
954 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
955 unsigned long rs_total;
956 /* number of resync blocks that failed in this run */
957 unsigned long rs_failed;
958 /* Syncer's start time [unit jiffies] */
959 unsigned long rs_start;
960 /* cumulated time in PausedSyncX state [unit jiffies] */
961 unsigned long rs_paused;
962 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
963 unsigned long rs_same_csum;
964 #define DRBD_SYNC_MARKS 8
965 #define DRBD_SYNC_MARK_STEP (3*HZ)
966 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
967 unsigned long rs_mark_left[DRBD_SYNC_MARKS];
968 /* marks's time [unit jiffies] */
969 unsigned long rs_mark_time[DRBD_SYNC_MARKS];
970 /* current index into rs_mark_{left,time} */
973 /* where does the admin want us to start? (sector) */
974 sector_t ov_start_sector;
975 /* where are we now? (sector) */
976 sector_t ov_position;
977 /* Start sector of out of sync range (to merge printk reporting). */
978 sector_t ov_last_oos_start;
979 /* size of out-of-sync range in sectors. */
980 sector_t ov_last_oos_size;
981 unsigned long ov_left; /* in bits */
983 struct drbd_bitmap *bitmap;
984 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
986 /* Used to track operations of resync... */
987 struct lru_cache *resync;
988 /* Number of locked elements in resync LRU */
989 unsigned int resync_locked;
990 /* resync extent number waiting for application requests */
991 unsigned int resync_wenr;
995 struct drbd_epoch *current_epoch;
996 spinlock_t epoch_lock;
998 enum write_ordering_e write_ordering;
999 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
1000 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
1001 struct list_head done_ee; /* need to send P_WRITE_ACK */
1002 struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */
1003 struct list_head net_ee; /* zero-copy network send in progress */
1005 int next_barrier_nr;
1006 struct list_head resync_reads;
1007 atomic_t pp_in_use; /* allocated from page pool */
1008 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */
1009 wait_queue_head_t ee_wait;
1010 struct page *md_io_page; /* one page buffer for md_io */
1011 struct mutex md_io_mutex; /* protects the md_io_buffer */
1013 wait_queue_head_t al_wait;
1014 struct lru_cache *act_log; /* activity log */
1015 unsigned int al_tr_number;
1017 int al_tr_pos; /* position of the next transaction in the journal */
1018 wait_queue_head_t seq_wait;
1019 atomic_t packet_seq;
1020 unsigned int peer_seq;
1021 spinlock_t peer_seq_lock;
1023 unsigned long comm_bm_set; /* communicated number of set bits. */
1024 struct bm_io_work bm_io_work;
1025 u64 ed_uuid; /* UUID of the exposed data */
1026 struct mutex own_state_mutex;
1027 struct mutex *state_mutex; /* either own_state_mutex or mdev->tconn->cstate_mutex */
1028 char congestion_reason; /* Why we where congested... */
1029 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
1030 atomic_t rs_sect_ev; /* for submitted resync data rate, both */
1031 int rs_last_sect_ev; /* counter to compare with */
1032 int rs_last_events; /* counter of read or write "events" (unit sectors)
1033 * on the lower level device when we last looked. */
1034 int c_sync_rate; /* current resync rate after syncer throttle magic */
1035 struct fifo_buffer rs_plan_s; /* correction values of resync planer */
1036 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
1037 int rs_planed; /* resync sectors already planned */
1038 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
1039 int peer_max_bio_size;
1040 int local_max_bio_size;
1043 static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
1045 return (struct drbd_conf *)idr_find(&minors, minor);
1048 static inline unsigned int mdev_to_minor(struct drbd_conf *mdev)
1053 static inline struct drbd_conf *vnr_to_mdev(struct drbd_tconn *tconn, int vnr)
1055 return (struct drbd_conf *)idr_find(&tconn->volumes, vnr);
1058 static inline int drbd_get_data_sock(struct drbd_tconn *tconn)
1060 mutex_lock(&tconn->data.mutex);
1061 if (!tconn->data.socket) {
1063 mutex_unlock(&tconn->data.mutex);
1069 static inline void drbd_put_data_sock(struct drbd_tconn *tconn)
1071 mutex_unlock(&tconn->data.mutex);
1075 * function declarations
1076 *************************/
1082 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1085 extern void drbd_init_set_defaults(struct drbd_conf *mdev);
1086 extern int drbd_thread_start(struct drbd_thread *thi);
1087 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1088 extern char *drbd_task_to_thread_name(struct drbd_tconn *tconn, struct task_struct *task);
1090 extern void drbd_thread_current_set_cpu(struct drbd_thread *thi);
1091 extern void drbd_calc_cpu_mask(struct drbd_tconn *tconn);
1093 #define drbd_thread_current_set_cpu(A) ({})
1094 #define drbd_calc_cpu_mask(A) ({})
1096 extern void drbd_free_resources(struct drbd_conf *mdev);
1097 extern void tl_release(struct drbd_tconn *, unsigned int barrier_nr,
1098 unsigned int set_size);
1099 extern void tl_clear(struct drbd_tconn *);
1100 extern void _tl_add_barrier(struct drbd_tconn *, struct drbd_tl_epoch *);
1101 extern void drbd_free_sock(struct drbd_tconn *tconn);
1102 extern int drbd_send(struct drbd_tconn *tconn, struct socket *sock,
1103 void *buf, size_t size, unsigned msg_flags);
1104 extern int drbd_send_all(struct drbd_tconn *, struct socket *, void *, size_t,
1107 extern int drbd_send_protocol(struct drbd_tconn *tconn);
1108 extern int drbd_send_uuids(struct drbd_conf *mdev);
1109 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
1110 extern void drbd_gen_and_send_sync_uuid(struct drbd_conf *mdev);
1111 extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags);
1112 extern int _conn_send_state_req(struct drbd_tconn *, int vnr, enum drbd_packet cmd,
1113 union drbd_state, union drbd_state);
1114 extern int _drbd_send_state(struct drbd_conf *mdev);
