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_HAND_SHAKE_M
= 0xfff1, /* First Packet on the MetaSock */
236 P_HAND_SHAKE_S
= 0xfff2, /* First Packet on the Socket */
238 P_HAND_SHAKE
= 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:
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"
558 struct p_header header
;
559 struct p_handshake handshake
;
561 struct p_block_ack block_ack
;
562 struct p_barrier barrier
;
563 struct p_barrier_ack barrier_ack
;
564 struct p_rs_param_89 rs_param_89
;
565 struct p_rs_param_95 rs_param_95
;
566 struct p_protocol protocol
;
567 struct p_sizes sizes
;
568 struct p_uuids uuids
;
569 struct p_state state
;
570 struct p_req_state req_state
;
571 struct p_req_state_reply req_state_reply
;
572 struct p_block_req block_req
;
573 struct p_delay_probe93 delay_probe93
;
574 struct p_rs_uuid rs_uuid
;
575 struct p_block_desc block_desc
;
578 /**********************************************************************/
579 enum drbd_thread_state
{
588 struct task_struct
*task
;
589 struct completion stop
;
590 enum drbd_thread_state t_state
;
591 int (*function
) (struct drbd_thread
*);
592 struct drbd_tconn
*tconn
;
597 static inline enum drbd_thread_state
get_t_state(struct drbd_thread
*thi
)
599 /* THINK testing the t_state seems to be uncritical in all cases
600 * (but thread_{start,stop}), so we can read it *without* the lock.
608 struct list_head list
;
609 int (*cb
)(struct drbd_work
*, int cancel
);
611 struct drbd_conf
*mdev
;
612 struct drbd_tconn
*tconn
;
616 #include "drbd_interval.h"
618 extern int drbd_wait_misc(struct drbd_conf
*, struct drbd_interval
*);
620 struct drbd_request
{
623 /* if local IO is not allowed, will be NULL.
624 * if local IO _is_ allowed, holds the locally submitted bio clone,
625 * or, after local IO completion, the ERR_PTR(error).
626 * see drbd_request_endio(). */
627 struct bio
*private_bio
;
629 struct drbd_interval i
;
630 unsigned int epoch
; /* barrier_nr */
632 /* barrier_nr: used to check on "completion" whether this req was in
633 * the current epoch, and we therefore have to close it,
634 * starting a new epoch...
637 struct list_head tl_requests
; /* ring list in the transfer log */
638 struct bio
*master_bio
; /* master bio pointer */
639 unsigned long rq_state
; /* see comments above _req_mod() */
641 unsigned long start_time
;
644 struct drbd_tl_epoch
{
646 struct list_head requests
; /* requests before */
647 struct drbd_tl_epoch
*next
; /* pointer to the next barrier */
648 unsigned int br_number
; /* the barriers identifier. */
649 int n_writes
; /* number of requests attached before this barrier */
653 struct list_head list
;
654 unsigned int barrier_nr
;
655 atomic_t epoch_size
; /* increased on every request added. */
656 atomic_t active
; /* increased on every req. added, and dec on every finished. */
660 /* drbd_epoch flag bits */
662 DE_HAVE_BARRIER_NUMBER
,
669 EV_CLEANUP
= 32, /* used as flag */
672 struct drbd_wq_barrier
{
674 struct completion done
;
682 struct drbd_peer_request
{
684 struct drbd_epoch
*epoch
; /* for writes */
686 atomic_t pending_bios
;
687 struct drbd_interval i
;
688 /* see comments on ee flag bits below */
692 struct digest_info
*digest
;
697 * While corresponding bios are in flight, the only modification will be
698 * set_bit WAS_ERROR, which has to be atomic.
699 * If no bios are in flight yet, or all have been completed,
700 * non-atomic modification to ee->flags is ok.
703 __EE_CALL_AL_COMPLETE_IO
,
704 __EE_MAY_SET_IN_SYNC
,
706 /* In case a barrier failed,
707 * we need to resubmit without the barrier flag. */
710 /* we may have several bios per peer request.
711 * if any of those fail, we set this flag atomically
712 * from the endio callback */
715 /* This ee has a pointer to a digest instead of a block id */
718 /* Conflicting local requests need to be restarted after this request */
719 __EE_RESTART_REQUESTS
,
721 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
722 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
723 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
724 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
725 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
726 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
728 /* flag bits per mdev */
730 CREATE_BARRIER
, /* next P_DATA is preceded by a P_BARRIER */
731 UNPLUG_QUEUED
, /* only relevant with kernel 2.4 */
732 UNPLUG_REMOTE
, /* sending a "UnplugRemote" could help */
733 MD_DIRTY
, /* current uuids and flags not yet on disk */
734 USE_DEGR_WFC_T
, /* degr-wfc-timeout instead of wfc-timeout. */
737 CRASHED_PRIMARY
, /* This node was a crashed primary.
738 * Gets cleared when the state.conn
739 * goes into C_CONNECTED state. */
740 NO_BARRIER_SUPP
, /* underlying block device doesn't implement barriers */
743 MD_NO_FUA
, /* Users wants us to not use FUA/FLUSH on meta data dev */
744 SUSPEND_IO
, /* suspend application io */
745 BITMAP_IO
, /* suspend application io;
746 once no more io in flight, start bitmap io */
747 BITMAP_IO_QUEUED
, /* Started bitmap IO */
748 GO_DISKLESS
, /* Disk is being detached, on io-error or admin request. */
749 WAS_IO_ERROR
, /* Local disk failed returned IO error */
750 RESYNC_AFTER_NEG
, /* Resync after online grow after the attach&negotiate finished. */
751 RESIZE_PENDING
, /* Size change detected locally, waiting for the response from
752 * the peer, if it changed there as well. */
753 NEW_CUR_UUID
, /* Create new current UUID when thawing IO */
754 AL_SUSPENDED
, /* Activity logging is currently suspended. */
755 AHEAD_TO_SYNC_SOURCE
, /* Ahead -> SyncSource queued */
756 B_RS_H_DONE
, /* Before resync handler done (already executed) */
759 struct drbd_bitmap
; /* opaque for drbd_conf */
761 /* definition of bits in bm_flags to be used in drbd_bm_lock
762 * and drbd_bitmap_io and friends. */
764 /* do we need to kfree, or vfree bm_pages? */
765 BM_P_VMALLOCED
= 0x10000, /* internal use only, will be masked out */
767 /* currently locked for bulk operation */
768 BM_LOCKED_MASK
= 0x7,
770 /* in detail, that is: */
775 /* (test bit, count bit) allowed (common case) */
776 BM_LOCKED_TEST_ALLOWED
= 0x3,
778 /* testing bits, as well as setting new bits allowed, but clearing bits
779 * would be unexpected. Used during bitmap receive. Setting new bits
780 * requires sending of "out-of-sync" information, though. */
781 BM_LOCKED_SET_ALLOWED
= 0x1,
783 /* clear is not expected while bitmap is locked for bulk operation */
787 /* TODO sort members for performance
788 * MAYBE group them further */
790 /* THINK maybe we actually want to use the default "event/%s" worker threads
791 * or similar in linux 2.6, which uses per cpu data and threads.
793 struct drbd_work_queue
{
795 struct semaphore s
; /* producers up it, worker down()s it */
796 spinlock_t q_lock
; /* to protect the list. */
800 struct drbd_work_queue work
;
802 struct socket
*socket
;
803 /* this way we get our
804 * send/receive buffers off the stack */
805 union p_polymorph sbuf
;
806 union p_polymorph rbuf
;
810 u64 md_offset
; /* sector offset to 'super' block */
812 u64 la_size_sect
; /* last agreed size, unit sectors */
818 s32 al_offset
; /* signed relative sector offset to al area */
819 s32 bm_offset
; /* signed relative sector offset to bitmap */
821 /* u32 al_nr_extents; important for restoring the AL
822 * is stored into ldev->dc.al_extents, which in turn
823 * gets applied to act_log->nr_elements
827 struct drbd_backing_dev
{
828 struct block_device
*backing_bdev
;
829 struct block_device
*md_bdev
;
831 struct disk_conf dc
; /* The user provided config... */
832 sector_t known_size
; /* last known size of that backing device */
836 struct drbd_conf
*mdev
;
837 struct completion event
;
845 int (*io_fn
)(struct drbd_conf
*mdev
);
846 void (*done
)(struct drbd_conf
*mdev
, int rv
);
849 enum write_ordering_e
{
857 unsigned int head_index
;
861 /* flag bits per tconn */
863 NET_CONGESTED
, /* The data socket is congested */
864 DISCARD_CONCURRENT
, /* Set on one node, cleared on the peer! */
865 SEND_PING
, /* whether asender should send a ping asap */
866 SIGNAL_ASENDER
, /* whether asender wants to be interrupted */
867 GOT_PING_ACK
, /* set when we receive a ping_ack packet, ping_wait gets woken */
870 CONFIG_PENDING
, /* serialization of (re)configuration requests.
