4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
6 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
7 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
8 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
10 drbd is free software; you can redistribute it and/or modify
11 it under the terms of the GNU General Public License as published by
12 the Free Software Foundation; either version 2, or (at your option)
15 drbd is distributed in the hope that it will be useful,
16 but WITHOUT ANY WARRANTY; without even the implied warranty of
17 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 GNU General Public License for more details.
20 You should have received a copy of the GNU General Public License
21 along with drbd; see the file COPYING. If not, write to
22 the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
29 #include <linux/compiler.h>
30 #include <linux/types.h>
31 #include <linux/version.h>
32 #include <linux/list.h>
33 #include <linux/sched.h>
34 #include <linux/bitops.h>
35 #include <linux/slab.h>
36 #include <linux/crypto.h>
37 #include <linux/ratelimit.h>
38 #include <linux/tcp.h>
39 #include <linux/mutex.h>
40 #include <linux/major.h>
41 #include <linux/blkdev.h>
42 #include <linux/genhd.h>
43 #include <linux/idr.h>
45 #include <linux/lru_cache.h>
46 #include <linux/prefetch.h>
47 #include <linux/drbd.h>
48 #include "drbd_state.h"
51 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
52 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
53 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
54 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
56 # define __protected_by(x)
57 # define __protected_read_by(x)
58 # define __protected_write_by(x)
59 # define __must_hold(x)
62 #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
64 /* module parameter, defined in drbd_main.c */
65 extern unsigned int minor_count
;
66 extern int disable_sendpage
;
68 extern unsigned int cn_idx
;
70 #ifdef CONFIG_DRBD_FAULT_INJECTION
71 extern int enable_faults
;
72 extern int fault_rate
;
73 extern int fault_devs
;
76 extern char usermode_helper
[];
79 /* I don't remember why XCPU ...
80 * This is used to wake the asender,
81 * and to interrupt sending the sending task
84 #define DRBD_SIG SIGXCPU
86 /* This is used to stop/restart our threads.
87 * Cannot use SIGTERM nor SIGKILL, since these
88 * are sent out by init on runlevel changes
89 * I choose SIGHUP for now.
91 #define DRBD_SIGKILL SIGHUP
93 #define ID_IN_SYNC (4711ULL)
94 #define ID_OUT_OF_SYNC (4712ULL)
95 #define ID_SYNCER (-1ULL)
97 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
103 /* to shorten dev_warn(DEV, "msg"); and relatives statements */
104 #define DEV (disk_to_dev(mdev->vdisk))
106 #define conn_printk(LEVEL, TCONN, FMT, ARGS...) \
107 printk(LEVEL "d-con %s: " FMT, TCONN->name , ## ARGS)
108 #define conn_alert(TCONN, FMT, ARGS...) conn_printk(KERN_ALERT, TCONN, FMT, ## ARGS)
109 #define conn_crit(TCONN, FMT, ARGS...) conn_printk(KERN_CRIT, TCONN, FMT, ## ARGS)
110 #define conn_err(TCONN, FMT, ARGS...) conn_printk(KERN_ERR, TCONN, FMT, ## ARGS)
111 #define conn_warn(TCONN, FMT, ARGS...) conn_printk(KERN_WARNING, TCONN, FMT, ## ARGS)
112 #define conn_notice(TCONN, FMT, ARGS...) conn_printk(KERN_NOTICE, TCONN, FMT, ## ARGS)
113 #define conn_info(TCONN, FMT, ARGS...) conn_printk(KERN_INFO, TCONN, FMT, ## ARGS)
114 #define conn_dbg(TCONN, FMT, ARGS...) conn_printk(KERN_DEBUG, TCONN, FMT, ## ARGS)
116 #define D_ASSERT(exp) if (!(exp)) \
117 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
120 * expect - Make an assertion
122 * Unlike the assert macro, this macro returns a boolean result.
124 #define expect(exp) ({ \
125 bool _bool = (exp); \
127 dev_err(DEV, "ASSERTION %s FAILED in %s\n", \
132 /* Defines to control fault insertion */
134 DRBD_FAULT_MD_WR
= 0, /* meta data write */
135 DRBD_FAULT_MD_RD
= 1, /* read */
136 DRBD_FAULT_RS_WR
= 2, /* resync */
137 DRBD_FAULT_RS_RD
= 3,
138 DRBD_FAULT_DT_WR
= 4, /* data */
139 DRBD_FAULT_DT_RD
= 5,
140 DRBD_FAULT_DT_RA
= 6, /* data read ahead */
141 DRBD_FAULT_BM_ALLOC
= 7, /* bitmap allocation */
142 DRBD_FAULT_AL_EE
= 8, /* alloc ee */
143 DRBD_FAULT_RECEIVE
= 9, /* Changes some bytes upon receiving a [rs]data block */
149 _drbd_insert_fault(struct drbd_conf
*mdev
, unsigned int type
);
152 drbd_insert_fault(struct drbd_conf
*mdev
, unsigned int type
) {
153 #ifdef CONFIG_DRBD_FAULT_INJECTION
155 (enable_faults
& (1<<type
)) &&
156 _drbd_insert_fault(mdev
, type
);
162 /* integer division, round _UP_ to the next integer */
163 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
164 /* usual integer division */
165 #define div_floor(A, B) ((A)/(B))
167 /* drbd_meta-data.c (still in drbd_main.c) */
168 /* 4th incarnation of the disk layout. */
169 #define DRBD_MD_MAGIC (DRBD_MAGIC+4)
171 extern struct drbd_conf
**minor_table
;
172 extern struct ratelimit_state drbd_ratelimit_state
;
176 /* receiver (data socket) */
178 P_DATA_REPLY
= 0x01, /* Response to P_DATA_REQUEST */
179 P_RS_DATA_REPLY
= 0x02, /* Response to P_RS_DATA_REQUEST */
182 P_BECOME_SYNC_TARGET
= 0x05,
183 P_BECOME_SYNC_SOURCE
= 0x06,
184 P_UNPLUG_REMOTE
= 0x07, /* Used at various times to hint the peer */
185 P_DATA_REQUEST
= 0x08, /* Used to ask for a data block */
186 P_RS_DATA_REQUEST
= 0x09, /* Used to ask for a data block for resync */
193 P_AUTH_CHALLENGE
= 0x10,
194 P_AUTH_RESPONSE
= 0x11,
195 P_STATE_CHG_REQ
= 0x12,
197 /* asender (meta socket */
200 P_RECV_ACK
= 0x15, /* Used in protocol B */
201 P_WRITE_ACK
= 0x16, /* Used in protocol C */
202 P_RS_WRITE_ACK
= 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
203 P_DISCARD_ACK
= 0x18, /* Used in proto C, two-primaries conflict detection */
204 P_NEG_ACK
= 0x19, /* Sent if local disk is unusable */
205 P_NEG_DREPLY
= 0x1a, /* Local disk is broken... */
206 P_NEG_RS_DREPLY
= 0x1b, /* Local disk is broken... */
207 P_BARRIER_ACK
= 0x1c,
208 P_STATE_CHG_REPLY
= 0x1d,
210 /* "new" commands, no longer fitting into the ordering scheme above */
212 P_OV_REQUEST
= 0x1e, /* data socket */
214 P_OV_RESULT
= 0x20, /* meta socket */
215 P_CSUM_RS_REQUEST
= 0x21, /* data socket */
216 P_RS_IS_IN_SYNC
= 0x22, /* meta socket */
217 P_SYNC_PARAM89
= 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
218 P_COMPRESSED_BITMAP
= 0x24, /* compressed or otherwise encoded bitmap transfer */
219 /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */
220 /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */
221 P_DELAY_PROBE
= 0x27, /* is used on BOTH sockets */
222 P_OUT_OF_SYNC
= 0x28, /* Mark as out of sync (Outrunning), data socket */
223 P_RS_CANCEL
= 0x29, /* meta: Used to cancel RS_DATA_REQUEST packet by SyncSource */
226 P_MAY_IGNORE
= 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
227 P_MAX_OPT_CMD
= 0x101,
229 /* special command ids for handshake */
231 P_HAND_SHAKE_M
= 0xfff1, /* First Packet on the MetaSock */
232 P_HAND_SHAKE_S
= 0xfff2, /* First Packet on the Socket */
234 P_HAND_SHAKE
= 0xfffe /* FIXED for the next century! */
237 extern const char *cmdname(enum drbd_packet cmd
);
239 /* for sending/receiving the bitmap,
240 * possibly in some encoding scheme */
243 * stores total bits and long words
244 * of the bitmap, so we don't need to
245 * call the accessor functions over and again. */
246 unsigned long bm_bits
;
247 unsigned long bm_words
;
248 /* during xfer, current position within the bitmap */
249 unsigned long bit_offset
;
250 unsigned long word_offset
;
252 /* statistics; index: (h->command == P_BITMAP) */
257 extern void INFO_bm_xfer_stats(struct drbd_conf
*mdev
,
258 const char *direction
, struct bm_xfer_ctx
*c
);
260 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx
*c
)
262 /* word_offset counts "native long words" (32 or 64 bit),
264 * Encoded packet may end at an unaligned bit offset.
265 * In case a fallback clear text packet is transmitted in
266 * between, we adjust this offset back to the last 64bit
267 * aligned "native long word", which makes coding and decoding
268 * the plain text bitmap much more convenient. */
269 #if BITS_PER_LONG == 64
270 c
->word_offset
= c
->bit_offset
>> 6;
271 #elif BITS_PER_LONG == 32
272 c
->word_offset
= c
->bit_offset
>> 5;
273 c
->word_offset
&= ~(1UL);
275 # error "unsupported BITS_PER_LONG"
280 #define __packed __attribute__((packed))
283 /* This is the layout for a packet on the wire.
284 * The byteorder is the network byte order.
285 * (except block_id and barrier fields.
286 * these are pointers to local structs
287 * and have no relevance for the partner,
288 * which just echoes them as received.)
290 * NOTE that the payload starts at a long aligned offset,
291 * regardless of 32 or 64 bit arch!