1115 extern int drbd_send_state(struct drbd_conf *mdev);
1116 extern int _conn_send_cmd(struct drbd_tconn *tconn, int vnr, struct drbd_socket *sock,
1117 enum drbd_packet cmd, struct p_header *h, size_t size,
1118 unsigned msg_flags);
1119 extern int conn_send_cmd(struct drbd_tconn *tconn, int vnr, struct drbd_socket *sock,
1120 enum drbd_packet cmd, struct p_header *h, size_t size);
1121 extern int conn_send_cmd2(struct drbd_tconn *tconn, enum drbd_packet cmd,
1122 char *data, size_t size);
1123 extern int drbd_send_sync_param(struct drbd_conf *mdev);
1124 extern void drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr,
1126 extern int drbd_send_ack(struct drbd_conf *, enum drbd_packet,
1127 struct drbd_peer_request *);
1128 extern void drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packet cmd,
1129 struct p_block_req *rp);
1130 extern void drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packet cmd,
1131 struct p_data *dp, int data_size);
1132 extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packet cmd,
1133 sector_t sector, int blksize, u64 block_id);
1134 extern int drbd_send_out_of_sync(struct drbd_conf *, struct drbd_request *);
1135 extern int drbd_send_block(struct drbd_conf *, enum drbd_packet,
1136 struct drbd_peer_request *);
1137 extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
1138 extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1139 sector_t sector, int size, u64 block_id);
1140 extern int drbd_send_drequest_csum(struct drbd_conf *mdev, sector_t sector,
1141 int size, void *digest, int digest_size,
1142 enum drbd_packet cmd);
1143 extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size);
1145 extern int drbd_send_bitmap(struct drbd_conf *mdev);
1146 extern void drbd_send_sr_reply(struct drbd_conf *mdev, enum drbd_state_rv retcode);
1147 extern int conn_send_sr_reply(struct drbd_tconn *tconn, enum drbd_state_rv retcode);
1148 extern void drbd_free_bc(struct drbd_backing_dev *ldev);
1149 extern void drbd_mdev_cleanup(struct drbd_conf *mdev);
1150 void drbd_print_uuids(struct drbd_conf *mdev, const char *text);
1152 extern void drbd_md_sync(struct drbd_conf *mdev);
1153 extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
1154 extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1155 extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1156 extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1157 extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1158 extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
1159 extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
1160 extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
1161 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1162 #ifndef DRBD_DEBUG_MD_SYNC
1163 extern void drbd_md_mark_dirty(struct drbd_conf *mdev);
1165 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1166 extern void drbd_md_mark_dirty_(struct drbd_conf *mdev,
1167 unsigned int line, const char *func);
1169 extern void drbd_queue_bitmap_io(struct drbd_conf *mdev,
1170 int (*io_fn)(struct drbd_conf *),
1171 void (*done)(struct drbd_conf *, int),
1172 char *why, enum bm_flag flags);
1173 extern int drbd_bitmap_io(struct drbd_conf *mdev,
1174 int (*io_fn)(struct drbd_conf *),
1175 char *why, enum bm_flag flags);
1176 extern int drbd_bmio_set_n_write(struct drbd_conf *mdev);
1177 extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev);
1178 extern void drbd_go_diskless(struct drbd_conf *mdev);
1179 extern void drbd_ldev_destroy(struct drbd_conf *mdev);
1182 We reserve a 128MB Block (4k aligned)
1183 * either at the end of the backing device
1184 * or on a separate meta data device. */
1186 /* The following numbers are sectors */
1187 /* Allows up to about 3.8TB, so if you want more,
1188 * you need to use the "flexible" meta data format. */
1189 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1190 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1191 #define MD_AL_SECTORS 64 /* = 32 kB on disk activity log ring buffer */
1192 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_SECTORS)
1194 /* we do all meta data IO in 4k blocks */
1195 #define MD_BLOCK_SHIFT 12
1196 #define MD_BLOCK_SIZE (1<<MD_BLOCK_SHIFT)
1198 /* One activity log extent represents 4M of storage */
1199 #define AL_EXTENT_SHIFT 22
1200 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1202 /* We could make these currently hardcoded constants configurable
1203 * variables at create-md time (or even re-configurable at runtime?).
1204 * Which will require some more changes to the DRBD "super block"
1207 * updates per transaction:
1208 * This many changes to the active set can be logged with one transaction.
1209 * This number is arbitrary.
1210 * context per transaction:
1211 * This many context extent numbers are logged with each transaction.
1212 * This number is resulting from the transaction block size (4k), the layout
1213 * of the transaction header, and the number of updates per transaction.
1214 * See drbd_actlog.c:struct al_transaction_on_disk
1216 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1217 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1219 #if BITS_PER_LONG == 32
1221 #define cpu_to_lel(A) cpu_to_le32(A)
1222 #define lel_to_cpu(A) le32_to_cpu(A)
1223 #elif BITS_PER_LONG == 64
1225 #define cpu_to_lel(A) cpu_to_le64(A)
1226 #define lel_to_cpu(A) le64_to_cpu(A)
1228 #error "LN2 of BITS_PER_LONG unknown!"
1232 /* 16MB sized 'bitmap extent' to track syncer usage */
1234 int rs_left; /* number of bits set (out of sync) in this extent. */
1235 int rs_failed; /* number of failed resync requests in this extent. */
1236 unsigned long flags;
1237 struct lc_element lce;
1240 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1241 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1242 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1246 * We need to store one bit for a block.
