1 /*
2 drbd_int.h
3
4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
5
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>.
9
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)
13 any later version.
14
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.
19
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.
23
24 */
25
26 #ifndef _DRBD_INT_H
27 #define _DRBD_INT_H
28
29 #include <crypto/hash.h>
30 #include <linux/compiler.h>
31 #include <linux/types.h>
32 #include <linux/list.h>
33 #include <linux/sched.h>
34 #include <linux/bitops.h>
35 #include <linux/slab.h>
36 #include <linux/ratelimit.h>
37 #include <linux/tcp.h>
38 #include <linux/mutex.h>
39 #include <linux/major.h>
40 #include <linux/blkdev.h>
41 #include <linux/backing-dev.h>
42 #include <linux/genhd.h>
43 #include <linux/idr.h>
44 #include <linux/dynamic_debug.h>
45 #include <net/tcp.h>
46 #include <linux/lru_cache.h>
47 #include <linux/prefetch.h>
48 #include <linux/drbd_genl_api.h>
49 #include <linux/drbd.h>
50 #include "drbd_strings.h"
51 #include "drbd_state.h"
52 #include "drbd_protocol.h"
53
54 #ifdef __CHECKER__
55 # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
56 # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
57 # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
58 # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
59 #else
60 # define __protected_by(x)
61 # define __protected_read_by(x)
62 # define __protected_write_by(x)
63 # define __must_hold(x)
64 #endif
65
66 /* module parameter, defined in drbd_main.c */
67 extern unsigned int minor_count;
68 extern bool disable_sendpage;
69 extern bool allow_oos;
70 void tl_abort_disk_io(struct drbd_device *device);
71
72 #ifdef CONFIG_DRBD_FAULT_INJECTION
73 extern int enable_faults;
74 extern int fault_rate;
75 extern int fault_devs;
76 #endif
77
78 extern char usermode_helper[];
79
80
81 /* This is used to stop/restart our threads.
82 * Cannot use SIGTERM nor SIGKILL, since these
83 * are sent out by init on runlevel changes
84 * I choose SIGHUP for now.
85 */
86 #define DRBD_SIGKILL SIGHUP
87
88 #define ID_IN_SYNC (4711ULL)
89 #define ID_OUT_OF_SYNC (4712ULL)
90 #define ID_SYNCER (-1ULL)
91
92 #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
93
94 struct drbd_device;
95 struct drbd_connection;
96
97 #define __drbd_printk_device(level, device, fmt, args...) \
98 dev_printk(level, disk_to_dev((device)->vdisk), fmt, ## args)
99 #define __drbd_printk_peer_device(level, peer_device, fmt, args...) \
100 dev_printk(level, disk_to_dev((peer_device)->device->vdisk), fmt, ## args)
101 #define __drbd_printk_resource(level, resource, fmt, args...) \
102 printk(level "drbd %s: " fmt, (resource)->name, ## args)
103 #define __drbd_printk_connection(level, connection, fmt, args...) \
104 printk(level "drbd %s: " fmt, (connection)->resource->name, ## args)
105
106 void drbd_printk_with_wrong_object_type(void);
107
108 #define __drbd_printk_if_same_type(obj, type, func, level, fmt, args...) \
109 (__builtin_types_compatible_p(typeof(obj), type) || \
110 __builtin_types_compatible_p(typeof(obj), const type)), \
111 func(level, (const type)(obj), fmt, ## args)
112
113 #define drbd_printk(level, obj, fmt, args...) \
114 __builtin_choose_expr( \
115 __drbd_printk_if_same_type(obj, struct drbd_device *, \
116 __drbd_printk_device, level, fmt, ## args), \
117 __builtin_choose_expr( \
118 __drbd_printk_if_same_type(obj, struct drbd_resource *, \
119 __drbd_printk_resource, level, fmt, ## args), \
120 __builtin_choose_expr( \
121 __drbd_printk_if_same_type(obj, struct drbd_connection *, \
122 __drbd_printk_connection, level, fmt, ## args), \
123 __builtin_choose_expr( \
124 __drbd_printk_if_same_type(obj, struct drbd_peer_device *, \
125 __drbd_printk_peer_device, level, fmt, ## args), \
126 drbd_printk_with_wrong_object_type()))))
127
128 #define drbd_dbg(obj, fmt, args...) \
129 drbd_printk(KERN_DEBUG, obj, fmt, ## args)
130 #define drbd_alert(obj, fmt, args...) \
131 drbd_printk(KERN_ALERT, obj, fmt, ## args)
132 #define drbd_err(obj, fmt, args...) \
133 drbd_printk(KERN_ERR, obj, fmt, ## args)
134 #define drbd_warn(obj, fmt, args...) \
135 drbd_printk(KERN_WARNING, obj, fmt, ## args)
136 #define drbd_info(obj, fmt, args...) \
137 drbd_printk(KERN_INFO, obj, fmt, ## args)
138 #define drbd_emerg(obj, fmt, args...) \
139 drbd_printk(KERN_EMERG, obj, fmt, ## args)
140
141 #define dynamic_drbd_dbg(device, fmt, args...) \
142 dynamic_dev_dbg(disk_to_dev(device->vdisk), fmt, ## args)
143
144 #define D_ASSERT(device, exp) do { \
145 if (!(exp)) \
146 drbd_err(device, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__); \
147 } while (0)
148
149 /**
150 * expect - Make an assertion
151 *
152 * Unlike the assert macro, this macro returns a boolean result.
153 */
154 #define expect(exp) ({ \
155 bool _bool = (exp); \
156 if (!_bool) \
157 drbd_err(device, "ASSERTION %s FAILED in %s\n", \
158 #exp, __func__); \
159 _bool; \
160 })
161
162 /* Defines to control fault insertion */
163 enum {
164 DRBD_FAULT_MD_WR = 0, /* meta data write */
165 DRBD_FAULT_MD_RD = 1, /* read */
166 DRBD_FAULT_RS_WR = 2, /* resync */
167 DRBD_FAULT_RS_RD = 3,
168 DRBD_FAULT_DT_WR = 4, /* data */
169 DRBD_FAULT_DT_RD = 5,
170 DRBD_FAULT_DT_RA = 6, /* data read ahead */
171 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */
172 DRBD_FAULT_AL_EE = 8, /* alloc ee */
173 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
174
175 DRBD_FAULT_MAX,
176 };
177
178 extern unsigned int
179 _drbd_insert_fault(struct drbd_device *device, unsigned int type);
180
181 static inline int
drbd_insert_fault(struct drbd_device * device,unsigned int type)182 drbd_insert_fault(struct drbd_device *device, unsigned int type) {
183 #ifdef CONFIG_DRBD_FAULT_INJECTION
184 return fault_rate &&
185 (enable_faults & (1<<type)) &&
186 _drbd_insert_fault(device, type);
187 #else
188 return 0;
189 #endif
190 }
191
192 /* integer division, round _UP_ to the next integer */
193 #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
194 /* usual integer division */
195 #define div_floor(A, B) ((A)/(B))
196
197 extern struct ratelimit_state drbd_ratelimit_state;
198 extern struct idr drbd_devices; /* RCU, updates: genl_lock() */
199 extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */
200
201 extern const char *cmdname(enum drbd_packet cmd);
202
203 /* for sending/receiving the bitmap,
204 * possibly in some encoding scheme */
205 struct bm_xfer_ctx {
206 /* "const"
207 * stores total bits and long words
208 * of the bitmap, so we don't need to
209 * call the accessor functions over and again. */
210 unsigned long bm_bits;
211 unsigned long bm_words;
212 /* during xfer, current position within the bitmap */
213 unsigned long bit_offset;
214 unsigned long word_offset;
215
216 /* statistics; index: (h->command == P_BITMAP) */
217 unsigned packets[2];
218 unsigned bytes[2];
219 };
220
221 extern void INFO_bm_xfer_stats(struct drbd_device *device,
222 const char *direction, struct bm_xfer_ctx *c);
223
bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx * c)224 static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
225 {
226 /* word_offset counts "native long words" (32 or 64 bit),
227 * aligned at 64 bit.
228 * Encoded packet may end at an unaligned bit offset.
229 * In case a fallback clear text packet is transmitted in
230 * between, we adjust this offset back to the last 64bit
231 * aligned "native long word", which makes coding and decoding
232 * the plain text bitmap much more convenient. */
233 #if BITS_PER_LONG == 64
234 c->word_offset = c->bit_offset >> 6;
235 #elif BITS_PER_LONG == 32
236 c->word_offset = c->bit_offset >> 5;
237 c->word_offset &= ~(1UL);
238 #else
239 # error "unsupported BITS_PER_LONG"
240 #endif
241 }
242
243 extern unsigned int drbd_header_size(struct drbd_connection *connection);
244
245 /**********************************************************************/
246 enum drbd_thread_state {
247 NONE,
248 RUNNING,
249 EXITING,
250 RESTARTING
251 };
252
253 struct drbd_thread {
254 spinlock_t t_lock;
255 struct task_struct *task;
256 struct completion stop;
257 enum drbd_thread_state t_state;
258 int (*function) (struct drbd_thread *);
259 struct drbd_resource *resource;
260 struct drbd_connection *connection;
261 int reset_cpu_mask;
262 const char *name;
263 };
264
get_t_state(struct drbd_thread * thi)265 static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
266 {
267 /* THINK testing the t_state seems to be uncritical in all cases
268 * (but thread_{start,stop}), so we can read it *without* the lock.
269 * --lge */
270
271 smp_rmb();
272 return thi->t_state;
273 }
274
275 struct drbd_work {
276 struct list_head list;
277 int (*cb)(struct drbd_work *, int cancel);
278 };
279
280 struct drbd_device_work {
281 struct drbd_work w;
282 struct drbd_device *device;
283 };
284
285 #include "drbd_interval.h"
286
287 extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *);
288
289 extern void lock_all_resources(void);
290 extern void unlock_all_resources(void);
291
292 struct drbd_request {
293 struct drbd_work w;
294 struct drbd_device *device;
295
296 /* if local IO is not allowed, will be NULL.
297 * if local IO _is_ allowed, holds the locally submitted bio clone,
298 * or, after local IO completion, the ERR_PTR(error).
299 * see drbd_request_endio(). */
300 struct bio *private_bio;
301
302 struct drbd_interval i;
303
304 /* epoch: used to check on "completion" whether this req was in
305 * the current epoch, and we therefore have to close it,
306 * causing a p_barrier packet to be send, starting a new epoch.
307 *
308 * This corresponds to "barrier" in struct p_barrier[_ack],
309 * and to "barrier_nr" in struct drbd_epoch (and various
310 * comments/function parameters/local variable names).
311 */
312 unsigned int epoch;
313
314 struct list_head tl_requests; /* ring list in the transfer log */
315 struct bio *master_bio; /* master bio pointer */
316
317 /* see struct drbd_device */
318 struct list_head req_pending_master_completion;
319 struct list_head req_pending_local;
320
321 /* for generic IO accounting */
322 unsigned long start_jif;
323
324 /* for DRBD internal statistics */
325
326 /* Minimal set of time stamps to determine if we wait for activity log
327 * transactions, local disk or peer. 32 bit "jiffies" are good enough,
328 * we don't expect a DRBD request to be stalled for several month.