296 u16 length
; /* bytes of data after this header */
299 /* Header for big packets, Used for data packets exceeding 64kB */
301 u16 magic
; /* use DRBD_MAGIC_BIG here */
303 u32 length
; /* Use only 24 bits of that. Ignore the highest 8 bit. */
308 struct p_header80 h80
;
309 struct p_header95 h95
;
315 * short commands, packets without payload, plain p_header:
318 * P_BECOME_SYNC_TARGET
319 * P_BECOME_SYNC_SOURCE
324 * commands with out-of-struct payload:
325 * P_BITMAP (no additional fields)
326 * P_DATA, P_DATA_REPLY (see p_data)
327 * P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
330 /* these defines must not be changed without changing the protocol version */
331 #define DP_HARDBARRIER 1 /* depricated */
332 #define DP_RW_SYNC 2 /* equals REQ_SYNC */
333 #define DP_MAY_SET_IN_SYNC 4
334 #define DP_UNPLUG 8 /* not used anymore */
335 #define DP_FUA 16 /* equals REQ_FUA */
336 #define DP_FLUSH 32 /* equals REQ_FLUSH */
337 #define DP_DISCARD 64 /* equals REQ_DISCARD */
340 struct p_header head
;
341 u64 sector
; /* 64 bits sector number */
342 u64 block_id
; /* to identify the request in protocol B&C */
348 * commands which share a struct:
350 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
351 * P_DISCARD_ACK (proto C, two-primaries conflict detection)
353 * P_DATA_REQUEST, P_RS_DATA_REQUEST
356 struct p_header head
;
365 struct p_header head
;
369 u32 pad
; /* to multiple of 8 Byte */
373 * commands with their own struct for additional fields:
382 struct p_header head
; /* Note: vnr will be ignored */
387 /* should be more than enough for future enhancements
388 * for now, feature_flags and the reserverd array shall be zero.
394 /* 80 bytes, FIXED for the next century */
397 struct p_header head
;
398 u32 barrier
; /* barrier number _handle_ only */
399 u32 pad
; /* to multiple of 8 Byte */
402 struct p_barrier_ack
{
403 struct p_header head
;
409 struct p_header head
;
412 /* Since protocol version 88 and higher. */
416 struct p_rs_param_89
{
417 struct p_header head
;
419 /* protocol version 89: */
420 char verify_alg
[SHARED_SECRET_MAX
];
421 char csums_alg
[SHARED_SECRET_MAX
];
424 struct p_rs_param_95
{
425 struct p_header head
;
427 char verify_alg
[SHARED_SECRET_MAX
];
428 char csums_alg
[SHARED_SECRET_MAX
];
435 enum drbd_conn_flags
{
441 struct p_header head
;
449 /* Since protocol version 87 and higher. */
450 char integrity_alg
[0];
455 struct p_header head
;
456 u64 uuid
[UI_EXTENDED_SIZE
];
460 struct p_header head
;
465 struct p_header head
;
466 u64 d_size
; /* size of disk */
467 u64 u_size
; /* user requested size */
468 u64 c_size
; /* current exported size */
469 u32 max_bio_size
; /* Maximal size of a BIO */
470 u16 queue_order_type
; /* not yet implemented in DRBD*/
471 u16 dds_flags
; /* use enum dds_flags here. */
475 struct p_header head
;
480 struct p_header head
;
485 struct p_req_state_reply
{
486 struct p_header head
;
490 struct p_drbd06_param
{
501 struct p_header head
;
507 struct p_block_desc
{
508 struct p_header head
;
511 u32 pad
; /* to multiple of 8 Byte */
514 /* Valid values for the encoding field.
515 * Bump proto version when changing this. */
516 enum drbd_bitmap_code
{
517 /* RLE_VLI_Bytes = 0,
518 * and other bit variants had been defined during
519 * algorithm evaluation. */
523 struct p_compressed_bm
{
524 struct p_header head
;
525 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
526 * (encoding & 0x80): polarity (set/unset) of first runlength
527 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
528 * used to pad up to head.length bytes
535 struct p_delay_probe93
{
536 struct p_header head
;
537 u32 seq_num
; /* sequence number to match the two probe packets */
538 u32 offset
; /* usecs the probe got sent after the reference time point */
541 /* DCBP: Drbd Compressed Bitmap Packet ... */
542 static inline enum drbd_bitmap_code
543 DCBP_get_code(struct p_compressed_bm
*p
)
545 return (enum drbd_bitmap_code
)(p
->encoding
& 0x0f);
549 DCBP_set_code(struct p_compressed_bm
*p
, enum drbd_bitmap_code code
)
552 p
->encoding
= (p
->encoding
& ~0xf) | code
;
556 DCBP_get_start(struct p_compressed_bm
*p
)
558 return (p
->encoding
& 0x80) != 0;
562 DCBP_set_start(struct p_compressed_bm
*p
, int set
)
564 p
->encoding
= (p
->encoding
& ~0x80) | (set
? 0x80 : 0);
568 DCBP_get_pad_bits(struct p_compressed_bm
*p
)
570 return (p
->encoding
>> 4) & 0x7;
574 DCBP_set_pad_bits(struct p_compressed_bm
*p
, int n
)
577 p
->encoding
= (p
->encoding
& (~0x7 << 4)) | (n
<< 4);
580 /* one bitmap packet, including the p_header,
581 * should fit within one _architecture independend_ page.
582 * so we need to use the fixed size 4KiB page size
583 * most architectures have used for a long time.
585 #define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header))
586 #define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
587 #define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm))
588 #if (PAGE_SIZE < 4096)
589 /* drbd_send_bitmap / receive_bitmap would break horribly */
590 #error "PAGE_SIZE too small"
594 struct p_header header
;
595 struct p_handshake handshake
;
597 struct p_block_ack block_ack
;
598 struct p_barrier barrier
;
599 struct p_barrier_ack barrier_ack
;
600 struct p_rs_param_89 rs_param_89
;
601 struct p_rs_param_95 rs_param_95
;
602 struct p_protocol protocol
;
603 struct p_sizes sizes
;
604 struct p_uuids uuids
;
605 struct p_state state
;
606 struct p_req_state req_state
;
607 struct p_req_state_reply req_state_reply
;
608 struct p_block_req block_req
;
609 struct p_delay_probe93 delay_probe93
;
610 struct p_rs_uuid rs_uuid
;
611 struct p_block_desc block_desc
;
614 /**********************************************************************/
615 enum drbd_thread_state
{
624 struct task_struct
*task
;
625 struct completion stop
;
626 enum drbd_thread_state t_state
;
627 int (*function
) (struct drbd_thread
*);
628 struct drbd_tconn
*tconn
;
633 static inline enum drbd_thread_state
get_t_state(struct drbd_thread
*thi
)
635 /* THINK testing the t_state seems to be uncritical in all cases
636 * (but thread_{start,stop}), so we can read it *without* the lock.
644 typedef int (*drbd_work_cb
)(struct drbd_work
*, int cancel
);
646 struct list_head list
;
649 struct drbd_conf
*mdev
;
650 struct drbd_tconn
*tconn
;
654 #include "drbd_interval.h"
656 struct drbd_request
{
659 /* if local IO is not allowed, will be NULL.
660 * if local IO _is_ allowed, holds the locally submitted bio clone,
661 * or, after local IO completion, the ERR_PTR(error).
662 * see drbd_endio_pri(). */
663 struct bio
*private_bio
;
665 struct drbd_interval i
;
666 unsigned int epoch
; /* barrier_nr */
668 /* barrier_nr: used to check on "completion" whether this req was in
669 * the current epoch, and we therefore have to close it,
670 * starting a new epoch...
673 struct list_head tl_requests
; /* ring list in the transfer log */
674 struct bio
*master_bio
; /* master bio pointer */
675 unsigned long rq_state
; /* see comments above _req_mod() */
677 unsigned long start_time
;
680 struct drbd_tl_epoch
{
682 struct list_head requests
; /* requests before */
683 struct drbd_tl_epoch
*next
; /* pointer to the next barrier */
684 unsigned int br_number
; /* the barriers identifier. */
685 int n_writes
; /* number of requests attached before this barrier */
689 struct list_head list
;
690 unsigned int barrier_nr
;
691 atomic_t epoch_size
; /* increased on every request added. */
692 atomic_t active
; /* increased on every req. added, and dec on every finished. */
696 /* drbd_epoch flag bits */
698 DE_HAVE_BARRIER_NUMBER
,
705 EV_CLEANUP
= 32, /* used as flag */
708 struct drbd_wq_barrier
{
710 struct completion done
;
718 struct drbd_peer_request
{
720 struct drbd_epoch
*epoch
; /* for writes */
722 atomic_t pending_bios
;
723 struct drbd_interval i
;
724 /* see comments on ee flag bits below */
728 struct digest_info
*digest
;
733 * While corresponding bios are in flight, the only modification will be
734 * set_bit WAS_ERROR, which has to be atomic.
735 * If no bios are in flight yet, or all have been completed,
736 * non-atomic modification to ee->flags is ok.
739 __EE_CALL_AL_COMPLETE_IO
,
740 __EE_MAY_SET_IN_SYNC
,
742 /* In case a barrier failed,
743 * we need to resubmit without the barrier flag. */
746 /* we may have several bios per peer request.
747 * if any of those fail, we set this flag atomically
748 * from the endio callback */
751 /* This ee has a pointer to a digest instead of a block id */
754 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
755 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
756 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
757 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
758 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
760 /* flag bits per mdev */
762 CREATE_BARRIER
, /* next P_DATA is preceded by a P_BARRIER */
763 UNPLUG_QUEUED
, /* only relevant with kernel 2.4 */
764 UNPLUG_REMOTE
, /* sending a "UnplugRemote" could help */
765 MD_DIRTY
, /* current uuids and flags not yet on disk */
766 USE_DEGR_WFC_T
, /* degr-wfc-timeout instead of wfc-timeout. */
767 CLUSTER_ST_CHANGE
, /* Cluster wide state change going on... */
770 CRASHED_PRIMARY
, /* This node was a crashed primary.