1247 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1248 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1249 * Bit 1 ==> local node thinks this block needs to be synced.
1252 #define SLEEP_TIME (HZ/10)
1254 /* We do bitmap IO in units of 4k blocks.
1255 * We also still have a hardcoded 4k per bit relation. */
1256 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1257 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1258 /* mostly arbitrarily set the represented size of one bitmap extent,
1259 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1260 * at 4k per bit resolution) */
1261 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
1262 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1264 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1265 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1268 /* thus many _storage_ sectors are described by one bit */
1269 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1270 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1271 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1273 /* bit to represented kilo byte conversion */
1274 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1276 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1277 * _storage_ sector is located in */
1278 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1280 /* how much _storage_ sectors we have per bitmap sector */
1281 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1282 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1284 /* in one sector of the bitmap, we have this many activity_log extents. */
1285 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1286 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1288 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1289 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1291 /* the extent in "PER_EXTENT" below is an activity log extent
1292 * we need that many (long words/bytes) to store the bitmap
1293 * of one AL_EXTENT_SIZE chunk of storage.
1294 * we can store the bitmap for that many AL_EXTENTS within
1295 * one sector of the _on_disk_ bitmap:
1296 * bit 0 bit 37 bit 38 bit (512*8)-1
1297 * ...|........|........|.. // ..|........|
1298 * sect. 0 `296 `304 ^(512*8*8)-1
1300 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1301 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1302 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1305 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1306 #define DRBD_MAX_SECTORS_BM \
1307 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1308 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1309 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1310 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1311 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1312 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1313 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1315 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1316 /* 16 TB in units of sectors */
1317 #if BITS_PER_LONG == 32
1318 /* adjust by one page worth of bitmap,
1319 * so we won't wrap around in drbd_bm_find_next_bit.
1320 * you should use 64bit OS for that much storage, anyways. */
1321 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1323 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1324 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1325 /* corresponds to (1UL << 38) bits right now. */
1329 /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE,
1330 * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte.
1331 * Since we may live in a mixed-platform cluster,
1332 * we limit us to a platform agnostic constant here for now.
1333 * A followup commit may allow even bigger BIO sizes,
1334 * once we thought that through. */
1335 #define DRBD_MAX_BIO_SIZE (1 << 20)
1336 #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1337 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1339 #define DRBD_MAX_BIO_SIZE_SAFE (1 << 12) /* Works always = 4k */
1341 #define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */
1343 extern int drbd_bm_init(struct drbd_conf *mdev);
1344 extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits);
1345 extern void drbd_bm_cleanup(struct drbd_conf *mdev);
1346 extern void drbd_bm_set_all(struct drbd_conf *mdev);
1347 extern void drbd_bm_clear_all(struct drbd_conf *mdev);
1348 /* set/clear/test only a few bits at a time */
1349 extern int drbd_bm_set_bits(
1350 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1351 extern int drbd_bm_clear_bits(
1352 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1353 extern int drbd_bm_count_bits(
1354 struct drbd_conf *mdev, const unsigned long s, const unsigned long e);
1355 /* bm_set_bits variant for use while holding drbd_bm_lock,
1356 * may process the whole bitmap in one go */
1357 extern void _drbd_bm_set_bits(struct drbd_conf *mdev,
1358 const unsigned long s, const unsigned long e);
1359 extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr);
1360 extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr);
1361 extern int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local);
1362 extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
1363 extern void drbd_bm_mark_for_writeout(struct drbd_conf *mdev, int page_nr);
1364 extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
1365 extern int drbd_bm_write_hinted(struct drbd_conf *mdev) __must_hold(local);
1366 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev,
1367 unsigned long al_enr);
1368 extern size_t drbd_bm_words(struct drbd_conf *mdev);
1369 extern unsigned long drbd_bm_bits(struct drbd_conf *mdev);
1370 extern sector_t drbd_bm_capacity(struct drbd_conf *mdev);
1372 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1373 extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1374 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1375 extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1376 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo);
1377 extern unsigned long _drbd_bm_total_weight(struct drbd_conf *mdev);
1378 extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev);
1379 extern int drbd_bm_rs_done(struct drbd_conf *mdev);
1380 /* for receive_bitmap */
1381 extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset,
1382 size_t number, unsigned long *buffer);
1383 /* for _drbd_send_bitmap */
1384 extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset,
1385 size_t number, unsigned long *buffer);
1387 extern void drbd_bm_lock(struct drbd_conf *mdev, char *why, enum bm_flag flags);
1388 extern void drbd_bm_unlock(struct drbd_conf *mdev);
1391 extern struct kmem_cache *drbd_request_cache;
1392 extern struct kmem_cache *drbd_ee_cache; /* peer requests */
1393 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1394 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1395 extern mempool_t *drbd_request_mempool;
1396 extern mempool_t *drbd_ee_mempool;
1398 /* drbd's page pool, used to buffer data received from the peer,
1399 * or data requested by the peer.
1401 * This does not have an emergency reserve.
1403 * When allocating from this pool, it first takes pages from the pool.
1404 * Only if the pool is depleted will try to allocate from the system.
1406 * The assumption is that pages taken from this pool will be processed,
1407 * and given back, "quickly", and then can be recycled, so we can avoid
1408 * frequent calls to alloc_page(), and still will be able to make progress even
1409 * under memory pressure.
1411 extern struct page *drbd_pp_pool;
1412 extern spinlock_t drbd_pp_lock;
1413 extern int drbd_pp_vacant;
1414 extern wait_queue_head_t drbd_pp_wait;
1416 /* We also need a standard (emergency-reserve backed) page pool
1417 * for meta data IO (activity log, bitmap).