871 * if set, also prevents the device from dying */
872 OBJECT_DYING
, /* device became unconfigured,
873 * but worker thread is still handling the cleanup.
874 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
875 * while this is set. */
876 CONN_DRY_RUN
, /* Expect disconnect after resync handshake. */
879 struct drbd_tconn
{ /* is a resource from the config file */
880 char *name
; /* Resource name */
881 struct list_head all_tconn
; /* linked on global drbd_tconns */
882 struct idr volumes
; /* <tconn, vnr> to mdev mapping */
883 enum drbd_conns cstate
; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
884 struct mutex cstate_mutex
; /* Protects graceful disconnects */
887 struct net_conf
*net_conf
; /* protected by get_net_conf() and put_net_conf() */
888 atomic_t net_cnt
; /* Users of net_conf */
889 wait_queue_head_t net_cnt_wait
;
890 wait_queue_head_t ping_wait
; /* Woken upon reception of a ping, and a state change */
891 struct res_opts res_opts
;
893 struct drbd_socket data
; /* data/barrier/cstate/parameter packets */
894 struct drbd_socket meta
; /* ping/ack (metadata) packets */
895 int agreed_pro_version
; /* actually used protocol version */
896 unsigned long last_received
; /* in jiffies, either socket */
897 unsigned int ko_count
;
900 struct drbd_tl_epoch
*unused_spare_tle
; /* for pre-allocation */
901 struct drbd_tl_epoch
*newest_tle
;
902 struct drbd_tl_epoch
*oldest_tle
;
903 struct list_head out_of_sequence_requests
;
905 struct crypto_hash
*cram_hmac_tfm
;
906 struct crypto_hash
*integrity_w_tfm
; /* to be used by the worker thread */
907 struct crypto_hash
*integrity_r_tfm
; /* to be used by the receiver thread */
908 struct crypto_hash
*csums_tfm
;
909 struct crypto_hash
*verify_tfm
;
914 struct drbd_thread receiver
;
915 struct drbd_thread worker
;
916 struct drbd_thread asender
;
917 cpumask_var_t cpu_mask
;
921 struct drbd_tconn
*tconn
;
922 int vnr
; /* volume number within the connection */
924 /* things that are stored as / read from meta data on disk */
927 /* configured by drbdsetup */
928 struct drbd_backing_dev
*ldev
__protected_by(local
);
930 sector_t p_size
; /* partner's disk size */
931 struct request_queue
*rq_queue
;
932 struct block_device
*this_bdev
;
933 struct gendisk
*vdisk
;
935 struct drbd_work resync_work
,
940 struct timer_list resync_timer
;
941 struct timer_list md_sync_timer
;
942 struct timer_list start_resync_timer
;
943 struct timer_list request_timer
;
944 #ifdef DRBD_DEBUG_MD_SYNC
948 } last_md_mark_dirty
;
951 /* Used after attach while negotiating new disk state. */
952 union drbd_state new_state_tmp
;
954 union drbd_state state
;
955 wait_queue_head_t misc_wait
;
956 wait_queue_head_t state_wait
; /* upon each state change. */
957 unsigned int send_cnt
;
958 unsigned int recv_cnt
;
959 unsigned int read_cnt
;
960 unsigned int writ_cnt
;
961 unsigned int al_writ_cnt
;
962 unsigned int bm_writ_cnt
;
963 atomic_t ap_bio_cnt
; /* Requests we need to complete */
964 atomic_t ap_pending_cnt
; /* AP data packets on the wire, ack expected */
965 atomic_t rs_pending_cnt
; /* RS request/data packets on the wire */
966 atomic_t unacked_cnt
; /* Need to send replys for */
967 atomic_t local_cnt
; /* Waiting for local completion */
969 /* Interval tree of pending local requests */
970 struct rb_root read_requests
;
971 struct rb_root write_requests
;
973 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
974 unsigned long rs_total
;
975 /* number of resync blocks that failed in this run */
976 unsigned long rs_failed
;
977 /* Syncer's start time [unit jiffies] */
978 unsigned long rs_start
;
979 /* cumulated time in PausedSyncX state [unit jiffies] */
980 unsigned long rs_paused
;
981 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
982 unsigned long rs_same_csum
;
983 #define DRBD_SYNC_MARKS 8
984 #define DRBD_SYNC_MARK_STEP (3*HZ)
985 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
986 unsigned long rs_mark_left
[DRBD_SYNC_MARKS
];
987 /* marks's time [unit jiffies] */
988 unsigned long rs_mark_time
[DRBD_SYNC_MARKS
];
989 /* current index into rs_mark_{left,time} */
992 /* where does the admin want us to start? (sector) */
993 sector_t ov_start_sector
;
994 /* where are we now? (sector) */
995 sector_t ov_position
;
996 /* Start sector of out of sync range (to merge printk reporting). */
997 sector_t ov_last_oos_start
;
998 /* size of out-of-sync range in sectors. */
999 sector_t ov_last_oos_size
;
1000 unsigned long ov_left
; /* in bits */
1002 struct drbd_bitmap
*bitmap
;
1003 unsigned long bm_resync_fo
; /* bit offset for drbd_bm_find_next */
1005 /* Used to track operations of resync... */
1006 struct lru_cache
*resync
;
1007 /* Number of locked elements in resync LRU */
1008 unsigned int resync_locked
;
1009 /* resync extent number waiting for application requests */
1010 unsigned int resync_wenr
;
1014 struct drbd_epoch
*current_epoch
;
1015 spinlock_t epoch_lock
;
1016 unsigned int epochs
;
1017 enum write_ordering_e write_ordering
;
1018 struct list_head active_ee
; /* IO in progress (P_DATA gets written to disk) */
1019 struct list_head sync_ee
; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
1020 struct list_head done_ee
; /* need to send P_WRITE_ACK */
1021 struct list_head read_ee
; /* [RS]P_DATA_REQUEST being read */
1022 struct list_head net_ee
; /* zero-copy network send in progress */
1024 int next_barrier_nr
;
1025 struct list_head resync_reads
;
1026 atomic_t pp_in_use
; /* allocated from page pool */
1027 atomic_t pp_in_use_by_net
; /* sendpage()d, still referenced by tcp */
1028 wait_queue_head_t ee_wait
;
1029 struct page
*md_io_page
; /* one page buffer for md_io */
1030 struct mutex md_io_mutex
; /* protects the md_io_buffer */
1032 wait_queue_head_t al_wait
;
1033 struct lru_cache
*act_log
; /* activity log */
1034 unsigned int al_tr_number
;
1036 int al_tr_pos
; /* position of the next transaction in the journal */
1037 wait_queue_head_t seq_wait
;
1038 atomic_t packet_seq
;
1039 unsigned int peer_seq
;
1040 spinlock_t peer_seq_lock
;
1042 unsigned long comm_bm_set
; /* communicated number of set bits. */
1043 struct bm_io_work bm_io_work
;
1044 u64 ed_uuid
; /* UUID of the exposed data */
1045 struct mutex own_state_mutex
;
1046 struct mutex
*state_mutex
; /* either own_state_mutex or mdev->tconn->cstate_mutex */
1047 char congestion_reason
; /* Why we where congested... */
1048 atomic_t rs_sect_in
; /* for incoming resync data rate, SyncTarget */
1049 atomic_t rs_sect_ev
; /* for submitted resync data rate, both */
1050 int rs_last_sect_ev
; /* counter to compare with */
1051 int rs_last_events
; /* counter of read or write "events" (unit sectors)
1052 * on the lower level device when we last looked. */
1053 int c_sync_rate
; /* current resync rate after syncer throttle magic */
1054 struct fifo_buffer rs_plan_s
; /* correction values of resync planer */
1055 int rs_in_flight
; /* resync sectors in flight (to proxy, in proxy and from proxy) */
1056 int rs_planed
; /* resync sectors already planned */
1057 atomic_t ap_in_flight
; /* App sectors in flight (waiting for ack) */
1058 int peer_max_bio_size
;
1059 int local_max_bio_size
;
1062 static inline struct drbd_conf
