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 PR |
47 | #include <linux/drbd.h> |
48 | #include "drbd_state.h" | |
a3603a6e | 49 | #include "drbd_protocol.h" |
b411b363 PR |
50 | |
51 | #ifdef __CHECKER__ | |
52 | # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr"))) | |
53 | # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read"))) | |
54 | # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write"))) | |
55 | # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call"))) | |
56 | #else | |
57 | # define __protected_by(x) | |
58 | # define __protected_read_by(x) | |
59 | # define __protected_write_by(x) | |
60 | # define __must_hold(x) | |
61 | #endif | |
62 | ||
63 | #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0) | |
64 | ||
65 | /* module parameter, defined in drbd_main.c */ | |
66 | extern unsigned int minor_count; | |
90ab5ee9 RR |
67 | extern bool disable_sendpage; |
68 | extern bool allow_oos; | |
b30ab791 | 69 | void tl_abort_disk_io(struct drbd_device *device); |
b411b363 PR |
70 | |
71 | #ifdef CONFIG_DRBD_FAULT_INJECTION | |
72 | extern int enable_faults; | |
73 | extern int fault_rate; | |
74 | extern int fault_devs; | |
75 | #endif | |
76 | ||
77 | extern char usermode_helper[]; | |
78 | ||
79 | ||
b411b363 PR |
80 | /* I don't remember why XCPU ... |
81 | * This is used to wake the asender, | |
82 | * and to interrupt sending the sending task | |
83 | * on disconnect. | |
84 | */ | |
85 | #define DRBD_SIG SIGXCPU | |
86 | ||
87 | /* This is used to stop/restart our threads. | |
88 | * Cannot use SIGTERM nor SIGKILL, since these | |
89 | * are sent out by init on runlevel changes | |
90 | * I choose SIGHUP for now. | |
91 | */ | |
92 | #define DRBD_SIGKILL SIGHUP | |
93 | ||
b411b363 PR |
94 | #define ID_IN_SYNC (4711ULL) |
95 | #define ID_OUT_OF_SYNC (4712ULL) | |
b411b363 | 96 | #define ID_SYNCER (-1ULL) |
579b57ed | 97 | |
4a23f264 | 98 | #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL) |
b411b363 | 99 | |
54761697 | 100 | struct drbd_device; |
bde89a9e | 101 | struct drbd_connection; |
b411b363 PR |
102 | |
103 | ||
104 | /* to shorten dev_warn(DEV, "msg"); and relatives statements */ | |
b30ab791 | 105 | #define DEV (disk_to_dev(device->vdisk)) |
b411b363 | 106 | |
60ae4966 | 107 | #define conn_printk(LEVEL, TCONN, FMT, ARGS...) \ |
77c556f6 | 108 | printk(LEVEL "d-con %s: " FMT, TCONN->resource->name , ## ARGS) |
60ae4966 PR |
109 | #define conn_alert(TCONN, FMT, ARGS...) conn_printk(KERN_ALERT, TCONN, FMT, ## ARGS) |
110 | #define conn_crit(TCONN, FMT, ARGS...) conn_printk(KERN_CRIT, TCONN, FMT, ## ARGS) | |
111 | #define conn_err(TCONN, FMT, ARGS...) conn_printk(KERN_ERR, TCONN, FMT, ## ARGS) | |
112 | #define conn_warn(TCONN, FMT, ARGS...) conn_printk(KERN_WARNING, TCONN, FMT, ## ARGS) | |
113 | #define conn_notice(TCONN, FMT, ARGS...) conn_printk(KERN_NOTICE, TCONN, FMT, ## ARGS) | |
114 | #define conn_info(TCONN, FMT, ARGS...) conn_printk(KERN_INFO, TCONN, FMT, ## ARGS) | |
115 | #define conn_dbg(TCONN, FMT, ARGS...) conn_printk(KERN_DEBUG, TCONN, FMT, ## ARGS) | |
116 | ||
b411b363 PR |
117 | #define D_ASSERT(exp) if (!(exp)) \ |
118 | dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__) | |
119 | ||
841ce241 AG |
120 | /** |
121 | * expect - Make an assertion | |
122 | * | |
123 | * Unlike the assert macro, this macro returns a boolean result. | |
124 | */ | |
125 | #define expect(exp) ({ \ | |
126 | bool _bool = (exp); \ | |
127 | if (!_bool) \ | |
128 | dev_err(DEV, "ASSERTION %s FAILED in %s\n", \ | |
129 | #exp, __func__); \ | |
130 | _bool; \ | |
131 | }) | |
b411b363 PR |
132 | |
133 | /* Defines to control fault insertion */ | |
134 | enum { | |
135 | DRBD_FAULT_MD_WR = 0, /* meta data write */ | |
136 | DRBD_FAULT_MD_RD = 1, /* read */ | |
137 | DRBD_FAULT_RS_WR = 2, /* resync */ | |
138 | DRBD_FAULT_RS_RD = 3, | |
139 | DRBD_FAULT_DT_WR = 4, /* data */ | |
140 | DRBD_FAULT_DT_RD = 5, | |
141 | DRBD_FAULT_DT_RA = 6, /* data read ahead */ | |
142 | DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */ | |
143 | DRBD_FAULT_AL_EE = 8, /* alloc ee */ | |
6b4388ac | 144 | DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */ |
b411b363 PR |
145 | |
146 | DRBD_FAULT_MAX, | |
147 | }; | |
148 | ||
b411b363 | 149 | extern unsigned int |
b30ab791 | 150 | _drbd_insert_fault(struct drbd_device *device, unsigned int type); |
0cf9d27e | 151 | |
b411b363 | 152 | static inline int |
b30ab791 | 153 | drbd_insert_fault(struct drbd_device *device, unsigned int type) { |
0cf9d27e | 154 | #ifdef CONFIG_DRBD_FAULT_INJECTION |
b411b363 PR |
155 | return fault_rate && |
156 | (enable_faults & (1<<type)) && | |
b30ab791 | 157 | _drbd_insert_fault(device, type); |
b411b363 | 158 | #else |
0cf9d27e | 159 | return 0; |
b411b363 | 160 | #endif |
0cf9d27e | 161 | } |
b411b363 PR |
162 | |
163 | /* integer division, round _UP_ to the next integer */ | |
164 | #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0)) | |
165 | /* usual integer division */ | |
166 | #define div_floor(A, B) ((A)/(B)) | |
167 | ||
b411b363 | 168 | extern struct ratelimit_state drbd_ratelimit_state; |
05a10ec7 | 169 | extern struct idr drbd_devices; /* RCU, updates: genl_lock() */ |
77c556f6 | 170 | extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */ |
b411b363 | 171 | |
d8763023 | 172 | extern const char *cmdname(enum drbd_packet cmd); |
b411b363 PR |
173 | |
174 | /* for sending/receiving the bitmap, | |
175 | * possibly in some encoding scheme */ | |
176 | struct bm_xfer_ctx { | |
177 | /* "const" | |
178 | * stores total bits and long words | |
179 | * of the bitmap, so we don't need to | |
180 | * call the accessor functions over and again. */ | |
181 | unsigned long bm_bits; | |
182 | unsigned long bm_words; | |
183 | /* during xfer, current position within the bitmap */ | |
184 | unsigned long bit_offset; | |
185 | unsigned long word_offset; | |
186 | ||
187 | /* statistics; index: (h->command == P_BITMAP) */ | |
188 | unsigned packets[2]; | |
189 | unsigned bytes[2]; | |
190 | }; | |
191 | ||
b30ab791 | 192 | extern void INFO_bm_xfer_stats(struct drbd_device *device, |
b411b363 PR |
193 | const char *direction, struct bm_xfer_ctx *c); |
194 | ||
195 | static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c) | |
196 | { | |
197 | /* word_offset counts "native long words" (32 or 64 bit), | |
198 | * aligned at 64 bit. | |
199 | * Encoded packet may end at an unaligned bit offset. | |
200 | * In case a fallback clear text packet is transmitted in | |
201 | * between, we adjust this offset back to the last 64bit | |
202 | * aligned "native long word", which makes coding and decoding | |
203 | * the plain text bitmap much more convenient. */ | |
204 | #if BITS_PER_LONG == 64 | |
205 | c->word_offset = c->bit_offset >> 6; | |
206 | #elif BITS_PER_LONG == 32 | |
207 | c->word_offset = c->bit_offset >> 5; | |
208 | c->word_offset &= ~(1UL); | |
209 | #else | |
210 | # error "unsupported BITS_PER_LONG" | |
211 | #endif | |
212 | } | |
213 | ||
bde89a9e | 214 | extern unsigned int drbd_header_size(struct drbd_connection *connection); |
b411b363 | 215 | |
b411b363 PR |
216 | /**********************************************************************/ |
217 | enum drbd_thread_state { | |
e77a0a5c AG |
218 | NONE, |
219 | RUNNING, | |
220 | EXITING, | |
221 | RESTARTING | |
b411b363 PR |
222 | }; |
223 | ||
224 | struct drbd_thread { | |
225 | spinlock_t t_lock; | |
226 | struct task_struct *task; | |
227 | struct completion stop; | |
228 | enum drbd_thread_state t_state; | |
229 | int (*function) (struct drbd_thread *); | |
bde89a9e | 230 | struct drbd_connection *connection; |
b411b363 | 231 | int reset_cpu_mask; |
bed879ae | 232 | char name[9]; |
b411b363 PR |
233 | }; |
234 | ||
235 | static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi) | |
236 | { | |
237 | /* THINK testing the t_state seems to be uncritical in all cases | |
238 | * (but thread_{start,stop}), so we can read it *without* the lock. | |
239 | * --lge */ | |
240 | ||
241 | smp_rmb(); | |
242 | return thi->t_state; | |
243 | } | |
244 | ||
b411b363 PR |
245 | struct drbd_work { |
246 | struct list_head list; | |
309a8348 | 247 | int (*cb)(struct drbd_work *, int cancel); |
00d56944 | 248 | union { |
b30ab791 | 249 | struct drbd_device *device; |