771 * Gets cleared when the state.conn
772 * goes into C_CONNECTED state. */
773 NO_BARRIER_SUPP
, /* underlying block device doesn't implement barriers */
776 MD_NO_FUA
, /* Users wants us to not use FUA/FLUSH on meta data dev */
777 SUSPEND_IO
, /* suspend application io */
778 BITMAP_IO
, /* suspend application io;
779 once no more io in flight, start bitmap io */
780 BITMAP_IO_QUEUED
, /* Started bitmap IO */
781 GO_DISKLESS
, /* Disk is being detached, on io-error or admin request. */
782 WAS_IO_ERROR
, /* Local disk failed returned IO error */
783 RESYNC_AFTER_NEG
, /* Resync after online grow after the attach&negotiate finished. */
784 CONFIG_PENDING
, /* serialization of (re)configuration requests.
785 * if set, also prevents the device from dying */
786 DEVICE_DYING
, /* device became unconfigured,
787 * but worker thread is still handling the cleanup.
788 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
789 * while this is set. */
790 RESIZE_PENDING
, /* Size change detected locally, waiting for the response from
791 * the peer, if it changed there as well. */
792 CONN_DRY_RUN
, /* Expect disconnect after resync handshake. */
793 NEW_CUR_UUID
, /* Create new current UUID when thawing IO */
794 AL_SUSPENDED
, /* Activity logging is currently suspended. */
795 AHEAD_TO_SYNC_SOURCE
, /* Ahead -> SyncSource queued */
796 B_RS_H_DONE
, /* Before resync handler done (already executed) */
799 struct drbd_bitmap
; /* opaque for drbd_conf */
801 /* definition of bits in bm_flags to be used in drbd_bm_lock
802 * and drbd_bitmap_io and friends. */
804 /* do we need to kfree, or vfree bm_pages? */
805 BM_P_VMALLOCED
= 0x10000, /* internal use only, will be masked out */
807 /* currently locked for bulk operation */
808 BM_LOCKED_MASK
= 0x7,
810 /* in detail, that is: */
815 /* (test bit, count bit) allowed (common case) */
816 BM_LOCKED_TEST_ALLOWED
= 0x3,
818 /* testing bits, as well as setting new bits allowed, but clearing bits
819 * would be unexpected. Used during bitmap receive. Setting new bits
820 * requires sending of "out-of-sync" information, though. */
821 BM_LOCKED_SET_ALLOWED
= 0x1,
823 /* clear is not expected while bitmap is locked for bulk operation */
827 /* TODO sort members for performance
828 * MAYBE group them further */
830 /* THINK maybe we actually want to use the default "event/%s" worker threads
831 * or similar in linux 2.6, which uses per cpu data and threads.
833 struct drbd_work_queue
{
835 struct semaphore s
; /* producers up it, worker down()s it */
836 spinlock_t q_lock
; /* to protect the list. */
840 struct drbd_work_queue work
;
842 struct socket
*socket
;
843 /* this way we get our
844 * send/receive buffers off the stack */
845 union p_polymorph sbuf
;
846 union p_polymorph rbuf
;
850 u64 md_offset
; /* sector offset to 'super' block */
852 u64 la_size_sect
; /* last agreed size, unit sectors */
858 s32 al_offset
; /* signed relative sector offset to al area */
859 s32 bm_offset
; /* signed relative sector offset to bitmap */
861 /* u32 al_nr_extents; important for restoring the AL
862 * is stored into sync_conf.al_extents, which in turn
863 * gets applied to act_log->nr_elements
867 /* for sync_conf and other types... */
868 #define NL_PACKET(name, number, fields) struct name { fields };
869 #define NL_INTEGER(pn,pr,member) int member;
870 #define NL_INT64(pn,pr,member) __u64 member;
871 #define NL_BIT(pn,pr,member) unsigned member:1;
872 #define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
873 #include "linux/drbd_nl.h"
875 struct drbd_backing_dev
{
876 struct block_device
*backing_bdev
;
877 struct block_device
*md_bdev
;
879 struct disk_conf dc
; /* The user provided config... */
880 sector_t known_size
; /* last known size of that backing device */
884 struct drbd_conf
*mdev
;
885 struct completion event
;
893 int (*io_fn
)(struct drbd_conf
*mdev
);
894 void (*done
)(struct drbd_conf
*mdev
, int rv
);
897 enum write_ordering_e
{
905 unsigned int head_index
;
909 /* flag bits per tconn */
911 NET_CONGESTED
, /* The data socket is congested */
912 DISCARD_CONCURRENT
, /* Set on one node, cleared on the peer! */
913 SEND_PING
, /* whether asender should send a ping asap */
914 SIGNAL_ASENDER
, /* whether asender wants to be interrupted */
915 GOT_PING_ACK
, /* set when we receive a ping_ack packet, ping_wait gets woken */
918 struct drbd_tconn
{ /* is a resource from the config file */
919 char *name
; /* Resource name */
920 struct list_head all_tconn
; /* List of all drbd_tconn, prot by global_state_lock */
921 struct idr volumes
; /* <tconn, vnr> to mdev mapping */
922 enum drbd_conns cstate
; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
925 struct net_conf
*net_conf
; /* protected by get_net_conf() and put_net_conf() */
926 atomic_t net_cnt
; /* Users of net_conf */
927 wait_queue_head_t net_cnt_wait
;
928 wait_queue_head_t ping_wait
; /* Woken upon reception of a ping, and a state change */
930 struct drbd_socket data
; /* data/barrier/cstate/parameter packets */
931 struct drbd_socket meta
; /* ping/ack (metadata) packets */
932 int agreed_pro_version
; /* actually used protocol version */
933 unsigned long last_received
; /* in jiffies, either socket */
934 unsigned int ko_count
;
937 struct drbd_tl_epoch
*unused_spare_tle
; /* for pre-allocation */
938 struct drbd_tl_epoch
*newest_tle
;
939 struct drbd_tl_epoch
*oldest_tle
;
940 struct list_head out_of_sequence_requests
;
942 struct crypto_hash
*cram_hmac_tfm
;
943 struct crypto_hash
*integrity_w_tfm
; /* to be used by the worker thread */
944 struct crypto_hash
*integrity_r_tfm
; /* to be used by the receiver thread */
949 struct drbd_thread receiver
;
950 struct drbd_thread worker
;
951 struct drbd_thread asender
;
952 cpumask_var_t cpu_mask
;
956 struct drbd_tconn
*tconn
;
957 int vnr
; /* volume number within the connection */
959 /* things that are stored as / read from meta data on disk */
962 /* configured by drbdsetup */
963 struct syncer_conf sync_conf
;
964 struct drbd_backing_dev
*ldev
__protected_by(local
);
966 sector_t p_size
; /* partner's disk size */
967 struct request_queue
*rq_queue
;
968 struct block_device
*this_bdev
;
969 struct gendisk
*vdisk
;
971 struct drbd_work resync_work
,
976 struct timer_list resync_timer
;
977 struct timer_list md_sync_timer
;
978 struct timer_list start_resync_timer
;
979 struct timer_list request_timer
;
980 #ifdef DRBD_DEBUG_MD_SYNC
984 } last_md_mark_dirty
;
987 /* Used after attach while negotiating new disk state. */
988 union drbd_state new_state_tmp
;
990 union drbd_state state
;
991 wait_queue_head_t misc_wait
;
992 wait_queue_head_t state_wait
; /* upon each state change. */
993 unsigned int send_cnt
;
994 unsigned int recv_cnt
;
995 unsigned int read_cnt
;
996 unsigned int writ_cnt
;
997 unsigned int al_writ_cnt
;
998 unsigned int bm_writ_cnt
;
999 atomic_t ap_bio_cnt
; /* Requests we need to complete */
1000 atomic_t ap_pending_cnt
; /* AP data packets on the wire, ack expected */
1001 atomic_t rs_pending_cnt
; /* RS request/data packets on the wire */
1002 atomic_t unacked_cnt
; /* Need to send replys for */
1003 atomic_t local_cnt
; /* Waiting for local completion */
1005 /* Interval tree of pending local requests */
1006 struct rb_root read_requests
;
1007 struct rb_root write_requests
;
1009 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
1010 unsigned long rs_total
;
1011 /* number of resync blocks that failed in this run */
1012 unsigned long rs_failed
;
1013 /* Syncer's start time [unit jiffies] */
1014 unsigned long rs_start
;
1015 /* cumulated time in PausedSyncX state [unit jiffies] */
1016 unsigned long rs_paused
;
1017 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
1018 unsigned long rs_same_csum
;
1019 #define DRBD_SYNC_MARKS 8
1020 #define DRBD_SYNC_MARK_STEP (3*HZ)
1021 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
1022 unsigned long rs_mark_left
[DRBD_SYNC_MARKS
];
1023 /* marks's time [unit jiffies] */
1024 unsigned long rs_mark_time
[DRBD_SYNC_MARKS
];
1025 /* current index into rs_mark_{left,time} */
1028 /* where does the admin want us to start? (sector) */
1029 sector_t ov_start_sector
;
1030 /* where are we now? (sector) */
1031 sector_t ov_position
;
1032 /* Start sector of out of sync range (to merge printk reporting). */
1033 sector_t ov_last_oos_start
;
1034 /* size of out-of-sync range in sectors. */
1035 sector_t ov_last_oos_size
;
1036 unsigned long ov_left
; /* in bits */
1037 struct crypto_hash
*csums_tfm
;
1038 struct crypto_hash
*verify_tfm
;
1040 struct drbd_bitmap
*bitmap
;
1041 unsigned long bm_resync_fo
; /* bit offset for drbd_bm_find_next */
1043 /* Used to track operations of resync... */
1044 struct lru_cache
*resync
;
1045 /* Number of locked elements in resync LRU */
1046 unsigned int resync_locked
;
1047 /* resync extent number waiting for application requests */
1048 unsigned int resync_wenr
;
1052 struct drbd_epoch
*current_epoch
;
1053 spinlock_t epoch_lock
;
1054 unsigned int epochs
;
1055 enum write_ordering_e write_ordering
;
1056 struct list_head active_ee
; /* IO in progress (P_DATA gets written to disk) */
1057 struct list_head sync_ee
; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
1058 struct list_head done_ee
; /* need to send P_WRITE_ACK */
1059 struct list_head read_ee
; /* [RS]P_DATA_REQUEST being read */
1060 struct list_head net_ee