1418 * We can keep it global, as long as it is used as "N pages at a time".
1419 * 128 should be plenty, currently we probably can get away with as few as 1.
1421 #define DRBD_MIN_POOL_PAGES 128
1422 extern mempool_t *drbd_md_io_page_pool;
1424 /* We also need to make sure we get a bio
1425 * when we need it for housekeeping purposes */
1426 extern struct bio_set *drbd_md_io_bio_set;
1427 /* to allocate from that set */
1428 extern struct bio *bio_alloc_drbd(gfp_t gfp_mask);
1430 extern rwlock_t global_state_lock;
1432 extern int conn_lowest_minor(struct drbd_tconn *tconn);
1433 enum drbd_ret_code conn_new_minor(struct drbd_tconn *tconn, unsigned int minor, int vnr);
1434 extern void drbd_free_mdev(struct drbd_conf *mdev);
1435 extern void drbd_delete_device(unsigned int minor);
1437 struct drbd_tconn *drbd_new_tconn(const char *name);
1438 extern void drbd_free_tconn(struct drbd_tconn *tconn);
1439 struct drbd_tconn *conn_by_name(const char *name);
1441 extern int proc_details;
1444 extern int __drbd_make_request(struct drbd_conf *, struct bio *, unsigned long);
1445 extern int drbd_make_request(struct request_queue *q, struct bio *bio);
1446 extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req);
1447 extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1448 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1452 extern int drbd_msg_put_info(const char *info);
1453 extern void drbd_suspend_io(struct drbd_conf *mdev);
1454 extern void drbd_resume_io(struct drbd_conf *mdev);
1455 extern char *ppsize(char *buf, unsigned long long size);
1456 extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int);
1457 enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
1458 extern enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local);
1459 extern void resync_after_online_grow(struct drbd_conf *);
1460 extern void drbd_reconsider_max_bio_size(struct drbd_conf *mdev);
1461 extern enum drbd_state_rv drbd_set_role(struct drbd_conf *mdev,
1462 enum drbd_role new_role,
1464 extern bool conn_try_outdate_peer(struct drbd_tconn *tconn);
1465 extern void conn_try_outdate_peer_async(struct drbd_tconn *tconn);
1466 extern int drbd_khelper(struct drbd_conf *mdev, char *cmd);
1469 extern int drbd_worker(struct drbd_thread *thi);
1470 extern int drbd_alter_sa(struct drbd_conf *mdev, int na);
1471 extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side);
1472 extern void resume_next_sg(struct drbd_conf *mdev);
1473 extern void suspend_other_sg(struct drbd_conf *mdev);
1474 extern int drbd_resync_finished(struct drbd_conf *mdev);
1475 /* maybe rather drbd_main.c ? */
1476 extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
1477 struct drbd_backing_dev *bdev, sector_t sector, int rw);
1478 extern void drbd_ov_out_of_sync_found(struct drbd_conf *, sector_t, int);
1479 extern void drbd_rs_controller_reset(struct drbd_conf *mdev);
1481 static inline void ov_out_of_sync_print(struct drbd_conf *mdev)
1483 if (mdev->ov_last_oos_size) {
1484 dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n",
1485 (unsigned long long)mdev->ov_last_oos_start,
1486 (unsigned long)mdev->ov_last_oos_size);
1488 mdev->ov_last_oos_size=0;
1492 extern void drbd_csum_bio(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
1493 extern void drbd_csum_ee(struct drbd_conf *, struct crypto_hash *,
1494 struct drbd_peer_request *, void *);
1495 /* worker callbacks */
1496 extern int w_read_retry_remote(struct drbd_work *, int);
1497 extern int w_e_end_data_req(struct drbd_work *, int);
1498 extern int w_e_end_rsdata_req(struct drbd_work *, int);
1499 extern int w_e_end_csum_rs_req(struct drbd_work *, int);
1500 extern int w_e_end_ov_reply(struct drbd_work *, int);
1501 extern int w_e_end_ov_req(struct drbd_work *, int);
1502 extern int w_ov_finished(struct drbd_work *, int);
1503 extern int w_resync_timer(struct drbd_work *, int);
1504 extern int w_send_write_hint(struct drbd_work *, int);
1505 extern int w_make_resync_request(struct drbd_work *, int);
1506 extern int w_send_dblock(struct drbd_work *, int);
1507 extern int w_send_barrier(struct drbd_work *, int);
1508 extern int w_send_read_req(struct drbd_work *, int);
1509 extern int w_prev_work_done(struct drbd_work *, int);
1510 extern int w_e_reissue(struct drbd_work *, int);
1511 extern int w_restart_disk_io(struct drbd_work *, int);
1512 extern int w_send_out_of_sync(struct drbd_work *, int);
1513 extern int w_start_resync(struct drbd_work *, int);
1515 extern void resync_timer_fn(unsigned long data);
1516 extern void start_resync_timer_fn(unsigned long data);
1518 /* drbd_receiver.c */
1519 extern int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector);
1520 extern int drbd_submit_peer_request(struct drbd_conf *,
1521 struct drbd_peer_request *, const unsigned,
1523 extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list);
1524 extern struct drbd_peer_request *drbd_alloc_ee(struct drbd_conf *,
1525 u64, sector_t, unsigned int,
1526 gfp_t) __must_hold(local);
1527 extern void drbd_free_some_ee(struct drbd_conf *, struct drbd_peer_request *,
1529 #define drbd_free_ee(m,e) drbd_free_some_ee(m, e, 0)
1530 #define drbd_free_net_ee(m,e) drbd_free_some_ee(m, e, 1)
1531 extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1532 struct list_head *head);
1533 extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1534 struct list_head *head);
1535 extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled);
1536 extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed);
1537 extern void conn_flush_workqueue(struct drbd_tconn *tconn);
1538 extern int drbd_connected(int vnr, void *p, void *data);
1539 static inline void drbd_flush_workqueue(struct drbd_conf *mdev)