bde89a9e | 250 | struct drbd_connection *connection; |
00d56944 | 251 | }; |
b411b363 PR |
252 | }; |
253 | ||
ace652ac AG |
254 | #include "drbd_interval.h" |
255 | ||
54761697 | 256 | extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *); |
7be8da07 | 257 | |
b411b363 PR |
258 | struct drbd_request { |
259 | struct drbd_work w; | |
b411b363 PR |
260 | |
261 | /* if local IO is not allowed, will be NULL. | |
262 | * if local IO _is_ allowed, holds the locally submitted bio clone, | |
263 | * or, after local IO completion, the ERR_PTR(error). | |
fcefa62e | 264 | * see drbd_request_endio(). */ |
b411b363 PR |
265 | struct bio *private_bio; |
266 | ||
ace652ac | 267 | struct drbd_interval i; |
b411b363 | 268 | |
b6dd1a89 | 269 | /* epoch: used to check on "completion" whether this req was in |
b411b363 | 270 | * the current epoch, and we therefore have to close it, |
b6dd1a89 LE |
271 | * causing a p_barrier packet to be send, starting a new epoch. |
272 | * | |
273 | * This corresponds to "barrier" in struct p_barrier[_ack], | |
274 | * and to "barrier_nr" in struct drbd_epoch (and various | |
275 | * comments/function parameters/local variable names). | |
b411b363 | 276 | */ |
b6dd1a89 | 277 | unsigned int epoch; |
b411b363 | 278 | |
b411b363 PR |
279 | struct list_head tl_requests; /* ring list in the transfer log */ |
280 | struct bio *master_bio; /* master bio pointer */ | |
b411b363 | 281 | unsigned long start_time; |
b411b363 | 282 | |
b406777e LE |
283 | /* once it hits 0, we may complete the master_bio */ |
284 | atomic_t completion_ref; | |
285 | /* once it hits 0, we may destroy this drbd_request object */ | |
286 | struct kref kref; | |
b411b363 | 287 | |
a0d856df | 288 | unsigned rq_state; /* see comments above _req_mod() */ |
b411b363 | 289 | }; |
b411b363 PR |
290 | |
291 | struct drbd_epoch { | |
bde89a9e | 292 | struct drbd_connection *connection; |
b411b363 PR |
293 | struct list_head list; |
294 | unsigned int barrier_nr; | |
295 | atomic_t epoch_size; /* increased on every request added. */ | |
296 | atomic_t active; /* increased on every req. added, and dec on every finished. */ | |
297 | unsigned long flags; | |
298 | }; | |
299 | ||
de0b2e69 RK |
300 | /* Prototype declaration of function defined in drbd_receiver.c */ |
301 | int drbdd_init(struct drbd_thread *); | |
302 | int drbd_asender(struct drbd_thread *); | |
303 | ||
b411b363 PR |
304 | /* drbd_epoch flag bits */ |
305 | enum { | |
b411b363 | 306 | DE_HAVE_BARRIER_NUMBER, |
b411b363 PR |
307 | }; |
308 | ||
309 | enum epoch_event { | |
310 | EV_PUT, | |
311 | EV_GOT_BARRIER_NR, | |
b411b363 | 312 | EV_BECAME_LAST, |
b411b363 PR |
313 | EV_CLEANUP = 32, /* used as flag */ |
314 | }; | |
315 | ||
b411b363 PR |
316 | struct drbd_wq_barrier { |
317 | struct drbd_work w; | |
318 | struct completion done; | |
319 | }; | |
320 | ||
321 | struct digest_info { | |
322 | int digest_size; | |
323 | void *digest; | |
324 | }; | |
325 | ||
f6ffca9f | 326 | struct drbd_peer_request { |
45bb912b | 327 | struct drbd_work w; |
85719573 | 328 | struct drbd_epoch *epoch; /* for writes */ |
45bb912b LE |
329 | struct page *pages; |
330 | atomic_t pending_bios; | |
010f6e67 | 331 | struct drbd_interval i; |
45bb912b LE |
332 | /* see comments on ee flag bits below */ |
333 | unsigned long flags; | |
85719573 PR |
334 | union { |
335 | u64 block_id; | |
336 | struct digest_info *digest; | |
337 | }; | |
45bb912b LE |
338 | }; |
339 | ||
340 | /* ee flag bits. | |
341 | * While corresponding bios are in flight, the only modification will be | |
342 | * set_bit WAS_ERROR, which has to be atomic. | |
343 | * If no bios are in flight yet, or all have been completed, | |
344 | * non-atomic modification to ee->flags is ok. | |
345 | */ | |
b411b363 PR |
346 | enum { |
347 | __EE_CALL_AL_COMPLETE_IO, | |
b411b363 | 348 | __EE_MAY_SET_IN_SYNC, |
45bb912b | 349 | |
45bb912b LE |
350 | /* In case a barrier failed, |
351 | * we need to resubmit without the barrier flag. */ | |
352 | __EE_RESUBMITTED, | |
353 | ||
6c852bec | 354 | /* we may have several bios per peer request. |
45bb912b LE |
355 | * if any of those fail, we set this flag atomically |
356 | * from the endio callback */ | |
357 | __EE_WAS_ERROR, | |
c36c3ced LE |
358 | |
359 | /* This ee has a pointer to a digest instead of a block id */ | |
360 | __EE_HAS_DIGEST, | |
7be8da07 AG |
361 | |
362 | /* Conflicting local requests need to be restarted after this request */ | |
363 | __EE_RESTART_REQUESTS, | |
303d1448 PR |
364 | |
365 | /* The peer wants a write ACK for this (wire proto C) */ | |
366 | __EE_SEND_WRITE_ACK, | |
302bdeae PR |
367 | |
368 | /* Is set when net_conf had two_primaries set while creating this peer_req */ | |
369 | __EE_IN_INTERVAL_TREE, | |
b411b363 PR |
370 | }; |
371 | #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO) | |
b411b363 | 372 | #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC) |
45bb912b LE |
373 | #define EE_RESUBMITTED (1<<__EE_RESUBMITTED) |
374 | #define EE_WAS_ERROR (1<<__EE_WAS_ERROR) | |
c36c3ced | 375 | #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST) |
7be8da07 | 376 | #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS) |
303d1448 | 377 | #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK) |
302bdeae | 378 | #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE) |
b411b363 | 379 | |
b30ab791 | 380 | /* flag bits per device */ |
b411b363 | 381 | enum { |
b411b363 PR |
382 | UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */ |
383 | MD_DIRTY, /* current uuids and flags not yet on disk */ | |
b411b363 | 384 | USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */ |
b411b363 PR |
385 | CL_ST_CHG_SUCCESS, |
386 | CL_ST_CHG_FAIL, | |
387 | CRASHED_PRIMARY, /* This node was a crashed primary. | |
388 | * Gets cleared when the state.conn | |
389 | * goes into C_CONNECTED state. */ | |
b411b363 PR |
390 | CONSIDER_RESYNC, |
391 | ||
a8a4e51e | 392 | MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */ |
b411b363 PR |
393 | SUSPEND_IO, /* suspend application io */ |
394 | BITMAP_IO, /* suspend application io; | |
395 | once no more io in flight, start bitmap io */ | |
396 | BITMAP_IO_QUEUED, /* Started bitmap IO */ | |
82f59cc6 | 397 | GO_DISKLESS, /* Disk is being detached, on io-error or admin request. */ |
a2a3c74f LE |
398 | WAS_IO_ERROR, /* Local disk failed, returned IO error */ |
399 | WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */ | |
383606e0 | 400 | FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */ |
b411b363 | 401 | RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */ |
b411b363 PR |
402 | RESIZE_PENDING, /* Size change detected locally, waiting for the response from |
403 | * the peer, if it changed there as well. */ | |
43a5182c | 404 | NEW_CUR_UUID, /* Create new current UUID when thawing IO */ |
0778286a | 405 | AL_SUSPENDED, /* Activity logging is currently suspended. */ |
370a43e7 | 406 | AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */ |
e64a3294 | 407 | B_RS_H_DONE, /* Before resync handler done (already executed) */ |
08b165ba | 408 | DISCARD_MY_DATA, /* discard_my_data flag per volume */ |
380207d0 | 409 | READ_BALANCE_RR, |
b411b363 PR |
410 | }; |
411 | ||
54761697 | 412 | struct drbd_bitmap; /* opaque for drbd_device */ |
b411b363 | 413 | |
20ceb2b2 LE |
414 | /* definition of bits in bm_flags to be used in drbd_bm_lock |
415 | * and drbd_bitmap_io and friends. */ | |
416 | enum bm_flag { | |
417 | /* do we need to kfree, or vfree bm_pages? */ | |
418 | BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */ | |
419 | ||
420 | /* currently locked for bulk operation */ | |
0e8488ad | 421 | BM_LOCKED_MASK = 0xf, |
20ceb2b2 LE |
422 | |
423 | /* in detail, that is: */ | |
424 | BM_DONT_CLEAR = 0x1, | |
425 | BM_DONT_SET = 0x2, | |
426 | BM_DONT_TEST = 0x4, | |
427 | ||
0e8488ad LE |
428 | /* so we can mark it locked for bulk operation, |
429 | * and still allow all non-bulk operations */ | |
430 | BM_IS_LOCKED = 0x8, | |
431 | ||
20ceb2b2 | 432 | /* (test bit, count bit) allowed (common case) */ |
0e8488ad | 433 | BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED, |
20ceb2b2 LE |
434 | |
435 | /* testing bits, as well as setting new bits allowed, but clearing bits | |