; /* zero-copy network send in progress */
1062 int next_barrier_nr
;
1063 struct list_head resync_reads
;
1064 atomic_t pp_in_use
; /* allocated from page pool */
1065 atomic_t pp_in_use_by_net
; /* sendpage()d, still referenced by tcp */
1066 wait_queue_head_t ee_wait
;
1067 struct page
*md_io_page
; /* one page buffer for md_io */
1068 struct page
*md_io_tmpp
; /* for logical_block_size != 512 */
1069 struct mutex md_io_mutex
; /* protects the md_io_buffer */
1071 wait_queue_head_t al_wait
;
1072 struct lru_cache
*act_log
; /* activity log */
1073 unsigned int al_tr_number
;
1075 int al_tr_pos
; /* position of the next transaction in the journal */
1076 wait_queue_head_t seq_wait
;
1077 atomic_t packet_seq
;
1078 unsigned int peer_seq
;
1079 spinlock_t peer_seq_lock
;
1081 unsigned long comm_bm_set
; /* communicated number of set bits. */
1082 struct bm_io_work bm_io_work
;
1083 u64 ed_uuid
; /* UUID of the exposed data */
1084 struct mutex state_mutex
;
1085 char congestion_reason
; /* Why we where congested... */
1086 atomic_t rs_sect_in
; /* for incoming resync data rate, SyncTarget */
1087 atomic_t rs_sect_ev
; /* for submitted resync data rate, both */
1088 int rs_last_sect_ev
; /* counter to compare with */
1089 int rs_last_events
; /* counter of read or write "events" (unit sectors)
1090 * on the lower level device when we last looked. */
1091 int c_sync_rate
; /* current resync rate after syncer throttle magic */
1092 struct fifo_buffer rs_plan_s
; /* correction values of resync planer */
1093 int rs_in_flight
; /* resync sectors in flight (to proxy, in proxy and from proxy) */
1094 int rs_planed
; /* resync sectors already planned */
1095 atomic_t ap_in_flight
; /* App sectors in flight (waiting for ack) */
1096 int peer_max_bio_size
;
1097 int local_max_bio_size
;
1100 static inline struct drbd_conf
*minor_to_mdev(unsigned int minor
)
1102 struct drbd_conf
*mdev
;
1104 mdev
= minor
< minor_count
? minor_table
[minor
] : NULL
;
1109 static inline unsigned int mdev_to_minor(struct drbd_conf
*mdev
)
1114 static inline struct drbd_conf
*vnr_to_mdev(struct drbd_tconn
*tconn
, int vnr
)
1116 return (struct drbd_conf
*)idr_find(&tconn
->volumes
, vnr
);
1119 /* returns 1 if it was successful,
1120 * returns 0 if there was no data socket.
1121 * so wherever you are going to use the data.socket, e.g. do
1122 * if (!drbd_get_data_sock(mdev->tconn))
1125 * drbd_get_data_sock(mdev->tconn);
1127 static inline int drbd_get_data_sock(struct drbd_tconn
*tconn
)
1129 mutex_lock(&tconn
->data
.mutex
);
1130 /* drbd_disconnect() could have called drbd_free_sock()
1131 * while we were waiting in down()... */
1132 if (unlikely(tconn
->data
.socket
== NULL
)) {
1133 mutex_unlock(&tconn
->data
.mutex
);
1139 static inline void drbd_put_data_sock(struct drbd_tconn
*tconn
)
1141 mutex_unlock(&tconn
->data
.mutex
);
1145 * function declarations
1146 *************************/
1152 DDSF_NO_RESYNC
= 2, /* Do not run a resync for the new space */
1155 extern void drbd_init_set_defaults(struct drbd_conf
*mdev
);
1156 extern int drbd_thread_start(struct drbd_thread
*thi
);
1157 extern void _drbd_thread_stop(struct drbd_thread
*thi
, int restart
, int wait
);
1158 extern char *drbd_task_to_thread_name(struct drbd_tconn
*tconn
, struct task_struct
*task
);
1160 extern void drbd_thread_current_set_cpu(struct drbd_thread
*thi
);
1161 extern void drbd_calc_cpu_mask(struct drbd_tconn
*tconn
);
1163 #define drbd_thread_current_set_cpu(A) ({})
1164 #define drbd_calc_cpu_mask(A) ({})
1166 extern void drbd_free_resources(struct drbd_conf
*mdev
);
1167 extern void tl_release(struct drbd_conf
*mdev
, unsigned int barrier_nr
,
1168 unsigned int set_size
);
1169 extern void tl_clear(struct drbd_conf
*mdev
);
1170 extern void _tl_add_barrier(struct drbd_conf
*, struct drbd_tl_epoch
*);
1171 extern void drbd_free_sock(struct drbd_tconn
*tconn
);
1172 extern int drbd_send(struct drbd_tconn
*tconn
, struct socket
*sock
,
1173 void *buf
, size_t size
, unsigned msg_flags
);
1174 extern int drbd_send_protocol(struct drbd_tconn
*tconn
);
1175 extern int drbd_send_uuids(struct drbd_conf
*mdev
);
1176 extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf
*mdev
);
1177 extern int drbd_gen_and_send_sync_uuid(struct drbd_conf
*mdev
);
1178 extern int drbd_send_sizes(struct drbd_conf
*mdev
, int trigger_reply
, enum dds_flags flags
);
1179 extern int _drbd_send_state(struct drbd_conf
*mdev
);
1180 extern int drbd_send_state(struct drbd_conf
*mdev
);
1181 extern int _conn_send_cmd(struct drbd_tconn
*tconn
, int vnr
, struct socket
*sock
,
1182 enum drbd_packet cmd
, struct p_header
*h
, size_t size
,
1183 unsigned msg_flags
);
1184 extern int conn_send_cmd(struct drbd_tconn
*tconn
, int vnr
, int use_data_socket
,
1185 enum drbd_packet cmd
, struct p_header
*h
, size_t size
);
1186 extern int conn_send_cmd2(struct drbd_tconn
*tconn
, enum drbd_packet cmd
,
1187 char *data
, size_t size
);
1188 #define USE_DATA_SOCKET 1
1189 #define USE_META_SOCKET 0
1190 extern int drbd_send_sync_param(struct drbd_conf
*mdev
, struct syncer_conf
*sc
);
1191 extern int drbd_send_b_ack(struct drbd_conf
*mdev
, u32 barrier_nr
,
1193 extern int drbd_send_ack(struct drbd_conf
*, enum drbd_packet
,
1194 struct drbd_peer_request
*);
1195 extern int drbd_send_ack_rp(struct drbd_conf
*mdev
, enum drbd_packet cmd
,
1196 struct p_block_req
*rp
);
1197 extern int drbd_send_ack_dp(struct drbd_conf
*mdev
, enum drbd_packet cmd
,
1198 struct p_data
*dp
, int data_size
);
1199 extern int drbd_send_ack_ex(struct drbd_conf
*mdev
, enum drbd_packet cmd
,
1200 sector_t sector
, int blksize
, u64 block_id
);
1201 extern int drbd_send_oos(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1202 extern int drbd_send_block(struct drbd_conf
*, enum drbd_packet
,
1203 struct drbd_peer_request
*);
1204 extern int drbd_send_dblock(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1205 extern int drbd_send_drequest(struct drbd_conf
*mdev
, int cmd
,
1206 sector_t sector
, int size
, u64 block_id
);
1207 extern int drbd_send_drequest_csum(struct drbd_conf
*mdev
, sector_t sector
,
1208 int size
, void *digest
, int digest_size
,
1209 enum drbd_packet cmd
);
1210 extern int drbd_send_ov_request(struct drbd_conf
*mdev
,sector_t sector
,int size
);
1212 extern int drbd_send_bitmap(struct drbd_conf
*mdev
);
1213 extern int _drbd_send_bitmap(struct drbd_conf
*mdev
);
1214 extern int drbd_send_sr_reply(struct drbd_conf
*mdev
, enum drbd_state_rv retcode
);
1215 extern void drbd_free_bc(struct drbd_backing_dev
*ldev
);
1216 extern void drbd_mdev_cleanup(struct drbd_conf
*mdev
);
1217 void drbd_print_uuids(struct drbd_conf
*mdev
, const char *text
);
1219 extern void drbd_md_sync(struct drbd_conf
*mdev
);
1220 extern int drbd_md_read(struct drbd_conf
*mdev
, struct drbd_backing_dev
*bdev
);
1221 extern void drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1222 extern void _drbd_uuid_set(struct drbd_conf
*mdev
, int idx
, u64 val
) __must_hold(local
);
1223 extern void drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1224 extern void _drbd_uuid_new_current(struct drbd_conf
*mdev
) __must_hold(local
);
1225 extern void drbd_uuid_set_bm(struct drbd_conf
*mdev
, u64 val
) __must_hold(local
);
1226 extern void drbd_md_set_flag(struct drbd_conf
*mdev
, int flags
) __must_hold(local
);
1227 extern void drbd_md_clear_flag(struct drbd_conf
*mdev
, int flags
)__must_hold(local
);
1228 extern int drbd_md_test_flag(struct drbd_backing_dev
*, int);
1229 #ifndef DRBD_DEBUG_MD_SYNC
1230 extern void drbd_md_mark_dirty(struct drbd_conf
*mdev
);
1232 #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1233 extern void drbd_md_mark_dirty_(struct drbd_conf
*mdev
,
1234 unsigned int line
, const char *func
);
1236 extern void drbd_queue_bitmap_io(struct drbd_conf
*mdev
,
1237 int (*io_fn
)(struct drbd_conf
*),
1238 void (*done
)(struct drbd_conf
*, int),
1239 char *why
, enum bm_flag flags
);
1240 extern int drbd_bitmap_io(struct drbd_conf
*mdev
,
1241 int (*io_fn
)(struct drbd_conf
*),
1242 char *why
, enum bm_flag flags
);
1243 extern int drbd_bmio_set_n_write(struct drbd_conf
*mdev
);
1244 extern int drbd_bmio_clear_n_write(struct drbd_conf
*mdev
);
1245 extern void drbd_go_diskless(struct drbd_conf
*mdev
);
1246 extern void drbd_ldev_destroy(struct drbd_conf
*mdev
);
1250 We reserve a 128MB Block (4k aligned)
1251 * either at the end of the backing device
1252 * or on a separate meta data device. */
1254 #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1255 /* The following numbers are sectors */
1256 #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1257 #define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */
1258 /* Allows up to about 3.8TB */
1259 #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1261 /* Since the smalles IO unit is usually 512 byte */
1262 #define MD_SECTOR_SHIFT 9
1263 #define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT)
1266 #define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1267 #define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */
1268 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1270 #if BITS_PER_LONG == 32
1272 #define cpu_to_lel(A) cpu_to_le32(A)
1273 #define lel_to_cpu(A) le32_to_cpu(A)
1274 #elif BITS_PER_LONG == 64
1276 #define cpu_to_lel(A) cpu_to_le64(A)
1277 #define lel_to_cpu(A) le64_to_cpu(A)
1279 #error "LN2 of BITS_PER_LONG unknown!"