329 */
330
331 /* before actual request processing */
332 unsigned long in_actlog_jif;
333
334 /* local disk */
335 unsigned long pre_submit_jif;
336
337 /* per connection */
338 unsigned long pre_send_jif;
339 unsigned long acked_jif;
340 unsigned long net_done_jif;
341
342 /* Possibly even more detail to track each phase:
343 * master_completion_jif
344 * how long did it take to complete the master bio
345 * (application visible latency)
346 * allocated_jif
347 * how long the master bio was blocked until we finally allocated
348 * a tracking struct
349 * in_actlog_jif
350 * how long did we wait for activity log transactions
351 *
352 * net_queued_jif
353 * when did we finally queue it for sending
354 * pre_send_jif
355 * when did we start sending it
356 * post_send_jif
357 * how long did we block in the network stack trying to send it
358 * acked_jif
359 * when did we receive (or fake, in protocol A) a remote ACK
360 * net_done_jif
361 * when did we receive final acknowledgement (P_BARRIER_ACK),
362 * or decide, e.g. on connection loss, that we do no longer expect
363 * anything from this peer for this request.
364 *
365 * pre_submit_jif
366 * post_sub_jif
367 * when did we start submiting to the lower level device,
368 * and how long did we block in that submit function
369 * local_completion_jif
370 * how long did it take the lower level device to complete this request
371 */
372
373
374 /* once it hits 0, we may complete the master_bio */
375 atomic_t completion_ref;
376 /* once it hits 0, we may destroy this drbd_request object */
377 struct kref kref;
378
379 unsigned rq_state; /* see comments above _req_mod() */
380 };
381
382 struct drbd_epoch {
383 struct drbd_connection *connection;
384 struct list_head list;
385 unsigned int barrier_nr;
386 atomic_t epoch_size; /* increased on every request added. */
387 atomic_t active; /* increased on every req. added, and dec on every finished. */
388 unsigned long flags;
389 };
390
391 /* Prototype declaration of function defined in drbd_receiver.c */
392 int drbdd_init(struct drbd_thread *);
393 int drbd_asender(struct drbd_thread *);
394
395 /* drbd_epoch flag bits */
396 enum {
397 DE_HAVE_BARRIER_NUMBER,
398 };
399
400 enum epoch_event {
401 EV_PUT,
402 EV_GOT_BARRIER_NR,
403 EV_BECAME_LAST,
404 EV_CLEANUP = 32, /* used as flag */
405 };
406
407 struct digest_info {
408 int digest_size;
409 void *digest;
410 };
411
412 struct drbd_peer_request {
413 struct drbd_work w;
414 struct drbd_peer_device *peer_device;
415 struct drbd_epoch *epoch; /* for writes */
416 struct page *pages;
417 atomic_t pending_bios;
418 struct drbd_interval i;
419 /* see comments on ee flag bits below */
420 unsigned long flags;
421 unsigned long submit_jif;
422 union {
423 u64 block_id;
424 struct digest_info *digest;
425 };
426 };
427
428 /* ee flag bits.
429 * While corresponding bios are in flight, the only modification will be
430 * set_bit WAS_ERROR, which has to be atomic.
431 * If no bios are in flight yet, or all have been completed,
432 * non-atomic modification to ee->flags is ok.
433 */
434 enum {
435 __EE_CALL_AL_COMPLETE_IO,
436 __EE_MAY_SET_IN_SYNC,
437
438 /* is this a TRIM aka REQ_DISCARD? */
439 __EE_IS_TRIM,
440 /* our lower level cannot handle trim,
441 * and we want to fall back to zeroout instead */
442 __EE_IS_TRIM_USE_ZEROOUT,
443
444 /* In case a barrier failed,
445 * we need to resubmit without the barrier flag. */
446 __EE_RESUBMITTED,
447
448 /* we may have several bios per peer request.
449 * if any of those fail, we set this flag atomically
450 * from the endio callback */
451 __EE_WAS_ERROR,
452
453 /* This ee has a pointer to a digest instead of a block id */
454 __EE_HAS_DIGEST,
455
456 /* Conflicting local requests need to be restarted after this request */
457 __EE_RESTART_REQUESTS,
458
459 /* The peer wants a write ACK for this (wire proto C) */
460 __EE_SEND_WRITE_ACK,
461
462 /* Is set when net_conf had two_primaries set while creating this peer_req */
463 __EE_IN_INTERVAL_TREE,
464
465 /* for debugfs: */
466 /* has this been submitted, or does it still wait for something else? */
467 __EE_SUBMITTED,
468
469 /* this is/was a write request */
470 __EE_WRITE,
471
472 /* this is/was a write same request */
473 __EE_WRITE_SAME,
474
475 /* this originates from application on peer
476 * (not some resync or verify or other DRBD internal request) */
477 __EE_APPLICATION,
478
479 /* If it contains only 0 bytes, send back P_RS_DEALLOCATED */
480 __EE_RS_THIN_REQ,
481 };
482 #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
483 #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
484 #define EE_IS_TRIM (1<<__EE_IS_TRIM)
485 #define EE_IS_TRIM_USE_ZEROOUT (1<<__EE_IS_TRIM_USE_ZEROOUT)
486 #define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
487 #define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
488 #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
489 #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
490 #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK)
491 #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE)
492 #define EE_SUBMITTED (1<<__EE_SUBMITTED)
493 #define EE_WRITE (1<<__EE_WRITE)
494 #define EE_WRITE_SAME (1<<__EE_WRITE_SAME)
495 #define EE_APPLICATION (1<<__EE_APPLICATION)
496 #define EE_RS_THIN_REQ (1<<__EE_RS_THIN_REQ)
497
498 /* flag bits per device */
499 enum {
500 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
501 MD_DIRTY, /* current uuids and flags not yet on disk */
502 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
503 CL_ST_CHG_SUCCESS,
504 CL_ST_CHG_FAIL,
505 CRASHED_PRIMARY, /* This node was a crashed primary.
506 * Gets cleared when the state.conn
507 * goes into C_CONNECTED state. */
508 CONSIDER_RESYNC,
509
510 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */
511
512 BITMAP_IO, /* suspend application io;
513 once no more io in flight, start bitmap io */
514 BITMAP_IO_QUEUED, /* Started bitmap IO */
515 WAS_IO_ERROR, /* Local disk failed, returned IO error */
516 WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */
517 FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */
518 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
519 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
520 * the peer, if it changed there as well. */
521 NEW_CUR_UUID, /* Create new current UUID when thawing IO */
522 AL_SUSPENDED, /* Activity logging is currently suspended. */
523 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */
524 B_RS_H_DONE, /* Before resync handler done (already executed) */
525 DISCARD_MY_DATA, /* discard_my_data flag per volume */
526 READ_BALANCE_RR,
527
528 FLUSH_PENDING, /* if set, device->flush_jif is when we submitted that flush
529 * from drbd_flush_after_epoch() */
530
531 /* cleared only after backing device related structures have been destroyed. */
532 GOING_DISKLESS, /* Disk is being detached, because of io-error, or admin request. */
533
534 /* to be used in drbd_device_post_work() */
535 GO_DISKLESS, /* tell worker to schedule cleanup before detach */
536 DESTROY_DISK, /* tell worker to close backing devices and destroy related structures. */
537 MD_SYNC, /* tell worker to call drbd_md_sync() */
538 RS_START, /* tell worker to start resync/OV */
539 RS_PROGRESS, /* tell worker that resync made significant progress */
540 RS_DONE, /* tell worker that resync is done */
541 };
542
543 struct drbd_bitmap; /* opaque for drbd_device */
544
545 /* definition of bits in bm_flags to be used in drbd_bm_lock
546 * and drbd_bitmap_io and friends. */
547 enum bm_flag {
548 /* currently locked for bulk operation */
549 BM_LOCKED_MASK = 0xf,
550
551 /* in detail, that is: */
552 BM_DONT_CLEAR = 0x1,
553 BM_DONT_SET = 0x2,
554 BM_DONT_TEST = 0x4,
555
556 /* so we can mark it locked for bulk operation,
557 * and still allow all non-bulk operations */
558 BM_IS_LOCKED = 0x8,
559
560 /* (test bit, count bit) allowed (common case) */
561 BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED,
562
563 /* testing bits, as well as setting new bits allowed, but clearing bits
564 * would be unexpected. Used during bitmap receive. Setting new bits
565 * requires sending of "out-of-sync" information, though. */
566 BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED,
567
568 /* for drbd_bm_write_copy_pages, everything is allowed,
569 * only concurrent bulk operations are locked out. */
570 BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED,
571 };
572
573 struct drbd_work_queue {
574 struct list_head q;
575 spinlock_t q_lock; /* to protect the list. */
576 wait_queue_head_t q_wait;
577 };
578
579 struct drbd_socket {
580 struct mutex mutex;
581 struct socket *socket;
582 /* this way we get our
583 * send/receive buffers off the stack */
584 void *sbuf;
585 void *rbuf;
586 };
587
588 struct drbd_md {
589 u64 md_offset; /* sector offset to 'super' block */
590
591 u64 la_size_sect; /* last agreed size, unit sectors */
592 spinlock_t uuid_lock;
593 u64 uuid[UI_SIZE];
594 u64 device_uuid;
595 u32 flags;
596 u32 md_size_sect;
597
598 s32 al_offset; /* signed relative sector offset to activity log */
599 s32 bm_offset; /* signed relative sector offset to bitmap */
600
601 /* cached value of bdev->disk_conf->meta_dev_idx (see below) */
602 s32 meta_dev_idx;
603
604 /* see al_tr_number_to_on_disk_sector() */
605 u32 al_stripes;
606 u32 al_stripe_size_4k;
607 u32 al_size_4k; /* cached product of the above */
608 };
609
610 struct drbd_backing_dev {
611 struct block_device *backing_bdev;
612 struct block_device *md_bdev;
613 struct drbd_md md;
614 struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */
615 sector_t known_size; /* last known size of that backing device */
616 };
617
618 struct drbd_md_io {
619 struct page *page;
620 unsigned long start_jif; /* last call to drbd_md_get_buffer */
621 unsigned long submit_jif; /* last _drbd_md_sync_page_io() submit */
622 const char *current_use;
623 atomic_t in_use;
624 unsigned int done;
625 int error;
626 };
627
628 struct bm_io_work {
629 struct drbd_work w;
630 char *why;
631 enum bm_flag flags;
632 int (*io_fn)(struct drbd_device *device);
633 void (*done)(struct drbd_device *device, int rv);
634 };
635
636 struct fifo_buffer {
637 unsigned int head_index;
638 unsigned int size;
639 int total; /* sum of all values */
640 int values[0];
641 };
642 extern struct fifo_buffer *fifo_alloc(int fifo_size);
643
644 /* flag bits per connection */
645 enum {
646 NET_CONGESTED, /* The data socket is congested */
647 RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */
648 SEND_PING,
649 GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */
650 CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */
651 CONN_WD_ST_CHG_OKAY,
652 CONN_WD_ST_CHG_FAIL,
653 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
654 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */
655 STATE_SENT, /* Do not change state/UUIDs while this is set */
656 CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
657 * pending, from drbd worker context.
658 * If set, bdi_write_congested() returns true,
659 * so shrink_page_list() would not recurse into,
660 * and potentially deadlock on, this drbd worker.