1541 conn_flush_workqueue(mdev->tconn);
1544 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1545 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1546 static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1547 char __user *optval, int optlen)
1550 if (level == SOL_SOCKET)
1551 err = sock_setsockopt(sock, level, optname, optval, optlen);
1553 err = sock->ops->setsockopt(sock, level, optname, optval,
1558 static inline void drbd_tcp_cork(struct socket *sock)
1561 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1562 (char __user *)&val, sizeof(val));
1565 static inline void drbd_tcp_uncork(struct socket *sock)
1568 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1569 (char __user *)&val, sizeof(val));
1572 static inline void drbd_tcp_nodelay(struct socket *sock)
1575 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1576 (char __user *)&val, sizeof(val));
1579 static inline void drbd_tcp_quickack(struct socket *sock)
1582 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1583 (char __user *)&val, sizeof(val));
1586 void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo);
1589 extern struct proc_dir_entry *drbd_proc;
1590 extern const struct file_operations drbd_proc_fops;
1591 extern const char *drbd_conn_str(enum drbd_conns s);
1592 extern const char *drbd_role_str(enum drbd_role s);
1595 extern void drbd_al_begin_io(struct drbd_conf *mdev, struct drbd_interval *i);
1596 extern void drbd_al_complete_io(struct drbd_conf *mdev, struct drbd_interval *i);
1597 extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector);
1598 extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1599 extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1600 extern void drbd_rs_cancel_all(struct drbd_conf *mdev);
1601 extern int drbd_rs_del_all(struct drbd_conf *mdev);
1602 extern void drbd_rs_failed_io(struct drbd_conf *mdev,
1603 sector_t sector, int size);
1604 extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *);
1605 extern void drbd_advance_rs_marks(struct drbd_conf *mdev, unsigned long still_to_go);
1606 extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
1607 int size, const char *file, const unsigned int line);
1608 #define drbd_set_in_sync(mdev, sector, size) \
1609 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1610 extern int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
1611 int size, const char *file, const unsigned int line);
1612 #define drbd_set_out_of_sync(mdev, sector, size) \
1613 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1614 extern void drbd_al_apply_to_bm(struct drbd_conf *mdev);
1615 extern void drbd_al_shrink(struct drbd_conf *mdev);
1618 /* state info broadcast */
1620 enum drbd_state_info_bcast_reason sib_reason;
1624 unsigned helper_exit_code;
1627 union drbd_state os;
1628 union drbd_state ns;
1632 void drbd_bcast_event(struct drbd_conf *mdev, const struct sib_info *sib);
1635 * inline helper functions
1636 *************************/
1638 /* see also page_chain_add and friends in drbd_receiver.c */
1639 static inline struct page *page_chain_next(struct page *page)
1641 return (struct page *)page_private(page);
1643 #define page_chain_for_each(page) \
1644 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1645 page = page_chain_next(page))
1646 #define page_chain_for_each_safe(page, n) \
1647 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1649 static inline int drbd_bio_has_active_page(struct bio *bio)
1651 struct bio_vec *bvec;
1654 __bio_for_each_segment(bvec, bio, i, 0) {
1655 if (page_count(bvec->bv_page) > 1)
1662 static inline int drbd_ee_has_active_page(struct drbd_peer_request *peer_req)
1664 struct page *page = peer_req->pages;
1665 page_chain_for_each(page) {
1666 if (page_count(page) > 1)
1672 static inline enum drbd_state_rv
1673 _drbd_set_state(struct drbd_conf *mdev, union drbd_state ns,
1674 enum chg_state_flags flags, struct completion *done)
1676 enum drbd_state_rv rv;
1678 read_lock(&global_state_lock);
1679 rv = __drbd_set_state(mdev, ns, flags, done);
1680 read_unlock(&global_state_lock);
1685 static inline union drbd_state drbd_read_state(struct drbd_conf *mdev)
1687 union drbd_state rv;
1694 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1695 static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach, const char *where)
1697 switch (mdev->ldev->dc.on_io_error) {
1700 if (__ratelimit(&drbd_ratelimit_state))
1701 dev_err(DEV, "Local IO failed in %s.\n", where);
1702 if (mdev->state.disk > D_INCONSISTENT)
1703 _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_HARD, NULL);
1706 /* NOTE fall through to detach case if forcedetach set */
1708 case EP_CALL_HELPER:
1709 set_bit(WAS_IO_ERROR, &mdev->flags);
1710 if (mdev->state.disk > D_FAILED) {
1711 _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL);
1713 "Local IO failed in %s. Detaching...\n", where);
1720 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1721 * @mdev: DRBD device.
1722 * @error: Error code passed to the IO completion callback
1723 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1725 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1727 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1728 static inline void drbd_chk_io_error_(struct drbd_conf *mdev,
1729 int error, int forcedetach, const char *where)
1732 unsigned long flags;
1733 spin_lock_irqsave(&mdev->tconn->req_lock, flags);
1734 __drbd_chk_io_error_(mdev, forcedetach, where);
1735 spin_unlock_irqrestore(&mdev->tconn->req_lock, flags);
1741 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1742 * @bdev: Meta data block device.
1744 * BTW, for internal meta data, this happens to be the maximum capacity
1745 * we could agree upon with our peer node.
