Commit | Line | Data |
---|---|---|
b411b363 PR |
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 <linux/compiler.h> | |
30 | #include <linux/types.h> | |
b411b363 PR |
31 | #include <linux/list.h> |
32 | #include <linux/sched.h> | |
33 | #include <linux/bitops.h> | |
34 | #include <linux/slab.h> | |
35 | #include <linux/crypto.h> | |
132cc538 | 36 | #include <linux/ratelimit.h> |
b411b363 PR |
37 | #include <linux/tcp.h> |
38 | #include <linux/mutex.h> | |
39 | #include <linux/major.h> | |
40 | #include <linux/blkdev.h> | |
41 | #include <linux/genhd.h> | |
062e879c | 42 | #include <linux/idr.h> |
b411b363 PR |
43 | #include <net/tcp.h> |
44 | #include <linux/lru_cache.h> | |
70c71606 | 45 | #include <linux/prefetch.h> |
3b98c0c2 | 46 | #include <linux/drbd_genl_api.h> |
b8907339 | 47 | #include <linux/drbd.h> |
d9f65229 | 48 | #include "drbd_strings.h" |
b8907339 | 49 | #include "drbd_state.h" |
a3603a6e | 50 | #include "drbd_protocol.h" |
b411b363 PR |
51 | |
52 | #ifdef __CHECKER__ | |
53 | # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr"))) | |
54 | # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read"))) | |
55 | # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write"))) | |
56 | # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call"))) | |
57 | #else | |
58 | # define __protected_by(x) | |
59 | # define __protected_read_by(x) | |
60 | # define __protected_write_by(x) | |
61 | # define __must_hold(x) | |
62 | #endif | |
63 | ||
b411b363 PR |
64 | /* module parameter, defined in drbd_main.c */ |
65 | extern unsigned int minor_count; | |
90ab5ee9 RR |
66 | extern bool disable_sendpage; |
67 | extern bool allow_oos; | |
b30ab791 | 68 | void tl_abort_disk_io(struct drbd_device *device); |
b411b363 PR |
69 | |
70 | #ifdef CONFIG_DRBD_FAULT_INJECTION | |
71 | extern int enable_faults; | |
72 | extern int fault_rate; | |
73 | extern int fault_devs; | |
74 | #endif | |
75 | ||
76 | extern char usermode_helper[]; | |
77 | ||
78 | ||
b411b363 PR |
79 | /* I don't remember why XCPU ... |
80 | * This is used to wake the asender, | |
81 | * and to interrupt sending the sending task | |
82 | * on disconnect. | |
83 | */ | |
84 | #define DRBD_SIG SIGXCPU | |
85 | ||
86 | /* This is used to stop/restart our threads. | |
87 | * Cannot use SIGTERM nor SIGKILL, since these | |
88 | * are sent out by init on runlevel changes | |
89 | * I choose SIGHUP for now. | |
90 | */ | |
91 | #define DRBD_SIGKILL SIGHUP | |
92 | ||
b411b363 PR |
93 | #define ID_IN_SYNC (4711ULL) |
94 | #define ID_OUT_OF_SYNC (4712ULL) | |
b411b363 | 95 | #define ID_SYNCER (-1ULL) |
579b57ed | 96 | |
4a23f264 | 97 | #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL) |
b411b363 | 98 | |
54761697 | 99 | struct drbd_device; |
bde89a9e | 100 | struct drbd_connection; |
b411b363 | 101 | |
3b52beff AG |
102 | #define __drbd_printk_device(level, device, fmt, args...) \ |
103 | dev_printk(level, disk_to_dev((device)->vdisk), fmt, ## args) | |
104 | #define __drbd_printk_peer_device(level, peer_device, fmt, args...) \ | |
105 | dev_printk(level, disk_to_dev((peer_device)->device->vdisk), fmt, ## args) | |
106 | #define __drbd_printk_resource(level, resource, fmt, args...) \ | |
107 | printk(level "drbd %s: " fmt, (resource)->name, ## args) | |
108 | #define __drbd_printk_connection(level, connection, fmt, args...) \ | |
109 | printk(level "drbd %s: " fmt, (connection)->resource->name, ## args) | |
110 | ||
111 | void drbd_printk_with_wrong_object_type(void); | |
112 | ||
113 | #define __drbd_printk_if_same_type(obj, type, func, level, fmt, args...) \ | |
114 | (__builtin_types_compatible_p(typeof(obj), type) || \ | |
115 | __builtin_types_compatible_p(typeof(obj), const type)), \ | |
116 | func(level, (const type)(obj), fmt, ## args) | |
117 | ||
118 | #define drbd_printk(level, obj, fmt, args...) \ | |
119 | __builtin_choose_expr( \ | |
120 | __drbd_printk_if_same_type(obj, struct drbd_device *, \ | |
121 | __drbd_printk_device, level, fmt, ## args), \ | |
122 | __builtin_choose_expr( \ | |
123 | __drbd_printk_if_same_type(obj, struct drbd_resource *, \ | |
124 | __drbd_printk_resource, level, fmt, ## args), \ | |
125 | __builtin_choose_expr( \ | |
126 | __drbd_printk_if_same_type(obj, struct drbd_connection *, \ | |
127 | __drbd_printk_connection, level, fmt, ## args), \ | |
128 | __builtin_choose_expr( \ | |
129 | __drbd_printk_if_same_type(obj, struct drbd_peer_device *, \ | |
130 | __drbd_printk_peer_device, level, fmt, ## args), \ | |
131 | drbd_printk_with_wrong_object_type())))) | |
132 | ||
133 | #define drbd_dbg(obj, fmt, args...) \ | |
134 | drbd_printk(KERN_DEBUG, obj, fmt, ## args) | |
135 | #define drbd_alert(obj, fmt, args...) \ | |
136 | drbd_printk(KERN_ALERT, obj, fmt, ## args) | |
137 | #define drbd_err(obj, fmt, args...) \ | |
138 | drbd_printk(KERN_ERR, obj, fmt, ## args) | |
139 | #define drbd_warn(obj, fmt, args...) \ | |
140 | drbd_printk(KERN_WARNING, obj, fmt, ## args) | |
141 | #define drbd_info(obj, fmt, args...) \ | |
142 | drbd_printk(KERN_INFO, obj, fmt, ## args) | |
143 | #define drbd_emerg(obj, fmt, args...) \ | |
144 | drbd_printk(KERN_EMERG, obj, fmt, ## args) | |
d0180171 AG |
145 | |
146 | #define dynamic_drbd_dbg(device, fmt, args...) \ | |
147 | dynamic_dev_dbg(disk_to_dev(device->vdisk), fmt, ## args) | |
b411b363 | 148 | |
0b0ba1ef AG |
149 | #define D_ASSERT(device, exp) do { \ |
150 | if (!(exp)) \ | |
151 | drbd_err(device, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__); \ | |
152 | } while (0) | |
b411b363 | 153 | |
841ce241 AG |
154 | /** |
155 | * expect - Make an assertion | |
156 | * | |
157 | * Unlike the assert macro, this macro returns a boolean result. | |
158 | */ | |
159 | #define expect(exp) ({ \ | |
160 | bool _bool = (exp); \ | |
161 | if (!_bool) \ | |
d0180171 | 162 | drbd_err(device, "ASSERTION %s FAILED in %s\n", \ |
841ce241 AG |
163 | #exp, __func__); \ |
164 | _bool; \ | |
165 | }) | |
b411b363 PR |
166 | |
167 | /* Defines to control fault insertion */ | |
168 | enum { | |
169 | DRBD_FAULT_MD_WR = 0, /* meta data write */ | |
170 | DRBD_FAULT_MD_RD = 1, /* read */ | |
171 | DRBD_FAULT_RS_WR = 2, /* resync */ | |
172 | DRBD_FAULT_RS_RD = 3, | |
173 | DRBD_FAULT_DT_WR = 4, /* data */ | |
174 | DRBD_FAULT_DT_RD = 5, | |
175 | DRBD_FAULT_DT_RA = 6, /* data read ahead */ | |
176 | DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */ | |
177 | DRBD_FAULT_AL_EE = 8, /* alloc ee */ | |
6b4388ac | 178 | DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */ |
b411b363 PR |
179 | |
180 | DRBD_FAULT_MAX, | |
181 | }; | |
182 | ||
b411b363 | 183 | extern unsigned int |
b30ab791 | 184 | _drbd_insert_fault(struct drbd_device *device, unsigned int type); |
0cf9d27e | 185 | |
b411b363 | 186 | static inline int |
b30ab791 | 187 | drbd_insert_fault(struct drbd_device *device, unsigned int type) { |
0cf9d27e | 188 | #ifdef CONFIG_DRBD_FAULT_INJECTION |
b411b363 PR |
189 | return fault_rate && |
190 | (enable_faults & (1<<type)) && | |
b30ab791 | 191 | _drbd_insert_fault(device, type); |
b411b363 | 192 | #else |
0cf9d27e | 193 | return 0; |
b411b363 | 194 | #endif |
0cf9d27e | 195 | } |
b411b363 PR |
196 | |
197 | /* integer division, round _UP_ to the next integer */ | |
198 | #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0)) | |
199 | /* usual integer division */ | |
200 | #define div_floor(A, B) ((A)/(B)) | |
201 | ||
b411b363 | 202 | extern struct ratelimit_state drbd_ratelimit_state; |
05a10ec7 | 203 | extern struct idr drbd_devices; /* RCU, updates: genl_lock() */ |
77c556f6 | 204 | extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */ |
b411b363 | 205 | |
d8763023 | 206 | extern const char *cmdname(enum drbd_packet cmd); |
b411b363 PR |
207 | |
208 | /* for sending/receiving the bitmap, | |
209 | * possibly in some encoding scheme */ | |
210 | struct bm_xfer_ctx { | |
211 | /* "const" | |
212 | * stores total bits and long words | |
213 | * of the bitmap, so we don't need to | |
214 | * call the accessor functions over and again. */ | |
215 | unsigned long bm_bits; | |
216 | unsigned long bm_words; | |
217 | /* during xfer, current position within the bitmap */ | |
218 | unsigned long bit_offset; | |
219 | unsigned long word_offset; | |
220 | ||
221 | /* statistics; index: (h->command == P_BITMAP) */ | |
222 | unsigned packets[2]; | |
223 | unsigned bytes[2]; | |
224 | }; | |
225 | ||
b30ab791 | 226 | extern void INFO_bm_xfer_stats(struct drbd_device *device, |
b411b363 PR |
227 | const char *direction, struct bm_xfer_ctx *c); |
228 | ||
229 | static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c) | |
230 | { | |
231 | /* word_offset counts "native long words" (32 or 64 bit), | |
232 | * aligned at 64 bit. | |
233 | * Encoded packet may end at an unaligned bit offset. | |
234 | * In case a fallback clear text packet is transmitted in | |
235 | * between, we adjust this offset back to the last 64bit | |
236 | * aligned "native long word", which makes coding and decoding | |
237 | * the plain text bitmap much more convenient. */ | |
238 | #if BITS_PER_LONG == 64 | |
239 | c->word_offset = c->bit_offset >> 6; | |
240 | #elif BITS_PER_LONG == 32 | |
241 | c->word_offset = c->bit_offset >> 5; | |
242 | c->word_offset &= ~(1UL); | |
243 | #else | |
244 | # error "unsupported BITS_PER_LONG" | |
245 | #endif | |
246 | } | |
247 | ||
bde89a9e | 248 | extern unsigned int drbd_header_size(struct drbd_connection *connection); |
b411b363 | 249 | |
b411b363 PR |
250 | /**********************************************************************/ |
251 | enum drbd_thread_state { | |
e77a0a5c AG |
252 | NONE, |
253 | RUNNING, | |
254 | EXITING, | |
255 | RESTARTING | |
b411b363 PR |
256 | }; |
257 | ||
258 | struct drbd_thread { | |
259 | spinlock_t t_lock; | |
260 | struct task_struct *task; | |
261 | struct completion stop; | |
262 | enum drbd_thread_state t_state; | |
263 | int (*function) (struct drbd_thread *); | |
2457b6d5 | 264 | struct drbd_resource *resource; |
bde89a9e | 265 | struct drbd_connection *connection; |
b411b363 | 266 | int reset_cpu_mask; |
c60b0251 | 267 | const char *name; |
b411b363 PR |
268 | }; |
269 | ||
270 | static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi) | |
271 | { | |
272 | /* THINK testing the t_state seems to be uncritical in all cases | |
273 | * (but thread_{start,stop}), so we can read it *without* the lock. | |
274 | * --lge */ | |
275 | ||
276 | smp_rmb(); | |
277 | return thi->t_state; | |
278 | } | |
279 | ||
b411b363 PR |
280 | struct drbd_work { |
281 | struct list_head list; | |
309a8348 | 282 | int (*cb)(struct drbd_work *, int cancel); |
84b8c06b AG |
283 | }; |
284 | ||
285 | struct drbd_device_work { | |
286 | struct drbd_work w; | |
287 | struct drbd_device *device; | |
b411b363 PR |
288 | }; |
289 | ||
ace652ac AG |
290 | #include "drbd_interval.h" |
291 | ||
54761697 | 292 | extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *); |
7be8da07 | 293 | |
b411b363 PR |
294 | struct drbd_request { |
295 | struct drbd_work w; | |
84b8c06b | 296 | struct drbd_device *device; |
b411b363 PR |
297 | |
298 | /* if local IO is not allowed, will be NULL. | |
299 | * if local IO _is_ allowed, holds the locally submitted bio clone, | |
300 | * or, after local IO completion, the ERR_PTR(error). | |
fcefa62e | 301 | * see drbd_request_endio(). */ |
b411b363 PR |
302 | struct bio *private_bio; |
303 | ||
ace652ac | 304 | struct drbd_interval i; |
b411b363 | 305 | |
b6dd1a89 | 306 | /* epoch: used to check on "completion" whether this req was in |
b411b363 | 307 | * the current epoch, and we therefore have to close it, |
b6dd1a89 LE |
308 | * causing a p_barrier packet to be send, starting a new epoch. |
309 | * | |
310 | * This corresponds to "barrier" in struct p_barrier[_ack], | |
311 | * and to "barrier_nr" in struct drbd_epoch (and various | |
312 | * comments/function parameters/local variable names). | |
b411b363 | 313 | */ |
b6dd1a89 | 314 | unsigned int epoch; |
b411b363 | 315 | |
b411b363 PR |
316 | struct list_head tl_requests; /* ring list in the transfer log */ |
317 | struct bio *master_bio; /* master bio pointer */ | |
e5f891b2 | 318 | |
844a6ae7 LE |
319 | /* see struct drbd_device */ |
320 | struct list_head req_pending_master_completion; | |
321 | struct list_head req_pending_local; | |
322 | ||
e5f891b2 LE |
323 | /* for generic IO accounting */ |
324 | unsigned long start_jif; | |
325 | ||
326 | /* for DRBD internal statistics */ | |
327 | ||
328 | /* Minimal set of time stamps to determine if we wait for activity log | |
329 | * transactions, local disk or peer. 32 bit "jiffies" are good enough, | |
330 | * we don't expect a DRBD request to be stalled for several month. | |
331 | */ | |
332 | ||
333 | /* before actual request processing */ | |
334 | unsigned long in_actlog_jif; | |
335 | ||
336 | /* local disk */ | |
337 | unsigned long pre_submit_jif; | |
338 | ||
339 | /* per connection */ | |
340 | unsigned long pre_send_jif; | |
341 | unsigned long acked_jif; | |
342 | unsigned long net_done_jif; | |
343 | ||
344 | /* Possibly even more detail to track each phase: | |
345 | * master_completion_jif | |
346 | * how long did it take to complete the master bio | |
347 | * (application visible latency) | |
348 | * allocated_jif | |
349 | * how long the master bio was blocked until we finally allocated | |
350 | * a tracking struct | |
351 | * in_actlog_jif | |
352 | * how long did we wait for activity log transactions | |
353 | * | |
354 | * net_queued_jif | |
355 | * when did we finally queue it for sending | |