661 */
662 DISCONNECT_SENT,
663
664 DEVICE_WORK_PENDING, /* tell worker that some device has pending work */
665 };
666
667 enum which_state { NOW, OLD = NOW, NEW };
668
669 struct drbd_resource {
670 char *name;
671 #ifdef CONFIG_DEBUG_FS
672 struct dentry *debugfs_res;
673 struct dentry *debugfs_res_volumes;
674 struct dentry *debugfs_res_connections;
675 struct dentry *debugfs_res_in_flight_summary;
676 #endif
677 struct kref kref;
678 struct idr devices; /* volume number to device mapping */
679 struct list_head connections;
680 struct list_head resources;
681 struct res_opts res_opts;
682 struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */
683 struct mutex adm_mutex; /* mutex to serialize administrative requests */
684 spinlock_t req_lock;
685
686 unsigned susp:1; /* IO suspended by user */
687 unsigned susp_nod:1; /* IO suspended because no data */
688 unsigned susp_fen:1; /* IO suspended because fence peer handler runs */
689
690 enum write_ordering_e write_ordering;
691
692 cpumask_var_t cpu_mask;
693 };
694
695 struct drbd_thread_timing_details
696 {
697 unsigned long start_jif;
698 void *cb_addr;
699 const char *caller_fn;
700 unsigned int line;
701 unsigned int cb_nr;
702 };
703
704 struct drbd_connection {
705 struct list_head connections;
706 struct drbd_resource *resource;
707 #ifdef CONFIG_DEBUG_FS
708 struct dentry *debugfs_conn;
709 struct dentry *debugfs_conn_callback_history;
710 struct dentry *debugfs_conn_oldest_requests;
711 #endif
712 struct kref kref;
713 struct idr peer_devices; /* volume number to peer device mapping */
714 enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
715 struct mutex cstate_mutex; /* Protects graceful disconnects */
716 unsigned int connect_cnt; /* Inc each time a connection is established */
717
718 unsigned long flags;
719 struct net_conf *net_conf; /* content protected by rcu */
720 wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */
721
722 struct sockaddr_storage my_addr;
723 int my_addr_len;
724 struct sockaddr_storage peer_addr;
725 int peer_addr_len;
726
727 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
728 struct drbd_socket meta; /* ping/ack (metadata) packets */
729 int agreed_pro_version; /* actually used protocol version */
730 u32 agreed_features;
731 unsigned long last_received; /* in jiffies, either socket */
732 unsigned int ko_count;
733
734 struct list_head transfer_log; /* all requests not yet fully processed */
735
736 struct crypto_shash *cram_hmac_tfm;
737 struct crypto_ahash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */
738 struct crypto_ahash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */
739 struct crypto_ahash *csums_tfm;
740 struct crypto_ahash *verify_tfm;
741 void *int_dig_in;
742 void *int_dig_vv;
743
744 /* receiver side */
745 struct drbd_epoch *current_epoch;
746 spinlock_t epoch_lock;
747 unsigned int epochs;
748 atomic_t current_tle_nr; /* transfer log epoch number */
749 unsigned current_tle_writes; /* writes seen within this tl epoch */
750
751 unsigned long last_reconnect_jif;
752 struct drbd_thread receiver;
753 struct drbd_thread worker;
754 struct drbd_thread ack_receiver;
755 struct workqueue_struct *ack_sender;
756
757 /* cached pointers,
758 * so we can look up the oldest pending requests more quickly.
759 * protected by resource->req_lock */
760 struct drbd_request *req_next; /* DRBD 9: todo.req_next */
761 struct drbd_request *req_ack_pending;
762 struct drbd_request *req_not_net_done;
763
764 /* sender side */
765 struct drbd_work_queue sender_work;
766
767 #define DRBD_THREAD_DETAILS_HIST 16
768 unsigned int w_cb_nr; /* keeps counting up */
769 unsigned int r_cb_nr; /* keeps counting up */
770 struct drbd_thread_timing_details w_timing_details[DRBD_THREAD_DETAILS_HIST];
771 struct drbd_thread_timing_details r_timing_details[DRBD_THREAD_DETAILS_HIST];
772
773 struct {
774 unsigned long last_sent_barrier_jif;
775
776 /* whether this sender thread
777 * has processed a single write yet. */
778 bool seen_any_write_yet;
779
780 /* Which barrier number to send with the next P_BARRIER */
781 int current_epoch_nr;
782
783 /* how many write requests have been sent
784 * with req->epoch == current_epoch_nr.
785 * If none, no P_BARRIER will be sent. */
786 unsigned current_epoch_writes;
787 } send;
788 };
789
has_net_conf(struct drbd_connection * connection)790 static inline bool has_net_conf(struct drbd_connection *connection)
791 {
792 bool has_net_conf;
793
794 rcu_read_lock();
795 has_net_conf = rcu_dereference(connection->net_conf);
796 rcu_read_unlock();
797
798 return has_net_conf;
799 }
800
801 void __update_timing_details(
802 struct drbd_thread_timing_details *tdp,
803 unsigned int *cb_nr,
804 void *cb,
805 const char *fn, const unsigned int line);
806
807 #define update_worker_timing_details(c, cb) \
808 __update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ )
809 #define update_receiver_timing_details(c, cb) \
810 __update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ )
811
812 struct submit_worker {
813 struct workqueue_struct *wq;
814 struct work_struct worker;
815
816 /* protected by ..->resource->req_lock */
817 struct list_head writes;
818 };
819
820 struct drbd_peer_device {
821 struct list_head peer_devices;
822 struct drbd_device *device;
823 struct drbd_connection *connection;
824 struct work_struct send_acks_work;
825 #ifdef CONFIG_DEBUG_FS
826 struct dentry *debugfs_peer_dev;
827 #endif
828 };
829
830 struct drbd_device {
831 struct drbd_resource *resource;
832 struct list_head peer_devices;
833 struct list_head pending_bitmap_io;
834
835 unsigned long flush_jif;
836 #ifdef CONFIG_DEBUG_FS
837 struct dentry *debugfs_minor;
838 struct dentry *debugfs_vol;
839 struct dentry *debugfs_vol_oldest_requests;
840 struct dentry *debugfs_vol_act_log_extents;
841 struct dentry *debugfs_vol_resync_extents;
842 struct dentry *debugfs_vol_data_gen_id;
843 struct dentry *debugfs_vol_ed_gen_id;
844 #endif
845
846 unsigned int vnr; /* volume number within the connection */
847 unsigned int minor; /* device minor number */
848
849 struct kref kref;
850
851 /* things that are stored as / read from meta data on disk */
852 unsigned long flags;
853
854 /* configured by drbdsetup */
855 struct drbd_backing_dev *ldev __protected_by(local);
856
857 sector_t p_size; /* partner's disk size */
858 struct request_queue *rq_queue;
859 struct block_device *this_bdev;
860 struct gendisk *vdisk;
861
862 unsigned long last_reattach_jif;
863 struct drbd_work resync_work;
864 struct drbd_work unplug_work;
865 struct timer_list resync_timer;
866 struct timer_list md_sync_timer;
867 struct timer_list start_resync_timer;
868 struct timer_list request_timer;
869
870 /* Used after attach while negotiating new disk state. */
871 union drbd_state new_state_tmp;
872
873 union drbd_dev_state state;
874 wait_queue_head_t misc_wait;
875 wait_queue_head_t state_wait; /* upon each state change. */
876 unsigned int send_cnt;
877 unsigned int recv_cnt;
878 unsigned int read_cnt;
879 unsigned int writ_cnt;
880 unsigned int al_writ_cnt;
881 unsigned int bm_writ_cnt;
882 atomic_t ap_bio_cnt; /* Requests we need to complete */
883 atomic_t ap_actlog_cnt; /* Requests waiting for activity log */
884 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
885 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
886 atomic_t unacked_cnt; /* Need to send replies for */
887 atomic_t local_cnt; /* Waiting for local completion */
888 atomic_t suspend_cnt;
889
890 /* Interval tree of pending local requests */
891 struct rb_root read_requests;
892 struct rb_root write_requests;
893
894 /* for statistics and timeouts */
895 /* [0] read, [1] write */
896 struct list_head pending_master_completion[2];
897 struct list_head pending_completion[2];
898
899 /* use checksums for *this* resync */
900 bool use_csums;
901 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
902 unsigned long rs_total;
903 /* number of resync blocks that failed in this run */
904 unsigned long rs_failed;
905 /* Syncer's start time [unit jiffies] */
906 unsigned long rs_start;
907 /* cumulated time in PausedSyncX state [unit jiffies] */
908 unsigned long rs_paused;
909 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
910 unsigned long rs_same_csum;
911 #define DRBD_SYNC_MARKS 8
912 #define DRBD_SYNC_MARK_STEP (3*HZ)
913 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
914 unsigned long rs_mark_left[DRBD_SYNC_MARKS];
915 /* marks's time [unit jiffies] */
916 unsigned long rs_mark_time[DRBD_SYNC_MARKS];
917 /* current index into rs_mark_{left,time} */
918 int rs_last_mark;
919 unsigned long rs_last_bcast; /* [unit jiffies] */
920
921 /* where does the admin want us to start? (sector) */
922 sector_t ov_start_sector;
923 sector_t ov_stop_sector;
924 /* where are we now? (sector) */
925 sector_t ov_position;
926 /* Start sector of out of sync range (to merge printk reporting). */
927 sector_t ov_last_oos_start;
928 /* size of out-of-sync range in sectors. */
929 sector_t ov_last_oos_size;
930 unsigned long ov_left; /* in bits */
931
932 struct drbd_bitmap *bitmap;
933 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
934
935 /* Used to track operations of resync... */
936 struct lru_cache *resync;
937 /* Number of locked elements in resync LRU */
938 unsigned int resync_locked;
939 /* resync extent number waiting for application requests */
940 unsigned int resync_wenr;
941
942 int open_cnt;
943 u64 *p_uuid;
944
945 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
946 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
947 struct list_head done_ee; /* need to send P_WRITE_ACK */
948 struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */
949 struct list_head net_ee; /* zero-copy network send in progress */
950
951 int next_barrier_nr;
952 struct list_head resync_reads;
953 atomic_t pp_in_use; /* allocated from page pool */
954 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */
955 wait_queue_head_t ee_wait;
956 struct drbd_md_io md_io;
957 spinlock_t al_lock;
958 wait_queue_head_t al_wait;
959 struct lru_cache *act_log; /* activity log */
960 unsigned int al_tr_number;
961 int al_tr_cycle;
962 wait_queue_head_t seq_wait;
963 atomic_t packet_seq;
964 unsigned int peer_seq;
965 spinlock_t peer_seq_lock;
966 unsigned long comm_bm_set; /* communicated number of set bits. */
967 struct bm_io_work bm_io_work;
968 u64 ed_uuid; /* UUID of the exposed data */
969 struct mutex own_state_mutex;
970 struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */
971 char congestion_reason; /* Why we where congested... */
972 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
973 atomic_t rs_sect_ev; /* for submitted resync data rate, both */
974 int rs_last_sect_ev; /* counter to compare with */
975 int rs_last_events; /* counter of read or write "events" (unit sectors)
976 * on the lower level device when we last looked. */
977 int c_sync_rate; /* current resync rate after syncer throttle magic */
978 struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */
979 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
980 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
981 unsigned int peer_max_bio_size;
982 unsigned int local_max_bio_size;
983
984 /* any requests that would block in drbd_make_request()
985 * are deferred to this single-threaded work queue */
986 struct submit_worker submit;
987 };
988
989 struct drbd_bm_aio_ctx {
990 struct drbd_device *device;
991 struct list_head list; /* on device->pending_bitmap_io */;
992 unsigned long start_jif;
993 atomic_t in_flight;
994 unsigned int done;
995 unsigned flags;
996 #define BM_AIO_COPY_PAGES 1
997 #define BM_AIO_WRITE_HINTED 2
998 #define BM_AIO_WRITE_ALL_PAGES 4