1283 /* 16MB sized 'bitmap extent' to track syncer usage */
1285 int rs_left
; /* number of bits set (out of sync) in this extent. */
1286 int rs_failed
; /* number of failed resync requests in this extent. */
1287 unsigned long flags
;
1288 struct lc_element lce
;
1291 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1292 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1293 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1297 * We need to store one bit for a block.
1298 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1299 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1300 * Bit 1 ==> local node thinks this block needs to be synced.
1303 #define SLEEP_TIME (HZ/10)
1305 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1306 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1307 /* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1308 * per sector of on disk bitmap */
1309 #define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */
1310 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1312 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1313 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1316 /* thus many _storage_ sectors are described by one bit */
1317 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1318 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1319 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1321 /* bit to represented kilo byte conversion */
1322 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1324 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1325 * _storage_ sector is located in */
1326 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1328 /* how much _storage_ sectors we have per bitmap sector */
1329 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1330 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1332 /* in one sector of the bitmap, we have this many activity_log extents. */
1333 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1334 #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1336 #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1337 #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1339 /* the extent in "PER_EXTENT" below is an activity log extent
1340 * we need that many (long words/bytes) to store the bitmap
1341 * of one AL_EXTENT_SIZE chunk of storage.
1342 * we can store the bitmap for that many AL_EXTENTS within
1343 * one sector of the _on_disk_ bitmap:
1344 * bit 0 bit 37 bit 38 bit (512*8)-1
1345 * ...|........|........|.. // ..|........|
1346 * sect. 0 `296 `304 ^(512*8*8)-1
1348 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1349 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1350 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1353 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1354 #define DRBD_MAX_SECTORS_BM \
1355 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1356 #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1357 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1358 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1359 #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1360 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1361 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1363 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1364 /* 16 TB in units of sectors */
1365 #if BITS_PER_LONG == 32
1366 /* adjust by one page worth of bitmap,
1367 * so we won't wrap around in drbd_bm_find_next_bit.
1368 * you should use 64bit OS for that much storage, anyways. */
1369 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1371 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1372 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1373 /* corresponds to (1UL << 38) bits right now. */
1378 #define DRBD_MAX_BIO_SIZE (1U<<(9+HT_SHIFT))
1379 #define DRBD_MAX_BIO_SIZE_SAFE (1 << 12) /* Works always = 4k */
1381 #define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */
1383 extern int drbd_bm_init(struct drbd_conf
*mdev
);
1384 extern int drbd_bm_resize(struct drbd_conf
*mdev
, sector_t sectors
, int set_new_bits
);
1385 extern void drbd_bm_cleanup(struct drbd_conf
*mdev
);
1386 extern void drbd_bm_set_all(struct drbd_conf
*mdev
);
1387 extern void drbd_bm_clear_all(struct drbd_conf
*mdev
);
1388 /* set/clear/test only a few bits at a time */
1389 extern int drbd_bm_set_bits(
1390 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1391 extern int drbd_bm_clear_bits(
1392 struct drbd_conf
*mdev
, unsigned long s
, unsigned long e
);
1393 extern int drbd_bm_count_bits(
1394 struct drbd_conf
*mdev
, const unsigned long s
, const unsigned long e
);
1395 /* bm_set_bits variant for use while holding drbd_bm_lock,
1396 * may process the whole bitmap in one go */
1397 extern void _drbd_bm_set_bits(struct drbd_conf
*mdev
,
1398 const unsigned long s
, const unsigned long e
);
1399 extern int drbd_bm_test_bit(struct drbd_conf
*mdev
, unsigned long bitnr
);
1400 extern int drbd_bm_e_weight(struct drbd_conf
*mdev
, unsigned long enr
);
1401 extern int drbd_bm_write_page(struct drbd_conf
*mdev
, unsigned int idx
) __must_hold(local
);
1402 extern int drbd_bm_read(struct drbd_conf
*mdev
) __must_hold(local
);
1403 extern int drbd_bm_write(struct drbd_conf
*mdev
) __must_hold(local
);
1404 extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf
*mdev
,
1405 unsigned long al_enr
);
1406 extern size_t drbd_bm_words(struct drbd_conf
*mdev
);
1407 extern unsigned long drbd_bm_bits(struct drbd_conf
*mdev
);
1408 extern sector_t
drbd_bm_capacity(struct drbd_conf
*mdev
);
1410 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1411 extern unsigned long drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1412 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1413 extern unsigned long _drbd_bm_find_next(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1414 extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf
*mdev
, unsigned long bm_fo
);
1415 extern unsigned long _drbd_bm_total_weight(struct drbd_conf
*mdev
);
1416 extern unsigned long drbd_bm_total_weight(struct drbd_conf
*mdev
);
1417 extern int drbd_bm_rs_done(struct drbd_conf
*mdev
);
1418 /* for receive_bitmap */
1419 extern void drbd_bm_merge_lel(struct drbd_conf
*mdev
, size_t offset
,
1420 size_t number
, unsigned long *buffer
);
1421 /* for _drbd_send_bitmap */
1422 extern void drbd_bm_get_lel(struct drbd_conf
*mdev
, size_t offset
,
1423 size_t number
, unsigned long *buffer
);
1425 extern void drbd_bm_lock(struct drbd_conf
*mdev
, char *why
, enum bm_flag flags
);
1426 extern void drbd_bm_unlock(struct drbd_conf
*mdev
);
1429 extern struct kmem_cache
*drbd_request_cache
;
1430 extern struct kmem_cache
*drbd_ee_cache
; /* peer requests */
1431 extern struct kmem_cache
*drbd_bm_ext_cache
; /* bitmap extents */
1432 extern struct kmem_cache
*drbd_al_ext_cache
; /* activity log extents */
1433 extern mempool_t
*drbd_request_mempool
;
1434 extern mempool_t
*drbd_ee_mempool
;
1436 extern struct page
*drbd_pp_pool
; /* drbd's page pool */
1437 extern spinlock_t drbd_pp_lock
;
1438 extern int drbd_pp_vacant
;
1439 extern wait_queue_head_t drbd_pp_wait
;
1441 extern rwlock_t global_state_lock
;
1443 extern struct drbd_conf
*drbd_new_device(unsigned int minor
);
1444 extern void drbd_free_mdev(struct drbd_conf
*mdev
);
1446 struct drbd_tconn
*drbd_new_tconn(char *name
);
1447 extern void drbd_free_tconn(struct drbd_tconn
*tconn
);
1449 extern int proc_details
;
1452 extern int drbd_make_request(struct request_queue
*q
, struct bio
*bio
);
1453 extern int drbd_read_remote(struct drbd_conf
*mdev
, struct drbd_request
*req
);
1454 extern int drbd_merge_bvec(struct request_queue
*q
, struct bvec_merge_data
*bvm
, struct bio_vec
*bvec
);
1455 extern int is_valid_ar_handle(struct drbd_request
*, sector_t
);
1459 extern void drbd_suspend_io(struct drbd_conf
*mdev
);
1460 extern void drbd_resume_io(struct drbd_conf
*mdev
);
1461 extern char *ppsize(char *buf
, unsigned long long size
);
1462 extern sector_t
drbd_new_dev_size(struct drbd_conf
*, struct drbd_backing_dev
*, int);
1463 enum determine_dev_size
{ dev_size_error
= -1, unchanged
= 0, shrunk
= 1, grew
= 2 };
1464 extern enum determine_dev_size
drbd_determine_dev_size(struct drbd_conf
*, enum dds_flags
) __must_hold(local
);
1465 extern void resync_after_online_grow(struct drbd_conf
*);
1466 extern void drbd_reconsider_max_bio_size(struct drbd_conf
*mdev
);
1467 extern enum drbd_state_rv
drbd_set_role(struct drbd_conf
*mdev
,
1468 enum drbd_role new_role
,
1470 extern enum drbd_disk_state
drbd_try_outdate_peer(struct drbd_conf
*mdev
);
1471 extern void drbd_try_outdate_peer_async(struct drbd_conf
*mdev
);
1472 extern int drbd_khelper(struct drbd_conf
*mdev
, char *cmd
);
1475 extern int drbd_worker(struct drbd_thread
*thi
);
1476 extern int drbd_alter_sa(struct drbd_conf
*mdev
, int na
);