356 | * pre_send_jif | |
357 | * when did we start sending it | |
358 | * post_send_jif | |
359 | * how long did we block in the network stack trying to send it | |
360 | * acked_jif | |
361 | * when did we receive (or fake, in protocol A) a remote ACK | |
362 | * net_done_jif | |
363 | * when did we receive final acknowledgement (P_BARRIER_ACK), | |
364 | * or decide, e.g. on connection loss, that we do no longer expect | |
365 | * anything from this peer for this request. | |
366 | * | |
367 | * pre_submit_jif | |
368 | * post_sub_jif | |
369 | * when did we start submiting to the lower level device, | |
370 | * and how long did we block in that submit function | |
371 | * local_completion_jif | |
372 | * how long did it take the lower level device to complete this request | |
373 | */ | |
374 | ||
b411b363 | 375 | |
b406777e LE |
376 | /* once it hits 0, we may complete the master_bio */ |
377 | atomic_t completion_ref; | |
378 | /* once it hits 0, we may destroy this drbd_request object */ | |
379 | struct kref kref; | |
b411b363 | 380 | |
a0d856df | 381 | unsigned rq_state; /* see comments above _req_mod() */ |
b411b363 | 382 | }; |
b411b363 PR |
383 | |
384 | struct drbd_epoch { | |
bde89a9e | 385 | struct drbd_connection *connection; |
b411b363 PR |
386 | struct list_head list; |
387 | unsigned int barrier_nr; | |
388 | atomic_t epoch_size; /* increased on every request added. */ | |
389 | atomic_t active; /* increased on every req. added, and dec on every finished. */ | |
390 | unsigned long flags; | |
391 | }; | |
392 | ||
de0b2e69 RK |
393 | /* Prototype declaration of function defined in drbd_receiver.c */ |
394 | int drbdd_init(struct drbd_thread *); | |
395 | int drbd_asender(struct drbd_thread *); | |
396 | ||
b411b363 PR |
397 | /* drbd_epoch flag bits */ |
398 | enum { | |
b411b363 | 399 | DE_HAVE_BARRIER_NUMBER, |
b411b363 PR |
400 | }; |
401 | ||
402 | enum epoch_event { | |
403 | EV_PUT, | |
404 | EV_GOT_BARRIER_NR, | |
b411b363 | 405 | EV_BECAME_LAST, |
b411b363 PR |
406 | EV_CLEANUP = 32, /* used as flag */ |
407 | }; | |
408 | ||
b411b363 PR |
409 | struct digest_info { |
410 | int digest_size; | |
411 | void *digest; | |
412 | }; | |
413 | ||
f6ffca9f | 414 | struct drbd_peer_request { |
a8cd15ba AG |
415 | struct drbd_work w; |
416 | struct drbd_peer_device *peer_device; | |
85719573 | 417 | struct drbd_epoch *epoch; /* for writes */ |
45bb912b LE |
418 | struct page *pages; |
419 | atomic_t pending_bios; | |
010f6e67 | 420 | struct drbd_interval i; |
45bb912b LE |
421 | /* see comments on ee flag bits below */ |
422 | unsigned long flags; | |
21ae5d7f | 423 | unsigned long submit_jif; |
85719573 PR |
424 | union { |
425 | u64 block_id; | |
426 | struct digest_info *digest; | |
427 | }; | |
45bb912b LE |
428 | }; |
429 | ||
430 | /* ee flag bits. | |
431 | * While corresponding bios are in flight, the only modification will be | |
432 | * set_bit WAS_ERROR, which has to be atomic. | |
433 | * If no bios are in flight yet, or all have been completed, | |
434 | * non-atomic modification to ee->flags is ok. | |
435 | */ | |
b411b363 PR |
436 | enum { |
437 | __EE_CALL_AL_COMPLETE_IO, | |
b411b363 | 438 | __EE_MAY_SET_IN_SYNC, |
45bb912b | 439 | |
a0fb3c47 LE |
440 | /* is this a TRIM aka REQ_DISCARD? */ |
441 | __EE_IS_TRIM, | |
442 | /* our lower level cannot handle trim, | |
443 | * and we want to fall back to zeroout instead */ | |
444 | __EE_IS_TRIM_USE_ZEROOUT, | |
445 | ||
45bb912b LE |
446 | /* In case a barrier failed, |
447 | * we need to resubmit without the barrier flag. */ | |
448 | __EE_RESUBMITTED, | |
449 | ||
6c852bec | 450 | /* we may have several bios per peer request. |
45bb912b LE |
451 | * if any of those fail, we set this flag atomically |
452 | * from the endio callback */ | |
453 | __EE_WAS_ERROR, | |
c36c3ced LE |
454 | |
455 | /* This ee has a pointer to a digest instead of a block id */ | |
456 | __EE_HAS_DIGEST, | |
7be8da07 AG |
457 | |
458 | /* Conflicting local requests need to be restarted after this request */ | |
459 | __EE_RESTART_REQUESTS, | |
303d1448 PR |
460 | |
461 | /* The peer wants a write ACK for this (wire proto C) */ | |
462 | __EE_SEND_WRITE_ACK, | |
302bdeae PR |
463 | |
464 | /* Is set when net_conf had two_primaries set while creating this peer_req */ | |
465 | __EE_IN_INTERVAL_TREE, | |
21ae5d7f LE |
466 | |
467 | /* for debugfs: */ | |
468 | /* has this been submitted, or does it still wait for something else? */ | |
469 | __EE_SUBMITTED, | |
470 | ||
471 | /* this is/was a write request */ | |
472 | __EE_WRITE, | |
473 | ||
474 | /* this originates from application on peer | |
475 | * (not some resync or verify or other DRBD internal request) */ | |
476 | __EE_APPLICATION, | |
b411b363 PR |
477 | }; |
478 | #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO) | |
b411b363 | 479 | #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC) |
a0fb3c47 LE |
480 | #define EE_IS_TRIM (1<<__EE_IS_TRIM) |
481 | #define EE_IS_TRIM_USE_ZEROOUT (1<<__EE_IS_TRIM_USE_ZEROOUT) | |
482 | #define EE_RESUBMITTED (1<<__EE_RESUBMITTED) | |
45bb912b | 483 | #define EE_WAS_ERROR (1<<__EE_WAS_ERROR) |
c36c3ced | 484 | #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST) |
7be8da07 | 485 | #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS) |
303d1448 | 486 | #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK) |
302bdeae | 487 | #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE) |
21ae5d7f LE |
488 | #define EE_SUBMITTED (1<<__EE_SUBMITTED) |
489 | #define EE_WRITE (1<<__EE_WRITE) | |
490 | #define EE_APPLICATION (1<<__EE_APPLICATION) | |
b411b363 | 491 | |
b30ab791 | 492 | /* flag bits per device */ |
b411b363 | 493 | enum { |
b411b363 PR |
494 | UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */ |
495 | MD_DIRTY, /* current uuids and flags not yet on disk */ | |
b411b363 | 496 | USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */ |
b411b363 PR |
497 | CL_ST_CHG_SUCCESS, |
498 | CL_ST_CHG_FAIL, | |
499 | CRASHED_PRIMARY, /* This node was a crashed primary. | |
500 | * Gets cleared when the state.conn | |
501 | * goes into C_CONNECTED state. */ | |
b411b363 PR |
502 | CONSIDER_RESYNC, |
503 | ||
a8a4e51e | 504 | MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */ |
5ab7d2c0 | 505 | |
b411b363 PR |
506 | SUSPEND_IO, /* suspend application io */ |
507 | BITMAP_IO, /* suspend application io; | |
508 | once no more io in flight, start bitmap io */ | |
509 | BITMAP_IO_QUEUED, /* Started bitmap IO */ | |
a2a3c74f LE |
510 | WAS_IO_ERROR, /* Local disk failed, returned IO error */ |
511 | WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */ | |
383606e0 | 512 | FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */ |
b411b363 | 513 | RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */ |
b411b363 PR |
514 | RESIZE_PENDING, /* Size change detected locally, waiting for the response from |
515 | * the peer, if it changed there as well. */ | |
43a5182c | 516 | NEW_CUR_UUID, /* Create new current UUID when thawing IO */ |
0778286a | 517 | AL_SUSPENDED, /* Activity logging is currently suspended. */ |
370a43e7 | 518 | AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */ |
e64a3294 | 519 | B_RS_H_DONE, /* Before resync handler done (already executed) */ |
08b165ba | 520 | DISCARD_MY_DATA, /* discard_my_data flag per volume */ |
380207d0 | 521 | READ_BALANCE_RR, |
e334f550 | 522 | |
f418815f LE |
523 | FLUSH_PENDING, /* if set, device->flush_jif is when we submitted that flush |
524 | * from drbd_flush_after_epoch() */ | |
525 | ||
e334f550 LE |
526 | /* cleared only after backing device related structures have been destroyed. */ |
527 | GOING_DISKLESS, /* Disk is being detached, because of io-error, or admin request. */ | |
528 | ||
529 | /* to be used in drbd_device_post_work() */ | |
530 | GO_DISKLESS, /* tell worker to schedule cleanup before detach */ | |
531 | DESTROY_DISK, /* tell worker to close backing devices and destroy related structures. */ | |
ac0acb9e LE |
532 | MD_SYNC, /* tell worker to call drbd_md_sync() */ |
533 | RS_START, /* tell worker to start resync/OV */ | |
e334f550 LE |
534 | RS_PROGRESS, /* tell worker that resync made significant progress */ |
535 | RS_DONE, /* tell worker that resync is done */ | |
b411b363 PR |
536 | }; |
537 | ||
54761697 | 538 | struct drbd_bitmap; /* opaque for drbd_device */ |
b411b363 | 539 | |
20ceb2b2 LE |
540 | /* definition of bits in bm_flags to be used in drbd_bm_lock |
541 | * and drbd_bitmap_io and friends. */ | |
542 | enum bm_flag { | |
543 | /* do we need to kfree, or vfree bm_pages? */ | |
544 | BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */ | |
545 | ||
546 | /* currently locked for bulk operation */ | |
0e8488ad | 547 | BM_LOCKED_MASK = 0xf, |
20ceb2b2 LE |
548 | |
549 | /* in detail, that is: */ | |
550 | BM_DONT_CLEAR = 0x1, | |
551 | BM_DONT_SET = 0x2, | |
552 | BM_DONT_TEST = 0x4, | |
553 | ||
0e8488ad LE |
554 | /* so we can mark it locked for bulk operation, |
555 | * and still allow all non-bulk operations */ | |
556 | BM_IS_LOCKED = 0x8, | |
557 | ||
20ceb2b2 | 558 | /* (test bit, count bit) allowed (common case) */ |
0e8488ad | 559 | BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED, |
20ceb2b2 LE |
560 | |
561 | /* testing bits, as well as setting new bits allowed, but clearing bits | |
562 | * would be unexpected. Used during bitmap receive. Setting new bits | |
563 | * requires sending of "out-of-sync" information, though. */ | |
0e8488ad | 564 | BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED, |
20ceb2b2 | 565 | |
0e8488ad LE |
566 | /* for drbd_bm_write_copy_pages, everything is allowed, |
567 | * only concurrent bulk operations are locked out. */ | |
568 | BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED, | |
20ceb2b2 LE |
569 | }; |
570 | ||
b411b363 PR |
571 | struct drbd_work_queue { |
572 | struct list_head q; | |
b411b363 | 573 | spinlock_t q_lock; /* to protect the list. */ |
8c0785a5 | 574 | wait_queue_head_t q_wait; |
b411b363 PR |
575 | }; |
576 | ||
577 | struct drbd_socket { | |
b411b363 PR |
578 | struct mutex mutex; |
579 | struct socket *socket; | |
580 | /* this way we get our | |
581 | * send/receive buffers off the stack */ | |
5a87d920 | 582 | void *sbuf; |
e6ef8a5c | 583 | void *rbuf; |
b411b363 PR |
584 | }; |
585 | ||
586 | struct drbd_md { | |
587 | u64 md_offset; /* sector offset to 'super' block */ | |
588 | ||
589 | u64 la_size_sect; /* last agreed size, unit sectors */ | |
9f2247bb | 590 | spinlock_t uuid_lock; |
b411b363 PR |
591 | u64 uuid[UI_SIZE]; |
592 | u64 device_uuid; | |
593 | u32 flags; | |
594 | u32 md_size_sect; | |
595 | ||
ae8bf312 | 596 | s32 al_offset; /* signed relative sector offset to activity log */ |
b411b363 | 597 | s32 bm_offset; /* signed relative sector offset to bitmap */ |
3a4d4eb3 LE |
598 | |
599 | /* cached value of bdev->disk_conf->meta_dev_idx (see below) */ | |
600 | s32 meta_dev_idx; | |
601 | ||
602 | /* see al_tr_number_to_on_disk_sector() */ | |
603 | u32 al_stripes; | |
604 | u32 al_stripe_size_4k; | |
605 | u32 al_size_4k; /* cached product of the above */ | |
b411b363 PR |
606 | }; |
607 | ||
b411b363 PR |
608 | struct drbd_backing_dev { |
609 | struct block_device *backing_bdev; | |
610 | struct block_device *md_bdev; | |
b411b363 | 611 | struct drbd_md md; |
0500813f | 612 | struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */ |
b411b363 PR |
613 | sector_t known_size; /* last known size of that backing device */ |
614 | }; | |
615 | ||
616 | struct drbd_md_io { | |
e37d2438 LE |
617 | struct page *page; |
618 | unsigned long start_jif; /* last call to drbd_md_get_buffer */ | |
619 | unsigned long submit_jif; /* last _drbd_md_sync_page_io() submit */ | |
620 | const char *current_use; | |
621 | atomic_t in_use; | |
0c464425 | 622 | unsigned int done; |
b411b363 PR |
623 | int error; |
624 | }; | |
625 | ||
626 | struct bm_io_work { | |
627 | struct drbd_work w; | |
628 | char *why; | |
20ceb2b2 | 629 | enum bm_flag flags; |
b30ab791 AG |
630 | int (*io_fn)(struct drbd_device *device); |
631 | void (*done)(struct drbd_device *device, int rv); | |
b411b363 PR |
632 | }; |
633 | ||
634 | enum write_ordering_e { | |
635 | WO_none, | |
636 | WO_drain_io, | |
637 | WO_bdev_flush, | |
b411b363 PR |
638 | }; |
639 | ||
778f271d | 640 | struct fifo_buffer { |
778f271d PR |
641 | unsigned int head_index; |
642 | unsigned int size; | |
9958c857 PR |
643 | int total; /* sum of all values */ |
644 | int values[0]; | |
778f271d | 645 | }; |
9958c857 | 646 | extern struct fifo_buffer *fifo_alloc(int fifo_size); |
778f271d | 647 | |
bde89a9e | 648 | /* flag bits per connection */ |
01a311a5 PR |
649 | enum { |
650 | NET_CONGESTED, /* The data socket is congested */ | |
427c0434 | 651 | RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */ |
e43ef195 | 652 | SEND_PING, /* whether asender should send a ping asap */ |
808e37b8 | 653 | SIGNAL_ASENDER, /* whether asender wants to be interrupted */ |
2a67d8b9 | 654 | GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */ |
4d0fc3fd | 655 | CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */ |
fc3b10a4 PR |
656 | CONN_WD_ST_CHG_OKAY, |
657 | CONN_WD_ST_CHG_FAIL, | |
8169e41b | 658 | CONN_DRY_RUN, /* Expect disconnect after resync handshake. */ |
6936fcb4 | 659 | CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */ |
a1096a6e | 660 | STATE_SENT, /* Do not change state/UUIDs while this is set */ |
6f3465ed LE |
661 | CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC) |