436 | * would be unexpected. Used during bitmap receive. Setting new bits | |
437 | * requires sending of "out-of-sync" information, though. */ | |
0e8488ad | 438 | BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED, |
20ceb2b2 | 439 | |
0e8488ad LE |
440 | /* for drbd_bm_write_copy_pages, everything is allowed, |
441 | * only concurrent bulk operations are locked out. */ | |
442 | BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED, | |
20ceb2b2 LE |
443 | }; |
444 | ||
b411b363 PR |
445 | struct drbd_work_queue { |
446 | struct list_head q; | |
b411b363 | 447 | spinlock_t q_lock; /* to protect the list. */ |
8c0785a5 | 448 | wait_queue_head_t q_wait; |
b411b363 PR |
449 | }; |
450 | ||
451 | struct drbd_socket { | |
b411b363 PR |
452 | struct mutex mutex; |
453 | struct socket *socket; | |
454 | /* this way we get our | |
455 | * send/receive buffers off the stack */ | |
5a87d920 | 456 | void *sbuf; |
e6ef8a5c | 457 | void *rbuf; |
b411b363 PR |
458 | }; |
459 | ||
460 | struct drbd_md { | |
461 | u64 md_offset; /* sector offset to 'super' block */ | |
462 | ||
463 | u64 la_size_sect; /* last agreed size, unit sectors */ | |
9f2247bb | 464 | spinlock_t uuid_lock; |
b411b363 PR |
465 | u64 uuid[UI_SIZE]; |
466 | u64 device_uuid; | |
467 | u32 flags; | |
468 | u32 md_size_sect; | |
469 | ||
ae8bf312 | 470 | s32 al_offset; /* signed relative sector offset to activity log */ |
b411b363 | 471 | s32 bm_offset; /* signed relative sector offset to bitmap */ |
3a4d4eb3 LE |
472 | |
473 | /* cached value of bdev->disk_conf->meta_dev_idx (see below) */ | |
474 | s32 meta_dev_idx; | |
475 | ||
476 | /* see al_tr_number_to_on_disk_sector() */ | |
477 | u32 al_stripes; | |
478 | u32 al_stripe_size_4k; | |
479 | u32 al_size_4k; /* cached product of the above */ | |
b411b363 PR |
480 | }; |
481 | ||
b411b363 PR |
482 | struct drbd_backing_dev { |
483 | struct block_device *backing_bdev; | |
484 | struct block_device *md_bdev; | |
b411b363 | 485 | struct drbd_md md; |
a6b32bc3 | 486 | struct disk_conf *disk_conf; /* RCU, for updates: first_peer_device(device)->connection->conf_update */ |
b411b363 PR |
487 | sector_t known_size; /* last known size of that backing device */ |
488 | }; | |
489 | ||
490 | struct drbd_md_io { | |
0c464425 | 491 | unsigned int done; |
b411b363 PR |
492 | int error; |
493 | }; | |
494 | ||
495 | struct bm_io_work { | |
496 | struct drbd_work w; | |
497 | char *why; | |
20ceb2b2 | 498 | enum bm_flag flags; |
b30ab791 AG |
499 | int (*io_fn)(struct drbd_device *device); |
500 | void (*done)(struct drbd_device *device, int rv); | |
b411b363 PR |
501 | }; |
502 | ||
503 | enum write_ordering_e { | |
504 | WO_none, | |
505 | WO_drain_io, | |
506 | WO_bdev_flush, | |
b411b363 PR |
507 | }; |
508 | ||
778f271d | 509 | struct fifo_buffer { |
778f271d PR |
510 | unsigned int head_index; |
511 | unsigned int size; | |
9958c857 PR |
512 | int total; /* sum of all values */ |
513 | int values[0]; | |
778f271d | 514 | }; |
9958c857 | 515 | extern struct fifo_buffer *fifo_alloc(int fifo_size); |
778f271d | 516 | |
bde89a9e | 517 | /* flag bits per connection */ |
01a311a5 PR |
518 | enum { |
519 | NET_CONGESTED, /* The data socket is congested */ | |
427c0434 | 520 | RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */ |
e43ef195 | 521 | SEND_PING, /* whether asender should send a ping asap */ |
808e37b8 | 522 | SIGNAL_ASENDER, /* whether asender wants to be interrupted */ |
2a67d8b9 | 523 | GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */ |
4d0fc3fd | 524 | CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */ |
fc3b10a4 PR |
525 | CONN_WD_ST_CHG_OKAY, |
526 | CONN_WD_ST_CHG_FAIL, | |
8169e41b | 527 | CONN_DRY_RUN, /* Expect disconnect after resync handshake. */ |
6936fcb4 | 528 | CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */ |
a1096a6e | 529 | STATE_SENT, /* Do not change state/UUIDs while this is set */ |
6f3465ed LE |
530 | CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC) |
531 | * pending, from drbd worker context. | |
532 | * If set, bdi_write_congested() returns true, | |
533 | * so shrink_page_list() would not recurse into, | |
534 | * and potentially deadlock on, this drbd worker. | |
535 | */ | |
b66623e3 | 536 | DISCONNECT_SENT, |
01a311a5 PR |
537 | }; |
538 | ||
77c556f6 AG |
539 | struct drbd_resource { |
540 | char *name; | |
541 | struct kref kref; | |
542 | struct list_head connections; | |
543 | struct list_head resources; | |
544 | }; | |
545 | ||
546 | struct drbd_connection { | |
547 | struct list_head connections; | |
548 | struct drbd_resource *resource; | |
9dc9fbb3 | 549 | struct kref kref; |
bde89a9e | 550 | struct idr volumes; /* <connection, vnr> to device mapping */ |
8410da8f | 551 | enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */ |
8e0af25f PR |
552 | unsigned susp:1; /* IO suspended by user */ |
553 | unsigned susp_nod:1; /* IO suspended because no data */ | |
554 | unsigned susp_fen:1; /* IO suspended because fence peer handler runs */ | |
8410da8f | 555 | struct mutex cstate_mutex; /* Protects graceful disconnects */ |
28e448bb | 556 | unsigned int connect_cnt; /* Inc each time a connection is established */ |
2111438b | 557 | |
062e879c | 558 | unsigned long flags; |
44ed167d | 559 | struct net_conf *net_conf; /* content protected by rcu */ |
a0095508 | 560 | struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */ |
91fd4dad | 561 | wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */ |
f399002e | 562 | struct res_opts res_opts; |
e42325a5 | 563 | |
089c075d AG |
564 | struct sockaddr_storage my_addr; |
565 | int my_addr_len; | |
566 | struct sockaddr_storage peer_addr; | |
567 | int peer_addr_len; | |
568 | ||
e42325a5 PR |
569 | struct drbd_socket data; /* data/barrier/cstate/parameter packets */ |
570 | struct drbd_socket meta; /* ping/ack (metadata) packets */ | |
31890f4a PR |
571 | int agreed_pro_version; /* actually used protocol version */ |
572 | unsigned long last_received; /* in jiffies, either socket */ | |
573 | unsigned int ko_count; | |
e6b3ea83 | 574 | |
87eeee41 | 575 | spinlock_t req_lock; |
b6dd1a89 LE |
576 | |
577 | struct list_head transfer_log; /* all requests not yet fully processed */ | |
87eeee41 | 578 | |
a0638456 | 579 | struct crypto_hash *cram_hmac_tfm; |
bde89a9e | 580 | struct crypto_hash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */ |
036b17ea | 581 | struct crypto_hash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */ |
f399002e LE |
582 | struct crypto_hash *csums_tfm; |
583 | struct crypto_hash *verify_tfm; | |
a0638456 PR |
584 | void *int_dig_in; |
585 | void *int_dig_vv; | |
586 | ||
b6dd1a89 | 587 | /* receiver side */ |
12038a3a PR |
588 | struct drbd_epoch *current_epoch; |
589 | spinlock_t epoch_lock; | |
590 | unsigned int epochs; | |
4b0007c0 | 591 | enum write_ordering_e write_ordering; |
b379c41e | 592 | atomic_t current_tle_nr; /* transfer log epoch number */ |
b6dd1a89 | 593 | unsigned current_tle_writes; /* writes seen within this tl epoch */ |
4b0007c0 | 594 | |
07be15b1 | 595 | unsigned long last_reconnect_jif; |
e6b3ea83 PR |
596 | struct drbd_thread receiver; |
597 | struct drbd_thread worker; | |
598 | struct drbd_thread asender; | |
80822284 | 599 | cpumask_var_t cpu_mask; |
b6dd1a89 LE |
600 | |
601 | /* sender side */ | |
d5b27b01 | 602 | struct drbd_work_queue sender_work; |
b6dd1a89 LE |
603 | |
604 | struct { | |
605 | /* whether this sender thread | |
606 | * has processed a single write yet. */ | |
607 | bool seen_any_write_yet; | |
608 | ||
609 | /* Which barrier number to send with the next P_BARRIER */ | |
610 | int current_epoch_nr; | |
611 | ||
612 | /* how many write requests have been sent | |
613 | * with req->epoch == current_epoch_nr. | |
614 | * If none, no P_BARRIER will be sent. */ | |
615 | unsigned current_epoch_writes; | |
616 | } send; | |
778f271d PR |
617 | }; |
618 | ||
113fef9e LE |
619 | struct submit_worker { |
620 | struct workqueue_struct *wq; | |
621 | struct work_struct worker; | |
622 | ||
623 | spinlock_t lock; | |
624 | struct list_head writes; | |
625 | }; | |
626 | ||
a6b32bc3 AG |
627 | struct drbd_peer_device { |
628 | struct list_head peer_devices; | |
629 | struct drbd_device *device; | |
bde89a9e | 630 | struct drbd_connection *connection; |
a6b32bc3 AG |
631 | }; |
632 | ||