999 #define BM_AIO_READ 8
1000 int error;
1001 struct kref kref;
1002 };
1003
1004 struct drbd_config_context {
1005 /* assigned from drbd_genlmsghdr */
1006 unsigned int minor;
1007 /* assigned from request attributes, if present */
1008 unsigned int volume;
1009 #define VOLUME_UNSPECIFIED (-1U)
1010 /* pointer into the request skb,
1011 * limited lifetime! */
1012 char *resource_name;
1013 struct nlattr *my_addr;
1014 struct nlattr *peer_addr;
1015
1016 /* reply buffer */
1017 struct sk_buff *reply_skb;
1018 /* pointer into reply buffer */
1019 struct drbd_genlmsghdr *reply_dh;
1020 /* resolved from attributes, if possible */
1021 struct drbd_device *device;
1022 struct drbd_resource *resource;
1023 struct drbd_connection *connection;
1024 };
1025
minor_to_device(unsigned int minor)1026 static inline struct drbd_device *minor_to_device(unsigned int minor)
1027 {
1028 return (struct drbd_device *)idr_find(&drbd_devices, minor);
1029 }
1030
first_peer_device(struct drbd_device * device)1031 static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device)
1032 {
1033 return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices);
1034 }
1035
1036 static inline struct drbd_peer_device *
conn_peer_device(struct drbd_connection * connection,int volume_number)1037 conn_peer_device(struct drbd_connection *connection, int volume_number)
1038 {
1039 return idr_find(&connection->peer_devices, volume_number);
1040 }
1041
1042 #define for_each_resource(resource, _resources) \
1043 list_for_each_entry(resource, _resources, resources)
1044
1045 #define for_each_resource_rcu(resource, _resources) \
1046 list_for_each_entry_rcu(resource, _resources, resources)
1047
1048 #define for_each_resource_safe(resource, tmp, _resources) \
1049 list_for_each_entry_safe(resource, tmp, _resources, resources)
1050
1051 #define for_each_connection(connection, resource) \
1052 list_for_each_entry(connection, &resource->connections, connections)
1053
1054 #define for_each_connection_rcu(connection, resource) \
1055 list_for_each_entry_rcu(connection, &resource->connections, connections)
1056
1057 #define for_each_connection_safe(connection, tmp, resource) \
1058 list_for_each_entry_safe(connection, tmp, &resource->connections, connections)
1059
1060 #define for_each_peer_device(peer_device, device) \
1061 list_for_each_entry(peer_device, &device->peer_devices, peer_devices)
1062
1063 #define for_each_peer_device_rcu(peer_device, device) \
1064 list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices)
1065
1066 #define for_each_peer_device_safe(peer_device, tmp, device) \
1067 list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices)
1068
device_to_minor(struct drbd_device * device)1069 static inline unsigned int device_to_minor(struct drbd_device *device)
1070 {
1071 return device->minor;
1072 }
1073
1074 /*
1075 * function declarations
1076 *************************/
1077
1078 /* drbd_main.c */
1079
1080 enum dds_flags {
1081 DDSF_FORCED = 1,
1082 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1083 };
1084
1085 extern void drbd_init_set_defaults(struct drbd_device *device);
1086 extern int drbd_thread_start(struct drbd_thread *thi);
1087 extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1088 #ifdef CONFIG_SMP
1089 extern void drbd_thread_current_set_cpu(struct drbd_thread *thi);
1090 #else
1091 #define drbd_thread_current_set_cpu(A) ({})
1092 #endif
1093 extern void tl_release(struct drbd_connection *, unsigned int barrier_nr,
1094 unsigned int set_size);
1095 extern void tl_clear(struct drbd_connection *);
1096 extern void drbd_free_sock(struct drbd_connection *connection);
1097 extern int drbd_send(struct drbd_connection *connection, struct socket *sock,
1098 void *buf, size_t size, unsigned msg_flags);
1099 extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t,
1100 unsigned);
1101
1102 extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd);
1103 extern int drbd_send_protocol(struct drbd_connection *connection);
1104 extern int drbd_send_uuids(struct drbd_peer_device *);
1105 extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *);
1106 extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *);
1107 extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags);
1108 extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s);
1109 extern int drbd_send_current_state(struct drbd_peer_device *);
1110 extern int drbd_send_sync_param(struct drbd_peer_device *);
1111 extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr,
1112 u32 set_size);
1113 extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet,
1114 struct drbd_peer_request *);
1115 extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet,
1116 struct p_block_req *rp);
1117 extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet,
1118 struct p_data *dp, int data_size);
1119 extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet,
1120 sector_t sector, int blksize, u64 block_id);
1121 extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *);
1122 extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet,
1123 struct drbd_peer_request *);
1124 extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req);
1125 extern int drbd_send_drequest(struct drbd_peer_device *, int cmd,
1126 sector_t sector, int size, u64 block_id);
1127 extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector,
1128 int size, void *digest, int digest_size,
1129 enum drbd_packet cmd);
1130 extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size);
1131
1132 extern int drbd_send_bitmap(struct drbd_device *device);
1133 extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode);
1134 extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode);
1135 extern int drbd_send_rs_deallocated(struct drbd_peer_device *, struct drbd_peer_request *);
1136 extern void drbd_backing_dev_free(struct drbd_device *device, struct drbd_backing_dev *ldev);
1137 extern void drbd_device_cleanup(struct drbd_device *device);
1138 void drbd_print_uuids(struct drbd_device *device, const char *text);
1139
1140 extern void conn_md_sync(struct drbd_connection *connection);
1141 extern void drbd_md_write(struct drbd_device *device, void *buffer);
1142 extern void drbd_md_sync(struct drbd_device *device);
1143 extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev);
1144 extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1145 extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1146 extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local);
1147 extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local);
1148 extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local);
1149 extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1150 extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local);
1151 extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local);
1152 extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1153 extern void drbd_md_mark_dirty(struct drbd_device *device);
1154 extern void drbd_queue_bitmap_io(struct drbd_device *device,
1155 int (*io_fn)(struct drbd_device *),
1156 void (*done)(struct drbd_device *, int),
1157 char *why, enum bm_flag flags);
1158 extern int drbd_bitmap_io(struct drbd_device *device,
1159 int (*io_fn)(struct drbd_device *),
1160 char *why, enum bm_flag flags);
1161 extern int drbd_bitmap_io_from_worker(struct drbd_device *device,
1162 int (*io_fn)(struct drbd_device *),
1163 char *why, enum bm_flag flags);
1164 extern int drbd_bmio_set_n_write(struct drbd_device *device) __must_hold(local);
1165 extern int drbd_bmio_clear_n_write(struct drbd_device *device) __must_hold(local);
1166
1167 /* Meta data layout
1168 *
1169 * We currently have two possible layouts.
1170 * Offsets in (512 byte) sectors.
1171 * external:
1172 * |----------- md_size_sect ------------------|
1173 * [ 4k superblock ][ activity log ][ Bitmap ]
1174 * | al_offset == 8 |
1175 * | bm_offset = al_offset + X |
1176 * ==> bitmap sectors = md_size_sect - bm_offset
1177 *
1178 * Variants:
1179 * old, indexed fixed size meta data:
1180 *
1181 * internal:
1182 * |----------- md_size_sect ------------------|
1183 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*]
1184 * | al_offset < 0 |
1185 * | bm_offset = al_offset - Y |
1186 * ==> bitmap sectors = Y = al_offset - bm_offset
1187 *
1188 * [padding*] are zero or up to 7 unused 512 Byte sectors to the
1189 * end of the device, so that the [4k superblock] will be 4k aligned.
1190 *
1191 * The activity log consists of 4k transaction blocks,
1192 * which are written in a ring-buffer, or striped ring-buffer like fashion,
1193 * which are writtensize used to be fixed 32kB,
1194 * but is about to become configurable.
1195 */
1196
1197 /* Our old fixed size meta data layout
1198 * allows up to about 3.8TB, so if you want more,
1199 * you need to use the "flexible" meta data format. */
1200 #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */
1201 #define MD_4kB_SECT 8
1202 #define MD_32kB_SECT 64
1203
1204 /* One activity log extent represents 4M of storage */
1205 #define AL_EXTENT_SHIFT 22
1206 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1207
1208 /* We could make these currently hardcoded constants configurable
1209 * variables at create-md time (or even re-configurable at runtime?).
1210 * Which will require some more changes to the DRBD "super block"
1211 * and attach code.
1212 *
1213 * updates per transaction:
1214 * This many changes to the active set can be logged with one transaction.
1215 * This number is arbitrary.
1216 * context per transaction:
1217 * This many context extent numbers are logged with each transaction.
1218 * This number is resulting from the transaction block size (4k), the layout
1219 * of the transaction header, and the number of updates per transaction.
1220 * See drbd_actlog.c:struct al_transaction_on_disk
1221 * */
1222 #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1223 #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1224
1225 #if BITS_PER_LONG == 32
1226 #define LN2_BPL 5
1227 #define cpu_to_lel(A) cpu_to_le32(A)
1228 #define lel_to_cpu(A) le32_to_cpu(A)
1229 #elif BITS_PER_LONG == 64
1230 #define LN2_BPL 6
1231 #define cpu_to_lel(A) cpu_to_le64(A)
1232 #define lel_to_cpu(A) le64_to_cpu(A)
1233 #else
1234 #error "LN2 of BITS_PER_LONG unknown!"
1235 #endif
1236
1237 /* resync bitmap */
1238 /* 16MB sized 'bitmap extent' to track syncer usage */
1239 struct bm_extent {
1240 int rs_left; /* number of bits set (out of sync) in this extent. */
1241 int rs_failed; /* number of failed resync requests in this extent. */
1242 unsigned long flags;
1243 struct lc_element lce;
1244 };
1245
1246 #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1247 #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1248 #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1249
1250 /* drbd_bitmap.c */
1251 /*
1252 * We need to store one bit for a block.
1253 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1254 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1255 * Bit 1 ==> local node thinks this block needs to be synced.
1256 */
1257
1258 #define SLEEP_TIME (HZ/10)
1259
1260 /* We do bitmap IO in units of 4k blocks.
1261 * We also still have a hardcoded 4k per bit relation. */
1262 #define BM_BLOCK_SHIFT 12 /* 4k per bit */
1263 #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1264 /* mostly arbitrarily set the represented size of one bitmap extent,
1265 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1266 * at 4k per bit resolution) */
1267 #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
1268 #define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1269
1270 #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1271 #error "HAVE YOU FIXED drbdmeta AS WELL??"