662 | * pending, from drbd worker context. | |
663 | * If set, bdi_write_congested() returns true, | |
664 | * so shrink_page_list() would not recurse into, | |
665 | * and potentially deadlock on, this drbd worker. | |
666 | */ | |
b66623e3 | 667 | DISCONNECT_SENT, |
e334f550 LE |
668 | |
669 | DEVICE_WORK_PENDING, /* tell worker that some device has pending work */ | |
01a311a5 PR |
670 | }; |
671 | ||
77c556f6 AG |
672 | struct drbd_resource { |
673 | char *name; | |
4d3d5aa8 LE |
674 | #ifdef CONFIG_DEBUG_FS |
675 | struct dentry *debugfs_res; | |
676 | struct dentry *debugfs_res_volumes; | |
677 | struct dentry *debugfs_res_connections; | |
678 | struct dentry *debugfs_res_in_flight_summary; | |
679 | #endif | |
77c556f6 | 680 | struct kref kref; |
803ea134 | 681 | struct idr devices; /* volume number to device mapping */ |
77c556f6 AG |
682 | struct list_head connections; |
683 | struct list_head resources; | |
eb6bea67 | 684 | struct res_opts res_opts; |
0500813f | 685 | struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */ |
9e276872 | 686 | struct mutex adm_mutex; /* mutex to serialize administrative requests */ |
0500813f | 687 | spinlock_t req_lock; |
6bbf53ca AG |
688 | |
689 | unsigned susp:1; /* IO suspended by user */ | |
690 | unsigned susp_nod:1; /* IO suspended because no data */ | |
691 | unsigned susp_fen:1; /* IO suspended because fence peer handler runs */ | |
625a6ba2 | 692 | |
e9526580 PR |
693 | enum write_ordering_e write_ordering; |
694 | ||
625a6ba2 | 695 | cpumask_var_t cpu_mask; |
77c556f6 AG |
696 | }; |
697 | ||
944410e9 LE |
698 | struct drbd_thread_timing_details |
699 | { | |
700 | unsigned long start_jif; | |
701 | void *cb_addr; | |
702 | const char *caller_fn; | |
703 | unsigned int line; | |
704 | unsigned int cb_nr; | |
705 | }; | |
706 | ||
77c556f6 AG |
707 | struct drbd_connection { |
708 | struct list_head connections; | |
709 | struct drbd_resource *resource; | |
4d3d5aa8 LE |
710 | #ifdef CONFIG_DEBUG_FS |
711 | struct dentry *debugfs_conn; | |
712 | struct dentry *debugfs_conn_callback_history; | |
713 | struct dentry *debugfs_conn_oldest_requests; | |
714 | #endif | |
9dc9fbb3 | 715 | struct kref kref; |
c06ece6b | 716 | struct idr peer_devices; /* volume number to peer device mapping */ |
8410da8f PR |
717 | enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */ |
718 | struct mutex cstate_mutex; /* Protects graceful disconnects */ | |
28e448bb | 719 | unsigned int connect_cnt; /* Inc each time a connection is established */ |
2111438b | 720 | |
062e879c | 721 | unsigned long flags; |
44ed167d | 722 | struct net_conf *net_conf; /* content protected by rcu */ |
91fd4dad | 723 | wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */ |
e42325a5 | 724 | |
089c075d AG |
725 | struct sockaddr_storage my_addr; |
726 | int my_addr_len; | |
727 | struct sockaddr_storage peer_addr; | |
728 | int peer_addr_len; | |
729 | ||
e42325a5 PR |
730 | struct drbd_socket data; /* data/barrier/cstate/parameter packets */ |
731 | struct drbd_socket meta; /* ping/ack (metadata) packets */ | |
31890f4a | 732 | int agreed_pro_version; /* actually used protocol version */ |
20c68fde | 733 | u32 agreed_features; |
31890f4a PR |
734 | unsigned long last_received; /* in jiffies, either socket */ |
735 | unsigned int ko_count; | |
e6b3ea83 | 736 | |
b6dd1a89 | 737 | struct list_head transfer_log; /* all requests not yet fully processed */ |
87eeee41 | 738 | |
a0638456 | 739 | struct crypto_hash *cram_hmac_tfm; |
bde89a9e | 740 | struct crypto_hash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */ |
036b17ea | 741 | struct crypto_hash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */ |
f399002e LE |
742 | struct crypto_hash *csums_tfm; |
743 | struct crypto_hash *verify_tfm; | |
a0638456 PR |
744 | void *int_dig_in; |
745 | void *int_dig_vv; | |
746 | ||
b6dd1a89 | 747 | /* receiver side */ |
12038a3a PR |
748 | struct drbd_epoch *current_epoch; |
749 | spinlock_t epoch_lock; | |
750 | unsigned int epochs; | |
b379c41e | 751 | atomic_t current_tle_nr; /* transfer log epoch number */ |
b6dd1a89 | 752 | unsigned current_tle_writes; /* writes seen within this tl epoch */ |
4b0007c0 | 753 | |
07be15b1 | 754 | unsigned long last_reconnect_jif; |
e6b3ea83 PR |
755 | struct drbd_thread receiver; |
756 | struct drbd_thread worker; | |
757 | struct drbd_thread asender; | |
b6dd1a89 | 758 | |
7753a4c1 LE |
759 | /* cached pointers, |
760 | * so we can look up the oldest pending requests more quickly. | |
761 | * protected by resource->req_lock */ | |
762 | struct drbd_request *req_next; /* DRBD 9: todo.req_next */ | |
763 | struct drbd_request *req_ack_pending; | |
764 | struct drbd_request *req_not_net_done; | |
765 | ||
b6dd1a89 | 766 | /* sender side */ |
d5b27b01 | 767 | struct drbd_work_queue sender_work; |
b6dd1a89 | 768 | |
944410e9 LE |
769 | #define DRBD_THREAD_DETAILS_HIST 16 |
770 | unsigned int w_cb_nr; /* keeps counting up */ | |
771 | unsigned int r_cb_nr; /* keeps counting up */ | |
772 | struct drbd_thread_timing_details w_timing_details[DRBD_THREAD_DETAILS_HIST]; | |
773 | struct drbd_thread_timing_details r_timing_details[DRBD_THREAD_DETAILS_HIST]; | |
774 | ||
b6dd1a89 LE |
775 | struct { |
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; | |
778f271d PR |
788 | }; |
789 | ||
944410e9 LE |
790 | void __update_timing_details( |
791 | struct drbd_thread_timing_details *tdp, | |
792 | unsigned int *cb_nr, | |
793 | void *cb, | |
794 | const char *fn, const unsigned int line); | |
795 | ||
796 | #define update_worker_timing_details(c, cb) \ | |
797 | __update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ ) | |
798 | #define update_receiver_timing_details(c, cb) \ | |
799 | __update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ ) | |
800 | ||
113fef9e LE |
801 | struct submit_worker { |
802 | struct workqueue_struct *wq; | |
803 | struct work_struct worker; | |
804 | ||
844a6ae7 | 805 | /* protected by ..->resource->req_lock */ |
113fef9e LE |
806 | struct list_head writes; |
807 | }; | |
808 | ||
a6b32bc3 AG |
809 | struct drbd_peer_device { |
810 | struct list_head peer_devices; | |
811 | struct drbd_device *device; | |
bde89a9e | 812 | struct drbd_connection *connection; |
4d3d5aa8 LE |
813 | #ifdef CONFIG_DEBUG_FS |
814 | struct dentry *debugfs_peer_dev; | |
815 | #endif | |
a6b32bc3 AG |
816 | }; |
817 | ||
818 | struct drbd_device { | |
d8628a86 | 819 | struct drbd_resource *resource; |
a6b32bc3 | 820 | struct list_head peer_devices; |
4ce49266 | 821 | struct list_head pending_bitmap_io; |
4d3d5aa8 LE |
822 | |
823 | unsigned long flush_jif; | |
824 | #ifdef CONFIG_DEBUG_FS | |
825 | struct dentry *debugfs_minor; | |
826 | struct dentry *debugfs_vol; | |
827 | struct dentry *debugfs_vol_oldest_requests; | |
828 | struct dentry *debugfs_vol_act_log_extents; | |
829 | struct dentry *debugfs_vol_resync_extents; | |
830 | struct dentry *debugfs_vol_data_gen_id; | |
831 | #endif | |
832 | ||
833 | unsigned int vnr; /* volume number within the connection */ | |
834 | unsigned int minor; /* device minor number */ | |
835 | ||
81fa2e67 | 836 | struct kref kref; |
2111438b | 837 | |
b411b363 PR |
838 | /* things that are stored as / read from meta data on disk */ |
839 | unsigned long flags; | |
840 | ||
841 | /* configured by drbdsetup */ | |
b411b363 PR |
842 | struct drbd_backing_dev *ldev __protected_by(local); |
843 | ||
844 | sector_t p_size; /* partner's disk size */ | |
845 | struct request_queue *rq_queue; | |
846 | struct block_device *this_bdev; | |
847 | struct gendisk *vdisk; | |
848 | ||
07be15b1 | 849 | unsigned long last_reattach_jif; |
84b8c06b AG |
850 | struct drbd_work resync_work; |
851 | struct drbd_work unplug_work; | |
b411b363 PR |
852 | struct timer_list resync_timer; |
853 | struct timer_list md_sync_timer; | |
370a43e7 | 854 | struct timer_list start_resync_timer; |
7fde2be9 | 855 | struct timer_list request_timer; |
b411b363 PR |
856 | |
857 | /* Used after attach while negotiating new disk state. */ | |
858 | union drbd_state new_state_tmp; | |
859 | ||
da9fbc27 | 860 | union drbd_dev_state state; |
b411b363 PR |
861 | wait_queue_head_t misc_wait; |
862 | wait_queue_head_t state_wait; /* upon each state change. */ | |
863 | unsigned int send_cnt; | |
864 | unsigned int recv_cnt; | |
865 | unsigned int read_cnt; | |
866 | unsigned int writ_cnt; | |
867 | unsigned int al_writ_cnt; | |
868 | unsigned int bm_writ_cnt; | |
869 | atomic_t ap_bio_cnt; /* Requests we need to complete */ | |
ad3fee79 | 870 | atomic_t ap_actlog_cnt; /* Requests waiting for activity log */ |
b411b363 PR |
871 | atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */ |
872 | atomic_t rs_pending_cnt; /* RS request/data packets on the wire */ | |
d942ae44 | 873 | atomic_t unacked_cnt; /* Need to send replies for */ |
b411b363 | 874 | atomic_t local_cnt; /* Waiting for local completion */ |
b2fb6dbe | 875 | |
dac1389c AG |
876 | /* Interval tree of pending local requests */ |
877 | struct rb_root read_requests; | |
de696716 | 878 | struct rb_root write_requests; |
b411b363 | 879 | |
844a6ae7 LE |
880 | /* for statistics and timeouts */ |
881 | /* [0] read, [1] write */ | |
882 | struct list_head pending_master_completion[2]; | |
883 | struct list_head pending_completion[2]; | |
884 | ||
aaaba345 LE |
885 | /* use checksums for *this* resync */ |
886 | bool use_csums; | |
4b0715f0 | 887 | /* blocks to resync in this run [unit BM_BLOCK_SIZE] */ |
b411b363 | 888 | unsigned long rs_total; |
4b0715f0 | 889 | /* number of resync blocks that failed in this run */ |
b411b363 PR |
890 | unsigned long rs_failed; |
891 | /* Syncer's start time [unit jiffies] */ | |
892 | unsigned long rs_start; | |
893 | /* cumulated time in PausedSyncX state [unit jiffies] */ | |
894 | unsigned long rs_paused; | |
1d7734a0 LE |
895 | /* skipped because csum was equal [unit BM_BLOCK_SIZE] */ |
896 | unsigned long rs_same_csum; | |
897 | #define DRBD_SYNC_MARKS 8 | |
898 | #define DRBD_SYNC_MARK_STEP (3*HZ) | |
b411b363 | 899 | /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */ |
1d7734a0 | 900 | unsigned long rs_mark_left[DRBD_SYNC_MARKS]; |
b411b363 | 901 | /* marks's time [unit jiffies] */ |
1d7734a0 LE |
902 | unsigned long rs_mark_time[DRBD_SYNC_MARKS]; |
903 | /* current index into rs_mark_{left,time} */ | |
904 | int rs_last_mark; | |
328e0f12 | 905 | unsigned long rs_last_bcast; /* [unit jiffies] */ |
b411b363 PR |
906 | |
907 | /* where does the admin want us to start? (sector) */ | |
908 | sector_t ov_start_sector; | |
02b91b55 | 909 | sector_t ov_stop_sector; |
b411b363 PR |
910 | /* where are we now? (sector) */ |
911 | sector_t ov_position; | |
912 | /* Start sector of out of sync range (to merge printk reporting). */ | |
913 | sector_t ov_last_oos_start; | |
914 | /* size of out-of-sync range in sectors. */ | |
915 | sector_t ov_last_oos_size; | |
916 | unsigned long ov_left; /* in bits */ | |
b411b363 | 917 | |
b411b363 PR |
918 | struct drbd_bitmap *bitmap; |
919 | unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */ | |
920 | ||
921 | /* Used to track operations of resync... */ | |
922 | struct lru_cache *resync; | |
923 | /* Number of locked elements in resync LRU */ | |
924 | unsigned int resync_locked; | |
925 | /* resync extent number waiting for application requests */ | |
926 | unsigned int resync_wenr; | |
927 | ||
928 | int open_cnt; | |
929 | u64 *p_uuid; | |
4b0007c0 | 930 | |
85719573 PR |
931 | struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */ |
932 | struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */ | |
18b75d75 AG |
933 | struct list_head done_ee; /* need to send P_WRITE_ACK */ |
934 | struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */ | |
b411b363 | 935 | struct list_head net_ee; /* zero-copy network send in progress */ |
b411b363 PR |
936 | |
937 | int next_barrier_nr; | |
b411b363 | 938 | struct list_head resync_reads; |
435f0740 LE |
939 | atomic_t pp_in_use; /* allocated from page pool */ |
940 | atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */ | |
b411b363 | 941 | wait_queue_head_t ee_wait; |
cc94c650 | 942 | struct drbd_md_io md_io; |
b411b363 PR |
943 | spinlock_t al_lock; |
944 | wait_queue_head_t al_wait; | |
945 | struct lru_cache *act_log; /* activity log */ | |
946 | unsigned int al_tr_number; | |
947 | int al_tr_cycle; | |
b411b363 PR |
948 | wait_queue_head_t seq_wait; |
949 | atomic_t packet_seq; | |
950 | unsigned int peer_seq; | |
951 | spinlock_t peer_seq_lock; | |
b411b363 | 952 | unsigned long comm_bm_set; /* communicated number of set bits. */ |
b411b363 PR |
953 | struct bm_io_work bm_io_work; |
954 | u64 ed_uuid; /* UUID of the exposed data */ | |
8410da8f | 955 | struct mutex own_state_mutex; |
a6b32bc3 | 956 | struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */ |
b411b363 | 957 | char congestion_reason; /* Why we where congested... */ |
1d7734a0 LE |
958 | atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */ |
959 | atomic_t rs_sect_ev; /* for submitted resync data rate, both */ | |
960 | int rs_last_sect_ev; /* counter to compare with */ | |
961 | int rs_last_events; /* counter of read or write "events" (unit sectors) | |
962 | * on the lower level device when we last looked. */ | |
963 | int c_sync_rate; /* current resync rate after syncer throttle magic */ | |
bde89a9e | 964 | struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */ |
778f271d | 965 | int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */ |
759fbdfb | 966 | atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */ |
db141b2f LE |
967 | unsigned int peer_max_bio_size; |
968 | unsigned int local_max_bio_size; | |
113fef9e LE |
969 | |
970 | /* any requests that would block in drbd_make_request() | |
971 | * are deferred to this single-threaded work queue */ | |
972 | struct submit_worker submit; | |
b411b363 PR |
973 | }; |
974 | ||
4ce49266 LE |
975 | struct drbd_bm_aio_ctx { |
976 | struct drbd_device *device; | |
977 | struct list_head list; /* on device->pending_bitmap_io */; | |
978 | unsigned long start_jif; | |
979 | atomic_t in_flight; | |
980 | unsigned int done; | |
981 | unsigned flags; | |
982 | #define BM_AIO_COPY_PAGES 1 | |
983 | #define BM_AIO_WRITE_HINTED 2 | |
984 | #define BM_AIO_WRITE_ALL_PAGES 4 | |
985 | #define BM_AIO_READ 8 | |
986 | int error; | |
987 | struct kref kref; | |
988 | }; | |
989 | ||
a910b123 LE |
990 | struct drbd_config_context { |
991 | /* assigned from drbd_genlmsghdr */ | |
992 | unsigned int minor; | |
993 | /* assigned from request attributes, if present */ | |
994 | unsigned int volume; | |
995 | #define VOLUME_UNSPECIFIED (-1U) | |
996 | /* pointer into the request skb, | |
997 | * limited lifetime! */ | |
998 | char *resource_name; | |
999 | struct nlattr *my_addr; | |
1000 | struct nlattr *peer_addr; | |
1001 | ||
1002 | /* reply buffer */ | |
1003 | struct sk_buff *reply_skb; | |
1004 | /* pointer into reply buffer */ | |
1005 | struct drbd_genlmsghdr *reply_dh; | |
1006 | /* resolved from attributes, if possible */ | |
1007 | struct drbd_device *device; | |
1008 | struct drbd_resource *resource; | |
1009 | struct drbd_connection *connection; | |
1010 | }; | |
1011 | ||
b30ab791 | 1012 | static inline struct drbd_device *minor_to_device(unsigned int minor) |
b411b363 | 1013 | { |
05a10ec7 | 1014 | return (struct drbd_device *)idr_find(&drbd_devices, minor); |
b411b363 PR |
1015 | } |
1016 | ||
a6b32bc3 AG |
1017 | static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device) |
1018 | { | |
ec4a3407 | 1019 | return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices); |
a6b32bc3 AG |
1020 | } |
1021 | ||
77c556f6 AG |
1022 | #define for_each_resource(resource, _resources) \ |
1023 | list_for_each_entry(resource, _resources, resources) | |
1024 | ||
1025 | #define for_each_resource_rcu(resource, _resources) \ | |
1026 | list_for_each_entry_rcu(resource, _resources, resources) | |
1027 | ||
1028 | #define for_each_resource_safe(resource, tmp, _resources) \ | |
1029 | list_for_each_entry_safe(resource, tmp, _resources, resources) | |
1030 | ||
1031 | #define for_each_connection(connection, resource) \ | |
1032 | list_for_each_entry(connection, &resource->connections, connections) | |
1033 | ||
1034 | #define for_each_connection_rcu(connection, resource) \ | |
1035 | list_for_each_entry_rcu(connection, &resource->connections, connections) | |
1036 | ||
1037 | #define for_each_connection_safe(connection, tmp, resource) \ | |
1038 | list_for_each_entry_safe(connection, tmp, &resource->connections, connections) | |
1039 | ||
a6b32bc3 AG |
1040 | #define for_each_peer_device(peer_device, device) \ |
1041 | list_for_each_entry(peer_device, &device->peer_devices, peer_devices) | |
1042 | ||
1043 | #define for_each_peer_device_rcu(peer_device, device) \ | |
1044 | list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices) | |
1045 | ||
1046 | #define for_each_peer_device_safe(peer_device, tmp, device) \ | |
1047 | list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices) | |
1048 | ||
b30ab791 | 1049 | static inline unsigned int device_to_minor(struct drbd_device *device) |
b411b363 | 1050 | { |
b30ab791 | 1051 | return device->minor; |
b411b363 PR |
1052 | } |
1053 | ||
b411b363 PR |
1054 | /* |
1055 | * function declarations | |
1056 | *************************/ | |
1057 | ||
1058 | /* drbd_main.c */ | |
1059 | ||
e89b591c PR |
1060 | enum dds_flags { |
1061 | DDSF_FORCED = 1, | |
1062 | DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */ | |
1063 | }; | |
1064 | ||
b30ab791 | 1065 | extern void drbd_init_set_defaults(struct drbd_device *device); |
b411b363 PR |
1066 | extern int drbd_thread_start(struct drbd_thread *thi); |
1067 | extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait); | |
1068 | #ifdef CONFIG_SMP | |
80822284 | 1069 | extern void drbd_thread_current_set_cpu(struct drbd_thread *thi); |
b411b363 PR |
1070 | #else |
1071 | #define drbd_thread_current_set_cpu(A) ({}) | |
b411b363 | 1072 | #endif |
bde89a9e | 1073 | extern void tl_release(struct drbd_connection *, unsigned int barrier_nr, |
b411b363 | 1074 | unsigned int set_size); |
bde89a9e AG |
1075 | extern void tl_clear(struct drbd_connection *); |
1076 | extern void drbd_free_sock(struct drbd_connection *connection); | |
1077 | extern int drbd_send(struct drbd_connection *connection, struct socket *sock, | |
bedbd2a5 | 1078 | void *buf, size_t size, unsigned msg_flags); |
bde89a9e | 1079 | extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t, |
fb708e40 AG |
1080 | unsigned); |
1081 | ||
bde89a9e AG |
1082 | extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd); |
1083 | extern int drbd_send_protocol(struct drbd_connection *connection); | |
69a22773 AG |
1084 | extern int drbd_send_uuids(struct drbd_peer_device *); |
1085 | extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *); | |
1086 | extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *); | |
1087 | extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags); | |
1088 | extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s); | |
1089 | extern int drbd_send_current_state(struct drbd_peer_device *); | |
1090 | extern int drbd_send_sync_param(struct drbd_peer_device *); | |
bde89a9e | 1091 | extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr, |
d4e67d7c | 1092 | u32 set_size); |
69a22773 | 1093 | extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet, |
f6ffca9f | 1094 | struct drbd_peer_request *); |
69a22773 | 1095 | extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet, |
a9a9994d | 1096 | struct p_block_req *rp); |
69a22773 | 1097 | extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet, |
a9a9994d | 1098 | struct p_data *dp, int data_size); |
69a22773 | 1099 | extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet, |
b411b363 | 1100 | sector_t sector, int blksize, u64 block_id); |
69a22773 AG |
1101 | extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *); |
1102 | extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet, | |
f6ffca9f | 1103 | struct drbd_peer_request *); |
69a22773 AG |
1104 | extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req); |
1105 | extern int drbd_send_drequest(struct drbd_peer_device *, int cmd, | |
b411b363 | 1106 | sector_t sector, int size, u64 block_id); |
69a22773 | 1107 | extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector, |
d8763023 AG |
1108 | int size, void *digest, int digest_size, |
1109 | enum drbd_packet cmd); | |
69a22773 | 1110 | extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size); |
b411b363 | 1111 | |
b30ab791 | 1112 | extern int drbd_send_bitmap(struct drbd_device *device); |
69a22773 | 1113 | extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode); |
bde89a9e | 1114 | extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode); |
28995af5 | 1115 | extern void drbd_free_ldev(struct drbd_backing_dev *ldev); |
b30ab791 AG |
1116 | extern void drbd_device_cleanup(struct drbd_device *device); |
1117 | void drbd_print_uuids(struct drbd_device *device, const char *text); | |
b411b363 | 1118 | |
bde89a9e | 1119 | extern void conn_md_sync(struct drbd_connection *connection); |
b30ab791 AG |
1120 | extern void drbd_md_write(struct drbd_device *device, void *buffer); |
1121 | extern void drbd_md_sync(struct drbd_device *device); | |
1122 | extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev); | |
1123 | extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); | |
1124 | extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); | |
1125 | extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local); | |
1126 | extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local); | |
1127 | extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local); | |
1128 | extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); | |
1129 | extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local); | |
1130 | extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local); | |
b411b363 | 1131 | extern int drbd_md_test_flag(struct drbd_backing_dev *, int); |
b30ab791 | 1132 | extern void drbd_md_mark_dirty(struct drbd_device *device); |
b30ab791 | 1133 | extern void drbd_queue_bitmap_io(struct drbd_device *device, |
54761697 AG |
1134 | int (*io_fn)(struct drbd_device *), |
1135 | void (*done)(struct drbd_device *, int), | |
20ceb2b2 | 1136 | char *why, enum bm_flag flags); |
b30ab791 | 1137 | extern int drbd_bitmap_io(struct drbd_device *device, |
54761697 | 1138 | int (*io_fn)(struct drbd_device *), |
20ceb2b2 | 1139 | char *why, enum bm_flag flags); |
b30ab791 | 1140 | extern int drbd_bitmap_io_from_worker(struct drbd_device *device, |
54761697 | 1141 | int (*io_fn)(struct drbd_device *), |
edc9f5eb | 1142 | char *why, enum bm_flag flags); |
8fe39aac PR |
1143 | extern int drbd_bmio_set_n_write(struct drbd_device *device) __must_hold(local); |
1144 | extern int drbd_bmio_clear_n_write(struct drbd_device *device) __must_hold(local); | |
b411b363 | 1145 | |
b411b363 | 1146 | /* Meta data layout |
ae8bf312 LE |
1147 | * |
1148 | * We currently have two possible layouts. | |
1149 | * Offsets in (512 byte) sectors. | |
1150 | * external: | |
1151 | * |----------- md_size_sect ------------------| | |
1152 | * [ 4k superblock ][ activity log ][ Bitmap ] | |
1153 | * | al_offset == 8 | | |
1154 | * | bm_offset = al_offset + X | | |
1155 | * ==> bitmap sectors = md_size_sect - bm_offset | |
1156 | * | |
1157 | * Variants: | |
1158 | * old, indexed fixed size meta data: | |
1159 | * | |
1160 | * internal: | |
1161 | * |----------- md_size_sect ------------------| | |
1162 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*] | |
1163 | * | al_offset < 0 | | |
1164 | * | bm_offset = al_offset - Y | | |
1165 | * ==> bitmap sectors = Y = al_offset - bm_offset | |
1166 | * | |
1167 | * [padding*] are zero or up to 7 unused 512 Byte sectors to the | |
1168 | * end of the device, so that the [4k superblock] will be 4k aligned. | |
1169 | * | |
1170 | * The activity log consists of 4k transaction blocks, | |
1171 | * which are written in a ring-buffer, or striped ring-buffer like fashion, | |
1172 | * which are writtensize used to be fixed 32kB, | |
1173 | * but is about to become configurable. | |
1174 | */ | |
b411b363 | 1175 | |
ae8bf312 LE |
1176 | /* Our old fixed size meta data layout |
1177 | * allows up to about 3.8TB, so if you want more, | |
7ad651b5 | 1178 | * you need to use the "flexible" meta data format. */ |
ae8bf312 LE |
1179 | #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */ |
1180 | #define MD_4kB_SECT 8 | |
1181 | #define MD_32kB_SECT 64 | |
7ad651b5 LE |
1182 | |
1183 | /* One activity log extent represents 4M of storage */ | |
1184 | #define AL_EXTENT_SHIFT 22 | |
b411b363 PR |
1185 | #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT) |
1186 | ||
7ad651b5 LE |
1187 | /* We could make these currently hardcoded constants configurable |
1188 | * variables at create-md time (or even re-configurable at runtime?). | |
1189 | * Which will require some more changes to the DRBD "super block" | |
1190 | * and attach code. | |
1191 | * | |
1192 | * updates per transaction: | |
1193 | * This many changes to the active set can be logged with one transaction. | |
1194 | * This number is arbitrary. | |
1195 | * context per transaction: | |
1196 | * This many context extent numbers are logged with each transaction. | |
1197 | * This number is resulting from the transaction block size (4k), the layout | |
1198 | * of the transaction header, and the number of updates per transaction. | |
1199 | * See drbd_actlog.c:struct al_transaction_on_disk | |
1200 | * */ | |
1201 | #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary | |
1202 | #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4 | |
1203 | ||
b411b363 PR |
1204 | #if BITS_PER_LONG == 32 |
1205 | #define LN2_BPL 5 | |
1206 | #define cpu_to_lel(A) cpu_to_le32(A) | |
1207 | #define lel_to_cpu(A) le32_to_cpu(A) | |
1208 | #elif BITS_PER_LONG == 64 | |
1209 | #define LN2_BPL 6 | |
1210 | #define cpu_to_lel(A) cpu_to_le64(A) | |
1211 | #define lel_to_cpu(A) le64_to_cpu(A) | |
1212 | #else | |
1213 | #error "LN2 of BITS_PER_LONG unknown!" | |
1214 | #endif | |
1215 | ||
1216 | /* resync bitmap */ | |
1217 | /* 16MB sized 'bitmap extent' to track syncer usage */ | |
1218 | struct bm_extent { | |
1219 | int rs_left; /* number of bits set (out of sync) in this extent. */ | |
1220 | int rs_failed; /* number of failed resync requests in this extent. */ | |
1221 | unsigned long flags; | |
1222 | struct lc_element lce; | |
1223 | }; | |
1224 | ||
1225 | #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */ | |
1226 | #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */ | |
e3555d85 | 1227 | #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */ |
b411b363 PR |
1228 | |
1229 | /* drbd_bitmap.c */ | |
1230 | /* | |
1231 | * We need to store one bit for a block. | |
1232 | * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap. | |
1233 | * Bit 0 ==> local node thinks this block is binary identical on both nodes | |