633 | struct drbd_device { | |
634 | struct list_head peer_devices; | |
2111438b | 635 | int vnr; /* volume number within the connection */ |
81fa2e67 | 636 | struct kref kref; |
2111438b | 637 | |
b411b363 PR |
638 | /* things that are stored as / read from meta data on disk */ |
639 | unsigned long flags; | |
640 | ||
641 | /* configured by drbdsetup */ | |
b411b363 PR |
642 | struct drbd_backing_dev *ldev __protected_by(local); |
643 | ||
644 | sector_t p_size; /* partner's disk size */ | |
645 | struct request_queue *rq_queue; | |
646 | struct block_device *this_bdev; | |
647 | struct gendisk *vdisk; | |
648 | ||
07be15b1 | 649 | unsigned long last_reattach_jif; |
b411b363 PR |
650 | struct drbd_work resync_work, |
651 | unplug_work, | |
e9e6f3ec | 652 | go_diskless, |
c4752ef1 PR |
653 | md_sync_work, |
654 | start_resync_work; | |
b411b363 PR |
655 | struct timer_list resync_timer; |
656 | struct timer_list md_sync_timer; | |
370a43e7 | 657 | struct timer_list start_resync_timer; |
7fde2be9 | 658 | struct timer_list request_timer; |
ee15b038 LE |
659 | #ifdef DRBD_DEBUG_MD_SYNC |
660 | struct { | |
661 | unsigned int line; | |
662 | const char* func; | |
663 | } last_md_mark_dirty; | |
664 | #endif | |
b411b363 PR |
665 | |
666 | /* Used after attach while negotiating new disk state. */ | |
667 | union drbd_state new_state_tmp; | |
668 | ||
da9fbc27 | 669 | union drbd_dev_state state; |
b411b363 PR |
670 | wait_queue_head_t misc_wait; |
671 | wait_queue_head_t state_wait; /* upon each state change. */ | |
672 | unsigned int send_cnt; | |
673 | unsigned int recv_cnt; | |
674 | unsigned int read_cnt; | |
675 | unsigned int writ_cnt; | |
676 | unsigned int al_writ_cnt; | |
677 | unsigned int bm_writ_cnt; | |
678 | atomic_t ap_bio_cnt; /* Requests we need to complete */ | |
679 | atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */ | |
680 | atomic_t rs_pending_cnt; /* RS request/data packets on the wire */ | |
d942ae44 | 681 | atomic_t unacked_cnt; /* Need to send replies for */ |
b411b363 | 682 | atomic_t local_cnt; /* Waiting for local completion */ |
b2fb6dbe | 683 | |
dac1389c AG |
684 | /* Interval tree of pending local requests */ |
685 | struct rb_root read_requests; | |
de696716 | 686 | struct rb_root write_requests; |
b411b363 | 687 | |
4b0715f0 | 688 | /* blocks to resync in this run [unit BM_BLOCK_SIZE] */ |
b411b363 | 689 | unsigned long rs_total; |
4b0715f0 | 690 | /* number of resync blocks that failed in this run */ |
b411b363 PR |
691 | unsigned long rs_failed; |
692 | /* Syncer's start time [unit jiffies] */ | |
693 | unsigned long rs_start; | |
694 | /* cumulated time in PausedSyncX state [unit jiffies] */ | |
695 | unsigned long rs_paused; | |
1d7734a0 LE |
696 | /* skipped because csum was equal [unit BM_BLOCK_SIZE] */ |
697 | unsigned long rs_same_csum; | |
698 | #define DRBD_SYNC_MARKS 8 | |
699 | #define DRBD_SYNC_MARK_STEP (3*HZ) | |
b411b363 | 700 | /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */ |
1d7734a0 | 701 | unsigned long rs_mark_left[DRBD_SYNC_MARKS]; |
b411b363 | 702 | /* marks's time [unit jiffies] */ |
1d7734a0 LE |
703 | unsigned long rs_mark_time[DRBD_SYNC_MARKS]; |
704 | /* current index into rs_mark_{left,time} */ | |
705 | int rs_last_mark; | |
328e0f12 | 706 | unsigned long rs_last_bcast; /* [unit jiffies] */ |
b411b363 PR |
707 | |
708 | /* where does the admin want us to start? (sector) */ | |
709 | sector_t ov_start_sector; | |
02b91b55 | 710 | sector_t ov_stop_sector; |
b411b363 PR |
711 | /* where are we now? (sector) */ |
712 | sector_t ov_position; | |
713 | /* Start sector of out of sync range (to merge printk reporting). */ | |
714 | sector_t ov_last_oos_start; | |
715 | /* size of out-of-sync range in sectors. */ | |
716 | sector_t ov_last_oos_size; | |
717 | unsigned long ov_left; /* in bits */ | |
b411b363 | 718 | |
b411b363 PR |
719 | struct drbd_bitmap *bitmap; |
720 | unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */ | |
721 | ||
722 | /* Used to track operations of resync... */ | |
723 | struct lru_cache *resync; | |
724 | /* Number of locked elements in resync LRU */ | |
725 | unsigned int resync_locked; | |
726 | /* resync extent number waiting for application requests */ | |
727 | unsigned int resync_wenr; | |
728 | ||
729 | int open_cnt; | |
730 | u64 *p_uuid; | |
4b0007c0 | 731 | |
85719573 PR |
732 | struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */ |
733 | struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */ | |
18b75d75 AG |
734 | struct list_head done_ee; /* need to send P_WRITE_ACK */ |
735 | struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */ | |
b411b363 | 736 | struct list_head net_ee; /* zero-copy network send in progress */ |
b411b363 PR |
737 | |
738 | int next_barrier_nr; | |
b411b363 | 739 | struct list_head resync_reads; |
435f0740 LE |
740 | atomic_t pp_in_use; /* allocated from page pool */ |
741 | atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */ | |
b411b363 PR |
742 | wait_queue_head_t ee_wait; |
743 | struct page *md_io_page; /* one page buffer for md_io */ | |
cc94c650 | 744 | struct drbd_md_io md_io; |
e1711731 | 745 | atomic_t md_io_in_use; /* protects the md_io, md_io_page and md_io_tmpp */ |
b411b363 PR |
746 | spinlock_t al_lock; |
747 | wait_queue_head_t al_wait; | |
748 | struct lru_cache *act_log; /* activity log */ | |
749 | unsigned int al_tr_number; | |
750 | int al_tr_cycle; | |
b411b363 PR |
751 | wait_queue_head_t seq_wait; |
752 | atomic_t packet_seq; | |
753 | unsigned int peer_seq; | |
754 | spinlock_t peer_seq_lock; | |
755 | unsigned int minor; | |
756 | unsigned long comm_bm_set; /* communicated number of set bits. */ | |
b411b363 PR |
757 | struct bm_io_work bm_io_work; |
758 | u64 ed_uuid; /* UUID of the exposed data */ | |
8410da8f | 759 | struct mutex own_state_mutex; |
a6b32bc3 | 760 | struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */ |
b411b363 | 761 | char congestion_reason; /* Why we where congested... */ |
1d7734a0 LE |
762 | atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */ |
763 | atomic_t rs_sect_ev; /* for submitted resync data rate, both */ | |
764 | int rs_last_sect_ev; /* counter to compare with */ | |
765 | int rs_last_events; /* counter of read or write "events" (unit sectors) | |
766 | * on the lower level device when we last looked. */ | |
767 | int c_sync_rate; /* current resync rate after syncer throttle magic */ | |
bde89a9e | 768 | struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */ |
778f271d | 769 | int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */ |
759fbdfb | 770 | atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */ |
db141b2f LE |
771 | unsigned int peer_max_bio_size; |
772 | unsigned int local_max_bio_size; | |
113fef9e LE |
773 | |
774 | /* any requests that would block in drbd_make_request() | |
775 | * are deferred to this single-threaded work queue */ | |
776 | struct submit_worker submit; | |
b411b363 PR |
777 | }; |
778 | ||
b30ab791 | 779 | static inline struct drbd_device *minor_to_device(unsigned int minor) |
b411b363 | 780 | { |
05a10ec7 | 781 | return (struct drbd_device *)idr_find(&drbd_devices, minor); |
b411b363 PR |
782 | } |
783 | ||
a6b32bc3 AG |
784 | static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device) |
785 | { | |
786 | return list_first_entry(&device->peer_devices, struct drbd_peer_device, peer_devices); | |
787 | } | |
788 | ||
77c556f6 AG |
789 | #define for_each_resource(resource, _resources) \ |
790 | list_for_each_entry(resource, _resources, resources) | |
791 | ||
792 | #define for_each_resource_rcu(resource, _resources) \ | |
793 | list_for_each_entry_rcu(resource, _resources, resources) | |
794 | ||
795 | #define for_each_resource_safe(resource, tmp, _resources) \ | |
796 | list_for_each_entry_safe(resource, tmp, _resources, resources) | |
797 | ||
798 | #define for_each_connection(connection, resource) \ | |
799 | list_for_each_entry(connection, &resource->connections, connections) | |
800 | ||
801 | #define for_each_connection_rcu(connection, resource) \ | |
802 | list_for_each_entry_rcu(connection, &resource->connections, connections) | |
803 | ||
804 | #define for_each_connection_safe(connection, tmp, resource) \ | |
805 | list_for_each_entry_safe(connection, tmp, &resource->connections, connections) | |