1747 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1749 switch (bdev->dc.meta_dev_idx) {
1750 case DRBD_MD_INDEX_INTERNAL:
1751 case DRBD_MD_INDEX_FLEX_INT:
1752 return bdev->md.md_offset + bdev->md.bm_offset;
1753 case DRBD_MD_INDEX_FLEX_EXT:
1755 return bdev->md.md_offset;
1760 * drbd_md_last_sector() - Return the last sector number of the meta data area
1761 * @bdev: Meta data block device.
1763 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1765 switch (bdev->dc.meta_dev_idx) {
1766 case DRBD_MD_INDEX_INTERNAL:
1767 case DRBD_MD_INDEX_FLEX_INT:
1768 return bdev->md.md_offset + MD_AL_OFFSET - 1;
1769 case DRBD_MD_INDEX_FLEX_EXT:
1771 return bdev->md.md_offset + bdev->md.md_size_sect;
1775 /* Returns the number of 512 byte sectors of the device */
1776 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1778 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1779 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0;
1783 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1784 * @bdev: Meta data block device.
1786 * returns the capacity we announce to out peer. we clip ourselves at the
1787 * various MAX_SECTORS, because if we don't, current implementation will
1788 * oops sooner or later
1790 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1793 switch (bdev->dc.meta_dev_idx) {
1794 case DRBD_MD_INDEX_INTERNAL:
1795 case DRBD_MD_INDEX_FLEX_INT:
1796 s = drbd_get_capacity(bdev->backing_bdev)
1797 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1798 drbd_md_first_sector(bdev))
1801 case DRBD_MD_INDEX_FLEX_EXT:
1802 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1803 drbd_get_capacity(bdev->backing_bdev));
1804 /* clip at maximum size the meta device can support */
1805 s = min_t(sector_t, s,
1806 BM_EXT_TO_SECT(bdev->md.md_size_sect
1807 - bdev->md.bm_offset));
1810 s = min_t(sector_t, DRBD_MAX_SECTORS,
1811 drbd_get_capacity(bdev->backing_bdev));
1817 * drbd_md_ss__() - Return the sector number of our meta data super block
1818 * @mdev: DRBD device.
1819 * @bdev: Meta data block device.
1821 static inline sector_t drbd_md_ss__(struct drbd_conf *mdev,
1822 struct drbd_backing_dev *bdev)
1824 switch (bdev->dc.meta_dev_idx) {
1825 default: /* external, some index */
1826 return MD_RESERVED_SECT * bdev->dc.meta_dev_idx;
1827 case DRBD_MD_INDEX_INTERNAL:
1828 /* with drbd08, internal meta data is always "flexible" */
1829 case DRBD_MD_INDEX_FLEX_INT:
1830 /* sizeof(struct md_on_disk_07) == 4k
1831 * position: last 4k aligned block of 4k size */
1832 if (!bdev->backing_bdev) {
1833 if (__ratelimit(&drbd_ratelimit_state)) {
1834 dev_err(DEV, "bdev->backing_bdev==NULL\n");
1839 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL)
1841 case DRBD_MD_INDEX_FLEX_EXT:
1847 drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w)
1849 unsigned long flags;
1850 spin_lock_irqsave(&q->q_lock, flags);
1851 list_add(&w->list, &q->q);
1852 up(&q->s); /* within the spinlock,
1853 see comment near end of drbd_worker() */
1854 spin_unlock_irqrestore(&q->q_lock, flags);
1858 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1860 unsigned long flags;
1861 spin_lock_irqsave(&q->q_lock, flags);
1862 list_add_tail(&w->list, &q->q);
1863 up(&q->s); /* within the spinlock,
1864 see comment near end of drbd_worker() */
1865 spin_unlock_irqrestore(&q->q_lock, flags);
1868 static inline void wake_asender(struct drbd_tconn *tconn)
1870 if (test_bit(SIGNAL_ASENDER, &tconn->flags))
1871 force_sig(DRBD_SIG, tconn->asender.task);
1874 static inline void request_ping(struct drbd_tconn *tconn)
1876 set_bit(SEND_PING, &tconn->flags);
1877 wake_asender(tconn);
1880 static inline int _drbd_send_cmd(struct drbd_conf *mdev, struct drbd_socket *sock,
1881 enum drbd_packet cmd, struct p_header *h, size_t size,
1884 return _conn_send_cmd(mdev->tconn, mdev->vnr, sock, cmd, h, size, msg_flags);
1887 static inline int drbd_send_cmd(struct drbd_conf *mdev, struct drbd_socket *sock,
1888 enum drbd_packet cmd, struct p_header *h, size_t size)
1890 return conn_send_cmd(mdev->tconn, mdev->vnr, sock, cmd, h, size);
1893 static inline int drbd_send_short_cmd(struct drbd_conf *mdev,
1894 enum drbd_packet cmd)
1897 return drbd_send_cmd(mdev, &mdev->tconn->data, cmd, &h, sizeof(h));
1900 extern int drbd_send_ping(struct drbd_tconn *tconn);
1901 extern int drbd_send_ping_ack(struct drbd_tconn *tconn);
1903 static inline int drbd_send_state_req(struct drbd_conf *mdev,
1904 union drbd_state mask, union drbd_state val)
1906 return _conn_send_state_req(mdev->tconn, mdev->vnr, P_STATE_CHG_REQ, mask, val);
1909 static inline int conn_send_state_req(struct drbd_tconn *tconn,
1910 union drbd_state mask, union drbd_state val)
1912 enum drbd_packet cmd = tconn->agreed_pro_version < 100 ? P_STATE_CHG_REQ : P_CONN_ST_CHG_REQ;
1913 return _conn_send_state_req(tconn, 0, cmd, mask, val);
1916 static inline void drbd_thread_stop(struct drbd_thread *thi)
1918 _drbd_thread_stop(thi, false, true);
1921 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
1923 _drbd_thread_stop(thi, false, false);
1926 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
1928 _drbd_thread_stop(thi, true, false);
1931 /* counts how many answer packets packets we expect from our peer,
1932 * for either explicit application requests,
1933 * or implicit barrier packets as necessary.