1272 #endif
1273
1274 /* thus many _storage_ sectors are described by one bit */
1275 #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1276 #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1277 #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1278
1279 /* bit to represented kilo byte conversion */
1280 #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1281
1282 /* in which _bitmap_ extent (resp. sector) the bit for a certain
1283 * _storage_ sector is located in */
1284 #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1285 #define BM_BIT_TO_EXT(x) ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1286
1287 /* first storage sector a bitmap extent corresponds to */
1288 #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1289 /* how much _storage_ sectors we have per bitmap extent */
1290 #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1291 /* how many bits are covered by one bitmap extent (resync extent) */
1292 #define BM_BITS_PER_EXT (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1293
1294 #define BM_BLOCKS_PER_BM_EXT_MASK (BM_BITS_PER_EXT - 1)
1295
1296
1297 /* in one sector of the bitmap, we have this many activity_log extents. */
1298 #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1299
1300 /* the extent in "PER_EXTENT" below is an activity log extent
1301 * we need that many (long words/bytes) to store the bitmap
1302 * of one AL_EXTENT_SIZE chunk of storage.
1303 * we can store the bitmap for that many AL_EXTENTS within
1304 * one sector of the _on_disk_ bitmap:
1305 * bit 0 bit 37 bit 38 bit (512*8)-1
1306 * ...|........|........|.. // ..|........|
1307 * sect. 0 `296 `304 ^(512*8*8)-1
1308 *
1309 #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1310 #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1311 #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1312 */
1313
1314 #define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1315 /* we have a certain meta data variant that has a fixed on-disk size of 128
1316 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity
1317 * log, leaving this many sectors for the bitmap.
1318 */
1319
1320 #define DRBD_MAX_SECTORS_FIXED_BM \
1321 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9)))
1322 #if !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1323 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1324 #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1325 #else
1326 #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM
1327 /* 16 TB in units of sectors */
1328 #if BITS_PER_LONG == 32
1329 /* adjust by one page worth of bitmap,
1330 * so we won't wrap around in drbd_bm_find_next_bit.
1331 * you should use 64bit OS for that much storage, anyways. */
1332 #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1333 #else
1334 /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1335 #define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1336 /* corresponds to (1UL << 38) bits right now. */
1337 #endif
1338 #endif
1339
1340 /* Estimate max bio size as 256 * PAGE_SIZE,
1341 * so for typical PAGE_SIZE of 4k, that is (1<<20) Byte.
1342 * Since we may live in a mixed-platform cluster,
1343 * we limit us to a platform agnostic constant here for now.
1344 * A followup commit may allow even bigger BIO sizes,
1345 * once we thought that through. */
1346 #define DRBD_MAX_BIO_SIZE (1U << 20)
1347 #if DRBD_MAX_BIO_SIZE > (BIO_MAX_PAGES << PAGE_SHIFT)
1348 #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1349 #endif
1350 #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */
1351
1352 #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */
1353 #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */
1354
1355 /* For now, don't allow more than half of what we can "activate" in one
1356 * activity log transaction to be discarded in one go. We may need to rework
1357 * drbd_al_begin_io() to allow for even larger discard ranges */
1358 #define DRBD_MAX_BATCH_BIO_SIZE (AL_UPDATES_PER_TRANSACTION/2*AL_EXTENT_SIZE)
1359 #define DRBD_MAX_BBIO_SECTORS (DRBD_MAX_BATCH_BIO_SIZE >> 9)
1360
1361 extern int drbd_bm_init(struct drbd_device *device);
1362 extern int drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits);
1363 extern void drbd_bm_cleanup(struct drbd_device *device);
1364 extern void drbd_bm_set_all(struct drbd_device *device);
1365 extern void drbd_bm_clear_all(struct drbd_device *device);
1366 /* set/clear/test only a few bits at a time */
1367 extern int drbd_bm_set_bits(
1368 struct drbd_device *device, unsigned long s, unsigned long e);
1369 extern int drbd_bm_clear_bits(
1370 struct drbd_device *device, unsigned long s, unsigned long e);
1371 extern int drbd_bm_count_bits(
1372 struct drbd_device *device, const unsigned long s, const unsigned long e);
1373 /* bm_set_bits variant for use while holding drbd_bm_lock,
1374 * may process the whole bitmap in one go */
1375 extern void _drbd_bm_set_bits(struct drbd_device *device,
1376 const unsigned long s, const unsigned long e);
1377 extern int drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr);
1378 extern int drbd_bm_e_weight(struct drbd_device *device, unsigned long enr);
1379 extern int drbd_bm_read(struct drbd_device *device) __must_hold(local);
1380 extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr);
1381 extern int drbd_bm_write(struct drbd_device *device) __must_hold(local);
1382 extern void drbd_bm_reset_al_hints(struct drbd_device *device) __must_hold(local);
1383 extern int drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local);
1384 extern int drbd_bm_write_lazy(struct drbd_device *device, unsigned upper_idx) __must_hold(local);
1385 extern int drbd_bm_write_all(struct drbd_device *device) __must_hold(local);
1386 extern int drbd_bm_write_copy_pages(struct drbd_device *device) __must_hold(local);
1387 extern size_t drbd_bm_words(struct drbd_device *device);
1388 extern unsigned long drbd_bm_bits(struct drbd_device *device);
1389 extern sector_t drbd_bm_capacity(struct drbd_device *device);
1390
1391 #define DRBD_END_OF_BITMAP (~(unsigned long)0)
1392 extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo);
1393 /* bm_find_next variants for use while you hold drbd_bm_lock() */
1394 extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo);
1395 extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo);
1396 extern unsigned long _drbd_bm_total_weight(struct drbd_device *device);
1397 extern unsigned long drbd_bm_total_weight(struct drbd_device *device);
1398 /* for receive_bitmap */
1399 extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset,
1400 size_t number, unsigned long *buffer);
1401 /* for _drbd_send_bitmap */
1402 extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset,
1403 size_t number, unsigned long *buffer);
1404
1405 extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags);
1406 extern void drbd_bm_unlock(struct drbd_device *device);
1407 /* drbd_main.c */
1408
1409 extern struct kmem_cache *drbd_request_cache;
1410 extern struct kmem_cache *drbd_ee_cache; /* peer requests */
1411 extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1412 extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1413 extern mempool_t *drbd_request_mempool;
1414 extern mempool_t *drbd_ee_mempool;
1415
1416 /* drbd's page pool, used to buffer data received from the peer,
1417 * or data requested by the peer.
1418 *
1419 * This does not have an emergency reserve.
1420 *
1421 * When allocating from this pool, it first takes pages from the pool.
1422 * Only if the pool is depleted will try to allocate from the system.
1423 *
1424 * The assumption is that pages taken from this pool will be processed,
1425 * and given back, "quickly", and then can be recycled, so we can avoid
1426 * frequent calls to alloc_page(), and still will be able to make progress even
1427 * under memory pressure.
1428 */
1429 extern struct page *drbd_pp_pool;
1430 extern spinlock_t drbd_pp_lock;
1431 extern int drbd_pp_vacant;
1432 extern wait_queue_head_t drbd_pp_wait;
1433
1434 /* We also need a standard (emergency-reserve backed) page pool
1435 * for meta data IO (activity log, bitmap).
1436 * We can keep it global, as long as it is used as "N pages at a time".
1437 * 128 should be plenty, currently we probably can get away with as few as 1.
1438 */
1439 #define DRBD_MIN_POOL_PAGES 128
1440 extern mempool_t *drbd_md_io_page_pool;
1441
1442 /* We also need to make sure we get a bio
1443 * when we need it for housekeeping purposes */
1444 extern struct bio_set *drbd_md_io_bio_set;
1445 /* to allocate from that set */
1446 extern struct bio *bio_alloc_drbd(gfp_t gfp_mask);
1447
1448 extern struct mutex resources_mutex;
1449
1450 extern int conn_lowest_minor(struct drbd_connection *connection);
1451 extern enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor);
1452 extern void drbd_destroy_device(struct kref *kref);
1453 extern void drbd_delete_device(struct drbd_device *device);
1454
1455 extern struct drbd_resource *drbd_create_resource(const char *name);
1456 extern void drbd_free_resource(struct drbd_resource *resource);
1457
1458 extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts);
1459 extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts);
1460 extern void drbd_destroy_connection(struct kref *kref);
1461 extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len,
1462 void *peer_addr, int peer_addr_len);
1463 extern struct drbd_resource *drbd_find_resource(const char *name);
1464 extern void drbd_destroy_resource(struct kref *kref);
1465 extern void conn_free_crypto(struct drbd_connection *connection);
1466
1467 extern int proc_details;
1468
1469 /* drbd_req */
1470 extern void do_submit(struct work_struct *ws);
1471 extern void __drbd_make_request(struct drbd_device *, struct bio *, unsigned long);
1472 extern blk_qc_t drbd_make_request(struct request_queue *q, struct bio *bio);
1473 extern int drbd_read_remote(struct drbd_device *device, struct drbd_request *req);
1474 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1475
1476
1477 /* drbd_nl.c */
1478
1479 extern struct mutex notification_mutex;
1480
1481 extern void drbd_suspend_io(struct drbd_device *device);
1482 extern void drbd_resume_io(struct drbd_device *device);
1483 extern char *ppsize(char *buf, unsigned long long size);
1484 extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int);
1485 enum determine_dev_size {
1486 DS_ERROR_SHRINK = -3,
1487 DS_ERROR_SPACE_MD = -2,
1488 DS_ERROR = -1,
1489 DS_UNCHANGED = 0,
1490 DS_SHRUNK = 1,
1491 DS_GREW = 2,
1492 DS_GREW_FROM_ZERO = 3,
1493 };
1494 extern enum determine_dev_size
1495 drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local);
1496 extern void resync_after_online_grow(struct drbd_device *);
1497 extern void drbd_reconsider_queue_parameters(struct drbd_device *device,
1498 struct drbd_backing_dev *bdev, struct o_qlim *o);
1499 extern enum drbd_state_rv drbd_set_role(struct drbd_device *device,
1500 enum drbd_role new_role,
1501 int force);
1502 extern bool conn_try_outdate_peer(struct drbd_connection *connection);
1503 extern void conn_try_outdate_peer_async(struct drbd_connection *connection);
1504 extern enum drbd_peer_state conn_khelper(struct drbd_connection *connection, char *cmd);
1505 extern int drbd_khelper(struct drbd_device *device, char *cmd);
1506
1507 /* drbd_worker.c */
1508 /* bi_end_io handlers */
1509 extern void drbd_md_endio(struct bio *bio);
1510 extern void drbd_peer_request_endio(struct bio *bio);
1511 extern void drbd_request_endio(struct bio *bio);
1512 extern int drbd_worker(struct drbd_thread *thi);
1513 enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor);