806 | ||
a6b32bc3 AG |
807 | #define for_each_peer_device(peer_device, device) \ |
808 | list_for_each_entry(peer_device, &device->peer_devices, peer_devices) | |
809 | ||
810 | #define for_each_peer_device_rcu(peer_device, device) \ | |
811 | list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices) | |
812 | ||
813 | #define for_each_peer_device_safe(peer_device, tmp, device) \ | |
814 | list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices) | |
815 | ||
b30ab791 | 816 | static inline unsigned int device_to_minor(struct drbd_device *device) |
b411b363 | 817 | { |
b30ab791 | 818 | return device->minor; |
b411b363 PR |
819 | } |
820 | ||
bde89a9e | 821 | static inline struct drbd_device *vnr_to_device(struct drbd_connection *connection, int vnr) |
b411b363 | 822 | { |
bde89a9e | 823 | return (struct drbd_device *)idr_find(&connection->volumes, vnr); |
b411b363 PR |
824 | } |
825 | ||
826 | /* | |
827 | * function declarations | |
828 | *************************/ | |
829 | ||
830 | /* drbd_main.c */ | |
831 | ||
e89b591c PR |
832 | enum dds_flags { |
833 | DDSF_FORCED = 1, | |
834 | DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */ | |
835 | }; | |
836 | ||
b30ab791 | 837 | extern void drbd_init_set_defaults(struct drbd_device *device); |
b411b363 PR |
838 | extern int drbd_thread_start(struct drbd_thread *thi); |
839 | extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait); | |
bde89a9e | 840 | extern char *drbd_task_to_thread_name(struct drbd_connection *connection, struct task_struct *task); |
b411b363 | 841 | #ifdef CONFIG_SMP |
80822284 | 842 | extern void drbd_thread_current_set_cpu(struct drbd_thread *thi); |
bde89a9e | 843 | extern void drbd_calc_cpu_mask(struct drbd_connection *connection); |
b411b363 PR |
844 | #else |
845 | #define drbd_thread_current_set_cpu(A) ({}) | |
846 | #define drbd_calc_cpu_mask(A) ({}) | |
847 | #endif | |
bde89a9e | 848 | extern void tl_release(struct drbd_connection *, unsigned int barrier_nr, |
b411b363 | 849 | unsigned int set_size); |
bde89a9e AG |
850 | extern void tl_clear(struct drbd_connection *); |
851 | extern void drbd_free_sock(struct drbd_connection *connection); | |
852 | extern int drbd_send(struct drbd_connection *connection, struct socket *sock, | |
bedbd2a5 | 853 | void *buf, size_t size, unsigned msg_flags); |
bde89a9e | 854 | extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t, |
fb708e40 AG |
855 | unsigned); |
856 | ||
bde89a9e AG |
857 | extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd); |
858 | extern int drbd_send_protocol(struct drbd_connection *connection); | |
b30ab791 AG |
859 | extern int drbd_send_uuids(struct drbd_device *device); |
860 | extern int drbd_send_uuids_skip_initial_sync(struct drbd_device *device); | |
861 | extern void drbd_gen_and_send_sync_uuid(struct drbd_device *device); | |
862 | extern int drbd_send_sizes(struct drbd_device *device, int trigger_reply, enum dds_flags flags); | |
863 | extern int drbd_send_state(struct drbd_device *device, union drbd_state s); | |
864 | extern int drbd_send_current_state(struct drbd_device *device); | |
865 | extern int drbd_send_sync_param(struct drbd_device *device); | |
bde89a9e | 866 | extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr, |
d4e67d7c | 867 | u32 set_size); |
54761697 | 868 | extern int drbd_send_ack(struct drbd_device *, enum drbd_packet, |
f6ffca9f | 869 | struct drbd_peer_request *); |
b30ab791 | 870 | extern void drbd_send_ack_rp(struct drbd_device *device, enum drbd_packet cmd, |
a9a9994d | 871 | struct p_block_req *rp); |
b30ab791 | 872 | extern void drbd_send_ack_dp(struct drbd_device *device, enum drbd_packet cmd, |
a9a9994d | 873 | struct p_data *dp, int data_size); |
b30ab791 | 874 | extern int drbd_send_ack_ex(struct drbd_device *device, enum drbd_packet cmd, |
b411b363 | 875 | sector_t sector, int blksize, u64 block_id); |
54761697 AG |
876 | extern int drbd_send_out_of_sync(struct drbd_device *, struct drbd_request *); |
877 | extern int drbd_send_block(struct drbd_device *, enum drbd_packet, | |
f6ffca9f | 878 | struct drbd_peer_request *); |
b30ab791 AG |
879 | extern int drbd_send_dblock(struct drbd_device *device, struct drbd_request *req); |
880 | extern int drbd_send_drequest(struct drbd_device *device, int cmd, | |
b411b363 | 881 | sector_t sector, int size, u64 block_id); |
b30ab791 | 882 | extern int drbd_send_drequest_csum(struct drbd_device *device, sector_t sector, |
d8763023 AG |
883 | int size, void *digest, int digest_size, |
884 | enum drbd_packet cmd); | |
b30ab791 | 885 | extern int drbd_send_ov_request(struct drbd_device *device, sector_t sector, int size); |
b411b363 | 886 | |
b30ab791 AG |
887 | extern int drbd_send_bitmap(struct drbd_device *device); |
888 | extern void drbd_send_sr_reply(struct drbd_device *device, enum drbd_state_rv retcode); | |
bde89a9e | 889 | extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode); |
b411b363 | 890 | extern void drbd_free_bc(struct drbd_backing_dev *ldev); |
b30ab791 AG |
891 | extern void drbd_device_cleanup(struct drbd_device *device); |
892 | void drbd_print_uuids(struct drbd_device *device, const char *text); | |
b411b363 | 893 | |
bde89a9e | 894 | extern void conn_md_sync(struct drbd_connection *connection); |
b30ab791 AG |
895 | extern void drbd_md_write(struct drbd_device *device, void *buffer); |
896 | extern void drbd_md_sync(struct drbd_device *device); | |
897 | extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev); | |
898 | extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); | |
899 | extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); | |
900 | extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local); | |
901 | extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local); | |
902 | extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local); | |
903 | extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); | |
904 | extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local); | |
905 | extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local); | |
b411b363 | 906 | extern int drbd_md_test_flag(struct drbd_backing_dev *, int); |
ee15b038 | 907 | #ifndef DRBD_DEBUG_MD_SYNC |
b30ab791 | 908 | extern void drbd_md_mark_dirty(struct drbd_device *device); |
ee15b038 LE |
909 | #else |
910 | #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ ) | |
b30ab791 | 911 | extern void drbd_md_mark_dirty_(struct drbd_device *device, |
ee15b038 LE |
912 | unsigned int line, const char *func); |
913 | #endif | |
b30ab791 | 914 | extern void drbd_queue_bitmap_io(struct drbd_device *device, |
54761697 AG |
915 | int (*io_fn)(struct drbd_device *), |
916 | void (*done)(struct drbd_device *, int), | |
20ceb2b2 | 917 | char *why, enum bm_flag flags); |
b30ab791 | 918 | extern int drbd_bitmap_io(struct drbd_device *device, |
54761697 | 919 | int (*io_fn)(struct drbd_device *), |
20ceb2b2 | 920 | char *why, enum bm_flag flags); |
b30ab791 | 921 | extern int drbd_bitmap_io_from_worker(struct drbd_device *device, |
54761697 | 922 | int (*io_fn)(struct drbd_device *), |
edc9f5eb | 923 | char *why, enum bm_flag flags); |
b30ab791 AG |
924 | extern int drbd_bmio_set_n_write(struct drbd_device *device); |
925 | extern int drbd_bmio_clear_n_write(struct drbd_device *device); | |
926 | extern void drbd_ldev_destroy(struct drbd_device *device); | |
b411b363 | 927 | |
b411b363 | 928 | /* Meta data layout |
ae8bf312 LE |
929 | * |
930 | * We currently have two possible layouts. | |
931 | * Offsets in (512 byte) sectors. | |
932 | * external: | |
933 | * |----------- md_size_sect ------------------| | |
934 | * [ 4k superblock ][ activity log ][ Bitmap ] | |
935 | * | al_offset == 8 | | |
936 | * | bm_offset = al_offset + X | | |
937 | * ==> bitmap sectors = md_size_sect - bm_offset | |
938 | * | |
939 | * Variants: | |
940 | * old, indexed fixed size meta data: | |
941 | * | |
942 | * internal: | |
943 | * |----------- md_size_sect ------------------| | |
944 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*] | |
945 | * | al_offset < 0 | | |
946 | * | bm_offset = al_offset - Y | | |
947 | * ==> bitmap sectors = Y = al_offset - bm_offset | |
948 | * | |
949 | * [padding*] are zero or up to 7 unused 512 Byte sectors to the | |