1234 | * Bit 1 ==> local node thinks this block needs to be synced. | |
1235 | */ | |
1236 | ||
8e26f9cc PR |
1237 | #define SLEEP_TIME (HZ/10) |
1238 | ||
45dfffeb LE |
1239 | /* We do bitmap IO in units of 4k blocks. |
1240 | * We also still have a hardcoded 4k per bit relation. */ | |
1241 | #define BM_BLOCK_SHIFT 12 /* 4k per bit */ | |
b411b363 | 1242 | #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT) |
45dfffeb LE |
1243 | /* mostly arbitrarily set the represented size of one bitmap extent, |
1244 | * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap | |
1245 | * at 4k per bit resolution) */ | |
1246 | #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */ | |
b411b363 PR |
1247 | #define BM_EXT_SIZE (1<<BM_EXT_SHIFT) |
1248 | ||
1249 | #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12) | |
1250 | #error "HAVE YOU FIXED drbdmeta AS WELL??" | |
1251 | #endif | |
1252 | ||
1253 | /* thus many _storage_ sectors are described by one bit */ | |
1254 | #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9)) | |
1255 | #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9)) | |
1256 | #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1) | |
1257 | ||
1258 | /* bit to represented kilo byte conversion */ | |
1259 | #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10)) | |
1260 | ||
1261 | /* in which _bitmap_ extent (resp. sector) the bit for a certain | |
1262 | * _storage_ sector is located in */ | |
1263 | #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9)) | |
5ab7d2c0 | 1264 | #define BM_BIT_TO_EXT(x) ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT)) |
b411b363 | 1265 | |
5ab7d2c0 | 1266 | /* first storage sector a bitmap extent corresponds to */ |
b411b363 | 1267 | #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9)) |
5ab7d2c0 | 1268 | /* how much _storage_ sectors we have per bitmap extent */ |
b411b363 | 1269 | #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1) |
5ab7d2c0 LE |
1270 | /* how many bits are covered by one bitmap extent (resync extent) */ |
1271 | #define BM_BITS_PER_EXT (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT)) | |
1272 | ||
1273 | #define BM_BLOCKS_PER_BM_EXT_MASK (BM_BITS_PER_EXT - 1) | |
1274 | ||
b411b363 PR |
1275 | |
1276 | /* in one sector of the bitmap, we have this many activity_log extents. */ | |
1277 | #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT)) | |
b411b363 | 1278 | |
b411b363 PR |
1279 | /* the extent in "PER_EXTENT" below is an activity log extent |
1280 | * we need that many (long words/bytes) to store the bitmap | |
1281 | * of one AL_EXTENT_SIZE chunk of storage. | |
1282 | * we can store the bitmap for that many AL_EXTENTS within | |
1283 | * one sector of the _on_disk_ bitmap: | |
1284 | * bit 0 bit 37 bit 38 bit (512*8)-1 | |
1285 | * ...|........|........|.. // ..|........| | |
1286 | * sect. 0 `296 `304 ^(512*8*8)-1 | |
1287 | * | |
1288 | #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG ) | |
1289 | #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128 | |
1290 | #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4 | |
1291 | */ | |
1292 | ||
1293 | #define DRBD_MAX_SECTORS_32 (0xffffffffLU) | |
ae8bf312 LE |
1294 | /* we have a certain meta data variant that has a fixed on-disk size of 128 |
1295 | * MiB, of which 4k are our "superblock", and 32k are the fixed size activity | |
1296 | * log, leaving this many sectors for the bitmap. | |
1297 | */ | |
1298 | ||
1299 | #define DRBD_MAX_SECTORS_FIXED_BM \ | |
1300 | ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9))) | |
1301 | #if !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32 | |
b411b363 PR |
1302 | #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32 |
1303 | #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32 | |
1304 | #else | |
ae8bf312 | 1305 | #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM |
b411b363 PR |
1306 | /* 16 TB in units of sectors */ |
1307 | #if BITS_PER_LONG == 32 | |
1308 | /* adjust by one page worth of bitmap, | |
1309 | * so we won't wrap around in drbd_bm_find_next_bit. | |
1310 | * you should use 64bit OS for that much storage, anyways. */ | |
1311 | #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff) | |
1312 | #else | |
4b0715f0 LE |
1313 | /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */ |
1314 | #define DRBD_MAX_SECTORS_FLEX (1UL << 51) | |
1315 | /* corresponds to (1UL << 38) bits right now. */ | |
b411b363 PR |
1316 | #endif |
1317 | #endif | |
1318 | ||
23361cf3 LE |
1319 | /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE, |
1320 | * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte. | |
1321 | * Since we may live in a mixed-platform cluster, | |
1322 | * we limit us to a platform agnostic constant here for now. | |
1323 | * A followup commit may allow even bigger BIO sizes, | |
1324 | * once we thought that through. */ | |
98683650 | 1325 | #define DRBD_MAX_BIO_SIZE (1U << 20) |
23361cf3 LE |
1326 | #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE |
1327 | #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE | |
1328 | #endif | |
db141b2f | 1329 | #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */ |
b411b363 | 1330 | |
98683650 PR |
1331 | #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */ |
1332 | #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */ | |
b411b363 | 1333 | |
a0fb3c47 LE |
1334 | /* For now, don't allow more than one activity log extent worth of data |
1335 | * to be discarded in one go. We may need to rework drbd_al_begin_io() | |
1336 | * to allow for even larger discard ranges */ | |
1337 | #define DRBD_MAX_DISCARD_SIZE AL_EXTENT_SIZE | |
1338 | #define DRBD_MAX_DISCARD_SECTORS (DRBD_MAX_DISCARD_SIZE >> 9) | |
1339 | ||
b30ab791 AG |
1340 | extern int drbd_bm_init(struct drbd_device *device); |
1341 | extern int drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits); | |
1342 | extern void drbd_bm_cleanup(struct drbd_device *device); | |
1343 | extern void drbd_bm_set_all(struct drbd_device *device); | |
1344 | extern void drbd_bm_clear_all(struct drbd_device *device); | |
4b0715f0 | 1345 | /* set/clear/test only a few bits at a time */ |
b411b363 | 1346 | extern int drbd_bm_set_bits( |
b30ab791 | 1347 | struct drbd_device *device, unsigned long s, unsigned long e); |
b411b363 | 1348 | extern int drbd_bm_clear_bits( |
b30ab791 | 1349 | struct drbd_device *device, unsigned long s, unsigned long e); |
4b0715f0 | 1350 | extern int drbd_bm_count_bits( |
b30ab791 | 1351 | struct drbd_device *device, const unsigned long s, const unsigned long e); |
4b0715f0 LE |
1352 | /* bm_set_bits variant for use while holding drbd_bm_lock, |
1353 | * may process the whole bitmap in one go */ | |
b30ab791 | 1354 | extern void _drbd_bm_set_bits(struct drbd_device *device, |
b411b363 | 1355 | const unsigned long s, const unsigned long e); |
b30ab791 AG |
1356 | extern int drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr); |
1357 | extern int drbd_bm_e_weight(struct drbd_device *device, unsigned long enr); | |
b30ab791 AG |
1358 | extern int drbd_bm_read(struct drbd_device *device) __must_hold(local); |
1359 | extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr); | |
1360 | extern int drbd_bm_write(struct drbd_device *device) __must_hold(local); | |
1361 | extern int drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local); | |
c7a58db4 | 1362 | extern int drbd_bm_write_lazy(struct drbd_device *device, unsigned upper_idx) __must_hold(local); |
b30ab791 AG |
1363 | extern int drbd_bm_write_all(struct drbd_device *device) __must_hold(local); |
1364 | extern int drbd_bm_write_copy_pages(struct drbd_device *device) __must_hold(local); | |
1365 | extern size_t drbd_bm_words(struct drbd_device *device); | |
1366 | extern unsigned long drbd_bm_bits(struct drbd_device *device); | |
1367 | extern sector_t drbd_bm_capacity(struct drbd_device *device); | |
4b0715f0 LE |
1368 | |
1369 | #define DRBD_END_OF_BITMAP (~(unsigned long)0) | |
b30ab791 | 1370 | extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); |
b411b363 | 1371 | /* bm_find_next variants for use while you hold drbd_bm_lock() */ |
b30ab791 AG |
1372 | extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); |
1373 | extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo); | |
1374 | extern unsigned long _drbd_bm_total_weight(struct drbd_device *device); | |
1375 | extern unsigned long drbd_bm_total_weight(struct drbd_device *device); | |
b411b363 | 1376 | /* for receive_bitmap */ |
b30ab791 | 1377 | extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset, |
b411b363 | 1378 | size_t number, unsigned long *buffer); |
19f843aa | 1379 | /* for _drbd_send_bitmap */ |
b30ab791 | 1380 | extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset, |
b411b363 PR |
1381 | size_t number, unsigned long *buffer); |
1382 | ||
b30ab791 AG |
1383 | extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags); |
1384 | extern void drbd_bm_unlock(struct drbd_device *device); | |
b411b363 PR |
1385 | /* drbd_main.c */ |
1386 | ||
1387 | extern struct kmem_cache *drbd_request_cache; | |
6c852bec | 1388 | extern struct kmem_cache *drbd_ee_cache; /* peer requests */ |
b411b363 PR |
1389 | extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */ |
1390 | extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */ | |
1391 | extern mempool_t *drbd_request_mempool; | |
1392 | extern mempool_t *drbd_ee_mempool; | |
1393 | ||
4281808f LE |
1394 | /* drbd's page pool, used to buffer data received from the peer, |
1395 | * or data requested by the peer. | |
1396 | * | |
1397 | * This does not have an emergency reserve. | |
1398 | * | |
1399 | * When allocating from this pool, it first takes pages from the pool. | |
1400 | * Only if the pool is depleted will try to allocate from the system. | |
1401 | * | |
1402 | * The assumption is that pages taken from this pool will be processed, | |
1403 | * and given back, "quickly", and then can be recycled, so we can avoid | |
1404 | * frequent calls to alloc_page(), and still will be able to make progress even | |
1405 | * under memory pressure. | |
1406 | */ | |
1407 | extern struct page *drbd_pp_pool; | |
b411b363 PR |
1408 | extern spinlock_t drbd_pp_lock; |
1409 | extern int drbd_pp_vacant; | |
1410 | extern wait_queue_head_t drbd_pp_wait; | |
1411 | ||
4281808f LE |
1412 | /* We also need a standard (emergency-reserve backed) page pool |
1413 | * for meta data IO (activity log, bitmap). | |
1414 | * We can keep it global, as long as it is used as "N pages at a time". | |
1415 | * 128 should be plenty, currently we probably can get away with as few as 1. | |
1416 | */ | |
1417 | #define DRBD_MIN_POOL_PAGES 128 | |
1418 | extern mempool_t *drbd_md_io_page_pool; | |
1419 | ||
9476f39d LE |
1420 | /* We also need to make sure we get a bio |
1421 | * when we need it for housekeeping purposes */ | |
1422 | extern struct bio_set *drbd_md_io_bio_set; | |
1423 | /* to allocate from that set */ | |
1424 | extern struct bio *bio_alloc_drbd(gfp_t gfp_mask); | |
1425 | ||
b411b363 PR |
1426 | extern rwlock_t global_state_lock; |
1427 | ||
bde89a9e | 1428 | extern int conn_lowest_minor(struct drbd_connection *connection); |
a910b123 | 1429 | extern enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor); |
05a10ec7 | 1430 | extern void drbd_destroy_device(struct kref *kref); |
a910b123 | 1431 | extern void drbd_delete_device(struct drbd_device *device); |
b411b363 | 1432 | |
77c556f6 AG |
1433 | extern struct drbd_resource *drbd_create_resource(const char *name); |
1434 | extern void drbd_free_resource(struct drbd_resource *resource); | |
1435 | ||
eb6bea67 | 1436 | extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts); |
bde89a9e | 1437 | extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts); |
05a10ec7 | 1438 | extern void drbd_destroy_connection(struct kref *kref); |
bde89a9e | 1439 | extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len, |
089c075d | 1440 | void *peer_addr, int peer_addr_len); |
4bc76048 | 1441 | extern struct drbd_resource *drbd_find_resource(const char *name); |
77c556f6 | 1442 | extern void drbd_destroy_resource(struct kref *kref); |
bde89a9e | 1443 | extern void conn_free_crypto(struct drbd_connection *connection); |
b411b363 PR |
1444 | |
1445 | extern int proc_details; | |
1446 | ||
1447 | /* drbd_req */ | |
113fef9e | 1448 | extern void do_submit(struct work_struct *ws); |
54761697 | 1449 | extern void __drbd_make_request(struct drbd_device *, struct bio *, unsigned long); |
5a7bbad2 | 1450 | extern void drbd_make_request(struct request_queue *q, struct bio *bio); |
b30ab791 | 1451 | extern int drbd_read_remote(struct drbd_device *device, struct drbd_request *req); |
b411b363 PR |
1452 | extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec); |
1453 | extern int is_valid_ar_handle(struct drbd_request *, sector_t); | |
1454 | ||
1455 | ||
1456 | /* drbd_nl.c */ | |
a910b123 | 1457 | extern int drbd_msg_put_info(struct sk_buff *skb, const char *info); |
b30ab791 AG |
1458 | extern void drbd_suspend_io(struct drbd_device *device); |
1459 | extern void drbd_resume_io(struct drbd_device *device); | |
b411b363 | 1460 | extern char *ppsize(char *buf, unsigned long long size); |
54761697 | 1461 | extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int); |
e96c9633 | 1462 | enum determine_dev_size { |
d752b269 PR |
1463 | DS_ERROR_SHRINK = -3, |
1464 | DS_ERROR_SPACE_MD = -2, | |
e96c9633 PR |
1465 | DS_ERROR = -1, |
1466 | DS_UNCHANGED = 0, | |
1467 | DS_SHRUNK = 1, | |
57737adc PR |
1468 | DS_GREW = 2, |