*minor_to_mdev(unsigned int minor
)
1064 return (struct drbd_conf
*)idr_find(&minors
, minor
);
1067 static inline unsigned int mdev_to_minor(struct drbd_conf
*mdev
)
1072 static inline struct drbd_conf
*vnr_to_mdev(struct drbd_tconn
*tconn
, int vnr
)
1074 return (struct drbd_conf
*)idr_find(&tconn
->volumes
, vnr
);
1077 static inline int drbd_get_data_sock(struct drbd_tconn
*tconn
)
1079 mutex_lock(&tconn
->data
.mutex
);
1080 if (!tconn
->data
.socket
) {
1082 mutex_unlock(&tconn
->data
.mutex
);
1088 static inline void drbd_put_data_sock(struct drbd_tconn
*tconn
)
1090 mutex_unlock(&tconn
->data
.mutex
);
1094 * function declarations
1095 *************************/
1101 DDSF_NO_RESYNC
= 2, /* Do not run a resync for the new space */
1104 extern void drbd_init_set_defaults(struct drbd_conf
*mdev
);
1105 extern int drbd_thread_start(struct drbd_thread
*thi
);
1106 extern void _drbd_thread_stop(struct drbd_thread
*thi
, int restart
, int wait
);
1107 extern char *drbd_task_to_thread_name(struct drbd_tconn
*tconn
, struct task_struct
*task
);
1109 extern void drbd_thread_current_set_cpu(struct drbd_thread
*thi
);
1110 extern void drbd_calc_cpu_mask(struct drbd_tconn
*tconn
);
1112 #define drbd_thread_current_set_cpu(A) ({})
1113 #define drbd_calc_cpu_mask(A) ({})
1115 extern void drbd_free_resources(struct drbd_conf
*mdev
);
1116 extern void tl_release(struct drbd_tconn
*, unsigned int barrier_nr
,
1117 unsigned int set_size
);
1118 extern void tl_clear(struct drbd_tconn
*);
1119 extern void _tl_add_barrier(struct drbd_tconn
*, struct drbd_tl_epoch
*);
1120 extern void drbd_free_sock(struct drbd_tconn
*tconn
);
1121 extern int drbd_send(struct drbd_tconn
*tconn
, struct socket
*sock
,
1122 void *buf
, size_t size
, unsigned msg_flags
);
1123 extern int drbd_send_all(struct drbd_tconn
*, struct socket
*, void *, size_t,
1126 extern int drbd_send_protocol(struct drbd_tconn
*tconn
);
1127 extern int drbd_send_uuids(struct drbd_conf
*mdev
);
1128 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf
*mdev
);
1129 extern void drbd_gen_and_send_sync_uuid(struct drbd_conf
*mdev
);
1130 extern int drbd_send_sizes(struct drbd_conf
*mdev
, int trigger_reply
, enum dds_flags flags
);
1131 extern int _conn_send_state_req(struct drbd_tconn
*, int vnr
, enum drbd_packet cmd
,
1132 union drbd_state
, union drbd_state
);
1133 extern int _drbd_send_state(struct drbd_conf
*mdev
);
1134 extern int drbd_send_state(struct drbd_conf
*mdev
);
1135 extern int _conn_send_cmd(struct drbd_tconn
*tconn
, int vnr
, struct socket
*sock
,
1136 enum drbd_packet cmd
, struct p_header
*h
, size_t size
,
1137 unsigned msg_flags
);
1138 extern int conn_send_cmd(struct drbd_tconn
*tconn
, int vnr
, struct drbd_socket
*sock
,
1139 enum drbd_packet cmd
, struct p_header
*h
, size_t size
);
1140 extern int conn_send_cmd2(struct drbd_tconn
*tconn
, enum drbd_packet cmd
,
1141 char *data
, size_t size
);
1142 extern int drbd_send_sync_param(struct drbd_conf
*mdev
);
1143 extern void drbd_send_b_ack(struct drbd_conf
*mdev
, u32 barrier_nr
,
1145 extern int drbd_send_ack(struct drbd_conf
*, enum drbd_packet
,
1146 struct drbd_peer_request
*);
1147 extern void drbd_send_ack_rp(struct drbd_conf
*mdev
, enum drbd_packet cmd
,
1148 struct p_block_req
*rp
);
1149 extern void drbd_send_ack_dp(struct drbd_conf
*mdev
, enum drbd_packet cmd
,
1150 struct p_data
*dp
, int data_size
);
1151 extern int drbd_send_ack_ex(struct drbd_conf
*mdev
, enum drbd_packet cmd
,
1152 sector_t sector
, int blksize
, u64 block_id
);
1153 extern int drbd_send_out_of_sync(struct drbd_conf
*, struct drbd_request
*);
1154 extern int drbd_send_block(struct drbd_conf
*, enum drbd_packet
,
1155 struct drbd_peer_request
*);
1156 extern int drbd_send_dblock(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1157 extern int drbd_send_drequest(struct drbd_conf
*mdev
, int cmd
,
1158 sector_t sector
, int size
, u64 block_id
);
1159 extern int drbd_send_drequest_csum(struct drbd_conf
*mdev
, sector_t sector
,
1160 int size
, void *digest
, int digest_size
,
1161 enum drbd_packet cmd
);
1162 extern int drbd_send_ov_request(struct drbd_conf
*mdev
,sector_t sector
,int size
);
1164 extern int drbd_send_bitmap(struct drbd_conf
*mdev
);
1165 extern void drbd_send_sr_reply(struct drbd_conf
*mdev
, enum drbd_state_rv retcode
);
1166 extern int conn_send_sr_reply(struct drbd_tconn
*tconn
, enum drbd_state_rv retcode
);
1167 extern void drbd_free_bc(struct drbd_backing_dev
*ldev
);
1168 extern void drbd_mdev_cleanup(struct drbd_conf
*mdev
);
1169 void drbd_print_uuids(struct drbd_conf
*mdev
, const char *text
);
1171 extern void drbd_md_sync(struct drbd_conf
*mdev
);
1172 extern int drbd_md_read(struct drbd_conf
*mdev
, struct drbd_backing_dev
*bdev
);
1173 extern void drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1174 extern void _drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1175 extern void drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1176 extern void _drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1177 extern void drbd_uuid_set_bm(struct drbd_conf
*mdev
, u64 val
) __must_hold(local
);
1178 extern void drbd_md_set_flag(struct drbd_conf
*mdev
, int flags
) __must_hold(local
);
1179 extern void drbd_md_clear_flag(struct drbd_conf
*mdev
, int flags
)__must_hold(local
);
1180 extern int drbd_md_test_flag(struct drbd_backing_dev
*, int);
1181 #ifndef DRBD_DEBUG_MD_SYNC
1182 extern void drbd_md_mark_dirty(struct drbd_conf
*mdev
);
1184 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1185 extern void drbd_md_mark_dirty_(struct drbd_conf
*mdev
,
1186 unsigned int line
, const char *func
);
1188 extern void drbd_queue_bitmap_io(struct drbd_conf
*mdev
,
1189 int (*io_fn
)(struct drbd_conf
*),
1190 void (*done
)(struct drbd_conf
*, int),
1191 char *why
, enum bm_flag flags
);
1192 extern int drbd_bitmap_io(struct drbd_conf
*mdev
,
1193 int (*io_fn
)(struct drbd_conf
*),
1194 char *why
, enum bm_flag flags
);
1195 extern int drbd_bmio_set_n_write(struct drbd_conf
*mdev
);
1196 extern int drbd_bmio_clear_n_write(struct drbd_conf
*mdev
);
1197 extern void drbd_go_diskless(struct drbd_conf
*mdev
);
1198 extern void drbd_ldev_destroy(struct drbd_conf
*mdev
);
1201 We reserve a 128MB Block (4k aligned)
1202 * either at the end of the backing device
1203 * or on a separate meta data device. */
1205 /* The following numbers are sectors */
1206 /* Allows up to about 3.8TB, so if you want more,
1207 * you need to use the "flexible" meta data format. */
1208 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1209 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1210 #define MD_AL_SECTORS 64 /* = 32 kB on disk activity log ring buffer */
1211 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_SECTORS)
1213 /* we do all meta data IO in 4k blocks */
1214 #define MD_BLOCK_SHIFT 12
1215 #define MD_BLOCK_SIZE (1<<MD_BLOCK_SHIFT)
1217 /* One activity log extent represents 4M of storage */
1218 #define AL_EXTENT_SHIFT 22
1219 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1221 /* We could make these currently hardcoded constants configurable
1222 * variables at create-md time (or even re-configurable at runtime?).