950 | * end of the device, so that the [4k superblock] will be 4k aligned. | |
951 | * | |
952 | * The activity log consists of 4k transaction blocks, | |
953 | * which are written in a ring-buffer, or striped ring-buffer like fashion, | |
954 | * which are writtensize used to be fixed 32kB, | |
955 | * but is about to become configurable. | |
956 | */ | |
b411b363 | 957 | |
ae8bf312 LE |
958 | /* Our old fixed size meta data layout |
959 | * allows up to about 3.8TB, so if you want more, | |
7ad651b5 | 960 | * you need to use the "flexible" meta data format. */ |
ae8bf312 LE |
961 | #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */ |
962 | #define MD_4kB_SECT 8 | |
963 | #define MD_32kB_SECT 64 | |
7ad651b5 LE |
964 | |
965 | /* One activity log extent represents 4M of storage */ | |
966 | #define AL_EXTENT_SHIFT 22 | |
b411b363 PR |
967 | #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT) |
968 | ||
7ad651b5 LE |
969 | /* We could make these currently hardcoded constants configurable |
970 | * variables at create-md time (or even re-configurable at runtime?). | |
971 | * Which will require some more changes to the DRBD "super block" | |
972 | * and attach code. | |
973 | * | |
974 | * updates per transaction: | |
975 | * This many changes to the active set can be logged with one transaction. | |
976 | * This number is arbitrary. | |
977 | * context per transaction: | |
978 | * This many context extent numbers are logged with each transaction. | |
979 | * This number is resulting from the transaction block size (4k), the layout | |
980 | * of the transaction header, and the number of updates per transaction. | |
981 | * See drbd_actlog.c:struct al_transaction_on_disk | |
982 | * */ | |
983 | #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary | |
984 | #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4 | |
985 | ||
b411b363 PR |
986 | #if BITS_PER_LONG == 32 |
987 | #define LN2_BPL 5 | |
988 | #define cpu_to_lel(A) cpu_to_le32(A) | |
989 | #define lel_to_cpu(A) le32_to_cpu(A) | |
990 | #elif BITS_PER_LONG == 64 | |
991 | #define LN2_BPL 6 | |
992 | #define cpu_to_lel(A) cpu_to_le64(A) | |
993 | #define lel_to_cpu(A) le64_to_cpu(A) | |
994 | #else | |
995 | #error "LN2 of BITS_PER_LONG unknown!" | |
996 | #endif | |
997 | ||
998 | /* resync bitmap */ | |
999 | /* 16MB sized 'bitmap extent' to track syncer usage */ | |
1000 | struct bm_extent { | |
1001 | int rs_left; /* number of bits set (out of sync) in this extent. */ | |
1002 | int rs_failed; /* number of failed resync requests in this extent. */ | |
1003 | unsigned long flags; | |
1004 | struct lc_element lce; | |
1005 | }; | |
1006 | ||
1007 | #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */ | |
1008 | #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */ | |
e3555d85 | 1009 | #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */ |
b411b363 PR |
1010 | |
1011 | /* drbd_bitmap.c */ | |
1012 | /* | |
1013 | * We need to store one bit for a block. | |
1014 | * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap. | |
1015 | * Bit 0 ==> local node thinks this block is binary identical on both nodes | |
1016 | * Bit 1 ==> local node thinks this block needs to be synced. | |
1017 | */ | |
1018 | ||
8e26f9cc PR |
1019 | #define SLEEP_TIME (HZ/10) |
1020 | ||
45dfffeb LE |
1021 | /* We do bitmap IO in units of 4k blocks. |
1022 | * We also still have a hardcoded 4k per bit relation. */ | |
1023 | #define BM_BLOCK_SHIFT 12 /* 4k per bit */ | |
b411b363 | 1024 | #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT) |
45dfffeb LE |
1025 | /* mostly arbitrarily set the represented size of one bitmap extent, |
1026 | * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap | |
1027 | * at 4k per bit resolution) */ | |
1028 | #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */ | |
b411b363 PR |
1029 | #define BM_EXT_SIZE (1<<BM_EXT_SHIFT) |
1030 | ||
1031 | #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12) | |
1032 | #error "HAVE YOU FIXED drbdmeta AS WELL??" | |
1033 | #endif | |
1034 | ||
1035 | /* thus many _storage_ sectors are described by one bit */ | |
1036 | #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9)) | |
1037 | #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9)) | |
1038 | #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1) | |
1039 | ||
1040 | /* bit to represented kilo byte conversion */ | |
1041 | #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10)) | |
1042 | ||
1043 | /* in which _bitmap_ extent (resp. sector) the bit for a certain | |
1044 | * _storage_ sector is located in */ | |
1045 | #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9)) | |
1046 | ||
1047 | /* how much _storage_ sectors we have per bitmap sector */ | |
1048 | #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9)) | |
1049 | #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1) | |
1050 | ||
1051 | /* in one sector of the bitmap, we have this many activity_log extents. */ | |
1052 | #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT)) | |
b411b363 PR |
1053 | |
1054 | #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT) | |
1055 | #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1) | |
1056 | ||
1057 | /* the extent in "PER_EXTENT" below is an activity log extent | |
1058 | * we need that many (long words/bytes) to store the bitmap | |
1059 | * of one AL_EXTENT_SIZE chunk of storage. | |
1060 | * we can store the bitmap for that many AL_EXTENTS within | |
1061 | * one sector of the _on_disk_ bitmap: | |
1062 | * bit 0 bit 37 bit 38 bit (512*8)-1 | |
1063 | * ...|........|........|.. // ..|........| | |
1064 | * sect. 0 `296 `304 ^(512*8*8)-1 | |
1065 | * | |
1066 | #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG ) | |
1067 | #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128 | |
1068 | #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4 | |
1069 | */ | |
1070 | ||
1071 | #define DRBD_MAX_SECTORS_32 (0xffffffffLU) | |
ae8bf312 LE |
1072 | /* we have a certain meta data variant that has a fixed on-disk size of 128 |
1073 | * MiB, of which 4k are our "superblock", and 32k are the fixed size activity | |
1074 | * log, leaving this many sectors for the bitmap. | |
1075 | */ | |
1076 | ||
1077 | #define DRBD_MAX_SECTORS_FIXED_BM \ | |
1078 | ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9))) | |
1079 | #if !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32 | |
b411b363 PR |
1080 | #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32 |
1081 | #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32 | |
1082 | #else | |
ae8bf312 | 1083 | #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM |
b411b363 PR |
1084 | /* 16 TB in units of sectors */ |
1085 | #if BITS_PER_LONG == 32 | |
1086 | /* adjust by one page worth of bitmap, | |
1087 | * so we won't wrap around in drbd_bm_find_next_bit. | |
1088 | * you should use 64bit OS for that much storage, anyways. */ | |
1089 | #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff) | |
1090 | #else | |
4b0715f0 LE |
1091 | /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */ |
1092 | #define DRBD_MAX_SECTORS_FLEX (1UL << 51) | |
1093 | /* corresponds to (1UL << 38) bits right now. */ | |
b411b363 PR |
1094 | #endif |
1095 | #endif | |
1096 | ||
23361cf3 LE |
1097 | /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE, |
1098 | * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte. | |
1099 | * Since we may live in a mixed-platform cluster, | |
1100 | * we limit us to a platform agnostic constant here for now. | |
1101 | * A followup commit may allow even bigger BIO sizes, | |
1102 | * once we thought that through. */ | |
98683650 | 1103 | #define DRBD_MAX_BIO_SIZE (1U << 20) |
23361cf3 LE |
1104 | #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE |
1105 | #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE | |
1106 | #endif | |
db141b2f | 1107 | #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */ |
b411b363 | 1108 | |
98683650 PR |
1109 | #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */ |
1110 | #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */ | |
b411b363 | 1111 | |
b30ab791 AG |
1112 | extern int drbd_bm_init(struct drbd_device *device); |
1113 | extern int drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits); | |