1936 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1937 * it is much easier and equally valid to count what we queue for the
1938 * worker, even before it actually was queued or send.
1939 * (drbd_make_request_common; recovery path on read io-error)
1941 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1942 * _req_mod(req, DATA_RECEIVED)
1943 * [from receive_DataReply]
1944 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1945 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1946 * for some reason it is NOT decreased in got_NegAck,
1947 * but in the resulting cleanup code from report_params.
1948 * we should try to remember the reason for that...
1949 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
1950 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
1951 * [from tl_clear_barrier]
1953 static inline void inc_ap_pending(struct drbd_conf *mdev)
1955 atomic_inc(&mdev->ap_pending_cnt);
1958 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \
1959 if (atomic_read(&mdev->which) < 0) \
1960 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
1962 atomic_read(&mdev->which))
1964 #define dec_ap_pending(mdev) _dec_ap_pending(mdev, __FUNCTION__, __LINE__)
1965 static inline void _dec_ap_pending(struct drbd_conf *mdev, const char *func, int line)
1967 if (atomic_dec_and_test(&mdev->ap_pending_cnt))
1968 wake_up(&mdev->misc_wait);
1969 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line);
1972 /* counts how many resync-related answers we still expect from the peer
1974 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1975 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
1976 * (or P_NEG_ACK with ID_SYNCER)
1978 static inline void inc_rs_pending(struct drbd_conf *mdev)
1980 atomic_inc(&mdev->rs_pending_cnt);
1983 #define dec_rs_pending(mdev) _dec_rs_pending(mdev, __FUNCTION__, __LINE__)
1984 static inline void _dec_rs_pending(struct drbd_conf *mdev, const char *func, int line)
1986 atomic_dec(&mdev->rs_pending_cnt);
1987 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line);
1990 /* counts how many answers we still need to send to the peer.
1992 * receive_Data unless protocol A;
1993 * we need to send a P_RECV_ACK (proto B)
1994 * or P_WRITE_ACK (proto C)
1995 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1996 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1997 * receive_Barrier_* we need to send a P_BARRIER_ACK
1999 static inline void inc_unacked(struct drbd_conf *mdev)
2001 atomic_inc(&mdev->unacked_cnt);
2004 #define dec_unacked(mdev) _dec_unacked(mdev, __FUNCTION__, __LINE__)
2005 static inline void _dec_unacked(struct drbd_conf *mdev, const char *func, int line)
2007 atomic_dec(&mdev->unacked_cnt);
2008 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
2011 #define sub_unacked(mdev, n) _sub_unacked(mdev, n, __FUNCTION__, __LINE__)
2012 static inline void _sub_unacked(struct drbd_conf *mdev, int n, const char *func, int line)
2014 atomic_sub(n, &mdev->unacked_cnt);
2015 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
2018 static inline void put_net_conf(struct drbd_tconn *tconn)
2020 if (atomic_dec_and_test(&tconn->net_cnt))
2021 wake_up(&tconn->net_cnt_wait);
2025 * get_net_conf() - Increase ref count on mdev->tconn->net_conf; Returns 0 if nothing there
2026 * @mdev: DRBD device.
2028 * You have to call put_net_conf() when finished working with mdev->tconn->net_conf.
2030 static inline int get_net_conf(struct drbd_tconn *tconn)
2034 atomic_inc(&tconn->net_cnt);
2035 have_net_conf = tconn->cstate >= C_UNCONNECTED;
2037 put_net_conf(tconn);
2038 return have_net_conf;
2042 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2045 * You have to call put_ldev() when finished working with mdev->ldev.
2047 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2048 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2050 static inline void put_ldev(struct drbd_conf *mdev)
2052 int i = atomic_dec_return(&mdev->local_cnt);
2054 /* This may be called from some endio handler,
2055 * so we must not sleep here. */
2060 if (mdev->state.disk == D_DISKLESS)
2061 /* even internal references gone, safe to destroy */
2062 drbd_ldev_destroy(mdev);
2063 if (mdev->state.disk == D_FAILED)
2064 /* all application IO references gone. */
2065 drbd_go_diskless(mdev);
2066 wake_up(&mdev->misc_wait);
2071 static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
2075 /* never get a reference while D_DISKLESS */
2076 if (mdev->state.disk == D_DISKLESS)
2079 atomic_inc(&mdev->local_cnt);
2080 io_allowed = (mdev->state.disk >= mins);
2086 extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins);
2089 /* you must have an "get_ldev" reference */
2090 static inline void drbd_get_syncer_progress(struct drbd_conf *mdev,
2091 unsigned long *bits_left, unsigned int *per_mil_done)
2093 /* this is to break it at compile time when we change that, in case we
2094 * want to support more than (1<<32) bits on a 32bit arch. */
2095 typecheck(unsigned long, mdev->rs_total);
2097 /* note: both rs_total and rs_left are in bits, i.e. in
2098 * units of BM_BLOCK_SIZE.
2099 * for the percentage, we don't care. */
2101 if (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T)
2102 *bits_left = mdev->ov_left;
2104 *bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
2105 /* >> 10 to prevent overflow,
2106 * +1 to prevent division by zero */
2107 if (*bits_left > mdev->rs_total) {
2108 /* doh. maybe a logic bug somewhere.