1477 extern void drbd_start_resync(struct drbd_conf
*mdev
, enum drbd_conns side
);
1478 extern void resume_next_sg(struct drbd_conf
*mdev
);
1479 extern void suspend_other_sg(struct drbd_conf
*mdev
);
1480 extern int drbd_resync_finished(struct drbd_conf
*mdev
);
1481 /* maybe rather drbd_main.c ? */
1482 extern int drbd_md_sync_page_io(struct drbd_conf
*mdev
,
1483 struct drbd_backing_dev
*bdev
, sector_t sector
, int rw
);
1484 extern void drbd_ov_oos_found(struct drbd_conf
*, sector_t
, int);
1485 extern void drbd_rs_controller_reset(struct drbd_conf
*mdev
);
1487 static inline void ov_oos_print(struct drbd_conf
*mdev
)
1489 if (mdev
->ov_last_oos_size
) {
1490 dev_err(DEV
, "Out of sync: start=%llu, size=%lu (sectors)\n",
1491 (unsigned long long)mdev
->ov_last_oos_start
,
1492 (unsigned long)mdev
->ov_last_oos_size
);
1494 mdev
->ov_last_oos_size
=0;
1498 extern void drbd_csum_bio(struct drbd_conf
*, struct crypto_hash
*, struct bio
*, void *);
1499 extern void drbd_csum_ee(struct drbd_conf
*, struct crypto_hash
*,
1500 struct drbd_peer_request
*, void *);
1501 /* worker callbacks */
1502 extern int w_req_cancel_conflict(struct drbd_work
*, int);
1503 extern int w_read_retry_remote(struct drbd_work
*, int);
1504 extern int w_e_end_data_req(struct drbd_work
*, int);
1505 extern int w_e_end_rsdata_req(struct drbd_work
*, int);
1506 extern int w_e_end_csum_rs_req(struct drbd_work
*, int);
1507 extern int w_e_end_ov_reply(struct drbd_work
*, int);
1508 extern int w_e_end_ov_req(struct drbd_work
*, int);
1509 extern int w_ov_finished(struct drbd_work
*, int);
1510 extern int w_resync_timer(struct drbd_work
*, int);
1511 extern int w_resume_next_sg(struct drbd_work
*, int);
1512 extern int w_send_write_hint(struct drbd_work
*, int);
1513 extern int w_send_dblock(struct drbd_work
*, int);
1514 extern int w_send_barrier(struct drbd_work
*, int);
1515 extern int w_send_read_req(struct drbd_work
*, int);
1516 extern int w_prev_work_done(struct drbd_work
*, int);
1517 extern int w_e_reissue(struct drbd_work
*, int);
1518 extern int w_restart_disk_io(struct drbd_work
*, int);
1519 extern int w_send_oos(struct drbd_work
*, int);
1520 extern int w_start_resync(struct drbd_work
*, int);
1522 extern void resync_timer_fn(unsigned long data
);
1523 extern void start_resync_timer_fn(unsigned long data
);
1525 /* drbd_receiver.c */
1526 extern int drbd_rs_should_slow_down(struct drbd_conf
*mdev
, sector_t sector
);
1527 extern int drbd_submit_ee(struct drbd_conf
*, struct drbd_peer_request
*,
1528 const unsigned, const int);
1529 extern int drbd_release_ee(struct drbd_conf
*mdev
, struct list_head
*list
);
1530 extern struct drbd_peer_request
*drbd_alloc_ee(struct drbd_conf
*,
1531 u64
, sector_t
, unsigned int,
1532 gfp_t
) __must_hold(local
);
1533 extern void drbd_free_some_ee(struct drbd_conf
*, struct drbd_peer_request
*,
1535 #define drbd_free_ee(m,e) drbd_free_some_ee(m, e, 0)
1536 #define drbd_free_net_ee(m,e) drbd_free_some_ee(m, e, 1)
1537 extern void drbd_wait_ee_list_empty(struct drbd_conf
*mdev
,
1538 struct list_head
*head
);
1539 extern void _drbd_wait_ee_list_empty(struct drbd_conf
*mdev
,
1540 struct list_head
*head
);
1541 extern void drbd_set_recv_tcq(struct drbd_conf
*mdev
, int tcq_enabled
);
1542 extern void _drbd_clear_done_ee(struct drbd_conf
*mdev
, struct list_head
*to_be_freed
);
1543 extern void drbd_flush_workqueue(struct drbd_conf
*mdev
);
1545 /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1546 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1547 static inline int drbd_setsockopt(struct socket
*sock
, int level
, int optname
,
1548 char __user
*optval
, int optlen
)
1551 if (level
== SOL_SOCKET
)
1552 err
= sock_setsockopt(sock
, level
, optname
, optval
, optlen
);
1554 err
= sock
->ops
->setsockopt(sock
, level
, optname
, optval
,
1559 static inline void drbd_tcp_cork(struct socket
*sock
)
1562 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1563 (char __user
*)&val
, sizeof(val
));
1566 static inline void drbd_tcp_uncork(struct socket
*sock
)
1569 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_CORK
,
1570 (char __user
*)&val
, sizeof(val
));
1573 static inline void drbd_tcp_nodelay(struct socket
*sock
)
1576 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_NODELAY
,
1577 (char __user
*)&val
, sizeof(val
));
1580 static inline void drbd_tcp_quickack(struct socket
*sock
)
1583 (void) drbd_setsockopt(sock
, SOL_TCP
, TCP_QUICKACK
,
1584 (char __user
*)&val
, sizeof(val
));
1587 void drbd_bump_write_ordering(struct drbd_conf
*mdev
, enum write_ordering_e wo
);
1590 extern struct proc_dir_entry
*drbd_proc
;
1591 extern const struct file_operations drbd_proc_fops
;
1592 extern const char *drbd_conn_str(enum drbd_conns s
);
1593 extern const char *drbd_role_str(enum drbd_role s
);
1596 extern void drbd_al_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1597 extern void drbd_al_complete_io(struct drbd_conf
*mdev
, sector_t sector
);
1598 extern void drbd_rs_complete_io(struct drbd_conf
*mdev
, sector_t sector
);
1599 extern int drbd_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1600 extern int drbd_try_rs_begin_io(struct drbd_conf
*mdev
, sector_t sector
);
1601 extern void drbd_rs_cancel_all(struct drbd_conf
*mdev
);
1602 extern int drbd_rs_del_all(struct drbd_conf
*mdev
);
1603 extern void drbd_rs_failed_io(struct drbd_conf
*mdev
,
1604 sector_t sector
, int size
);
1605 extern int drbd_al_read_log(struct drbd_conf
*mdev
, struct drbd_backing_dev
*);
1606 extern void drbd_advance_rs_marks(struct drbd_conf
*mdev
, unsigned long still_to_go
);
1607 extern void __drbd_set_in_sync(struct drbd_conf
*mdev
, sector_t sector
,
1608 int size
, const char *file
, const unsigned int line
);
1609 #define drbd_set_in_sync(mdev, sector, size) \
1610 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1611 extern int __drbd_set_out_of_sync(struct drbd_conf
*mdev
, sector_t sector
,
1612 int size
, const char *file
, const unsigned int line
);
1613 #define drbd_set_out_of_sync(mdev, sector, size) \
1614 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1615 extern void drbd_al_apply_to_bm(struct drbd_conf
*mdev
);
1616 extern void drbd_al_shrink(struct drbd_conf
*mdev
);
1621 void drbd_nl_cleanup(void);
1622 int __init
drbd_nl_init(void);
1623 void drbd_bcast_state(struct drbd_conf
*mdev
, union drbd_state
);
1624 void drbd_bcast_sync_progress(struct drbd_conf
*mdev
);
1625 void drbd_bcast_ee(struct drbd_conf
*, const char *, const int, const char *,
1626 const char *, const struct drbd_peer_request
*);
1630 * inline helper functions
1631 *************************/
1633 /* see also page_chain_add and friends in drbd_receiver.c */
1634 static inline struct page
*page_chain_next(struct page
*page
)
1636 return (struct page
*)page_private(page
);
1638 #define page_chain_for_each(page) \
1639 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1640 page = page_chain_next(page))
1641 #define page_chain_for_each_safe(page, n) \
1642 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1644 static inline int drbd_bio_has_active_page(struct bio
*bio
)
1646 struct bio_vec
*bvec
;
1649 __bio_for_each_segment(bvec
, bio
, i
, 0) {
1650 if (page_count(bvec
->bv_page
) > 1)
1657 static inline int drbd_ee_has_active_page(struct drbd_peer_request
*peer_req
)
1659 struct page
*page
= peer_req
->pages
;
1660 page_chain_for_each(page
) {
1661 if (page_count(page
) > 1)
1672 static inline void drbd_state_lock(struct drbd_conf
*mdev
)
1674 wait_event(mdev
->misc_wait
,
1675 !test_and_set_bit(CLUSTER_ST_CHANGE
, &mdev
->flags
));
1678 static inline void drbd_state_unlock(struct drbd_conf
*mdev
)
1680 clear_bit(CLUSTER_ST_CHANGE
, &mdev
->flags
);
1681 wake_up(&mdev
->misc_wait
);
1684 static inline enum drbd_state_rv
1685 _drbd_set_state(struct drbd_conf
*mdev
, union drbd_state ns
,
1686 enum chg_state_flags flags
, struct completion
*done
)
1688 enum drbd_state_rv rv
;
1690 read_lock(&global_state_lock
);
1691 rv
= __drbd_set_state(mdev
, ns
, flags
, done
);
1692 read_unlock(&global_state_lock
);
1697 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1698 static inline void __drbd_chk_io_error_(struct drbd_conf
*mdev
, int forcedetach
, const char *where
)
1700 switch (mdev
->ldev
->dc
.on_io_error
) {
1703 if (__ratelimit(&drbd_ratelimit_state
))
1704 dev_err(DEV
, "Local IO failed in %s.\n", where
);
1705 if (mdev
->state
.disk
> D_INCONSISTENT
)
1706 _drbd_set_state(_NS(mdev
, disk
, D_INCONSISTENT
), CS_HARD
, NULL
);
1709 /* NOTE fall through to detach case if forcedetach set */
1711 case EP_CALL_HELPER
:
1712 set_bit(WAS_IO_ERROR
, &mdev
->flags
);
1713 if (mdev
->state
.disk
> D_FAILED
) {
1714 _drbd_set_state(_NS(mdev
, disk
, D_FAILED
), CS_HARD
, NULL
);
1716 "Local IO failed in %s. Detaching...\n", where
);
1723 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1724 * @mdev: DRBD device.