1223 * Which will require some more changes to the DRBD "super block"
1226 * updates per transaction:
1227 * This many changes to the active set can be logged with one transaction.
1228 * This number is arbitrary.
1229 * context per transaction:
1230 * This many context extent numbers are logged with each transaction.
1231 * This number is resulting from the transaction block size (4k), the layout
1232 * of the transaction header, and the number of updates per transaction.
1233 * See drbd_actlog.c:struct al_transaction_on_disk
1235 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1236 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1238 #if BITS_PER_LONG == 32
1240 #define cpu_to_lel(A) cpu_to_le32(A)
1241 #define lel_to_cpu(A) le32_to_cpu(A)
1242 #elif BITS_PER_LONG == 64
1244 #define cpu_to_lel(A) cpu_to_le64(A)
1245 #define lel_to_cpu(A) le64_to_cpu(A)
1247 #error "LN2 of BITS_PER_LONG unknown!"
1251 /* 16MB sized 'bitmap extent' to track syncer usage */
1253 int rs_left
; /* number of bits set (out of sync) in this extent. */
1254 int rs_failed
; /* number of failed resync requests in this extent. */
1255 unsigned long flags
;
1256 struct lc_element lce
;
1259 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1260 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1261 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1265 * We need to store one bit for a block.
1266 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1267 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1268 * Bit 1 ==> local node thinks this block needs to be synced.
1271 #define SLEEP_TIME (HZ/10)
1273 /* We do bitmap IO in units of 4k blocks.
1274 * We also still have a hardcoded 4k per bit relation. */
1275 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1276 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1277 /* mostly arbitrarily set the represented size of one bitmap extent,
1278 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1279 * at 4k per bit resolution) */
1280 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
1281 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1283 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1284 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1287 /* thus many _storage_ sectors are described by one bit */
1288 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1289 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1290 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1292 /* bit to represented kilo byte conversion */
1293 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1295 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1296 * _storage_ sector is located in */
1297 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1299 /* how much _storage_ sectors we have per bitmap sector */
1300 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1301 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1303 /* in one sector of the bitmap, we have this many activity_log extents. */
1304 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1305 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1307 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1308 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1310 /* the extent in "PER_EXTENT" below is an activity log extent
1311 * we need that many (long words/bytes) to store the bitmap
1312 * of one AL_EXTENT_SIZE chunk of storage.
1313 * we can store the bitmap for that many AL_EXTENTS within
1314 * one sector of the _on_disk_ bitmap:
1315 * bit 0 bit 37 bit 38 bit (512*8)-1
1316 * ...|........|........|.. // ..|........|
1317 * sect. 0 `296 `304 ^(512*8*8)-1
1319 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1320 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1321 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1324 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1325 #define DRBD_MAX_SECTORS_BM \
1326 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1327 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1328 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1329 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1330 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1331 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1332 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1334 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1335 /* 16 TB in units of sectors */
1336 #if BITS_PER_LONG == 32
1337 /* adjust by one page worth of bitmap,
1338 * so we won't wrap around in drbd_bm_find_next_bit.
1339 * you should use 64bit OS for that much storage, anyways. */
1340 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1342 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1343 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1344 /* corresponds to (1UL << 38) bits right now. */
1349 #define DRBD_MAX_BIO_SIZE (1U<<(9+HT_SHIFT))
1350 #define DRBD_MAX_BIO_SIZE_SAFE (1 << 12) /* Works always = 4k */
1352 #define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */
1354 extern int drbd_bm_init(struct drbd_conf
*mdev
);
1355 extern int drbd_bm_resize(struct drbd_conf
*mdev
, sector_t sectors
, int set_new_bits
);
1356 extern void drbd_bm_cleanup(struct drbd_conf
*mdev
);
1357 extern void drbd_bm_set_all(struct drbd_conf
*mdev
);
1358 extern void drbd_bm_clear_all(struct drbd_conf
*mdev
);
1359 /* set/clear/test only a few bits at a time */
1360 extern int drbd_bm_set_bits(
1361 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1362 extern int drbd_bm_clear_bits(
1363 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1364 extern int drbd_bm_count_bits(
1365 struct drbd_conf
*mdev
, const unsigned long s
, const unsigned long e
);
1366 /* bm_set_bits variant for use while holding drbd_bm_lock,
1367 * may process the whole bitmap in one go */
1368 extern void _drbd_bm_set_bits(struct drbd_conf
*mdev
,
1369 const unsigned long s
, const unsigned long e
);
1370 extern int drbd_bm_test_bit(struct drbd_conf
*mdev
, unsigned long bitnr
);
1371 extern int drbd_bm_e_weight(struct drbd_conf
*mdev
, unsigned long enr
);
1372 extern int drbd_bm_write_page(struct drbd_conf
*mdev
, unsigned int idx
) __must_hold(local
);
1373 extern int drbd_bm_read(struct drbd_conf
*mdev
) __must_hold(local
);
1374 extern void drbd_bm_mark_for_writeout(struct drbd_conf
*mdev
, int page_nr
);
1375 extern int drbd_bm_write(struct drbd_conf
*mdev
) __must_hold(local
);
1376 extern int drbd_bm_write_hinted(struct drbd_conf
*mdev
) __must_hold(local
);
1377 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf
*mdev
,
1378 unsigned long al_enr
);
1379 extern size_t drbd_bm_words(struct drbd_conf
*mdev
);
1380 extern unsigned long drbd_bm_bits(struct drbd_conf
*mdev
);
1381 extern sector_t
drbd_bm_capacity(struct drbd_conf
*mdev
);
1383 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1384 extern unsigned long drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1385 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1386 extern unsigned long _drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1387 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1388 extern unsigned long _drbd_bm_total_weight(struct drbd_conf
*mdev
);
1389 extern unsigned long drbd_bm_total_weight(struct drbd_conf
*mdev
);
1390 extern int drbd_bm_rs_done(struct drbd_conf
*mdev
);
1391 /* for receive_bitmap */
1392 extern void drbd_bm_merge_lel(struct drbd_conf
*mdev
, size_t offset
,
1393 size_t number
, unsigned long *buffer
);
1394 /* for _drbd_send_bitmap */
1395 extern void drbd_bm_get_lel(struct drbd_conf
*mdev
, size_t offset
,
1396 size_t number
, unsigned long *buffer
);
1398 extern void drbd_bm_lock(struct drbd_conf
*mdev
, char *why
, enum bm_flag flags
);
1399 extern void drbd_bm_unlock(struct drbd_conf
*mdev
);
1402 extern struct kmem_cache
*drbd_request_cache
;
1403 extern struct kmem_cache
*drbd_ee_cache
; /* peer requests */
1404 extern struct kmem_cache
*drbd_bm_ext_cache
; /* bitmap extents */
1405 extern struct kmem_cache
*drbd_al_ext_cache
; /* activity log extents */
1406 extern mempool_t
*drbd_request_mempool
;
1407 extern mempool_t
*drbd_ee_mempool
;
1409 /* drbd's page pool, used to buffer data received from the peer,
1410 * or data requested by the peer.
1412 * This does not have an emergency reserve.
1414 * When allocating from this pool, it first takes pages from the pool.
1415 * Only if the pool is depleted will try to allocate from the system.
1417 * The assumption is that pages taken from this pool will be processed,
1418 * and given back, "quickly", and then can be recycled, so we can avoid
1419 * frequent calls to alloc_page(), and still will be able to make progress even
1420 * under memory pressure.
1422 extern struct page
*drbd_pp_pool
;
1423 extern spinlock_t drbd_pp_lock
;
1424 extern int drbd_pp_vacant
;
1425 extern wait_queue_head_t drbd_pp_wait
;
1427 /* We also need a standard (emergency-reserve backed) page pool
1428 * for meta data IO (activity log, bitmap).