1514 void drbd_resync_after_changed(struct drbd_device *device);
1515 extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side);
1516 extern void resume_next_sg(struct drbd_device *device);
1517 extern void suspend_other_sg(struct drbd_device *device);
1518 extern int drbd_resync_finished(struct drbd_device *device);
1519 /* maybe rather drbd_main.c ? */
1520 extern void *drbd_md_get_buffer(struct drbd_device *device, const char *intent);
1521 extern void drbd_md_put_buffer(struct drbd_device *device);
1522 extern int drbd_md_sync_page_io(struct drbd_device *device,
1523 struct drbd_backing_dev *bdev, sector_t sector, int op);
1524 extern void drbd_ov_out_of_sync_found(struct drbd_device *, sector_t, int);
1525 extern void wait_until_done_or_force_detached(struct drbd_device *device,
1526 struct drbd_backing_dev *bdev, unsigned int *done);
1527 extern void drbd_rs_controller_reset(struct drbd_device *device);
1528
ov_out_of_sync_print(struct drbd_device * device)1529 static inline void ov_out_of_sync_print(struct drbd_device *device)
1530 {
1531 if (device->ov_last_oos_size) {
1532 drbd_err(device, "Out of sync: start=%llu, size=%lu (sectors)\n",
1533 (unsigned long long)device->ov_last_oos_start,
1534 (unsigned long)device->ov_last_oos_size);
1535 }
1536 device->ov_last_oos_size = 0;
1537 }
1538
1539
1540 extern void drbd_csum_bio(struct crypto_ahash *, struct bio *, void *);
1541 extern void drbd_csum_ee(struct crypto_ahash *, struct drbd_peer_request *, void *);
1542 /* worker callbacks */
1543 extern int w_e_end_data_req(struct drbd_work *, int);
1544 extern int w_e_end_rsdata_req(struct drbd_work *, int);
1545 extern int w_e_end_csum_rs_req(struct drbd_work *, int);
1546 extern int w_e_end_ov_reply(struct drbd_work *, int);
1547 extern int w_e_end_ov_req(struct drbd_work *, int);
1548 extern int w_ov_finished(struct drbd_work *, int);
1549 extern int w_resync_timer(struct drbd_work *, int);
1550 extern int w_send_write_hint(struct drbd_work *, int);
1551 extern int w_send_dblock(struct drbd_work *, int);
1552 extern int w_send_read_req(struct drbd_work *, int);
1553 extern int w_e_reissue(struct drbd_work *, int);
1554 extern int w_restart_disk_io(struct drbd_work *, int);
1555 extern int w_send_out_of_sync(struct drbd_work *, int);
1556 extern int w_start_resync(struct drbd_work *, int);
1557
1558 extern void resync_timer_fn(unsigned long data);
1559 extern void start_resync_timer_fn(unsigned long data);
1560
1561 extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req);
1562
1563 /* drbd_receiver.c */
1564 extern int drbd_issue_discard_or_zero_out(struct drbd_device *device,
1565 sector_t start, unsigned int nr_sectors, bool discard);
1566 extern int drbd_receiver(struct drbd_thread *thi);
1567 extern int drbd_ack_receiver(struct drbd_thread *thi);
1568 extern void drbd_send_ping_wf(struct work_struct *ws);
1569 extern void drbd_send_acks_wf(struct work_struct *ws);
1570 extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device);
1571 extern bool drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector,
1572 bool throttle_if_app_is_waiting);
1573 extern int drbd_submit_peer_request(struct drbd_device *,
1574 struct drbd_peer_request *, const unsigned,
1575 const unsigned, const int);
1576 extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *);
1577 extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64,
1578 sector_t, unsigned int,
1579 unsigned int,
1580 gfp_t) __must_hold(local);
1581 extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *,
1582 int);
1583 #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1584 #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
1585 extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool);
1586 extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
1587 extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
1588 extern int drbd_connected(struct drbd_peer_device *);
1589
drbd_tcp_cork(struct socket * sock)1590 static inline void drbd_tcp_cork(struct socket *sock)
1591 {
1592 int val = 1;
1593 (void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
1594 (char*)&val, sizeof(val));
1595 }
1596
drbd_tcp_uncork(struct socket * sock)1597 static inline void drbd_tcp_uncork(struct socket *sock)
1598 {
1599 int val = 0;
1600 (void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
1601 (char*)&val, sizeof(val));
1602 }
1603
drbd_tcp_nodelay(struct socket * sock)1604 static inline void drbd_tcp_nodelay(struct socket *sock)
1605 {
1606 int val = 1;
1607 (void) kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1608 (char*)&val, sizeof(val));
1609 }
1610
drbd_tcp_quickack(struct socket * sock)1611 static inline void drbd_tcp_quickack(struct socket *sock)
1612 {
1613 int val = 2;
1614 (void) kernel_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1615 (char*)&val, sizeof(val));
1616 }
1617
1618 /* sets the number of 512 byte sectors of our virtual device */
drbd_set_my_capacity(struct drbd_device * device,sector_t size)1619 static inline void drbd_set_my_capacity(struct drbd_device *device,
1620 sector_t size)
1621 {
1622 /* set_capacity(device->this_bdev->bd_disk, size); */
1623 set_capacity(device->vdisk, size);
1624 device->this_bdev->bd_inode->i_size = (loff_t)size << 9;
1625 }
1626
1627 /*
1628 * used to submit our private bio
1629 */
drbd_generic_make_request(struct drbd_device * device,int fault_type,struct bio * bio)1630 static inline void drbd_generic_make_request(struct drbd_device *device,
1631 int fault_type, struct bio *bio)
1632 {
1633 __release(local);
1634 if (!bio->bi_bdev) {
1635 drbd_err(device, "drbd_generic_make_request: bio->bi_bdev == NULL\n");
1636 bio->bi_error = -ENODEV;
1637 bio_endio(bio);
1638 return;
1639 }
1640
1641 if (drbd_insert_fault(device, fault_type))
1642 bio_io_error(bio);
1643 else
1644 generic_make_request(bio);
1645 }
1646
1647 void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev,
1648 enum write_ordering_e wo);
1649
1650 /* drbd_proc.c */
1651 extern struct proc_dir_entry *drbd_proc;
1652 extern const struct file_operations drbd_proc_fops;
1653
1654 /* drbd_actlog.c */
1655 extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i);
1656 extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i);
1657 extern void drbd_al_begin_io_commit(struct drbd_device *device);
1658 extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i);
1659 extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i);
1660 extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i);
1661 extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector);
1662 extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector);
1663 extern int drbd_try_rs_begin_io(struct drbd_device *device, sector_t sector);
1664 extern void drbd_rs_cancel_all(struct drbd_device *device);
1665 extern int drbd_rs_del_all(struct drbd_device *device);
1666 extern void drbd_rs_failed_io(struct drbd_device *device,
1667 sector_t sector, int size);
1668 extern void drbd_advance_rs_marks(struct drbd_device *device, unsigned long still_to_go);
1669
1670 enum update_sync_bits_mode { RECORD_RS_FAILED, SET_OUT_OF_SYNC, SET_IN_SYNC };
1671 extern int __drbd_change_sync(struct drbd_device *device, sector_t sector, int size,
1672 enum update_sync_bits_mode mode);
1673 #define drbd_set_in_sync(device, sector, size) \
1674 __drbd_change_sync(device, sector, size, SET_IN_SYNC)
1675 #define drbd_set_out_of_sync(device, sector, size) \
1676 __drbd_change_sync(device, sector, size, SET_OUT_OF_SYNC)
1677 #define drbd_rs_failed_io(device, sector, size) \
1678 __drbd_change_sync(device, sector, size, RECORD_RS_FAILED)
1679 extern void drbd_al_shrink(struct drbd_device *device);
1680 extern int drbd_al_initialize(struct drbd_device *, void *);
1681
1682 /* drbd_nl.c */
1683 /* state info broadcast */
1684 struct sib_info {
1685 enum drbd_state_info_bcast_reason sib_reason;
1686 union {
1687 struct {
1688 char *helper_name;
1689 unsigned helper_exit_code;
1690 };
1691 struct {
1692 union drbd_state os;
1693 union drbd_state ns;
1694 };
1695 };
1696 };
1697 void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib);
1698
1699 extern void notify_resource_state(struct sk_buff *,
1700 unsigned int,
1701 struct drbd_resource *,
1702 struct resource_info *,
1703 enum drbd_notification_type);
1704 extern void notify_device_state(struct sk_buff *,
1705 unsigned int,
1706 struct drbd_device *,
1707 struct device_info *,
1708 enum drbd_notification_type);
1709 extern void notify_connection_state(struct sk_buff *,
1710 unsigned int,
1711 struct drbd_connection *,
1712 struct connection_info *,
1713 enum drbd_notification_type);
1714 extern void notify_peer_device_state(struct sk_buff *,
1715 unsigned int,
1716 struct drbd_peer_device *,
1717 struct peer_device_info *,
1718 enum drbd_notification_type);
1719 extern void notify_helper(enum drbd_notification_type, struct drbd_device *,
1720 struct drbd_connection *, const char *, int);
1721
1722 /*
1723 * inline helper functions
1724 *************************/
1725
1726 /* see also page_chain_add and friends in drbd_receiver.c */
page_chain_next(struct page * page)1727 static inline struct page *page_chain_next(struct page *page)
1728 {
1729 return (struct page *)page_private(page);
1730 }
1731 #define page_chain_for_each(page) \
1732 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1733 page = page_chain_next(page))
1734 #define page_chain_for_each_safe(page, n) \
1735 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1736
1737
drbd_peer_req_has_active_page(struct drbd_peer_request * peer_req)1738 static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req)
1739 {
1740 struct page *page = peer_req->pages;
1741 page_chain_for_each(page) {
1742 if (page_count(page) > 1)
1743 return 1;
1744 }
1745 return 0;
1746 }
1747
drbd_read_state(struct drbd_device * device)1748 static inline union drbd_state drbd_read_state(struct drbd_device *device)
1749 {
1750 struct drbd_resource *resource = device->resource;
1751 union drbd_state rv;
1752
1753 rv.i = device->state.i;
1754 rv.susp = resource->susp;
1755 rv.susp_nod = resource->susp_nod;
1756 rv.susp_fen = resource->susp_fen;
1757
1758 return rv;
1759 }
1760
1761 enum drbd_force_detach_flags {
1762 DRBD_READ_ERROR,
1763 DRBD_WRITE_ERROR,
1764 DRBD_META_IO_ERROR,
1765 DRBD_FORCE_DETACH,
1766 };
1767
1768 #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
__drbd_chk_io_error_(struct drbd_device * device,enum drbd_force_detach_flags df,const char * where)1769 static inline void __drbd_chk_io_error_(struct drbd_device *device,
1770 enum drbd_force_detach_flags df,
1771 const char *where)
1772 {
1773 enum drbd_io_error_p ep;
1774
1775 rcu_read_lock();
1776 ep = rcu_dereference(device->ldev->disk_conf)->on_io_error;
1777 rcu_read_unlock();
1778 switch (ep) {
1779 case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */
1780 if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) {
1781 if (__ratelimit(&drbd_ratelimit_state))
1782 drbd_err(device, "Local IO failed in %s.\n", where);
1783 if (device->state.disk > D_INCONSISTENT)
1784 _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL);
1785 break;
1786 }
1787 /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */
1788 case EP_DETACH:
1789 case EP_CALL_HELPER:
1790 /* Remember whether we saw a READ or WRITE error.
1791 *
1792 * Recovery of the affected area for WRITE failure is covered
1793 * by the activity log.
1794 * READ errors may fall outside that area though. Certain READ
1795 * errors can be "healed" by writing good data to the affected
1796 * blocks, which triggers block re-allocation in lower layers.