1114 | extern void drbd_bm_cleanup(struct drbd_device *device); | |
1115 | extern void drbd_bm_set_all(struct drbd_device *device); | |
1116 | extern void drbd_bm_clear_all(struct drbd_device *device); | |
4b0715f0 | 1117 | /* set/clear/test only a few bits at a time */ |
b411b363 | 1118 | extern int drbd_bm_set_bits( |
b30ab791 | 1119 | struct drbd_device *device, unsigned long s, unsigned long e); |
b411b363 | 1120 | extern int drbd_bm_clear_bits( |
b30ab791 | 1121 | struct drbd_device *device, unsigned long s, unsigned long e); |
4b0715f0 | 1122 | extern int drbd_bm_count_bits( |
b30ab791 | 1123 | struct drbd_device *device, const unsigned long s, const unsigned long e); |
4b0715f0 LE |
1124 | /* bm_set_bits variant for use while holding drbd_bm_lock, |
1125 | * may process the whole bitmap in one go */ | |
b30ab791 | 1126 | extern void _drbd_bm_set_bits(struct drbd_device *device, |
b411b363 | 1127 | const unsigned long s, const unsigned long e); |
b30ab791 AG |
1128 | extern int drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr); |
1129 | extern int drbd_bm_e_weight(struct drbd_device *device, unsigned long enr); | |
1130 | extern int drbd_bm_write_page(struct drbd_device *device, unsigned int idx) __must_hold(local); | |
1131 | extern int drbd_bm_read(struct drbd_device *device) __must_hold(local); | |
1132 | extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr); | |
1133 | extern int drbd_bm_write(struct drbd_device *device) __must_hold(local); | |
1134 | extern int drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local); | |
1135 | extern int drbd_bm_write_all(struct drbd_device *device) __must_hold(local); | |
1136 | extern int drbd_bm_write_copy_pages(struct drbd_device *device) __must_hold(local); | |
1137 | extern size_t drbd_bm_words(struct drbd_device *device); | |
1138 | extern unsigned long drbd_bm_bits(struct drbd_device *device); | |
1139 | extern sector_t drbd_bm_capacity(struct drbd_device *device); | |
4b0715f0 LE |
1140 | |
1141 | #define DRBD_END_OF_BITMAP (~(unsigned long)0) | |
b30ab791 | 1142 | extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); |
b411b363 | 1143 | /* bm_find_next variants for use while you hold drbd_bm_lock() */ |
b30ab791 AG |
1144 | extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); |
1145 | extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo); | |
1146 | extern unsigned long _drbd_bm_total_weight(struct drbd_device *device); | |
1147 | extern unsigned long drbd_bm_total_weight(struct drbd_device *device); | |
1148 | extern int drbd_bm_rs_done(struct drbd_device *device); | |
b411b363 | 1149 | /* for receive_bitmap */ |
b30ab791 | 1150 | extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset, |
b411b363 | 1151 | size_t number, unsigned long *buffer); |
19f843aa | 1152 | /* for _drbd_send_bitmap */ |
b30ab791 | 1153 | extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset, |
b411b363 PR |
1154 | size_t number, unsigned long *buffer); |
1155 | ||
b30ab791 AG |
1156 | extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags); |
1157 | extern void drbd_bm_unlock(struct drbd_device *device); | |
b411b363 PR |
1158 | /* drbd_main.c */ |
1159 | ||
1160 | extern struct kmem_cache *drbd_request_cache; | |
6c852bec | 1161 | extern struct kmem_cache *drbd_ee_cache; /* peer requests */ |
b411b363 PR |
1162 | extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */ |
1163 | extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */ | |
1164 | extern mempool_t *drbd_request_mempool; | |
1165 | extern mempool_t *drbd_ee_mempool; | |
1166 | ||
4281808f LE |
1167 | /* drbd's page pool, used to buffer data received from the peer, |
1168 | * or data requested by the peer. | |
1169 | * | |
1170 | * This does not have an emergency reserve. | |
1171 | * | |
1172 | * When allocating from this pool, it first takes pages from the pool. | |
1173 | * Only if the pool is depleted will try to allocate from the system. | |
1174 | * | |
1175 | * The assumption is that pages taken from this pool will be processed, | |
1176 | * and given back, "quickly", and then can be recycled, so we can avoid | |
1177 | * frequent calls to alloc_page(), and still will be able to make progress even | |
1178 | * under memory pressure. | |
1179 | */ | |
1180 | extern struct page *drbd_pp_pool; | |
b411b363 PR |
1181 | extern spinlock_t drbd_pp_lock; |
1182 | extern int drbd_pp_vacant; | |
1183 | extern wait_queue_head_t drbd_pp_wait; | |
1184 | ||
4281808f LE |
1185 | /* We also need a standard (emergency-reserve backed) page pool |
1186 | * for meta data IO (activity log, bitmap). | |
1187 | * We can keep it global, as long as it is used as "N pages at a time". | |
1188 | * 128 should be plenty, currently we probably can get away with as few as 1. | |
1189 | */ | |
1190 | #define DRBD_MIN_POOL_PAGES 128 | |
1191 | extern mempool_t *drbd_md_io_page_pool; | |
1192 | ||
9476f39d LE |
1193 | /* We also need to make sure we get a bio |
1194 | * when we need it for housekeeping purposes */ | |
1195 | extern struct bio_set *drbd_md_io_bio_set; | |
1196 | /* to allocate from that set */ | |
1197 | extern struct bio *bio_alloc_drbd(gfp_t gfp_mask); | |
1198 | ||
b411b363 PR |
1199 | extern rwlock_t global_state_lock; |
1200 | ||
bde89a9e | 1201 | extern int conn_lowest_minor(struct drbd_connection *connection); |
a6b32bc3 | 1202 | enum drbd_ret_code drbd_create_minor(struct drbd_connection *connection, unsigned int minor, int vnr); |
05a10ec7 | 1203 | extern void drbd_destroy_device(struct kref *kref); |
b411b363 | 1204 | |
77c556f6 AG |
1205 | extern struct drbd_resource *drbd_create_resource(const char *name); |
1206 | extern void drbd_free_resource(struct drbd_resource *resource); | |
1207 | ||
bde89a9e AG |
1208 | extern int set_resource_options(struct drbd_connection *connection, struct res_opts *res_opts); |
1209 | extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts); | |
05a10ec7 | 1210 | extern void drbd_destroy_connection(struct kref *kref); |
bde89a9e AG |
1211 | struct drbd_connection *conn_get_by_name(const char *name); |
1212 | extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len, | |
089c075d | 1213 | void *peer_addr, int peer_addr_len); |
77c556f6 | 1214 | extern void drbd_destroy_resource(struct kref *kref); |
bde89a9e | 1215 | extern void conn_free_crypto(struct drbd_connection *connection); |
b411b363 PR |
1216 | |
1217 | extern int proc_details; | |
1218 | ||
1219 | /* drbd_req */ | |
113fef9e | 1220 | extern void do_submit(struct work_struct *ws); |
54761697 | 1221 | extern void __drbd_make_request(struct drbd_device *, struct bio *, unsigned long); |
5a7bbad2 | 1222 | extern void drbd_make_request(struct request_queue *q, struct bio *bio); |
b30ab791 | 1223 | extern int drbd_read_remote(struct drbd_device *device, struct drbd_request *req); |
b411b363 PR |
1224 | extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec); |
1225 | extern int is_valid_ar_handle(struct drbd_request *, sector_t); | |
1226 | ||
1227 | ||
1228 | /* drbd_nl.c */ | |
8432b314 | 1229 | extern int drbd_msg_put_info(const char *info); |
b30ab791 AG |
1230 | extern void drbd_suspend_io(struct drbd_device *device); |
1231 | extern void drbd_resume_io(struct drbd_device *device); | |
b411b363 | 1232 | extern char *ppsize(char *buf, unsigned long long size); |
54761697 | 1233 | extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int); |
e96c9633 | 1234 | enum determine_dev_size { |
d752b269 PR |
1235 | DS_ERROR_SHRINK = -3, |
1236 | DS_ERROR_SPACE_MD = -2, | |
e96c9633 PR |
1237 | DS_ERROR = -1, |
1238 | DS_UNCHANGED = 0, | |
1239 | DS_SHRUNK = 1, | |
57737adc PR |
1240 | DS_GREW = 2, |
1241 | DS_GREW_FROM_ZERO = 3, | |
e96c9633 | 1242 | }; |
d752b269 | 1243 | extern enum determine_dev_size |
54761697 AG |
1244 | drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local); |
1245 | extern void resync_after_online_grow(struct drbd_device *); | |
b30ab791 AG |
1246 | extern void drbd_reconsider_max_bio_size(struct drbd_device *device); |
1247 | extern enum drbd_state_rv drbd_set_role(struct drbd_device *device, | |
bf885f8a AG |
1248 | enum drbd_role new_role, |
1249 | int force); | |
bde89a9e AG |
1250 | extern bool conn_try_outdate_peer(struct drbd_connection *connection); |