1469 | DS_GREW_FROM_ZERO = 3, | |
e96c9633 | 1470 | }; |
d752b269 | 1471 | extern enum determine_dev_size |
54761697 AG |
1472 | drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local); |
1473 | extern void resync_after_online_grow(struct drbd_device *); | |
8fe39aac | 1474 | extern void drbd_reconsider_max_bio_size(struct drbd_device *device, struct drbd_backing_dev *bdev); |
b30ab791 | 1475 | extern enum drbd_state_rv drbd_set_role(struct drbd_device *device, |
bf885f8a AG |
1476 | enum drbd_role new_role, |
1477 | int force); | |
bde89a9e AG |
1478 | extern bool conn_try_outdate_peer(struct drbd_connection *connection); |
1479 | extern void conn_try_outdate_peer_async(struct drbd_connection *connection); | |
b30ab791 | 1480 | extern int drbd_khelper(struct drbd_device *device, char *cmd); |
b411b363 PR |
1481 | |
1482 | /* drbd_worker.c */ | |
d40e5671 | 1483 | /* bi_end_io handlers */ |
ed15b795 | 1484 | extern void drbd_md_endio(struct bio *bio, int error); |
d40e5671 PR |
1485 | extern void drbd_peer_request_endio(struct bio *bio, int error); |
1486 | extern void drbd_request_endio(struct bio *bio, int error); | |
b411b363 | 1487 | extern int drbd_worker(struct drbd_thread *thi); |
b30ab791 AG |
1488 | enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor); |
1489 | void drbd_resync_after_changed(struct drbd_device *device); | |
1490 | extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side); | |
1491 | extern void resume_next_sg(struct drbd_device *device); | |
1492 | extern void suspend_other_sg(struct drbd_device *device); | |
1493 | extern int drbd_resync_finished(struct drbd_device *device); | |
b411b363 | 1494 | /* maybe rather drbd_main.c ? */ |
e37d2438 | 1495 | extern void *drbd_md_get_buffer(struct drbd_device *device, const char *intent); |
b30ab791 AG |
1496 | extern void drbd_md_put_buffer(struct drbd_device *device); |
1497 | extern int drbd_md_sync_page_io(struct drbd_device *device, | |
b411b363 | 1498 | struct drbd_backing_dev *bdev, sector_t sector, int rw); |
54761697 | 1499 | extern void drbd_ov_out_of_sync_found(struct drbd_device *, sector_t, int); |
b30ab791 | 1500 | extern void wait_until_done_or_force_detached(struct drbd_device *device, |
44edfb0d | 1501 | struct drbd_backing_dev *bdev, unsigned int *done); |
b30ab791 | 1502 | extern void drbd_rs_controller_reset(struct drbd_device *device); |
b411b363 | 1503 | |
b30ab791 | 1504 | static inline void ov_out_of_sync_print(struct drbd_device *device) |
b411b363 | 1505 | { |
b30ab791 | 1506 | if (device->ov_last_oos_size) { |
d0180171 | 1507 | drbd_err(device, "Out of sync: start=%llu, size=%lu (sectors)\n", |
b30ab791 AG |
1508 | (unsigned long long)device->ov_last_oos_start, |
1509 | (unsigned long)device->ov_last_oos_size); | |
b411b363 | 1510 | } |
b30ab791 | 1511 | device->ov_last_oos_size = 0; |
b411b363 PR |
1512 | } |
1513 | ||
1514 | ||
79a3c8d3 AG |
1515 | extern void drbd_csum_bio(struct crypto_hash *, struct bio *, void *); |
1516 | extern void drbd_csum_ee(struct crypto_hash *, struct drbd_peer_request *, void *); | |
b411b363 | 1517 | /* worker callbacks */ |
99920dc5 AG |
1518 | extern int w_e_end_data_req(struct drbd_work *, int); |
1519 | extern int w_e_end_rsdata_req(struct drbd_work *, int); | |
1520 | extern int w_e_end_csum_rs_req(struct drbd_work *, int); | |
1521 | extern int w_e_end_ov_reply(struct drbd_work *, int); | |
1522 | extern int w_e_end_ov_req(struct drbd_work *, int); | |
1523 | extern int w_ov_finished(struct drbd_work *, int); | |
1524 | extern int w_resync_timer(struct drbd_work *, int); | |
1525 | extern int w_send_write_hint(struct drbd_work *, int); | |
99920dc5 | 1526 | extern int w_send_dblock(struct drbd_work *, int); |
99920dc5 | 1527 | extern int w_send_read_req(struct drbd_work *, int); |
99920dc5 AG |
1528 | extern int w_e_reissue(struct drbd_work *, int); |
1529 | extern int w_restart_disk_io(struct drbd_work *, int); | |
8f7bed77 | 1530 | extern int w_send_out_of_sync(struct drbd_work *, int); |
99920dc5 | 1531 | extern int w_start_resync(struct drbd_work *, int); |
b411b363 PR |
1532 | |
1533 | extern void resync_timer_fn(unsigned long data); | |
370a43e7 | 1534 | extern void start_resync_timer_fn(unsigned long data); |
b411b363 | 1535 | |
a0fb3c47 LE |
1536 | extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req); |
1537 | ||
b411b363 | 1538 | /* drbd_receiver.c */ |
753c6191 AG |
1539 | extern int drbd_receiver(struct drbd_thread *thi); |
1540 | extern int drbd_asender(struct drbd_thread *thi); | |
e8299874 | 1541 | extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device); |
ad3fee79 LE |
1542 | extern bool drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector, |
1543 | bool throttle_if_app_is_waiting); | |
54761697 | 1544 | extern int drbd_submit_peer_request(struct drbd_device *, |
fbe29dec AG |
1545 | struct drbd_peer_request *, const unsigned, |
1546 | const int); | |
54761697 | 1547 | extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *); |
69a22773 | 1548 | extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64, |
0db55363 | 1549 | sector_t, unsigned int, |
a0fb3c47 | 1550 | bool, |
0db55363 | 1551 | gfp_t) __must_hold(local); |
54761697 | 1552 | extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *, |
3967deb1 AG |
1553 | int); |
1554 | #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0) | |
1555 | #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1) | |
69a22773 | 1556 | extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool); |
b30ab791 AG |
1557 | extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled); |
1558 | extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed); | |
69a22773 | 1559 | extern int drbd_connected(struct drbd_peer_device *); |
b411b363 | 1560 | |
ed439848 LE |
1561 | /* Yes, there is kernel_setsockopt, but only since 2.6.18. |
1562 | * So we have our own copy of it here. */ | |
b411b363 | 1563 | static inline int drbd_setsockopt(struct socket *sock, int level, int optname, |
ed439848 | 1564 | char *optval, int optlen) |
b411b363 | 1565 | { |
ed439848 LE |
1566 | mm_segment_t oldfs = get_fs(); |
1567 | char __user *uoptval; | |
b411b363 | 1568 | int err; |
ed439848 LE |
1569 | |
1570 | uoptval = (char __user __force *)optval; | |
1571 | ||
1572 | set_fs(KERNEL_DS); | |
b411b363 | 1573 | if (level == SOL_SOCKET) |
ed439848 | 1574 | err = sock_setsockopt(sock, level, optname, uoptval, optlen); |
b411b363 | 1575 | else |
ed439848 | 1576 | err = sock->ops->setsockopt(sock, level, optname, uoptval, |
b411b363 | 1577 | optlen); |
ed439848 | 1578 | set_fs(oldfs); |
b411b363 PR |
1579 | return err; |
1580 | } | |
1581 | ||
1582 | static inline void drbd_tcp_cork(struct socket *sock) | |
1583 | { | |
ed439848 | 1584 | int val = 1; |
b411b363 | 1585 | (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK, |
ed439848 | 1586 | (char*)&val, sizeof(val)); |
b411b363 PR |
1587 | } |
1588 | ||
1589 | static inline void drbd_tcp_uncork(struct socket *sock) | |
1590 | { | |
ed439848 | 1591 | int val = 0; |
b411b363 | 1592 | (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK, |
ed439848 | 1593 | (char*)&val, sizeof(val)); |
b411b363 PR |
1594 | } |
1595 | ||
1596 | static inline void drbd_tcp_nodelay(struct socket *sock) | |
1597 | { | |
ed439848 | 1598 | int val = 1; |
b411b363 | 1599 | (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY, |
ed439848 | 1600 | (char*)&val, sizeof(val)); |
b411b363 PR |
1601 | } |
1602 | ||
1603 | static inline void drbd_tcp_quickack(struct socket *sock) | |
1604 | { | |
ed439848 | 1605 | int val = 2; |
b411b363 | 1606 | (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK, |
ed439848 | 1607 | (char*)&val, sizeof(val)); |
b411b363 PR |
1608 | } |
1609 | ||
d40e5671 PR |
1610 | /* sets the number of 512 byte sectors of our virtual device */ |
1611 | static inline void drbd_set_my_capacity(struct drbd_device *device, | |
1612 | sector_t size) | |
1613 | { | |
1614 | /* set_capacity(device->this_bdev->bd_disk, size); */ | |
1615 | set_capacity(device->vdisk, size); | |
1616 | device->this_bdev->bd_inode->i_size = (loff_t)size << 9; | |
1617 | } | |
1618 | ||
1619 | /* | |
1620 | * used to submit our private bio | |
1621 | */ | |
1622 | static inline void drbd_generic_make_request(struct drbd_device *device, | |
1623 | int fault_type, struct bio *bio) | |
1624 | { | |
1625 | __release(local); | |
1626 | if (!bio->bi_bdev) { | |
f88c5d90 | 1627 | drbd_err(device, "drbd_generic_make_request: bio->bi_bdev == NULL\n"); |
d40e5671 PR |
1628 | bio_endio(bio, -ENODEV); |
1629 | return; | |
1630 | } | |
1631 | ||
1632 | if (drbd_insert_fault(device, fault_type)) | |
1633 | bio_endio(bio, -EIO); | |
1634 | else | |
1635 | generic_make_request(bio); | |
1636 | } | |
1637 | ||
8fe39aac PR |
1638 | void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev, |
1639 | enum write_ordering_e wo); | |
b411b363 PR |
1640 | |
1641 | /* drbd_proc.c */ | |
1642 | extern struct proc_dir_entry *drbd_proc; | |
7d4e9d09 | 1643 | extern const struct file_operations drbd_proc_fops; |
b411b363 PR |
1644 | extern const char *drbd_conn_str(enum drbd_conns s); |
1645 | extern const char *drbd_role_str(enum drbd_role s); | |
1646 | ||
1647 | /* drbd_actlog.c */ | |
e4d7d6f4 | 1648 | extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i); |
b30ab791 | 1649 | extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i); |
4dd726f0 | 1650 | extern void drbd_al_begin_io_commit(struct drbd_device *device); |
b30ab791 | 1651 | extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i); |
4dd726f0 | 1652 | extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i); |
b30ab791 AG |
1653 | extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i); |
1654 | extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector); | |
1655 | extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector); | |
1656 | extern int drbd_try_rs_begin_io(struct drbd_device *device, sector_t sector); | |
1657 | extern void drbd_rs_cancel_all(struct drbd_device *device); | |
1658 | extern int drbd_rs_del_all(struct drbd_device *device); | |
1659 | extern void drbd_rs_failed_io(struct drbd_device *device, | |
b411b363 | 1660 | sector_t sector, int size); |
b30ab791 | 1661 | extern void drbd_advance_rs_marks(struct drbd_device *device, unsigned long still_to_go); |
5ab7d2c0 LE |
1662 | |
1663 | enum update_sync_bits_mode { RECORD_RS_FAILED, SET_OUT_OF_SYNC, SET_IN_SYNC }; | |
1664 | extern int __drbd_change_sync(struct drbd_device *device, sector_t sector, int size, | |
179e20b8 | 1665 | enum update_sync_bits_mode mode); |
b30ab791 | 1666 | #define drbd_set_in_sync(device, sector, size) \ |
179e20b8 | 1667 | __drbd_change_sync(device, sector, size, SET_IN_SYNC) |
b30ab791 | 1668 | #define drbd_set_out_of_sync(device, sector, size) \ |
179e20b8 | 1669 | __drbd_change_sync(device, sector, size, SET_OUT_OF_SYNC) |
5ab7d2c0 | 1670 | #define drbd_rs_failed_io(device, sector, size) \ |
179e20b8 | 1671 | __drbd_change_sync(device, sector, size, RECORD_RS_FAILED) |
b30ab791 | 1672 | extern void drbd_al_shrink(struct drbd_device *device); |
54761697 | 1673 | extern int drbd_initialize_al(struct drbd_device *, void *); |
b411b363 | 1674 | |
b411b363 | 1675 | /* drbd_nl.c */ |
3b98c0c2 LE |
1676 | /* state info broadcast */ |
1677 | struct sib_info { | |
1678 | enum drbd_state_info_bcast_reason sib_reason; | |
1679 | union { | |
1680 | struct { | |
1681 | char *helper_name; | |
1682 | unsigned helper_exit_code; | |
1683 | }; | |
1684 | struct { | |
1685 | union drbd_state os; | |
1686 | union drbd_state ns; | |
1687 | }; | |
1688 | }; | |
1689 | }; | |
b30ab791 | 1690 | void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib); |
b411b363 PR |
1691 | |
1692 | /* | |
1693 | * inline helper functions | |
1694 | *************************/ | |
1695 | ||
45bb912b LE |
1696 | /* see also page_chain_add and friends in drbd_receiver.c */ |
1697 | static inline struct page *page_chain_next(struct page *page) | |
1698 | { | |
1699 | return (struct page *)page_private(page); | |
1700 | } | |
1701 | #define page_chain_for_each(page) \ | |
1702 | for (; page && ({ prefetch(page_chain_next(page)); 1; }); \ | |
1703 | page = page_chain_next(page)) | |
1704 | #define page_chain_for_each_safe(page, n) \ | |
1705 | for (; page && ({ n = page_chain_next(page); 1; }); page = n) | |
1706 | ||
45bb912b | 1707 | |
045417f7 | 1708 | static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req) |
45bb912b | 1709 | { |
db830c46 | 1710 | struct page *page = peer_req->pages; |
45bb912b LE |
1711 | page_chain_for_each(page) { |
1712 | if (page_count(page) > 1) | |
1713 | return 1; | |
1714 | } | |
1715 | return 0; | |
1716 | } | |
1717 | ||
bf885f8a | 1718 | static inline enum drbd_state_rv |
b30ab791 | 1719 | _drbd_set_state(struct drbd_device *device, union drbd_state ns, |
bf885f8a | 1720 | enum chg_state_flags flags, struct completion *done) |
b411b363 | 1721 | { |
bf885f8a | 1722 | enum drbd_state_rv rv; |
b411b363 PR |
1723 | |
1724 | read_lock(&global_state_lock); | |
b30ab791 | 1725 | rv = __drbd_set_state(device, ns, flags, done); |
b411b363 PR |
1726 | read_unlock(&global_state_lock); |
1727 | ||
1728 | return rv; | |
1729 | } | |
1730 | ||
b30ab791 | 1731 | static inline union drbd_state drbd_read_state(struct drbd_device *device) |
b411b363 | 1732 | { |
6bbf53ca | 1733 | struct drbd_resource *resource = device->resource; |
78bae59b PR |
1734 | union drbd_state rv; |
1735 | ||
b30ab791 | 1736 | rv.i = device->state.i; |
6bbf53ca AG |