1429 * We can keep it global, as long as it is used as "N pages at a time".
1430 * 128 should be plenty, currently we probably can get away with as few as 1.
1432 #define DRBD_MIN_POOL_PAGES 128
1433 extern mempool_t
*drbd_md_io_page_pool
;
1435 /* We also need to make sure we get a bio
1436 * when we need it for housekeeping purposes */
1437 extern struct bio_set
*drbd_md_io_bio_set
;
1438 /* to allocate from that set */
1439 extern struct bio
*bio_alloc_drbd(gfp_t gfp_mask
);
1441 extern rwlock_t global_state_lock
;
1443 extern int conn_lowest_minor(struct drbd_tconn
*tconn
);
1444 enum drbd_ret_code
conn_new_minor(struct drbd_tconn
*tconn
, unsigned int minor
, int vnr
);
1445 extern void drbd_free_mdev(struct drbd_conf
*mdev
);
1446 extern void drbd_delete_device(unsigned int minor
);
1448 struct drbd_tconn
*drbd_new_tconn(const char *name
);
1449 extern void drbd_free_tconn(struct drbd_tconn
*tconn
);
1450 struct drbd_tconn
*conn_by_name(const char *name
);
1452 extern int proc_details
;
1455 extern int __drbd_make_request(struct drbd_conf
*, struct bio
*, unsigned long);
1456 extern int drbd_make_request(struct request_queue
*q
, struct bio
*bio
);
1457 extern int drbd_read_remote(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1458 extern int drbd_merge_bvec(struct request_queue
*q
, struct bvec_merge_data
*bvm
, struct bio_vec
*bvec
);
1459 extern int is_valid_ar_handle(struct drbd_request
*, sector_t
);
1463 extern int drbd_msg_put_info(const char *info
);
1464 extern void drbd_suspend_io(struct drbd_conf
*mdev
);
1465 extern void drbd_resume_io(struct drbd_conf
*mdev
);
1466 extern char *ppsize(char *buf
, unsigned long long size
);
1467 extern sector_t
drbd_new_dev_size(struct drbd_conf
*, struct drbd_backing_dev
*, int);
1468 enum determine_dev_size
{ dev_size_error
= -1, unchanged
= 0, shrunk
= 1, grew
= 2 };
1469 extern enum determine_dev_size
drbd_determine_dev_size(struct drbd_conf
*, enum dds_flags
) __must_hold(local
);
1470 extern void resync_after_online_grow(struct drbd_conf
*);
1471 extern void drbd_reconsider_max_bio_size(struct drbd_conf
*mdev
);
1472 extern enum drbd_state_rv
drbd_set_role(struct drbd_conf
*mdev
,
1473 enum drbd_role new_role
,
1475 extern enum drbd_disk_state
drbd_try_outdate_peer(struct drbd_conf
*mdev
);
1476 extern void drbd_try_outdate_peer_async(struct drbd_conf
*mdev
);
1477 extern int drbd_khelper(struct drbd_conf
*mdev
, char *cmd
);
1480 extern int drbd_worker(struct drbd_thread
*thi
);
1481 extern int drbd_alter_sa(struct drbd_conf
*mdev
, int na
);
1482 extern void drbd_start_resync(struct drbd_conf
*mdev
, enum drbd_conns side
);
1483 extern void resume_next_sg(struct drbd_conf
*mdev
);
1484 extern void suspend_other_sg(struct drbd_conf
*mdev
);
1485 extern int drbd_resync_finished(struct drbd_conf
*mdev
);
1486 /* maybe rather drbd_main.c ? */
1487 extern int drbd_md_sync_page_io(struct drbd_conf
*mdev
,
1488 struct drbd_backing_dev
*bdev
, sector_t sector
, int rw
);
1489 extern void drbd_ov_out_of_sync_found(struct drbd_conf
*, sector_t
, int);
1490 extern void drbd_rs_controller_reset(struct drbd_conf
*mdev
);
1492 static inline void ov_out_of_sync_print(struct drbd_conf
*mdev
)
1494 if (mdev
->ov_last_oos_size
) {
1495 dev_err(DEV
, "Out of sync: start=%llu, size=%lu (sectors)\n",
1496 (unsigned long long)mdev
->ov_last_oos_start
,
1497 (unsigned long)mdev
->ov_last_oos_size
);
1499 mdev
->ov_last_oos_size
=0;
1503 extern void drbd_csum_bio(struct drbd_conf
*, struct crypto_hash
*, struct bio
*, void *);
1504 extern void drbd_csum_ee(struct drbd_conf
*, struct crypto_hash
*,
1505 struct drbd_peer_request
*, void *);
1506 /* worker callbacks */
1507 extern int w_read_retry_remote(struct drbd_work
*, int);
1508 extern int w_e_end_data_req(struct drbd_work
*, int);
1509 extern int w_e_end_rsdata_req(struct drbd_work
*, int);
1510 extern int w_e_end_csum_rs_req(struct drbd_work
*, int);
1511 extern int w_e_end_ov_reply(struct drbd_work
*, int);
1512 extern int w_e_end_ov_req(struct drbd_work
*, int);
1513 extern int w_ov_finished(struct drbd_work
*, int);
1514 extern int w_resync_timer(struct drbd_work
*, int);
1515 extern int w_send_write_hint(struct drbd_work
*, int);
1516 extern int w_make_resync_request(struct drbd_work
*, int);
1517 extern int w_send_dblock(struct drbd_work
*, int);
1518 extern int w_send_barrier(struct drbd_work
*, int);
1519 extern int w_send_read_req(struct drbd_work
*, int);
1520 extern int w_prev_work_done(struct drbd_work
*, int);
1521 extern int w_e_reissue(struct drbd_work
*, int);
1522 extern int w_restart_disk_io(struct drbd_work
*, int);
1523 extern int w_send_out_of_sync(struct drbd_work
*, int);
1524 extern int w_start_resync(struct drbd_work
*, int);
1526 extern void resync_timer_fn(unsigned long data
);
1527 extern void start_resync_timer_fn(unsigned long data
);
1529 /* drbd_receiver.c */
1530 extern int drbd_rs_should_slow_down(struct drbd_conf
*mdev
, sector_t sector
);
1531 extern int drbd_submit_peer_request(struct drbd_conf
*,
1532 struct drbd_peer_request
*, const unsigned,
1534 extern int drbd_release_ee(struct drbd_conf
*mdev
, struct list_head
*list
);
1535 extern struct drbd_peer_request
*drbd_alloc_ee(struct drbd_conf
*,
1536 u64
, sector_t
, unsigned int,
1537 gfp_t
) __must_hold(local
);
1538 extern void drbd_free_some_ee(struct drbd_conf
*, struct drbd_peer_request
*,
1540 #define drbd_free_ee(m,e) drbd_free_some_ee(m, e, 0)
1541 #define drbd_free_net_ee(m,e) drbd_free_some_ee(m, e, 1)
1542 extern void drbd_wait_ee_list_empty(struct drbd_conf
*mdev
,
1543 struct list_head
*head
);
1544 extern void _drbd_wait_ee_list_empty(struct drbd_conf
*mdev
,
1545 struct list_head
*head
);
1546 extern void drbd_set_recv_tcq(struct drbd_conf
*mdev
, int tcq_enabled
);
1547 extern void _drbd_clear_done_ee(struct drbd_conf
*mdev
, struct list_head
*to_be_freed
);
1548 extern void conn_flush_workqueue(struct drbd_tconn
*tconn
);
1549 extern int drbd_connected(int vnr
, void *p
, void *data
);
1550 static inline void drbd_flush_workqueue(struct drbd_conf
*mdev
)
1552 conn_flush_workqueue(mdev
->tconn
);
1555 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1556 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1557 static inline int drbd_setsockopt(struct socket
*sock
, int level
, int optname
,
1558 char __user
*optval
, int optlen
)
1561 if (level
== SOL_SOCKET
)
1562 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
1564 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1569 static inline void drbd_tcp_cork(struct socket
*sock
)
1572 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1573 (char __user
*)&val
, sizeof(val
));
1576 static inline void drbd_tcp_uncork(struct socket
*sock
)
1579 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1580 (char __user
*)&val
, sizeof(val
));
1583 static inline void drbd_tcp_nodelay(struct socket
*sock
)
1586 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_NODELAY
,
1587 (char __user
*)&val
, sizeof(val
));
1590 static inline void drbd_tcp_quickack(struct socket
*sock
)