1797 *
1798 * If we can not write the bitmap after a READ error,
1799 * we may need to trigger a full sync (see w_go_diskless()).
1800 *
1801 * Force-detach is not really an IO error, but rather a
1802 * desperate measure to try to deal with a completely
1803 * unresponsive lower level IO stack.
1804 * Still it should be treated as a WRITE error.
1805 *
1806 * Meta IO error is always WRITE error:
1807 * we read meta data only once during attach,
1808 * which will fail in case of errors.
1809 */
1810 set_bit(WAS_IO_ERROR, &device->flags);
1811 if (df == DRBD_READ_ERROR)
1812 set_bit(WAS_READ_ERROR, &device->flags);
1813 if (df == DRBD_FORCE_DETACH)
1814 set_bit(FORCE_DETACH, &device->flags);
1815 if (device->state.disk > D_FAILED) {
1816 _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL);
1817 drbd_err(device,
1818 "Local IO failed in %s. Detaching...\n", where);
1819 }
1820 break;
1821 }
1822 }
1823
1824 /**
1825 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1826 * @device: DRBD device.
1827 * @error: Error code passed to the IO completion callback
1828 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1829 *
1830 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1831 */
1832 #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
drbd_chk_io_error_(struct drbd_device * device,int error,enum drbd_force_detach_flags forcedetach,const char * where)1833 static inline void drbd_chk_io_error_(struct drbd_device *device,
1834 int error, enum drbd_force_detach_flags forcedetach, const char *where)
1835 {
1836 if (error) {
1837 unsigned long flags;
1838 spin_lock_irqsave(&device->resource->req_lock, flags);
1839 __drbd_chk_io_error_(device, forcedetach, where);
1840 spin_unlock_irqrestore(&device->resource->req_lock, flags);
1841 }
1842 }
1843
1844
1845 /**
1846 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1847 * @bdev: Meta data block device.
1848 *
1849 * BTW, for internal meta data, this happens to be the maximum capacity
1850 * we could agree upon with our peer node.
1851 */
drbd_md_first_sector(struct drbd_backing_dev * bdev)1852 static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1853 {
1854 switch (bdev->md.meta_dev_idx) {
1855 case DRBD_MD_INDEX_INTERNAL:
1856 case DRBD_MD_INDEX_FLEX_INT:
1857 return bdev->md.md_offset + bdev->md.bm_offset;
1858 case DRBD_MD_INDEX_FLEX_EXT:
1859 default:
1860 return bdev->md.md_offset;
1861 }
1862 }
1863
1864 /**
1865 * drbd_md_last_sector() - Return the last sector number of the meta data area
1866 * @bdev: Meta data block device.
1867 */
drbd_md_last_sector(struct drbd_backing_dev * bdev)1868 static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1869 {
1870 switch (bdev->md.meta_dev_idx) {
1871 case DRBD_MD_INDEX_INTERNAL:
1872 case DRBD_MD_INDEX_FLEX_INT:
1873 return bdev->md.md_offset + MD_4kB_SECT -1;
1874 case DRBD_MD_INDEX_FLEX_EXT:
1875 default:
1876 return bdev->md.md_offset + bdev->md.md_size_sect -1;
1877 }
1878 }
1879
1880 /* Returns the number of 512 byte sectors of the device */
drbd_get_capacity(struct block_device * bdev)1881 static inline sector_t drbd_get_capacity(struct block_device *bdev)
1882 {
1883 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1884 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0;
1885 }
1886
1887 /**
1888 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1889 * @bdev: Meta data block device.
1890 *
1891 * returns the capacity we announce to out peer. we clip ourselves at the
1892 * various MAX_SECTORS, because if we don't, current implementation will
1893 * oops sooner or later
1894 */
drbd_get_max_capacity(struct drbd_backing_dev * bdev)1895 static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1896 {
1897 sector_t s;
1898
1899 switch (bdev->md.meta_dev_idx) {
1900 case DRBD_MD_INDEX_INTERNAL:
1901 case DRBD_MD_INDEX_FLEX_INT:
1902 s = drbd_get_capacity(bdev->backing_bdev)
1903 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1904 drbd_md_first_sector(bdev))
1905 : 0;
1906 break;
1907 case DRBD_MD_INDEX_FLEX_EXT:
1908 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1909 drbd_get_capacity(bdev->backing_bdev));
1910 /* clip at maximum size the meta device can support */
1911 s = min_t(sector_t, s,
1912 BM_EXT_TO_SECT(bdev->md.md_size_sect
1913 - bdev->md.bm_offset));
1914 break;
1915 default:
1916 s = min_t(sector_t, DRBD_MAX_SECTORS,
1917 drbd_get_capacity(bdev->backing_bdev));
1918 }
1919 return s;
1920 }
1921
1922 /**
1923 * drbd_md_ss() - Return the sector number of our meta data super block
1924 * @bdev: Meta data block device.
1925 */
drbd_md_ss(struct drbd_backing_dev * bdev)1926 static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev)
1927 {
1928 const int meta_dev_idx = bdev->md.meta_dev_idx;
1929
1930 if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT)
1931 return 0;
1932
1933 /* Since drbd08, internal meta data is always "flexible".
1934 * position: last 4k aligned block of 4k size */
1935 if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL ||
1936 meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)
1937 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8;
1938
1939 /* external, some index; this is the old fixed size layout */
1940 return MD_128MB_SECT * bdev->md.meta_dev_idx;
1941 }
1942
1943 static inline void
drbd_queue_work(struct drbd_work_queue * q,struct drbd_work * w)1944 drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1945 {
1946 unsigned long flags;
1947 spin_lock_irqsave(&q->q_lock, flags);
1948 list_add_tail(&w->list, &q->q);
1949 spin_unlock_irqrestore(&q->q_lock, flags);
1950 wake_up(&q->q_wait);
1951 }
1952
1953 static inline void
drbd_queue_work_if_unqueued(struct drbd_work_queue * q,struct drbd_work * w)1954 drbd_queue_work_if_unqueued(struct drbd_work_queue *q, struct drbd_work *w)
1955 {
1956 unsigned long flags;
1957 spin_lock_irqsave(&q->q_lock, flags);
1958 if (list_empty_careful(&w->list))
1959 list_add_tail(&w->list, &q->q);
1960 spin_unlock_irqrestore(&q->q_lock, flags);
1961 wake_up(&q->q_wait);
1962 }
1963
1964 static inline void
drbd_device_post_work(struct drbd_device * device,int work_bit)1965 drbd_device_post_work(struct drbd_device *device, int work_bit)
1966 {
1967 if (!test_and_set_bit(work_bit, &device->flags)) {
1968 struct drbd_connection *connection =
1969 first_peer_device(device)->connection;
1970 struct drbd_work_queue *q = &connection->sender_work;
1971 if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags))
1972 wake_up(&q->q_wait);
1973 }
1974 }
1975
1976 extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue);
1977
1978 /* To get the ack_receiver out of the blocking network stack,
1979 * so it can change its sk_rcvtimeo from idle- to ping-timeout,
1980 * and send a ping, we need to send a signal.
1981 * Which signal we send is irrelevant. */
wake_ack_receiver(struct drbd_connection * connection)1982 static inline void wake_ack_receiver(struct drbd_connection *connection)
1983 {
1984 struct task_struct *task = connection->ack_receiver.task;
1985 if (task && get_t_state(&connection->ack_receiver) == RUNNING)
1986 force_sig(SIGXCPU, task);
1987 }
1988
request_ping(struct drbd_connection * connection)1989 static inline void request_ping(struct drbd_connection *connection)
1990 {
1991 set_bit(SEND_PING, &connection->flags);
1992 wake_ack_receiver(connection);
1993 }
1994
1995 extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *);
1996 extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *);
1997 extern int conn_send_command(struct drbd_connection *, struct drbd_socket *,
1998 enum drbd_packet, unsigned int, void *,
1999 unsigned int);
2000 extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *,
2001 enum drbd_packet, unsigned int, void *,
2002 unsigned int);
2003
2004 extern int drbd_send_ping(struct drbd_connection *connection);
2005 extern int drbd_send_ping_ack(struct drbd_connection *connection);
2006 extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state);
2007 extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state);
2008
drbd_thread_stop(struct drbd_thread * thi)2009 static inline void drbd_thread_stop(struct drbd_thread *thi)
2010 {
2011 _drbd_thread_stop(thi, false, true);
2012 }
2013
drbd_thread_stop_nowait(struct drbd_thread * thi)2014 static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
2015 {
2016 _drbd_thread_stop(thi, false, false);
2017 }
2018
drbd_thread_restart_nowait(struct drbd_thread * thi)2019 static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
2020 {
2021 _drbd_thread_stop(thi, true, false);
2022 }
2023
2024 /* counts how many answer packets packets we expect from our peer,
2025 * for either explicit application requests,
2026 * or implicit barrier packets as necessary.
2027 * increased:
2028 * w_send_barrier
2029 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
2030 * it is much easier and equally valid to count what we queue for the
2031 * worker, even before it actually was queued or send.
2032 * (drbd_make_request_common; recovery path on read io-error)
2033 * decreased:
2034 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
2035 * _req_mod(req, DATA_RECEIVED)
2036 * [from receive_DataReply]
2037 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
2038 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
2039 * for some reason it is NOT decreased in got_NegAck,
2040 * but in the resulting cleanup code from report_params.
2041 * we should try to remember the reason for that...
2042 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
2043 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
2044 * [from tl_clear_barrier]
2045 */
inc_ap_pending(struct drbd_device * device)2046 static inline void inc_ap_pending(struct drbd_device *device)
2047 {
2048 atomic_inc(&device->ap_pending_cnt);
2049 }
2050
2051 #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \
2052 if (atomic_read(&device->which) < 0) \
2053 drbd_err(device, "in %s:%d: " #which " = %d < 0 !\n", \
2054 func, line, \
2055 atomic_read(&device->which))
2056
2057 #define dec_ap_pending(device) _dec_ap_pending(device, __func__, __LINE__)
_dec_ap_pending(struct drbd_device * device,const char * func,int line)2058 static inline void _dec_ap_pending(struct drbd_device *device, const char *func, int line)
2059 {
2060 if (atomic_dec_and_test(&device->ap_pending_cnt))
2061 wake_up(&device->misc_wait);
2062 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line);
2063 }
2064
2065 /* counts how many resync-related answers we still expect from the peer
2066 * increase decrease
2067 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
2068 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
2069 * (or P_NEG_ACK with ID_SYNCER)
2070 */
inc_rs_pending(struct drbd_device * device)2071 static inline void inc_rs_pending(struct drbd_device *device)
2072 {
2073 atomic_inc(&device->rs_pending_cnt);
2074 }
2075
2076 #define dec_rs_pending(device) _dec_rs_pending(device, __func__, __LINE__)
_dec_rs_pending(struct drbd_device * device,const char * func,int line)2077 static inline void _dec_rs_pending(struct drbd_device *device, const char *func, int line)
2078 {
2079 atomic_dec(&device->rs_pending_cnt);
2080 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line);
2081 }
2082
2083 /* counts how many answers we still need to send to the peer.