1737 | rv.susp = resource->susp; |
1738 | rv.susp_nod = resource->susp_nod; | |
1739 | rv.susp_fen = resource->susp_fen; | |
78bae59b PR |
1740 | |
1741 | return rv; | |
b411b363 PR |
1742 | } |
1743 | ||
383606e0 | 1744 | enum drbd_force_detach_flags { |
a2a3c74f LE |
1745 | DRBD_READ_ERROR, |
1746 | DRBD_WRITE_ERROR, | |
383606e0 LE |
1747 | DRBD_META_IO_ERROR, |
1748 | DRBD_FORCE_DETACH, | |
1749 | }; | |
1750 | ||
b411b363 | 1751 | #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__) |
b30ab791 | 1752 | static inline void __drbd_chk_io_error_(struct drbd_device *device, |
a2a3c74f | 1753 | enum drbd_force_detach_flags df, |
383606e0 | 1754 | const char *where) |
b411b363 | 1755 | { |
daeda1cc PR |
1756 | enum drbd_io_error_p ep; |
1757 | ||
1758 | rcu_read_lock(); | |
b30ab791 | 1759 | ep = rcu_dereference(device->ldev->disk_conf)->on_io_error; |
daeda1cc PR |
1760 | rcu_read_unlock(); |
1761 | switch (ep) { | |
1762 | case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */ | |
a2a3c74f | 1763 | if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) { |
7383506c | 1764 | if (__ratelimit(&drbd_ratelimit_state)) |
d0180171 | 1765 | drbd_err(device, "Local IO failed in %s.\n", where); |
b30ab791 AG |
1766 | if (device->state.disk > D_INCONSISTENT) |
1767 | _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL); | |
b411b363 PR |
1768 | break; |
1769 | } | |
a2a3c74f | 1770 | /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */ |
b411b363 PR |
1771 | case EP_DETACH: |
1772 | case EP_CALL_HELPER: | |
a2a3c74f LE |
1773 | /* Remember whether we saw a READ or WRITE error. |
1774 | * | |
1775 | * Recovery of the affected area for WRITE failure is covered | |
1776 | * by the activity log. | |
1777 | * READ errors may fall outside that area though. Certain READ | |
1778 | * errors can be "healed" by writing good data to the affected | |
1779 | * blocks, which triggers block re-allocation in lower layers. | |
1780 | * | |
1781 | * If we can not write the bitmap after a READ error, | |
1782 | * we may need to trigger a full sync (see w_go_diskless()). | |
1783 | * | |
1784 | * Force-detach is not really an IO error, but rather a | |
1785 | * desperate measure to try to deal with a completely | |
1786 | * unresponsive lower level IO stack. | |
1787 | * Still it should be treated as a WRITE error. | |
1788 | * | |
1789 | * Meta IO error is always WRITE error: | |
1790 | * we read meta data only once during attach, | |
1791 | * which will fail in case of errors. | |
1792 | */ | |
b30ab791 | 1793 | set_bit(WAS_IO_ERROR, &device->flags); |
a2a3c74f | 1794 | if (df == DRBD_READ_ERROR) |
b30ab791 | 1795 | set_bit(WAS_READ_ERROR, &device->flags); |
a2a3c74f | 1796 | if (df == DRBD_FORCE_DETACH) |
b30ab791 AG |
1797 | set_bit(FORCE_DETACH, &device->flags); |
1798 | if (device->state.disk > D_FAILED) { | |
1799 | _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL); | |
d0180171 | 1800 | drbd_err(device, |
82f59cc6 | 1801 | "Local IO failed in %s. Detaching...\n", where); |
b411b363 PR |
1802 | } |
1803 | break; | |
1804 | } | |
1805 | } | |
1806 | ||
1807 | /** | |
1808 | * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers | |
b30ab791 | 1809 | * @device: DRBD device. |
b411b363 PR |
1810 | * @error: Error code passed to the IO completion callback |
1811 | * @forcedetach: Force detach. I.e. the error happened while accessing the meta data | |
1812 | * | |
1813 | * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED) | |
1814 | */ | |
1815 | #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__) | |
b30ab791 | 1816 | static inline void drbd_chk_io_error_(struct drbd_device *device, |
383606e0 | 1817 | int error, enum drbd_force_detach_flags forcedetach, const char *where) |
b411b363 PR |
1818 | { |
1819 | if (error) { | |
1820 | unsigned long flags; | |
0500813f | 1821 | spin_lock_irqsave(&device->resource->req_lock, flags); |
b30ab791 | 1822 | __drbd_chk_io_error_(device, forcedetach, where); |
0500813f | 1823 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
b411b363 PR |
1824 | } |
1825 | } | |
1826 | ||
1827 | ||
1828 | /** | |
1829 | * drbd_md_first_sector() - Returns the first sector number of the meta data area | |
1830 | * @bdev: Meta data block device. | |
1831 | * | |
1832 | * BTW, for internal meta data, this happens to be the maximum capacity | |
1833 | * we could agree upon with our peer node. | |
1834 | */ | |
68e41a43 | 1835 | static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev) |
b411b363 | 1836 | { |
68e41a43 | 1837 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
1838 | case DRBD_MD_INDEX_INTERNAL: |
1839 | case DRBD_MD_INDEX_FLEX_INT: | |
1840 | return bdev->md.md_offset + bdev->md.bm_offset; | |
1841 | case DRBD_MD_INDEX_FLEX_EXT: | |
1842 | default: | |
1843 | return bdev->md.md_offset; | |
1844 | } | |
1845 | } | |
1846 | ||
1847 | /** | |
1848 | * drbd_md_last_sector() - Return the last sector number of the meta data area | |
1849 | * @bdev: Meta data block device. | |
1850 | */ | |
1851 | static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev) | |
1852 | { | |
68e41a43 | 1853 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
1854 | case DRBD_MD_INDEX_INTERNAL: |
1855 | case DRBD_MD_INDEX_FLEX_INT: | |
ae8bf312 | 1856 | return bdev->md.md_offset + MD_4kB_SECT -1; |
b411b363 PR |
1857 | case DRBD_MD_INDEX_FLEX_EXT: |
1858 | default: | |
ae8bf312 | 1859 | return bdev->md.md_offset + bdev->md.md_size_sect -1; |
b411b363 PR |
1860 | } |
1861 | } | |
1862 | ||
1863 | /* Returns the number of 512 byte sectors of the device */ | |
1864 | static inline sector_t drbd_get_capacity(struct block_device *bdev) | |
1865 | { | |
1866 | /* return bdev ? get_capacity(bdev->bd_disk) : 0; */ | |
77304d2a | 1867 | return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0; |
b411b363 PR |
1868 | } |
1869 | ||
1870 | /** | |
1871 | * drbd_get_max_capacity() - Returns the capacity we announce to out peer | |
1872 | * @bdev: Meta data block device. | |
1873 | * | |
1874 | * returns the capacity we announce to out peer. we clip ourselves at the | |
1875 | * various MAX_SECTORS, because if we don't, current implementation will | |
1876 | * oops sooner or later | |
1877 | */ | |
1878 | static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev) | |
1879 | { | |
1880 | sector_t s; | |
daeda1cc | 1881 | |
68e41a43 | 1882 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
1883 | case DRBD_MD_INDEX_INTERNAL: |
1884 | case DRBD_MD_INDEX_FLEX_INT: | |
1885 | s = drbd_get_capacity(bdev->backing_bdev) | |
1886 | ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX, | |
68e41a43 | 1887 | drbd_md_first_sector(bdev)) |
b411b363 PR |
1888 | : 0; |
1889 | break; | |
1890 | case DRBD_MD_INDEX_FLEX_EXT: | |
1891 | s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX, | |
1892 | drbd_get_capacity(bdev->backing_bdev)); | |
1893 | /* clip at maximum size the meta device can support */ | |
1894 | s = min_t(sector_t, s, | |
1895 | BM_EXT_TO_SECT(bdev->md.md_size_sect | |
1896 | - bdev->md.bm_offset)); | |
1897 | break; | |
1898 | default: | |
1899 | s = min_t(sector_t, DRBD_MAX_SECTORS, | |
1900 | drbd_get_capacity(bdev->backing_bdev)); | |
1901 | } | |
1902 | return s; | |
1903 | } | |
1904 | ||
1905 | /** | |
3a4d4eb3 | 1906 | * drbd_md_ss() - Return the sector number of our meta data super block |
b411b363 PR |
1907 | * @bdev: Meta data block device. |
1908 | */ | |
3a4d4eb3 | 1909 | static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev) |
b411b363 | 1910 | { |
3a4d4eb3 | 1911 | const int meta_dev_idx = bdev->md.meta_dev_idx; |
daeda1cc | 1912 | |
3a4d4eb3 LE |
1913 | if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT) |
1914 | return 0; | |
daeda1cc | 1915 | |
3a4d4eb3 LE |
1916 | /* Since drbd08, internal meta data is always "flexible". |
1917 | * position: last 4k aligned block of 4k size */ | |
1918 | if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL || | |
1919 | meta_dev_idx == DRBD_MD_INDEX_FLEX_INT) | |
ae8bf312 | 1920 | return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8; |
3a4d4eb3 LE |
1921 | |
1922 | /* external, some index; this is the old fixed size layout */ | |
1923 | return MD_128MB_SECT * bdev->md.meta_dev_idx; | |
b411b363 PR |
1924 | } |
1925 | ||
b411b363 PR |
1926 | static inline void |
1927 | drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w) | |
1928 | { | |
1929 | unsigned long flags; | |
1930 | spin_lock_irqsave(&q->q_lock, flags); | |
1931 | list_add_tail(&w->list, &q->q); | |
b411b363 | 1932 | spin_unlock_irqrestore(&q->q_lock, flags); |
8c0785a5 | 1933 | wake_up(&q->q_wait); |
b411b363 PR |
1934 | } |
1935 | ||
15e26f6a LE |
1936 | static inline void |
1937 | drbd_queue_work_if_unqueued(struct drbd_work_queue *q, struct drbd_work *w) | |
1938 | { | |
1939 | unsigned long flags; | |
1940 | spin_lock_irqsave(&q->q_lock, flags); | |
1941 | if (list_empty_careful(&w->list)) | |
1942 | list_add_tail(&w->list, &q->q); | |
1943 | spin_unlock_irqrestore(&q->q_lock, flags); | |
1944 | wake_up(&q->q_wait); | |
1945 | } | |
1946 | ||
e334f550 LE |
1947 | static inline void |
1948 | drbd_device_post_work(struct drbd_device *device, int work_bit) | |
1949 | { | |
1950 | if (!test_and_set_bit(work_bit, &device->flags)) { | |
1951 | struct drbd_connection *connection = | |
1952 | first_peer_device(device)->connection; | |
1953 | struct drbd_work_queue *q = &connection->sender_work; | |
1954 | if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags)) | |
1955 | wake_up(&q->q_wait); | |
1956 | } | |
1957 | } | |
1958 | ||
b5043c5e AG |
1959 | extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue); |
1960 | ||
bde89a9e | 1961 | static inline void wake_asender(struct drbd_connection *connection) |
b411b363 | 1962 | { |
bde89a9e AG |
1963 | if (test_bit(SIGNAL_ASENDER, &connection->flags)) |
1964 | force_sig(DRBD_SIG, connection->asender.task); | |
b411b363 PR |
1965 | } |
1966 | ||
bde89a9e | 1967 | static inline void request_ping(struct drbd_connection *connection) |
b411b363 | 1968 | { |
bde89a9e AG |
1969 | set_bit(SEND_PING, &connection->flags); |
1970 | wake_asender(connection); | |
b411b363 PR |
1971 | } |
1972 | ||
bde89a9e | 1973 | extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *); |
69a22773 | 1974 | extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *); |
bde89a9e | 1975 | extern int conn_send_command(struct drbd_connection *, struct drbd_socket *, |
dba58587 AG |
1976 | enum drbd_packet, unsigned int, void *, |
1977 | unsigned int); | |
69a22773 | 1978 | extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *, |
dba58587 AG |
1979 | enum drbd_packet, unsigned int, void *, |
1980 | unsigned int); | |
b411b363 | 1981 | |
bde89a9e AG |
1982 | extern int drbd_send_ping(struct drbd_connection *connection); |
1983 | extern int drbd_send_ping_ack(struct drbd_connection *connection); | |
69a22773 | 1984 | extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state); |
bde89a9e | 1985 | extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state); |
b411b363 PR |
1986 | |
1987 | static inline void drbd_thread_stop(struct drbd_thread *thi) | |
1988 | { | |
81e84650 | 1989 | _drbd_thread_stop(thi, false, true); |
b411b363 PR |
1990 | } |
1991 | ||
1992 | static inline void drbd_thread_stop_nowait(struct drbd_thread *thi) | |
1993 | { | |
81e84650 | 1994 | _drbd_thread_stop(thi, false, false); |
b411b363 PR |
1995 | } |
1996 | ||
1997 | static inline void drbd_thread_restart_nowait(struct drbd_thread *thi) | |
1998 | { | |
81e84650 | 1999 | _drbd_thread_stop(thi, true, false); |
b411b363 PR |
2000 | } |
2001 | ||
2002 | /* counts how many answer packets packets we expect from our peer, | |
2003 | * for either explicit application requests, | |
2004 | * or implicit barrier packets as necessary. | |
2005 | * increased: | |
2006 | * w_send_barrier | |
8554df1c | 2007 | * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ); |
b411b363 PR |
2008 | * it is much easier and equally valid to count what we queue for the |
2009 | * worker, even before it actually was queued or send. | |
2010 | * (drbd_make_request_common; recovery path on read io-error) | |
2011 | * decreased: | |
2012 | * got_BarrierAck (respective tl_clear, tl_clear_barrier) | |
8554df1c | 2013 | * _req_mod(req, DATA_RECEIVED) |
b411b363 | 2014 | * [from receive_DataReply] |
8554df1c | 2015 | * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED) |
b411b363 PR |
2016 | * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)] |
2017 | * for some reason it is NOT decreased in got_NegAck, | |
2018 | * but in the resulting cleanup code from report_params. | |
2019 | * we should try to remember the reason for that... | |
8554df1c AG |
2020 | * _req_mod(req, SEND_FAILED or SEND_CANCELED) |
2021 | * _req_mod(req, CONNECTION_LOST_WHILE_PENDING) | |
b411b363 PR |
2022 | * [from tl_clear_barrier] |
2023 | */ | |
b30ab791 | 2024 | static inline void inc_ap_pending(struct drbd_device *device) |
b411b363 | 2025 | { |
b30ab791 | 2026 | atomic_inc(&device->ap_pending_cnt); |
b411b363 PR |
2027 | } |
2028 | ||
49559d87 | 2029 | #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \ |
b30ab791 | 2030 | if (atomic_read(&device->which) < 0) \ |
d0180171 | 2031 | drbd_err(device, "in %s:%d: " #which " = %d < 0 !\n", \ |
49559d87 | 2032 | func, line, \ |
b30ab791 | 2033 | atomic_read(&device->which)) |
b411b363 | 2034 | |
659b2e3b | 2035 | #define dec_ap_pending(device) _dec_ap_pending(device, __func__, __LINE__) |
b30ab791 | 2036 | static inline void _dec_ap_pending(struct drbd_device *device, const char *func, int line) |