1593 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_QUICKACK
,
1594 (char __user
*)&val
, sizeof(val
));
1597 void drbd_bump_write_ordering(struct drbd_conf
*mdev
, enum write_ordering_e wo
);
1600 extern struct proc_dir_entry
*drbd_proc
;
1601 extern const struct file_operations drbd_proc_fops
;
1602 extern const char *drbd_conn_str(enum drbd_conns s
);
1603 extern const char *drbd_role_str(enum drbd_role s
);
1606 extern void drbd_al_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1607 extern void drbd_al_complete_io(struct drbd_conf
*mdev
, sector_t sector
);
1608 extern void drbd_rs_complete_io(struct drbd_conf
*mdev
, sector_t sector
);
1609 extern int drbd_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1610 extern int drbd_try_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1611 extern void drbd_rs_cancel_all(struct drbd_conf
*mdev
);
1612 extern int drbd_rs_del_all(struct drbd_conf
*mdev
);
1613 extern void drbd_rs_failed_io(struct drbd_conf
*mdev
,
1614 sector_t sector
, int size
);
1615 extern int drbd_al_read_log(struct drbd_conf
*mdev
, struct drbd_backing_dev
*);
1616 extern void drbd_advance_rs_marks(struct drbd_conf
*mdev
, unsigned long still_to_go
);
1617 extern void __drbd_set_in_sync(struct drbd_conf
*mdev
, sector_t sector
,
1618 int size
, const char *file
, const unsigned int line
);
1619 #define drbd_set_in_sync(mdev, sector, size) \
1620 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1621 extern int __drbd_set_out_of_sync(struct drbd_conf
*mdev
, sector_t sector
,
1622 int size
, const char *file
, const unsigned int line
);
1623 #define drbd_set_out_of_sync(mdev, sector, size) \
1624 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1625 extern void drbd_al_apply_to_bm(struct drbd_conf
*mdev
);
1626 extern void drbd_al_shrink(struct drbd_conf
*mdev
);
1629 /* state info broadcast */
1631 enum drbd_state_info_bcast_reason sib_reason
;
1635 unsigned helper_exit_code
;
1638 union drbd_state os
;
1639 union drbd_state ns
;
1643 void drbd_bcast_event(struct drbd_conf
*mdev
, const struct sib_info
*sib
);
1646 * inline helper functions
1647 *************************/
1649 /* see also page_chain_add and friends in drbd_receiver.c */
1650 static inline struct page
*page_chain_next(struct page
*page
)
1652 return (struct page
*)page_private(page
);
1654 #define page_chain_for_each(page) \
1655 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1656 page = page_chain_next(page))
1657 #define page_chain_for_each_safe(page, n) \
1658 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1660 static inline int drbd_bio_has_active_page(struct bio
*bio
)
1662 struct bio_vec
*bvec
;
1665 __bio_for_each_segment(bvec
, bio
, i
, 0) {
1666 if (page_count(bvec
->bv_page
) > 1)
1673 static inline int drbd_ee_has_active_page(struct drbd_peer_request
*peer_req
)
1675 struct page
*page
= peer_req
->pages
;
1676 page_chain_for_each(page
) {
1677 if (page_count(page
) > 1)
1683 static inline enum drbd_state_rv
1684 _drbd_set_state(struct drbd_conf
*mdev
, union drbd_state ns
,
1685 enum chg_state_flags flags
, struct completion
*done
)
1687 enum drbd_state_rv rv
;
1689 read_lock(&global_state_lock
);
1690 rv
= __drbd_set_state(mdev
, ns
, flags
, done
);
1691 read_unlock(&global_state_lock
);
1696 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1697 static inline void __drbd_chk_io_error_(struct drbd_conf
*mdev
, int forcedetach
, const char *where
)
1699 switch (mdev
->ldev
->dc
.on_io_error
) {
1702 if (__ratelimit(&drbd_ratelimit_state
))
1703 dev_err(DEV
, "Local IO failed in %s.\n", where
);
1704 if (mdev
->state
.disk
> D_INCONSISTENT
)
1705 _drbd_set_state(_NS(mdev
, disk
, D_INCONSISTENT
), CS_HARD
, NULL
);
1708 /* NOTE fall through to detach case if forcedetach set */
1710 case EP_CALL_HELPER
:
1711 set_bit(WAS_IO_ERROR
, &mdev
->flags
);
1712 if (mdev
->state
.disk
> D_FAILED
) {
1713 _drbd_set_state(_NS(mdev
, disk
, D_FAILED
), CS_HARD
, NULL
);
1715 "Local IO failed in %s. Detaching...\n", where
);
1722 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1723 * @mdev: DRBD device.
1724 * @error: Error code passed to the IO completion callback
1725 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1727 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1729 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1730 static inline void drbd_chk_io_error_(struct drbd_conf
*mdev
,
1731 int error
, int forcedetach
, const char *where
)
1734 unsigned long flags
;
1735 spin_lock_irqsave(&mdev
->tconn
->req_lock
, flags
);
1736 __drbd_chk_io_error_(mdev
, forcedetach
, where
);
1737 spin_unlock_irqrestore(&mdev
->tconn
->req_lock
, flags
);
1743 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1744 * @bdev: Meta data block device.
1746 * BTW, for internal meta data, this happens to be the maximum capacity
1747 * we could agree upon with our peer node.
1749 static inline sector_t
drbd_md_first_sector(struct drbd_backing_dev
*bdev
)
1751 switch (bdev
->dc
.meta_dev_idx
) {
1752 case DRBD_MD_INDEX_INTERNAL
:
1753 case DRBD_MD_INDEX_FLEX_INT
:
1754 return bdev
->md
.md_offset
+ bdev
->md
.bm_offset
;
1755 case DRBD_MD_INDEX_FLEX_EXT
:
1757 return bdev
->md
.md_offset
;
1762 * drbd_md_last_sector() - Return the last sector number of the meta data area
1763 * @bdev: Meta data block device.
1765 static inline sector_t
drbd_md_last_sector(struct drbd_backing_dev
*bdev
)
1767 switch (bdev
->dc
.meta_dev_idx
) {
1768 case DRBD_MD_INDEX_INTERNAL
:
1769 case DRBD_MD_INDEX_FLEX_INT
:
1770 return bdev
->md
.md_offset
+ MD_AL_OFFSET
- 1;
1771 case DRBD_MD_INDEX_FLEX_EXT
:
1773 return bdev
->md
.md_offset
+ bdev
->md
.md_size_sect
;
1777 /* Returns the number of 512 byte sectors of the device */
1778 static inline sector_t
drbd_get_capacity(struct block_device
*bdev
)
1780 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1781 return bdev
? i_size_read(bdev
->bd_inode
) >> 9 : 0;
1785 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1786 * @bdev: Meta data block device.
1788 * returns the capacity we announce to out peer. we clip ourselves at the
1789 * various MAX_SECTORS, because if we don't, current implementation will
1790 * oops sooner or later
1792 static inline sector_t
drbd_get_max_capacity(struct drbd_backing_dev
*bdev
)
1795 switch (bdev
->dc
.meta_dev_idx
) {
1796 case DRBD_MD_INDEX_INTERNAL
:
1797 case DRBD_MD_INDEX_FLEX_INT
:
1798 s
= drbd_get_capacity(bdev
->backing_bdev
)
1799 ? min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1800 drbd_md_first_sector(bdev
))
1803 case DRBD_MD_INDEX_FLEX_EXT
:
1804 s
= min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1805 drbd_get_capacity(bdev
->backing_bdev
));
1806 /* clip at maximum size the meta device can support */
1807 s
= min_t(sector_t
, s
,
1808 BM_EXT_TO_SECT(bdev
->md
.md_size_sect
1809 - bdev
->md
.bm_offset
));
1812 s
= min_t(sector_t
, DRBD_MAX_SECTORS
,
1813 drbd_get_capacity(bdev
->backing_bdev
));
1819 * drbd_md_ss__() - Return the sector number of our meta data super block
1820 * @mdev: DRBD device.
1821 * @bdev: Meta data block device.