2109 * may also be just a race condition
2110 * between this and a disconnect during sync.
2111 * for now, just prevent in-kernel buffer overflow.
2114 dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2115 drbd_conn_str(mdev->state.conn),
2116 *bits_left, mdev->rs_total, mdev->rs_failed);
2119 /* Make sure the division happens in long context.
2120 * We allow up to one petabyte storage right now,
2121 * at a granularity of 4k per bit that is 2**38 bits.
2122 * After shift right and multiplication by 1000,
2123 * this should still fit easily into a 32bit long,
2124 * so we don't need a 64bit division on 32bit arch.
2125 * Note: currently we don't support such large bitmaps on 32bit
2126 * arch anyways, but no harm done to be prepared for it here.
2128 unsigned int shift = mdev->rs_total >= (1ULL << 32) ? 16 : 10;
2129 unsigned long left = *bits_left >> shift;
2130 unsigned long total = 1UL + (mdev->rs_total >> shift);
2131 unsigned long tmp = 1000UL - left * 1000UL/total;
2132 *per_mil_done = tmp;
2137 /* this throttles on-the-fly application requests
2138 * according to max_buffers settings;
2139 * maybe re-implement using semaphores? */
2140 static inline int drbd_get_max_buffers(struct drbd_conf *mdev)
2142 int mxb = 1000000; /* arbitrary limit on open requests */
2143 if (get_net_conf(mdev->tconn)) {
2144 mxb = mdev->tconn->net_conf->max_buffers;
2145 put_net_conf(mdev->tconn);
2150 static inline int drbd_state_is_stable(struct drbd_conf *mdev)
2152 union drbd_state s = mdev->state;
2154 /* DO NOT add a default clause, we want the compiler to warn us
2155 * for any newly introduced state we may have forgotten to add here */
2157 switch ((enum drbd_conns)s.conn) {
2158 /* new io only accepted when there is no connection, ... */
2160 case C_WF_CONNECTION:
2161 /* ... or there is a well established connection. */
2167 case C_PAUSED_SYNC_S:
2168 case C_PAUSED_SYNC_T:
2171 /* transitional states, IO allowed */
2172 case C_DISCONNECTING:
2176 case C_NETWORK_FAILURE:
2177 case C_PROTOCOL_ERROR:
2179 case C_WF_REPORT_PARAMS:
2180 case C_STARTING_SYNC_S:
2181 case C_STARTING_SYNC_T:
2184 /* Allow IO in BM exchange states with new protocols */
2186 if (mdev->tconn->agreed_pro_version < 96)
2190 /* no new io accepted in these states */
2192 case C_WF_SYNC_UUID:
2198 switch ((enum drbd_disk_state)s.disk) {
2200 case D_INCONSISTENT:
2204 /* disk state is stable as well. */
2207 /* no new io accepted during tansitional states */
2220 static inline int is_susp(union drbd_state s)
2222 return s.susp || s.susp_nod || s.susp_fen;
2225 static inline bool may_inc_ap_bio(struct drbd_conf *mdev)
2227 int mxb = drbd_get_max_buffers(mdev);
2229 if (is_susp(mdev->state))
2231 if (test_bit(SUSPEND_IO, &mdev->flags))
2234 /* to avoid potential deadlock or bitmap corruption,
2235 * in various places, we only allow new application io
2236 * to start during "stable" states. */
2238 /* no new io accepted when attaching or detaching the disk */
2239 if (!drbd_state_is_stable(mdev))
2242 /* since some older kernels don't have atomic_add_unless,
2243 * and we are within the spinlock anyways, we have this workaround. */
2244 if (atomic_read(&mdev->ap_bio_cnt) > mxb)
2246 if (test_bit(BITMAP_IO, &mdev->flags))
2251 static inline bool inc_ap_bio_cond(struct drbd_conf *mdev)
2255 spin_lock_irq(&mdev->tconn->req_lock);
2256 rv = may_inc_ap_bio(mdev);
2258 atomic_inc(&mdev->ap_bio_cnt);
2259 spin_unlock_irq(&mdev->tconn->req_lock);
2264 static inline void inc_ap_bio(struct drbd_conf *mdev)
2267 * as long as the device is suspended
2268 * until the bitmap is no longer on the fly during connection
2269 * handshake as long as we would exeed the max_buffer limit.
2271 * to avoid races with the reconnect code,
2272 * we need to atomic_inc within the spinlock. */
2274 wait_event(mdev->misc_wait, inc_ap_bio_cond(mdev));
2277 static inline void dec_ap_bio(struct drbd_conf *mdev)
2279 int mxb = drbd_get_max_buffers(mdev);
2280 int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt);
2282 D_ASSERT(ap_bio >= 0);
2283 /* this currently does wake_up for every dec_ap_bio!
2284 * maybe rather introduce some type of hysteresis?
2285 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2287 wake_up(&mdev->misc_wait);
2288 if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) {
2289 if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags))
2290 drbd_queue_work(&mdev->tconn->data.work, &mdev->bm_io_work.w);
2294 static inline int drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val)
2296 int changed = mdev->ed_uuid != val;
2297 mdev->ed_uuid = val;
2301 static inline int drbd_queue_order_type(struct drbd_conf *mdev)
2303 /* sorry, we currently have no working implementation
2304 * of distributed TCQ stuff */
2305 #ifndef QUEUE_ORDERED_NONE
2306 #define QUEUE_ORDERED_NONE 0
2308 return QUEUE_ORDERED_NONE;
2311 static inline void drbd_md_flush(struct drbd_conf *mdev)
2315 if (test_bit(MD_NO_FUA, &mdev->flags))
2318 r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL);
2320 set_bit(MD_NO_FUA, &mdev->flags);
2321 dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r);