49559d87 | 2037 | { |
b30ab791 AG |
2038 | if (atomic_dec_and_test(&device->ap_pending_cnt)) |
2039 | wake_up(&device->misc_wait); | |
49559d87 PR |
2040 | ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line); |
2041 | } | |
b411b363 PR |
2042 | |
2043 | /* counts how many resync-related answers we still expect from the peer | |
2044 | * increase decrease | |
2045 | * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY) | |
25985edc | 2046 | * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER) |
b411b363 PR |
2047 | * (or P_NEG_ACK with ID_SYNCER) |
2048 | */ | |
b30ab791 | 2049 | static inline void inc_rs_pending(struct drbd_device *device) |
b411b363 | 2050 | { |
b30ab791 | 2051 | atomic_inc(&device->rs_pending_cnt); |
b411b363 PR |
2052 | } |
2053 | ||
659b2e3b | 2054 | #define dec_rs_pending(device) _dec_rs_pending(device, __func__, __LINE__) |
b30ab791 | 2055 | static inline void _dec_rs_pending(struct drbd_device *device, const char *func, int line) |
49559d87 | 2056 | { |
b30ab791 | 2057 | atomic_dec(&device->rs_pending_cnt); |
49559d87 PR |
2058 | ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line); |
2059 | } | |
b411b363 PR |
2060 | |
2061 | /* counts how many answers we still need to send to the peer. | |
2062 | * increased on | |
2063 | * receive_Data unless protocol A; | |
2064 | * we need to send a P_RECV_ACK (proto B) | |
2065 | * or P_WRITE_ACK (proto C) | |
2066 | * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK | |
2067 | * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA | |
2068 | * receive_Barrier_* we need to send a P_BARRIER_ACK | |
2069 | */ | |
b30ab791 | 2070 | static inline void inc_unacked(struct drbd_device *device) |
b411b363 | 2071 | { |
b30ab791 | 2072 | atomic_inc(&device->unacked_cnt); |
b411b363 PR |
2073 | } |
2074 | ||
659b2e3b | 2075 | #define dec_unacked(device) _dec_unacked(device, __func__, __LINE__) |
b30ab791 | 2076 | static inline void _dec_unacked(struct drbd_device *device, const char *func, int line) |
b411b363 | 2077 | { |
b30ab791 | 2078 | atomic_dec(&device->unacked_cnt); |
49559d87 | 2079 | ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line); |
b411b363 PR |
2080 | } |
2081 | ||
659b2e3b | 2082 | #define sub_unacked(device, n) _sub_unacked(device, n, __func__, __LINE__) |
b30ab791 | 2083 | static inline void _sub_unacked(struct drbd_device *device, int n, const char *func, int line) |
b411b363 | 2084 | { |
b30ab791 | 2085 | atomic_sub(n, &device->unacked_cnt); |
49559d87 | 2086 | ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line); |
b411b363 PR |
2087 | } |
2088 | ||
5ab7d2c0 LE |
2089 | static inline bool is_sync_state(enum drbd_conns connection_state) |
2090 | { | |
2091 | return | |
2092 | (connection_state == C_SYNC_SOURCE | |
2093 | || connection_state == C_SYNC_TARGET | |
2094 | || connection_state == C_PAUSED_SYNC_S | |
2095 | || connection_state == C_PAUSED_SYNC_T); | |
2096 | } | |
2097 | ||
b411b363 | 2098 | /** |
b30ab791 | 2099 | * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev |
d1b80853 AG |
2100 | * @_device: DRBD device. |
2101 | * @_min_state: Minimum device state required for success. | |
b411b363 | 2102 | * |
b30ab791 | 2103 | * You have to call put_ldev() when finished working with device->ldev. |
b411b363 | 2104 | */ |
d1b80853 AG |
2105 | #define get_ldev_if_state(_device, _min_state) \ |
2106 | (_get_ldev_if_state((_device), (_min_state)) ? \ | |
2107 | ({ __acquire(x); true; }) : false) | |
2108 | #define get_ldev(_device) get_ldev_if_state(_device, D_INCONSISTENT) | |
b411b363 | 2109 | |
b30ab791 | 2110 | static inline void put_ldev(struct drbd_device *device) |
b411b363 | 2111 | { |
11f8b2b6 | 2112 | enum drbd_disk_state disk_state = device->state.disk; |
ba3c6fb8 LE |
2113 | /* We must check the state *before* the atomic_dec becomes visible, |
2114 | * or we have a theoretical race where someone hitting zero, | |
2115 | * while state still D_FAILED, will then see D_DISKLESS in the | |
2116 | * condition below and calling into destroy, where he must not, yet. */ | |
b30ab791 | 2117 | int i = atomic_dec_return(&device->local_cnt); |
9a0d9d03 LE |
2118 | |
2119 | /* This may be called from some endio handler, | |
2120 | * so we must not sleep here. */ | |
2121 | ||
b411b363 | 2122 | __release(local); |
0b0ba1ef | 2123 | D_ASSERT(device, i >= 0); |
e9e6f3ec | 2124 | if (i == 0) { |
11f8b2b6 | 2125 | if (disk_state == D_DISKLESS) |
82f59cc6 | 2126 | /* even internal references gone, safe to destroy */ |
e334f550 | 2127 | drbd_device_post_work(device, DESTROY_DISK); |
11f8b2b6 | 2128 | if (disk_state == D_FAILED) |
82f59cc6 | 2129 | /* all application IO references gone. */ |
e334f550 LE |
2130 | if (!test_and_set_bit(GOING_DISKLESS, &device->flags)) |
2131 | drbd_device_post_work(device, GO_DISKLESS); | |
b30ab791 | 2132 | wake_up(&device->misc_wait); |
e9e6f3ec | 2133 | } |
b411b363 PR |
2134 | } |
2135 | ||
2136 | #ifndef __CHECKER__ | |
b30ab791 | 2137 | static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins) |
b411b363 PR |
2138 | { |
2139 | int io_allowed; | |
2140 | ||
82f59cc6 | 2141 | /* never get a reference while D_DISKLESS */ |
b30ab791 | 2142 | if (device->state.disk == D_DISKLESS) |
82f59cc6 LE |
2143 | return 0; |
2144 | ||
b30ab791 AG |
2145 | atomic_inc(&device->local_cnt); |
2146 | io_allowed = (device->state.disk >= mins); | |
b411b363 | 2147 | if (!io_allowed) |
b30ab791 | 2148 | put_ldev(device); |
b411b363 PR |
2149 | return io_allowed; |
2150 | } | |
2151 | #else | |
b30ab791 | 2152 | extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins); |
b411b363 PR |
2153 | #endif |
2154 | ||
b411b363 PR |
2155 | /* this throttles on-the-fly application requests |
2156 | * according to max_buffers settings; | |
2157 | * maybe re-implement using semaphores? */ | |
b30ab791 | 2158 | static inline int drbd_get_max_buffers(struct drbd_device *device) |
b411b363 | 2159 | { |
44ed167d PR |
2160 | struct net_conf *nc; |
2161 | int mxb; | |
2162 | ||
2163 | rcu_read_lock(); | |
a6b32bc3 | 2164 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d PR |
2165 | mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */ |
2166 | rcu_read_unlock(); | |
2167 | ||
b411b363 PR |
2168 | return mxb; |
2169 | } | |
2170 | ||
b30ab791 | 2171 | static inline int drbd_state_is_stable(struct drbd_device *device) |
b411b363 | 2172 | { |
b30ab791 | 2173 | union drbd_dev_state s = device->state; |
b411b363 PR |
2174 | |
2175 | /* DO NOT add a default clause, we want the compiler to warn us | |
2176 | * for any newly introduced state we may have forgotten to add here */ | |
2177 | ||
2178 | switch ((enum drbd_conns)s.conn) { | |
2179 | /* new io only accepted when there is no connection, ... */ | |
2180 | case C_STANDALONE: | |
2181 | case C_WF_CONNECTION: | |
2182 | /* ... or there is a well established connection. */ | |
2183 | case C_CONNECTED: | |
2184 | case C_SYNC_SOURCE: | |
2185 | case C_SYNC_TARGET: | |
2186 | case C_VERIFY_S: | |
2187 | case C_VERIFY_T: | |
2188 | case C_PAUSED_SYNC_S: | |
2189 | case C_PAUSED_SYNC_T: | |
67531718 PR |
2190 | case C_AHEAD: |
2191 | case C_BEHIND: | |
3719094e | 2192 | /* transitional states, IO allowed */ |
b411b363 PR |
2193 | case C_DISCONNECTING: |
2194 | case C_UNCONNECTED: | |
2195 | case C_TIMEOUT: | |
2196 | case C_BROKEN_PIPE: | |
2197 | case C_NETWORK_FAILURE: | |
2198 | case C_PROTOCOL_ERROR: | |
2199 | case C_TEAR_DOWN: | |
2200 | case C_WF_REPORT_PARAMS: | |
2201 | case C_STARTING_SYNC_S: | |
2202 | case C_STARTING_SYNC_T: | |
3719094e PR |
2203 | break; |
2204 | ||
2205 | /* Allow IO in BM exchange states with new protocols */ | |
b411b363 | 2206 | case C_WF_BITMAP_S: |
a6b32bc3 | 2207 | if (first_peer_device(device)->connection->agreed_pro_version < 96) |
3719094e PR |
2208 | return 0; |
2209 | break; | |
2210 | ||
2211 | /* no new io accepted in these states */ | |
b411b363 PR |
2212 | case C_WF_BITMAP_T: |
2213 | case C_WF_SYNC_UUID: | |
2214 | case C_MASK: | |
2215 | /* not "stable" */ | |
2216 | return 0; | |
2217 | } | |
2218 | ||
2219 | switch ((enum drbd_disk_state)s.disk) { | |
2220 | case D_DISKLESS: | |
2221 | case D_INCONSISTENT: | |
2222 | case D_OUTDATED: | |
2223 | case D_CONSISTENT: | |
2224 | case D_UP_TO_DATE: | |
5ca1de03 | 2225 | case D_FAILED: |
b411b363 PR |
2226 | /* disk state is stable as well. */ |
2227 | break; | |
2228 | ||
d942ae44 | 2229 | /* no new io accepted during transitional states */ |
b411b363 | 2230 | case D_ATTACHING: |
b411b363 PR |
2231 | case D_NEGOTIATING: |
2232 | case D_UNKNOWN: | |
2233 | case D_MASK: | |
2234 | /* not "stable" */ | |
2235 | return 0; | |
2236 | } | |
2237 | ||
2238 | return 1; | |
2239 | } | |
2240 | ||
b30ab791 | 2241 | static inline int drbd_suspended(struct drbd_device *device) |
fb22c402 | 2242 | { |
6bbf53ca | 2243 | struct drbd_resource *resource = device->resource; |
8e0af25f | 2244 | |
6bbf53ca | 2245 | return resource->susp || resource->susp_fen || resource->susp_nod; |
fb22c402 PR |
2246 | } |
2247 | ||
b30ab791 | 2248 | static inline bool may_inc_ap_bio(struct drbd_device *device) |
b411b363 | 2249 | { |
b30ab791 | 2250 | int mxb = drbd_get_max_buffers(device); |
b411b363 | 2251 | |
b30ab791 | 2252 | if (drbd_suspended(device)) |
1b881ef7 | 2253 | return false; |
b30ab791 | 2254 | if (test_bit(SUSPEND_IO, &device->flags)) |
1b881ef7 | 2255 | return false; |
b411b363 PR |
2256 | |
2257 | /* to avoid potential deadlock or bitmap corruption, | |
2258 | * in various places, we only allow new application io | |
2259 | * to start during "stable" states. */ | |
2260 | ||
2261 | /* no new io accepted when attaching or detaching the disk */ | |
b30ab791 | 2262 | if (!drbd_state_is_stable(device)) |
1b881ef7 | 2263 | return false; |
b411b363 PR |
2264 | |
2265 | /* since some older kernels don't have atomic_add_unless, | |
2266 | * and we are within the spinlock anyways, we have this workaround. */ | |
b30ab791 | 2267 | if (atomic_read(&device->ap_bio_cnt) > mxb) |
1b881ef7 | 2268 | return false; |
b30ab791 | 2269 | if (test_bit(BITMAP_IO, &device->flags)) |
1b881ef7 AG |
2270 | return false; |
2271 | return true; | |
b411b363 PR |
2272 | } |
2273 | ||
b30ab791 | 2274 | static inline bool inc_ap_bio_cond(struct drbd_device *device) |
b411b363 | 2275 | { |
1b881ef7 | 2276 | bool rv = false; |
8869d683 | 2277 | |
0500813f | 2278 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 2279 | rv = may_inc_ap_bio(device); |
8869d683 | 2280 | if (rv) |
b30ab791 | 2281 | atomic_inc(&device->ap_bio_cnt); |
0500813f | 2282 | spin_unlock_irq(&device->resource->req_lock); |
8869d683 PR |
2283 | |
2284 | return rv; | |
2285 | } | |
b411b363 | 2286 | |
b30ab791 | 2287 | static inline void inc_ap_bio(struct drbd_device *device) |
8869d683 | 2288 | { |
b411b363 PR |
2289 | /* we wait here |
2290 | * as long as the device is suspended | |
2291 | * until the bitmap is no longer on the fly during connection | |
d942ae44 | 2292 | * handshake as long as we would exceed the max_buffer limit. |
b411b363 PR |
2293 | * |
2294 | * to avoid races with the reconnect code, | |
2295 | * we need to atomic_inc within the spinlock. */ | |
2296 | ||
b30ab791 | 2297 | wait_event(device->misc_wait, inc_ap_bio_cond(device)); |
b411b363 PR |
2298 | } |
2299 | ||
b30ab791 | 2300 | static inline void dec_ap_bio(struct drbd_device *device) |
b411b363 | 2301 | { |
b30ab791 AG |
2302 | int mxb = drbd_get_max_buffers(device); |
2303 | int ap_bio = atomic_dec_return(&device->ap_bio_cnt); | |
b411b363 | 2304 | |
0b0ba1ef | 2305 | D_ASSERT(device, ap_bio >= 0); |
7ee1fb93 | 2306 | |
b30ab791 AG |
2307 | if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) { |
2308 | if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags)) | |
84b8c06b AG |
2309 | drbd_queue_work(&first_peer_device(device)-> |
2310 | connection->sender_work, | |
2311 | &device->bm_io_work.w); | |
7ee1fb93 LE |
2312 | } |
2313 | ||
b411b363 PR |
2314 | /* this currently does wake_up for every dec_ap_bio! |
2315 | * maybe rather introduce some type of hysteresis? | |
2316 | * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */ | |
2317 | if (ap_bio < mxb) | |
b30ab791 | 2318 | wake_up(&device->misc_wait); |
b411b363 PR |
2319 | } |
2320 | ||
b30ab791 | 2321 | static inline bool verify_can_do_stop_sector(struct drbd_device *device) |
58ffa580 | 2322 | { |
a6b32bc3 AG |
2323 | return first_peer_device(device)->connection->agreed_pro_version >= 97 && |
2324 | first_peer_device(device)->connection->agreed_pro_version != 100; | |
58ffa580 LE |
2325 | } |
2326 | ||
b30ab791 | 2327 | static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val) |
b411b363 | 2328 | { |
b30ab791 AG |
2329 | int changed = device->ed_uuid != val; |
2330 | device->ed_uuid = val; | |
62b0da3a | 2331 | return changed; |
b411b363 PR |
2332 | } |
2333 | ||
b30ab791 | 2334 | static inline int drbd_queue_order_type(struct drbd_device *device) |
b411b363 PR |
2335 | { |
2336 | /* sorry, we currently have no working implementation | |
2337 | * of distributed TCQ stuff */ | |
2338 | #ifndef QUEUE_ORDERED_NONE | |
2339 | #define QUEUE_ORDERED_NONE 0 | |
2340 | #endif | |
2341 | return QUEUE_ORDERED_NONE; | |
2342 | } | |
2343 | ||
77c556f6 AG |
2344 | static inline struct drbd_connection *first_connection(struct drbd_resource *resource) |
2345 | { | |
ec4a3407 | 2346 | return list_first_entry_or_null(&resource->connections, |
77c556f6 AG |
2347 | struct drbd_connection, connections); |
2348 | } | |
2349 | ||
b411b363 | 2350 | #endif |