1823 static inline sector_t
drbd_md_ss__(struct drbd_conf
*mdev
,
1824 struct drbd_backing_dev
*bdev
)
1826 switch (bdev
->dc
.meta_dev_idx
) {
1827 default: /* external, some index */
1828 return MD_RESERVED_SECT
* bdev
->dc
.meta_dev_idx
;
1829 case DRBD_MD_INDEX_INTERNAL
:
1830 /* with drbd08, internal meta data is always "flexible" */
1831 case DRBD_MD_INDEX_FLEX_INT
:
1832 /* sizeof(struct md_on_disk_07) == 4k
1833 * position: last 4k aligned block of 4k size */
1834 if (!bdev
->backing_bdev
) {
1835 if (__ratelimit(&drbd_ratelimit_state
)) {
1836 dev_err(DEV
, "bdev->backing_bdev==NULL\n");
1841 return (drbd_get_capacity(bdev
->backing_bdev
) & ~7ULL)
1843 case DRBD_MD_INDEX_FLEX_EXT
:
1849 drbd_queue_work_front(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1851 unsigned long flags
;
1852 spin_lock_irqsave(&q
->q_lock
, flags
);
1853 list_add(&w
->list
, &q
->q
);
1854 up(&q
->s
); /* within the spinlock,
1855 see comment near end of drbd_worker() */
1856 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1860 drbd_queue_work(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1862 unsigned long flags
;
1863 spin_lock_irqsave(&q
->q_lock
, flags
);
1864 list_add_tail(&w
->list
, &q
->q
);
1865 up(&q
->s
); /* within the spinlock,
1866 see comment near end of drbd_worker() */
1867 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1870 static inline void wake_asender(struct drbd_tconn
*tconn
)
1872 if (test_bit(SIGNAL_ASENDER
, &tconn
->flags
))
1873 force_sig(DRBD_SIG
, tconn
->asender
.task
);
1876 static inline void request_ping(struct drbd_tconn
*tconn
)
1878 set_bit(SEND_PING
, &tconn
->flags
);
1879 wake_asender(tconn
);
1882 static inline int _drbd_send_cmd(struct drbd_conf
*mdev
, struct socket
*sock
,
1883 enum drbd_packet cmd
, struct p_header
*h
, size_t size
,
1886 return _conn_send_cmd(mdev
->tconn
, mdev
->vnr
, sock
, cmd
, h
, size
, msg_flags
);
1889 static inline int drbd_send_cmd(struct drbd_conf
*mdev
, struct drbd_socket
*sock
,
1890 enum drbd_packet cmd
, struct p_header
*h
, size_t size
)
1892 return conn_send_cmd(mdev
->tconn
, mdev
->vnr
, sock
, cmd
, h
, size
);
1895 static inline int drbd_send_short_cmd(struct drbd_conf
*mdev
,
1896 enum drbd_packet cmd
)
1899 return drbd_send_cmd(mdev
, &mdev
->tconn
->data
, cmd
, &h
, sizeof(h
));
1902 extern int drbd_send_ping(struct drbd_tconn
*tconn
);
1903 extern int drbd_send_ping_ack(struct drbd_tconn
*tconn
);
1905 static inline int drbd_send_state_req(struct drbd_conf
*mdev
,
1906 union drbd_state mask
, union drbd_state val
)
1908 return _conn_send_state_req(mdev
->tconn
, mdev
->vnr
, P_STATE_CHG_REQ
, mask
, val
);
1911 static inline int conn_send_state_req(struct drbd_tconn
*tconn
,
1912 union drbd_state mask
, union drbd_state val
)
1914 enum drbd_packet cmd
= tconn
->agreed_pro_version
< 100 ? P_STATE_CHG_REQ
: P_CONN_ST_CHG_REQ
;
1915 return _conn_send_state_req(tconn
, 0, cmd
, mask
, val
);
1918 static inline void drbd_thread_stop(struct drbd_thread
*thi
)
1920 _drbd_thread_stop(thi
, false, true);
1923 static inline void drbd_thread_stop_nowait(struct drbd_thread
*thi
)
1925 _drbd_thread_stop(thi
, false, false);
1928 static inline void drbd_thread_restart_nowait(struct drbd_thread
*thi
)
1930 _drbd_thread_stop(thi
, true, false);
1933 /* counts how many answer packets packets we expect from our peer,
1934 * for either explicit application requests,
1935 * or implicit barrier packets as necessary.
1938 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1939 * it is much easier and equally valid to count what we queue for the
1940 * worker, even before it actually was queued or send.
1941 * (drbd_make_request_common; recovery path on read io-error)
1943 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1944 * _req_mod(req, DATA_RECEIVED)
1945 * [from receive_DataReply]
1946 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1947 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1948 * for some reason it is NOT decreased in got_NegAck,
1949 * but in the resulting cleanup code from report_params.
1950 * we should try to remember the reason for that...
1951 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
1952 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
1953 * [from tl_clear_barrier]
1955 static inline void inc_ap_pending(struct drbd_conf
*mdev
)
1957 atomic_inc(&mdev
->ap_pending_cnt
);
1960 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \
1961 if (atomic_read(&mdev->which) < 0) \
1962 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
1964 atomic_read(&mdev->which))
1966 #define dec_ap_pending(mdev) _dec_ap_pending(mdev, __FUNCTION__, __LINE__)
1967 static inline void _dec_ap_pending(struct drbd_conf
*mdev
, const char *func
, int line
)
1969 if (atomic_dec_and_test(&mdev
->ap_pending_cnt
))
1970 wake_up(&mdev
->misc_wait
);
1971 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt
, func
, line
);
1974 /* counts how many resync-related answers we still expect from the peer
1976 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1977 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
1978 * (or P_NEG_ACK with ID_SYNCER)
1980 static inline void inc_rs_pending(struct drbd_conf
*mdev
)
1982 atomic_inc(&mdev
->rs_pending_cnt
);
1985 #define dec_rs_pending(mdev) _dec_rs_pending(mdev, __FUNCTION__, __LINE__)
1986 static inline void _dec_rs_pending(struct drbd_conf
*mdev
, const char *func
, int line
)
1988 atomic_dec(&mdev
->rs_pending_cnt
);
1989 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt
, func
, line
);
1992 /* counts how many answers we still need to send to the peer.
1994 * receive_Data unless protocol A;
1995 * we need to send a P_RECV_ACK (proto B)
1996 * or P_WRITE_ACK (proto C)
1997 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1998 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1999 * receive_Barrier_* we need to send a P_BARRIER_ACK
2001 static inline void inc_unacked(struct drbd_conf
*mdev
)
2003 atomic_inc(&mdev
->unacked_cnt
);
2006 #define dec_unacked(mdev) _dec_unacked(mdev, __FUNCTION__, __LINE__)
2007 static inline void _dec_unacked(struct drbd_conf
*mdev
, const char *func
, int line
)
2009 atomic_dec(&mdev
->unacked_cnt
);
2010 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt
, func
, line
);
2013 #define sub_unacked(mdev, n) _sub_unacked(mdev, n, __FUNCTION__, __LINE__)
2014 static inline void _sub_unacked(struct drbd_conf
*mdev
, int n
, const char *func
, int line
)
2016 atomic_sub(n
, &mdev
->unacked_cnt
);
2017 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt
, func
, line
);
2020 static inline void put_net_conf(struct drbd_tconn
*tconn
)
2022 if (atomic_dec_and_test(&tconn
->net_cnt
))
2023 wake_up(&tconn
->net_cnt_wait
);
2027 * get_net_conf() - Increase ref count on mdev->tconn->net_conf; Returns 0 if nothing there
2028 * @mdev: DRBD device.
2030 * You have to call put_net_conf() when finished working with mdev->tconn->net_conf.
2032 static inline int get_net_conf(struct drbd_tconn
*tconn
)
2036 atomic_inc(&tconn
->net_cnt
);
2037 have_net_conf
= tconn
->cstate
>= C_UNCONNECTED
;
2039 put_net_conf(tconn
);
2040 return have_net_conf
;
2044 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2047 * You have to call put_ldev() when finished working with mdev->ldev.