1725 * @error: Error code passed to the IO completion callback
1726 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1728 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1730 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1731 static inline void drbd_chk_io_error_(struct drbd_conf
*mdev
,
1732 int error
, int forcedetach
, const char *where
)
1735 unsigned long flags
;
1736 spin_lock_irqsave(&mdev
->tconn
->req_lock
, flags
);
1737 __drbd_chk_io_error_(mdev
, forcedetach
, where
);
1738 spin_unlock_irqrestore(&mdev
->tconn
->req_lock
, flags
);
1744 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1745 * @bdev: Meta data block device.
1747 * BTW, for internal meta data, this happens to be the maximum capacity
1748 * we could agree upon with our peer node.
1750 static inline sector_t
drbd_md_first_sector(struct drbd_backing_dev
*bdev
)
1752 switch (bdev
->dc
.meta_dev_idx
) {
1753 case DRBD_MD_INDEX_INTERNAL
:
1754 case DRBD_MD_INDEX_FLEX_INT
:
1755 return bdev
->md
.md_offset
+ bdev
->md
.bm_offset
;
1756 case DRBD_MD_INDEX_FLEX_EXT
:
1758 return bdev
->md
.md_offset
;
1763 * drbd_md_last_sector() - Return the last sector number of the meta data area
1764 * @bdev: Meta data block device.
1766 static inline sector_t
drbd_md_last_sector(struct drbd_backing_dev
*bdev
)
1768 switch (bdev
->dc
.meta_dev_idx
) {
1769 case DRBD_MD_INDEX_INTERNAL
:
1770 case DRBD_MD_INDEX_FLEX_INT
:
1771 return bdev
->md
.md_offset
+ MD_AL_OFFSET
- 1;
1772 case DRBD_MD_INDEX_FLEX_EXT
:
1774 return bdev
->md
.md_offset
+ bdev
->md
.md_size_sect
;
1778 /* Returns the number of 512 byte sectors of the device */
1779 static inline sector_t
drbd_get_capacity(struct block_device
*bdev
)
1781 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1782 return bdev
? i_size_read(bdev
->bd_inode
) >> 9 : 0;
1786 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1787 * @bdev: Meta data block device.
1789 * returns the capacity we announce to out peer. we clip ourselves at the
1790 * various MAX_SECTORS, because if we don't, current implementation will
1791 * oops sooner or later
1793 static inline sector_t
drbd_get_max_capacity(struct drbd_backing_dev
*bdev
)
1796 switch (bdev
->dc
.meta_dev_idx
) {
1797 case DRBD_MD_INDEX_INTERNAL
:
1798 case DRBD_MD_INDEX_FLEX_INT
:
1799 s
= drbd_get_capacity(bdev
->backing_bdev
)
1800 ? min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1801 drbd_md_first_sector(bdev
))
1804 case DRBD_MD_INDEX_FLEX_EXT
:
1805 s
= min_t(sector_t
, DRBD_MAX_SECTORS_FLEX
,
1806 drbd_get_capacity(bdev
->backing_bdev
));
1807 /* clip at maximum size the meta device can support */
1808 s
= min_t(sector_t
, s
,
1809 BM_EXT_TO_SECT(bdev
->md
.md_size_sect
1810 - bdev
->md
.bm_offset
));
1813 s
= min_t(sector_t
, DRBD_MAX_SECTORS
,
1814 drbd_get_capacity(bdev
->backing_bdev
));
1820 * drbd_md_ss__() - Return the sector number of our meta data super block
1821 * @mdev: DRBD device.
1822 * @bdev: Meta data block device.
1824 static inline sector_t
drbd_md_ss__(struct drbd_conf
*mdev
,
1825 struct drbd_backing_dev
*bdev
)
1827 switch (bdev
->dc
.meta_dev_idx
) {
1828 default: /* external, some index */
1829 return MD_RESERVED_SECT
* bdev
->dc
.meta_dev_idx
;
1830 case DRBD_MD_INDEX_INTERNAL
:
1831 /* with drbd08, internal meta data is always "flexible" */
1832 case DRBD_MD_INDEX_FLEX_INT
:
1833 /* sizeof(struct md_on_disk_07) == 4k
1834 * position: last 4k aligned block of 4k size */
1835 if (!bdev
->backing_bdev
) {
1836 if (__ratelimit(&drbd_ratelimit_state
)) {
1837 dev_err(DEV
, "bdev->backing_bdev==NULL\n");
1842 return (drbd_get_capacity(bdev
->backing_bdev
) & ~7ULL)
1844 case DRBD_MD_INDEX_FLEX_EXT
:
1850 drbd_queue_work_front(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1852 unsigned long flags
;
1853 spin_lock_irqsave(&q
->q_lock
, flags
);
1854 list_add(&w
->list
, &q
->q
);
1855 up(&q
->s
); /* within the spinlock,
1856 see comment near end of drbd_worker() */
1857 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1861 drbd_queue_work(struct drbd_work_queue
*q
, struct drbd_work
*w
)
1863 unsigned long flags
;
1864 spin_lock_irqsave(&q
->q_lock
, flags
);
1865 list_add_tail(&w
->list
, &q
->q
);
1866 up(&q
->s
); /* within the spinlock,
1867 see comment near end of drbd_worker() */
1868 spin_unlock_irqrestore(&q
->q_lock
, flags
);
1871 static inline void wake_asender(struct drbd_tconn
*tconn
)
1873 if (test_bit(SIGNAL_ASENDER
, &tconn
->flags
))
1874 force_sig(DRBD_SIG
, tconn
->asender
.task
);
1877 static inline void request_ping(struct drbd_tconn
*tconn
)
1879 set_bit(SEND_PING
, &tconn
->flags
);
1880 wake_asender(tconn
);
1883 static inline int _drbd_send_cmd(struct drbd_conf
*mdev
, struct socket
*sock
,
1884 enum drbd_packet cmd
, struct p_header
*h
, size_t size
,
1887 return _conn_send_cmd(mdev
->tconn
, mdev
->vnr
, sock
, cmd
, h
, size
, msg_flags
);
1890 static inline int drbd_send_cmd(struct drbd_conf
*mdev
, int use_data_socket
,
1891 enum drbd_packet cmd
, struct p_header
*h
, size_t size
)
1893 return conn_send_cmd(mdev
->tconn
, mdev
->vnr
, use_data_socket
, cmd
, h
, size
);
1896 static inline int drbd_send_short_cmd(struct drbd_conf
*mdev
,
1897 enum drbd_packet cmd
)
1900 return drbd_send_cmd(mdev
, USE_DATA_SOCKET
, cmd
, &h
, sizeof(h
));
1903 static inline int drbd_send_ping(struct drbd_tconn
*tconn
)
1906 return conn_send_cmd(tconn
, 0, USE_META_SOCKET
, P_PING
, &h
, sizeof(h
));
1909 static inline int drbd_send_ping_ack(struct drbd_tconn
*tconn
)
1912 return conn_send_cmd(tconn
, 0, USE_META_SOCKET
, P_PING_ACK
, &h
, sizeof(h
));
1915 static inline void drbd_thread_stop(struct drbd_thread
*thi
)
1917 _drbd_thread_stop(thi
, false, true);
1920 static inline void drbd_thread_stop_nowait(struct drbd_thread
*thi
)
1922 _drbd_thread_stop(thi
, false, false);
1925 static inline void drbd_thread_restart_nowait(struct drbd_thread
*thi
)
1927 _drbd_thread_stop(thi
, true, false);
1930 /* counts how many answer packets packets we expect from our peer,
1931 * for either explicit application requests,
1932 * or implicit barrier packets as necessary.
1935 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
1936 * it is much easier and equally valid to count what we queue for the
1937 * worker, even before it actually was queued or send.
1938 * (drbd_make_request_common; recovery path on read io-error)
1940 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
1941 * _req_mod(req, DATA_RECEIVED)
1942 * [from receive_DataReply]
1943 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
1944 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1945 * for some reason it is NOT decreased in got_NegAck,
1946 * but in the resulting cleanup code from report_params.
1947 * we should try to remember the reason for that...
1948 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
1949 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
1950 * [from tl_clear_barrier]
1952 static inline void inc_ap_pending(struct drbd_conf
*mdev
)
1954 atomic_inc(&mdev
->ap_pending_cnt
);
1957 #define ERR_IF_CNT_IS_NEGATIVE(which) \
1958 if (atomic_read(&mdev->which) < 0) \
1959 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
1960 __func__ , __LINE__ , \
1961 atomic_read(&mdev->which))
1963 #define dec_ap_pending(mdev) do { \
1964 typecheck(struct drbd_conf *, mdev); \
1965 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \
1966 wake_up(&mdev->misc_wait); \
1967 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
1969 /* counts how many resync-related answers we still expect from the peer
1971 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
1972 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
1973 * (or P_NEG_ACK with ID_SYNCER)
1975 static inline void inc_rs_pending(struct drbd_conf
*mdev
)
1977 atomic_inc(&mdev
->rs_pending_cnt
);
1980 #define dec_rs_pending(mdev) do { \
1981 typecheck(struct drbd_conf *, mdev); \
1982 atomic_dec(&mdev->rs_pending_cnt); \
1983 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
1985 /* counts how many answers we still need to send to the peer.