2084 * increased on
2085 * receive_Data unless protocol A;
2086 * we need to send a P_RECV_ACK (proto B)
2087 * or P_WRITE_ACK (proto C)
2088 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2089 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2090 * receive_Barrier_* we need to send a P_BARRIER_ACK
2091 */
inc_unacked(struct drbd_device * device)2092 static inline void inc_unacked(struct drbd_device *device)
2093 {
2094 atomic_inc(&device->unacked_cnt);
2095 }
2096
2097 #define dec_unacked(device) _dec_unacked(device, __func__, __LINE__)
_dec_unacked(struct drbd_device * device,const char * func,int line)2098 static inline void _dec_unacked(struct drbd_device *device, const char *func, int line)
2099 {
2100 atomic_dec(&device->unacked_cnt);
2101 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
2102 }
2103
2104 #define sub_unacked(device, n) _sub_unacked(device, n, __func__, __LINE__)
_sub_unacked(struct drbd_device * device,int n,const char * func,int line)2105 static inline void _sub_unacked(struct drbd_device *device, int n, const char *func, int line)
2106 {
2107 atomic_sub(n, &device->unacked_cnt);
2108 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
2109 }
2110
is_sync_target_state(enum drbd_conns connection_state)2111 static inline bool is_sync_target_state(enum drbd_conns connection_state)
2112 {
2113 return connection_state == C_SYNC_TARGET ||
2114 connection_state == C_PAUSED_SYNC_T;
2115 }
2116
is_sync_source_state(enum drbd_conns connection_state)2117 static inline bool is_sync_source_state(enum drbd_conns connection_state)
2118 {
2119 return connection_state == C_SYNC_SOURCE ||
2120 connection_state == C_PAUSED_SYNC_S;
2121 }
2122
is_sync_state(enum drbd_conns connection_state)2123 static inline bool is_sync_state(enum drbd_conns connection_state)
2124 {
2125 return is_sync_source_state(connection_state) ||
2126 is_sync_target_state(connection_state);
2127 }
2128
2129 /**
2130 * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev
2131 * @_device: DRBD device.
2132 * @_min_state: Minimum device state required for success.
2133 *
2134 * You have to call put_ldev() when finished working with device->ldev.
2135 */
2136 #define get_ldev_if_state(_device, _min_state) \
2137 (_get_ldev_if_state((_device), (_min_state)) ? \
2138 ({ __acquire(x); true; }) : false)
2139 #define get_ldev(_device) get_ldev_if_state(_device, D_INCONSISTENT)
2140
put_ldev(struct drbd_device * device)2141 static inline void put_ldev(struct drbd_device *device)
2142 {
2143 enum drbd_disk_state disk_state = device->state.disk;
2144 /* We must check the state *before* the atomic_dec becomes visible,
2145 * or we have a theoretical race where someone hitting zero,
2146 * while state still D_FAILED, will then see D_DISKLESS in the
2147 * condition below and calling into destroy, where he must not, yet. */
2148 int i = atomic_dec_return(&device->local_cnt);
2149
2150 /* This may be called from some endio handler,
2151 * so we must not sleep here. */
2152
2153 __release(local);
2154 D_ASSERT(device, i >= 0);
2155 if (i == 0) {
2156 if (disk_state == D_DISKLESS)
2157 /* even internal references gone, safe to destroy */
2158 drbd_device_post_work(device, DESTROY_DISK);
2159 if (disk_state == D_FAILED)
2160 /* all application IO references gone. */
2161 if (!test_and_set_bit(GOING_DISKLESS, &device->flags))
2162 drbd_device_post_work(device, GO_DISKLESS);
2163 wake_up(&device->misc_wait);
2164 }
2165 }
2166
2167 #ifndef __CHECKER__
_get_ldev_if_state(struct drbd_device * device,enum drbd_disk_state mins)2168 static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins)
2169 {
2170 int io_allowed;
2171
2172 /* never get a reference while D_DISKLESS */
2173 if (device->state.disk == D_DISKLESS)
2174 return 0;
2175
2176 atomic_inc(&device->local_cnt);
2177 io_allowed = (device->state.disk >= mins);
2178 if (!io_allowed)
2179 put_ldev(device);
2180 return io_allowed;
2181 }
2182 #else
2183 extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins);
2184 #endif
2185
2186 /* this throttles on-the-fly application requests
2187 * according to max_buffers settings;
2188 * maybe re-implement using semaphores? */
drbd_get_max_buffers(struct drbd_device * device)2189 static inline int drbd_get_max_buffers(struct drbd_device *device)
2190 {
2191 struct net_conf *nc;
2192 int mxb;
2193
2194 rcu_read_lock();
2195 nc = rcu_dereference(first_peer_device(device)->connection->net_conf);
2196 mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */
2197 rcu_read_unlock();
2198
2199 return mxb;
2200 }
2201
drbd_state_is_stable(struct drbd_device * device)2202 static inline int drbd_state_is_stable(struct drbd_device *device)
2203 {
2204 union drbd_dev_state s = device->state;
2205
2206 /* DO NOT add a default clause, we want the compiler to warn us
2207 * for any newly introduced state we may have forgotten to add here */
2208
2209 switch ((enum drbd_conns)s.conn) {
2210 /* new io only accepted when there is no connection, ... */
2211 case C_STANDALONE:
2212 case C_WF_CONNECTION:
2213 /* ... or there is a well established connection. */
2214 case C_CONNECTED:
2215 case C_SYNC_SOURCE:
2216 case C_SYNC_TARGET:
2217 case C_VERIFY_S:
2218 case C_VERIFY_T:
2219 case C_PAUSED_SYNC_S:
2220 case C_PAUSED_SYNC_T:
2221 case C_AHEAD:
2222 case C_BEHIND:
2223 /* transitional states, IO allowed */
2224 case C_DISCONNECTING:
2225 case C_UNCONNECTED:
2226 case C_TIMEOUT:
2227 case C_BROKEN_PIPE:
2228 case C_NETWORK_FAILURE:
2229 case C_PROTOCOL_ERROR:
2230 case C_TEAR_DOWN:
2231 case C_WF_REPORT_PARAMS:
2232 case C_STARTING_SYNC_S:
2233 case C_STARTING_SYNC_T:
2234 break;
2235
2236 /* Allow IO in BM exchange states with new protocols */
2237 case C_WF_BITMAP_S:
2238 if (first_peer_device(device)->connection->agreed_pro_version < 96)
2239 return 0;
2240 break;
2241
2242 /* no new io accepted in these states */
2243 case C_WF_BITMAP_T:
2244 case C_WF_SYNC_UUID:
2245 case C_MASK:
2246 /* not "stable" */
2247 return 0;
2248 }
2249
2250 switch ((enum drbd_disk_state)s.disk) {
2251 case D_DISKLESS:
2252 case D_INCONSISTENT:
2253 case D_OUTDATED:
2254 case D_CONSISTENT:
2255 case D_UP_TO_DATE:
2256 case D_FAILED:
2257 /* disk state is stable as well. */
2258 break;
2259
2260 /* no new io accepted during transitional states */
2261 case D_ATTACHING:
2262 case D_NEGOTIATING:
2263 case D_UNKNOWN:
2264 case D_MASK:
2265 /* not "stable" */
2266 return 0;
2267 }
2268
2269 return 1;
2270 }
2271
drbd_suspended(struct drbd_device * device)2272 static inline int drbd_suspended(struct drbd_device *device)
2273 {
2274 struct drbd_resource *resource = device->resource;
2275
2276 return resource->susp || resource->susp_fen || resource->susp_nod;
2277 }
2278
may_inc_ap_bio(struct drbd_device * device)2279 static inline bool may_inc_ap_bio(struct drbd_device *device)
2280 {
2281 int mxb = drbd_get_max_buffers(device);
2282
2283 if (drbd_suspended(device))
2284 return false;
2285 if (atomic_read(&device->suspend_cnt))
2286 return false;
2287
2288 /* to avoid potential deadlock or bitmap corruption,
2289 * in various places, we only allow new application io
2290 * to start during "stable" states. */
2291
2292 /* no new io accepted when attaching or detaching the disk */
2293 if (!drbd_state_is_stable(device))
2294 return false;
2295
2296 /* since some older kernels don't have atomic_add_unless,
2297 * and we are within the spinlock anyways, we have this workaround. */
2298 if (atomic_read(&device->ap_bio_cnt) > mxb)
2299 return false;
2300 if (test_bit(BITMAP_IO, &device->flags))
2301 return false;
2302 return true;
2303 }
2304
inc_ap_bio_cond(struct drbd_device * device)2305 static inline bool inc_ap_bio_cond(struct drbd_device *device)
2306 {
2307 bool rv = false;
2308
2309 spin_lock_irq(&device->resource->req_lock);
2310 rv = may_inc_ap_bio(device);
2311 if (rv)
2312 atomic_inc(&device->ap_bio_cnt);
2313 spin_unlock_irq(&device->resource->req_lock);
2314
2315 return rv;
2316 }
2317
inc_ap_bio(struct drbd_device * device)2318 static inline void inc_ap_bio(struct drbd_device *device)
2319 {
2320 /* we wait here
2321 * as long as the device is suspended
2322 * until the bitmap is no longer on the fly during connection
2323 * handshake as long as we would exceed the max_buffer limit.
2324 *
2325 * to avoid races with the reconnect code,
2326 * we need to atomic_inc within the spinlock. */
2327
2328 wait_event(device->misc_wait, inc_ap_bio_cond(device));
2329 }
2330
dec_ap_bio(struct drbd_device * device)2331 static inline void dec_ap_bio(struct drbd_device *device)
2332 {
2333 int mxb = drbd_get_max_buffers(device);
2334 int ap_bio = atomic_dec_return(&device->ap_bio_cnt);
2335
2336 D_ASSERT(device, ap_bio >= 0);
2337
2338 if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) {
2339 if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags))
2340 drbd_queue_work(&first_peer_device(device)->
2341 connection->sender_work,
2342 &device->bm_io_work.w);
2343 }
2344
2345 /* this currently does wake_up for every dec_ap_bio!
2346 * maybe rather introduce some type of hysteresis?
2347 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2348 if (ap_bio < mxb)
2349 wake_up(&device->misc_wait);
2350 }
2351
verify_can_do_stop_sector(struct drbd_device * device)2352 static inline bool verify_can_do_stop_sector(struct drbd_device *device)
2353 {
2354 return first_peer_device(device)->connection->agreed_pro_version >= 97 &&
2355 first_peer_device(device)->connection->agreed_pro_version != 100;
2356 }
2357
drbd_set_ed_uuid(struct drbd_device * device,u64 val)2358 static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val)
2359 {
2360 int changed = device->ed_uuid != val;
2361 device->ed_uuid = val;
2362 return changed;
2363 }
2364
drbd_queue_order_type(struct drbd_device * device)2365 static inline int drbd_queue_order_type(struct drbd_device *device)
2366 {
2367 /* sorry, we currently have no working implementation
2368 * of distributed TCQ stuff */
2369 #ifndef QUEUE_ORDERED_NONE
2370 #define QUEUE_ORDERED_NONE 0
2371 #endif
2372 return QUEUE_ORDERED_NONE;
2373 }
2374
first_connection(struct drbd_resource * resource)2375 static inline struct drbd_connection *first_connection(struct drbd_resource *resource)
2376 {
2377 return list_first_entry_or_null(&resource->connections,
2378 struct drbd_connection, connections);
2379 }
2380
2381 #endif
2382