2049 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2050 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2052 static inline void put_ldev(struct drbd_conf
*mdev
)
2054 int i
= atomic_dec_return(&mdev
->local_cnt
);
2056 /* This may be called from some endio handler,
2057 * so we must not sleep here. */
2062 if (mdev
->state
.disk
== D_DISKLESS
)
2063 /* even internal references gone, safe to destroy */
2064 drbd_ldev_destroy(mdev
);
2065 if (mdev
->state
.disk
== D_FAILED
)
2066 /* all application IO references gone. */
2067 drbd_go_diskless(mdev
);
2068 wake_up(&mdev
->misc_wait
);
2073 static inline int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
)
2077 /* never get a reference while D_DISKLESS */
2078 if (mdev
->state
.disk
== D_DISKLESS
)
2081 atomic_inc(&mdev
->local_cnt
);
2082 io_allowed
= (mdev
->state
.disk
>= mins
);
2088 extern int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
);
2091 /* you must have an "get_ldev" reference */
2092 static inline void drbd_get_syncer_progress(struct drbd_conf
*mdev
,
2093 unsigned long *bits_left
, unsigned int *per_mil_done
)
2095 /* this is to break it at compile time when we change that, in case we
2096 * want to support more than (1<<32) bits on a 32bit arch. */
2097 typecheck(unsigned long, mdev
->rs_total
);
2099 /* note: both rs_total and rs_left are in bits, i.e. in
2100 * units of BM_BLOCK_SIZE.
2101 * for the percentage, we don't care. */
2103 if (mdev
->state
.conn
== C_VERIFY_S
|| mdev
->state
.conn
== C_VERIFY_T
)
2104 *bits_left
= mdev
->ov_left
;
2106 *bits_left
= drbd_bm_total_weight(mdev
) - mdev
->rs_failed
;
2107 /* >> 10 to prevent overflow,
2108 * +1 to prevent division by zero */
2109 if (*bits_left
> mdev
->rs_total
) {
2110 /* doh. maybe a logic bug somewhere.
2111 * may also be just a race condition
2112 * between this and a disconnect during sync.
2113 * for now, just prevent in-kernel buffer overflow.
2116 dev_warn(DEV
, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2117 drbd_conn_str(mdev
->state
.conn
),
2118 *bits_left
, mdev
->rs_total
, mdev
->rs_failed
);
2121 /* Make sure the division happens in long context.
2122 * We allow up to one petabyte storage right now,
2123 * at a granularity of 4k per bit that is 2**38 bits.
2124 * After shift right and multiplication by 1000,
2125 * this should still fit easily into a 32bit long,
2126 * so we don't need a 64bit division on 32bit arch.
2127 * Note: currently we don't support such large bitmaps on 32bit
2128 * arch anyways, but no harm done to be prepared for it here.
2130 unsigned int shift
= mdev
->rs_total
>= (1ULL << 32) ? 16 : 10;
2131 unsigned long left
= *bits_left
>> shift
;
2132 unsigned long total
= 1UL + (mdev
->rs_total
>> shift
);
2133 unsigned long tmp
= 1000UL - left
* 1000UL/total
;
2134 *per_mil_done
= tmp
;
2139 /* this throttles on-the-fly application requests
2140 * according to max_buffers settings;
2141 * maybe re-implement using semaphores? */
2142 static inline int drbd_get_max_buffers(struct drbd_conf
*mdev
)
2144 int mxb
= 1000000; /* arbitrary limit on open requests */
2145 if (get_net_conf(mdev
->tconn
)) {
2146 mxb
= mdev
->tconn
->net_conf
->max_buffers
;
2147 put_net_conf(mdev
->tconn
);
2152 static inline int drbd_state_is_stable(struct drbd_conf
*mdev
)
2154 union drbd_state s
= mdev
->state
;
2156 /* DO NOT add a default clause, we want the compiler to warn us
2157 * for any newly introduced state we may have forgotten to add here */
2159 switch ((enum drbd_conns
)s
.conn
) {
2160 /* new io only accepted when there is no connection, ... */
2162 case C_WF_CONNECTION
:
2163 /* ... or there is a well established connection. */
2169 case C_PAUSED_SYNC_S
:
2170 case C_PAUSED_SYNC_T
:
2173 /* transitional states, IO allowed */
2174 case C_DISCONNECTING
:
2178 case C_NETWORK_FAILURE
:
2179 case C_PROTOCOL_ERROR
:
2181 case C_WF_REPORT_PARAMS
:
2182 case C_STARTING_SYNC_S
:
2183 case C_STARTING_SYNC_T
:
2186 /* Allow IO in BM exchange states with new protocols */
2188 if (mdev
->tconn
->agreed_pro_version
< 96)
2192 /* no new io accepted in these states */
2194 case C_WF_SYNC_UUID
:
2200 switch ((enum drbd_disk_state
)s
.disk
) {
2202 case D_INCONSISTENT
:
2206 /* disk state is stable as well. */
2209 /* no new io accepted during tansitional states */
2222 static inline int is_susp(union drbd_state s
)
2224 return s
.susp
|| s
.susp_nod
|| s
.susp_fen
;
2227 static inline bool may_inc_ap_bio(struct drbd_conf
*mdev
)
2229 int mxb
= drbd_get_max_buffers(mdev
);
2231 if (is_susp(mdev
->state
))
2233 if (test_bit(SUSPEND_IO
, &mdev
->flags
))
2236 /* to avoid potential deadlock or bitmap corruption,
2237 * in various places, we only allow new application io
2238 * to start during "stable" states. */
2240 /* no new io accepted when attaching or detaching the disk */
2241 if (!drbd_state_is_stable(mdev
))
2244 /* since some older kernels don't have atomic_add_unless,
2245 * and we are within the spinlock anyways, we have this workaround. */
2246 if (atomic_read(&mdev
->ap_bio_cnt
) > mxb
)
2248 if (test_bit(BITMAP_IO
, &mdev
->flags
))
2253 static inline bool inc_ap_bio_cond(struct drbd_conf
*mdev
, int count
)
2257 spin_lock_irq(&mdev
->tconn
->req_lock
);
2258 rv
= may_inc_ap_bio(mdev
);
2260 atomic_add(count
, &mdev
->ap_bio_cnt
);
2261 spin_unlock_irq(&mdev
->tconn
->req_lock
);
2266 static inline void inc_ap_bio(struct drbd_conf
*mdev
, int count
)
2269 * as long as the device is suspended
2270 * until the bitmap is no longer on the fly during connection
2271 * handshake as long as we would exeed the max_buffer limit.
2273 * to avoid races with the reconnect code,
2274 * we need to atomic_inc within the spinlock. */
2276 wait_event(mdev
->misc_wait
, inc_ap_bio_cond(mdev
, count
));
2279 static inline void dec_ap_bio(struct drbd_conf
*mdev
)
2281 int mxb
= drbd_get_max_buffers(mdev
);
2282 int ap_bio
= atomic_dec_return(&mdev
->ap_bio_cnt
);
2284 D_ASSERT(ap_bio
>= 0);
2285 /* this currently does wake_up for every dec_ap_bio!
2286 * maybe rather introduce some type of hysteresis?
2287 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2289 wake_up(&mdev
->misc_wait
);
2290 if (ap_bio
== 0 && test_bit(BITMAP_IO
, &mdev
->flags
)) {
2291 if (!test_and_set_bit(BITMAP_IO_QUEUED
, &mdev
->flags
))
2292 drbd_queue_work(&mdev
->tconn
->data
.work
, &mdev
->bm_io_work
.w
);
2296 static inline int drbd_set_ed_uuid(struct drbd_conf
*mdev
, u64 val
)
2298 int changed
= mdev
->ed_uuid
!= val
;
2299 mdev
->ed_uuid
= val
;
2303 static inline int drbd_queue_order_type(struct drbd_conf
*mdev
)
2305 /* sorry, we currently have no working implementation
2306 * of distributed TCQ stuff */
2307 #ifndef QUEUE_ORDERED_NONE
2308 #define QUEUE_ORDERED_NONE 0
2310 return QUEUE_ORDERED_NONE
;
2313 static inline void drbd_md_flush(struct drbd_conf
*mdev
)
2317 if (test_bit(MD_NO_FUA
, &mdev
->flags
))
2320 r
= blkdev_issue_flush(mdev
->ldev
->md_bdev
, GFP_KERNEL
, NULL
);
2322 set_bit(MD_NO_FUA
, &mdev
->flags
);
2323 dev_err(DEV
, "meta data flush failed with status %d, disabling md-flushes\n", r
);