1987 * receive_Data unless protocol A;
1988 * we need to send a P_RECV_ACK (proto B)
1989 * or P_WRITE_ACK (proto C)
1990 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1991 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1992 * receive_Barrier_* we need to send a P_BARRIER_ACK
1994 static inline void inc_unacked(struct drbd_conf
*mdev
)
1996 atomic_inc(&mdev
->unacked_cnt
);
1999 #define dec_unacked(mdev) do { \
2000 typecheck(struct drbd_conf *, mdev); \
2001 atomic_dec(&mdev->unacked_cnt); \
2002 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2004 #define sub_unacked(mdev, n) do { \
2005 typecheck(struct drbd_conf *, mdev); \
2006 atomic_sub(n, &mdev->unacked_cnt); \
2007 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2010 static inline void put_net_conf(struct drbd_tconn
*tconn
)
2012 if (atomic_dec_and_test(&tconn
->net_cnt
))
2013 wake_up(&tconn
->net_cnt_wait
);
2017 * get_net_conf() - Increase ref count on mdev->tconn->net_conf; Returns 0 if nothing there
2018 * @mdev: DRBD device.
2020 * You have to call put_net_conf() when finished working with mdev->tconn->net_conf.
2022 static inline int get_net_conf(struct drbd_tconn
*tconn
)
2026 atomic_inc(&tconn
->net_cnt
);
2027 have_net_conf
= tconn
->cstate
>= C_UNCONNECTED
;
2029 put_net_conf(tconn
);
2030 return have_net_conf
;
2034 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2037 * You have to call put_ldev() when finished working with mdev->ldev.
2039 #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2040 #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2042 static inline void put_ldev(struct drbd_conf
*mdev
)
2044 int i
= atomic_dec_return(&mdev
->local_cnt
);
2046 /* This may be called from some endio handler,
2047 * so we must not sleep here. */
2052 if (mdev
->state
.disk
== D_DISKLESS
)
2053 /* even internal references gone, safe to destroy */
2054 drbd_ldev_destroy(mdev
);
2055 if (mdev
->state
.disk
== D_FAILED
)
2056 /* all application IO references gone. */
2057 drbd_go_diskless(mdev
);
2058 wake_up(&mdev
->misc_wait
);
2063 static inline int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
)
2067 /* never get a reference while D_DISKLESS */
2068 if (mdev
->state
.disk
== D_DISKLESS
)
2071 atomic_inc(&mdev
->local_cnt
);
2072 io_allowed
= (mdev
->state
.disk
>= mins
);
2078 extern int _get_ldev_if_state(struct drbd_conf
*mdev
, enum drbd_disk_state mins
);
2081 /* you must have an "get_ldev" reference */
2082 static inline void drbd_get_syncer_progress(struct drbd_conf
*mdev
,
2083 unsigned long *bits_left
, unsigned int *per_mil_done
)
2085 /* this is to break it at compile time when we change that, in case we
2086 * want to support more than (1<<32) bits on a 32bit arch. */
2087 typecheck(unsigned long, mdev
->rs_total
);
2089 /* note: both rs_total and rs_left are in bits, i.e. in
2090 * units of BM_BLOCK_SIZE.
2091 * for the percentage, we don't care. */
2093 if (mdev
->state
.conn
== C_VERIFY_S
|| mdev
->state
.conn
== C_VERIFY_T
)
2094 *bits_left
= mdev
->ov_left
;
2096 *bits_left
= drbd_bm_total_weight(mdev
) - mdev
->rs_failed
;
2097 /* >> 10 to prevent overflow,
2098 * +1 to prevent division by zero */
2099 if (*bits_left
> mdev
->rs_total
) {
2100 /* doh. maybe a logic bug somewhere.
2101 * may also be just a race condition
2102 * between this and a disconnect during sync.
2103 * for now, just prevent in-kernel buffer overflow.
2106 dev_warn(DEV
, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2107 drbd_conn_str(mdev
->state
.conn
),
2108 *bits_left
, mdev
->rs_total
, mdev
->rs_failed
);
2111 /* Make sure the division happens in long context.
2112 * We allow up to one petabyte storage right now,
2113 * at a granularity of 4k per bit that is 2**38 bits.
2114 * After shift right and multiplication by 1000,
2115 * this should still fit easily into a 32bit long,
2116 * so we don't need a 64bit division on 32bit arch.
2117 * Note: currently we don't support such large bitmaps on 32bit
2118 * arch anyways, but no harm done to be prepared for it here.
2120 unsigned int shift
= mdev
->rs_total
>= (1ULL << 32) ? 16 : 10;
2121 unsigned long left
= *bits_left
>> shift
;
2122 unsigned long total
= 1UL + (mdev
->rs_total
>> shift
);
2123 unsigned long tmp
= 1000UL - left
* 1000UL/total
;
2124 *per_mil_done
= tmp
;
2129 /* this throttles on-the-fly application requests
2130 * according to max_buffers settings;
2131 * maybe re-implement using semaphores? */
2132 static inline int drbd_get_max_buffers(struct drbd_conf
*mdev
)
2134 int mxb
= 1000000; /* arbitrary limit on open requests */
2135 if (get_net_conf(mdev
->tconn
)) {
2136 mxb
= mdev
->tconn
->net_conf
->max_buffers
;
2137 put_net_conf(mdev
->tconn
);
2142 static inline int drbd_state_is_stable(struct drbd_conf
*mdev
)
2144 union drbd_state s
= mdev
->state
;
2146 /* DO NOT add a default clause, we want the compiler to warn us
2147 * for any newly introduced state we may have forgotten to add here */
2149 switch ((enum drbd_conns
)s
.conn
) {
2150 /* new io only accepted when there is no connection, ... */
2152 case C_WF_CONNECTION
:
2153 /* ... or there is a well established connection. */
2159 case C_PAUSED_SYNC_S
:
2160 case C_PAUSED_SYNC_T
:
2163 /* transitional states, IO allowed */
2164 case C_DISCONNECTING
:
2168 case C_NETWORK_FAILURE
:
2169 case C_PROTOCOL_ERROR
:
2171 case C_WF_REPORT_PARAMS
:
2172 case C_STARTING_SYNC_S
:
2173 case C_STARTING_SYNC_T
:
2176 /* Allow IO in BM exchange states with new protocols */
2178 if (mdev
->tconn
->agreed_pro_version
< 96)
2182 /* no new io accepted in these states */
2184 case C_WF_SYNC_UUID
:
2190 switch ((enum drbd_disk_state
)s
.disk
) {
2192 case D_INCONSISTENT
:
2196 /* disk state is stable as well. */
2199 /* no new io accepted during tansitional states */
2212 static inline int is_susp(union drbd_state s
)
2214 return s
.susp
|| s
.susp_nod
|| s
.susp_fen
;
2217 static inline bool may_inc_ap_bio(struct drbd_conf
*mdev
)
2219 int mxb
= drbd_get_max_buffers(mdev
);
2221 if (is_susp(mdev
->state
))
2223 if (test_bit(SUSPEND_IO
, &mdev
->flags
))
2226 /* to avoid potential deadlock or bitmap corruption,
2227 * in various places, we only allow new application io
2228 * to start during "stable" states. */
2230 /* no new io accepted when attaching or detaching the disk */
2231 if (!drbd_state_is_stable(mdev
))
2234 /* since some older kernels don't have atomic_add_unless,
2235 * and we are within the spinlock anyways, we have this workaround. */
2236 if (atomic_read(&mdev
->ap_bio_cnt
) > mxb
)
2238 if (test_bit(BITMAP_IO
, &mdev
->flags
))
2243 static inline bool inc_ap_bio_cond(struct drbd_conf
*mdev
, int count
)
2247 spin_lock_irq(&mdev
->tconn
->req_lock
);
2248 rv
= may_inc_ap_bio(mdev
);
2250 atomic_add(count
, &mdev
->ap_bio_cnt
);
2251 spin_unlock_irq(&mdev
->tconn
->req_lock
);
2256 static inline void inc_ap_bio(struct drbd_conf
*mdev
, int count
)
2259 * as long as the device is suspended
2260 * until the bitmap is no longer on the fly during connection
2261 * handshake as long as we would exeed the max_buffer limit.
2263 * to avoid races with the reconnect code,
2264 * we need to atomic_inc within the spinlock. */
2266 wait_event(mdev
->misc_wait
, inc_ap_bio_cond(mdev
, count
));
2269 static inline void dec_ap_bio(struct drbd_conf
*mdev
)
2271 int mxb
= drbd_get_max_buffers(mdev
);
2272 int ap_bio
= atomic_dec_return(&mdev
->ap_bio_cnt
);
2274 D_ASSERT(ap_bio
>= 0);
2275 /* this currently does wake_up for every dec_ap_bio!
2276 * maybe rather introduce some type of hysteresis?
2277 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2279 wake_up(&mdev
->misc_wait
);
2280 if (ap_bio
== 0 && test_bit(BITMAP_IO
, &mdev
->flags
)) {
2281 if (!test_and_set_bit(BITMAP_IO_QUEUED
, &mdev
->flags
))
2282 drbd_queue_work(&mdev
->tconn
->data
.work
, &mdev
->bm_io_work
.w
);
2286 static inline int drbd_set_ed_uuid(struct drbd_conf
*mdev
, u64 val
)
2288 int changed
= mdev
->ed_uuid
!= val
;
2289 mdev
->ed_uuid
= val
;
2293 static inline int drbd_queue_order_type(struct drbd_conf
*mdev
)
2295 /* sorry, we currently have no working implementation
2296 * of distributed TCQ stuff */
2297 #ifndef QUEUE_ORDERED_NONE
2298 #define QUEUE_ORDERED_NONE 0
2300 return QUEUE_ORDERED_NONE
;
2303 static inline void drbd_md_flush(struct drbd_conf
*mdev
)
2307 if (test_bit(MD_NO_FUA
, &mdev
->flags
))
2310 r
= blkdev_issue_flush(mdev
->ldev
->md_bdev
, GFP_KERNEL
, NULL
);
2312 set_bit(MD_NO_FUA
, &mdev
->flags
);
2313 dev_err(DEV
, "meta data flush failed with status %d, disabling md-flushes\n", r
);