1251 | extern void conn_try_outdate_peer_async(struct drbd_connection *connection); | |
b30ab791 | 1252 | extern int drbd_khelper(struct drbd_device *device, char *cmd); |
b411b363 PR |
1253 | |
1254 | /* drbd_worker.c */ | |
1255 | extern int drbd_worker(struct drbd_thread *thi); | |
b30ab791 AG |
1256 | enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor); |
1257 | void drbd_resync_after_changed(struct drbd_device *device); | |
1258 | extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side); | |
1259 | extern void resume_next_sg(struct drbd_device *device); | |
1260 | extern void suspend_other_sg(struct drbd_device *device); | |
1261 | extern int drbd_resync_finished(struct drbd_device *device); | |
b411b363 | 1262 | /* maybe rather drbd_main.c ? */ |
b30ab791 AG |
1263 | extern void *drbd_md_get_buffer(struct drbd_device *device); |
1264 | extern void drbd_md_put_buffer(struct drbd_device *device); | |
1265 | extern int drbd_md_sync_page_io(struct drbd_device *device, | |
b411b363 | 1266 | struct drbd_backing_dev *bdev, sector_t sector, int rw); |
54761697 | 1267 | extern void drbd_ov_out_of_sync_found(struct drbd_device *, sector_t, int); |
b30ab791 | 1268 | extern void wait_until_done_or_force_detached(struct drbd_device *device, |
44edfb0d | 1269 | struct drbd_backing_dev *bdev, unsigned int *done); |
b30ab791 | 1270 | extern void drbd_rs_controller_reset(struct drbd_device *device); |
b411b363 | 1271 | |
b30ab791 | 1272 | static inline void ov_out_of_sync_print(struct drbd_device *device) |
b411b363 | 1273 | { |
b30ab791 | 1274 | if (device->ov_last_oos_size) { |
b411b363 | 1275 | dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n", |
b30ab791 AG |
1276 | (unsigned long long)device->ov_last_oos_start, |
1277 | (unsigned long)device->ov_last_oos_size); | |
b411b363 | 1278 | } |
b30ab791 | 1279 | device->ov_last_oos_size = 0; |
b411b363 PR |
1280 | } |
1281 | ||
1282 | ||
54761697 AG |
1283 | extern void drbd_csum_bio(struct drbd_device *, struct crypto_hash *, struct bio *, void *); |
1284 | extern void drbd_csum_ee(struct drbd_device *, struct crypto_hash *, | |
f6ffca9f | 1285 | struct drbd_peer_request *, void *); |
b411b363 | 1286 | /* worker callbacks */ |
99920dc5 AG |
1287 | extern int w_e_end_data_req(struct drbd_work *, int); |
1288 | extern int w_e_end_rsdata_req(struct drbd_work *, int); | |
1289 | extern int w_e_end_csum_rs_req(struct drbd_work *, int); | |
1290 | extern int w_e_end_ov_reply(struct drbd_work *, int); | |
1291 | extern int w_e_end_ov_req(struct drbd_work *, int); | |
1292 | extern int w_ov_finished(struct drbd_work *, int); | |
1293 | extern int w_resync_timer(struct drbd_work *, int); | |
1294 | extern int w_send_write_hint(struct drbd_work *, int); | |
1295 | extern int w_make_resync_request(struct drbd_work *, int); | |
1296 | extern int w_send_dblock(struct drbd_work *, int); | |
99920dc5 AG |
1297 | extern int w_send_read_req(struct drbd_work *, int); |
1298 | extern int w_prev_work_done(struct drbd_work *, int); | |
1299 | extern int w_e_reissue(struct drbd_work *, int); | |
1300 | extern int w_restart_disk_io(struct drbd_work *, int); | |
8f7bed77 | 1301 | extern int w_send_out_of_sync(struct drbd_work *, int); |
99920dc5 | 1302 | extern int w_start_resync(struct drbd_work *, int); |
b411b363 PR |
1303 | |
1304 | extern void resync_timer_fn(unsigned long data); | |
370a43e7 | 1305 | extern void start_resync_timer_fn(unsigned long data); |
b411b363 PR |
1306 | |
1307 | /* drbd_receiver.c */ | |
b30ab791 | 1308 | extern int drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector); |
54761697 | 1309 | extern int drbd_submit_peer_request(struct drbd_device *, |
fbe29dec AG |
1310 | struct drbd_peer_request *, const unsigned, |
1311 | const int); | |
54761697 AG |
1312 | extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *); |
1313 | extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_device *, u64, | |
0db55363 AG |
1314 | sector_t, unsigned int, |
1315 | gfp_t) __must_hold(local); | |
54761697 | 1316 | extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *, |
3967deb1 AG |
1317 | int); |
1318 | #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0) | |
1319 | #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1) | |
54761697 | 1320 | extern struct page *drbd_alloc_pages(struct drbd_device *, unsigned int, bool); |
b30ab791 AG |
1321 | extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled); |
1322 | extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed); | |
bde89a9e | 1323 | extern void conn_flush_workqueue(struct drbd_connection *connection); |
b30ab791 AG |
1324 | extern int drbd_connected(struct drbd_device *device); |
1325 | static inline void drbd_flush_workqueue(struct drbd_device *device) | |
0e29d163 | 1326 | { |
a6b32bc3 | 1327 | conn_flush_workqueue(first_peer_device(device)->connection); |
0e29d163 | 1328 | } |
b411b363 | 1329 | |
ed439848 LE |
1330 | /* Yes, there is kernel_setsockopt, but only since 2.6.18. |
1331 | * So we have our own copy of it here. */ | |
b411b363 | 1332 | static inline int drbd_setsockopt(struct socket *sock, int level, int optname, |
ed439848 | 1333 | char *optval, int optlen) |
b411b363 | 1334 | { |
ed439848 LE |
1335 | mm_segment_t oldfs = get_fs(); |
1336 | char __user *uoptval; | |
b411b363 | 1337 | int err; |
ed439848 LE |
1338 | |
1339 | uoptval = (char __user __force *)optval; | |
1340 | ||
1341 | set_fs(KERNEL_DS); | |
b411b363 | 1342 | if (level == SOL_SOCKET) |
ed439848 | 1343 | err = sock_setsockopt(sock, level, optname, uoptval, optlen); |
b411b363 | 1344 | else |
ed439848 | 1345 | err = sock->ops->setsockopt(sock, level, optname, uoptval, |
b411b363 | 1346 | optlen); |
ed439848 | 1347 | set_fs(oldfs); |
b411b363 PR |
1348 | return err; |
1349 | } | |
1350 | ||
1351 | static inline void drbd_tcp_cork(struct socket *sock) | |
1352 | { | |
ed439848 | 1353 | int val = 1; |
b411b363 | 1354 | (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK, |
ed439848 | 1355 | (char*)&val, sizeof(val)); |
b411b363 PR |
1356 | } |
1357 | ||
1358 | static inline void drbd_tcp_uncork(struct socket *sock) | |
1359 | { | |
ed439848 | 1360 | int val = 0; |
b411b363 | 1361 | (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK, |
ed439848 | 1362 | (char*)&val, sizeof(val)); |
b411b363 PR |
1363 | } |
1364 | ||
1365 | static inline void drbd_tcp_nodelay(struct socket *sock) | |
1366 | { | |
ed439848 | 1367 | int val = 1; |
b411b363 | 1368 | (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY, |
ed439848 | 1369 | (char*)&val, sizeof(val)); |
b411b363 PR |
1370 | } |
1371 | ||
1372 | static inline void drbd_tcp_quickack(struct socket *sock) | |
1373 | { | |
ed439848 | 1374 | int val = 2; |
b411b363 | 1375 | (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK, |
ed439848 | 1376 | (char*)&val, sizeof(val)); |
b411b363 PR |
1377 | } |
1378 | ||
bde89a9e | 1379 | void drbd_bump_write_ordering(struct drbd_connection *connection, enum write_ordering_e wo); |
b411b363 PR |
1380 | |
1381 | /* drbd_proc.c */ | |
1382 | extern struct proc_dir_entry *drbd_proc; | |
7d4e9d09 | 1383 | extern const struct file_operations drbd_proc_fops; |
b411b363 PR |
1384 | extern const char *drbd_conn_str(enum drbd_conns s); |
1385 | extern const char *drbd_role_str(enum drbd_role s); | |
1386 | ||
1387 | /* drbd_actlog.c */ | |
b30ab791 AG |
1388 | extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i); |
1389 | extern void drbd_al_begin_io_commit(struct drbd_device *device, bool delegate); | |
1390 | extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i); | |
1391 | extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i, bool delegate); | |
1392 | extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i); | |
1393 | extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector); | |
1394 | extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector); | |
1395 | extern int drbd_try_rs_begin_io(struct drbd_device *device, sector_t sector); | |
1396 | extern void drbd_rs_cancel_all(struct drbd_device *device); | |
1397 | extern int drbd_rs_del_all(struct drbd_device *device); | |
1398 | extern void drbd_rs_failed_io(struct drbd_device *device, | |
b411b363 | 1399 | sector_t sector, int size); |
b30ab791 AG |
1400 | extern void drbd_advance_rs_marks(struct drbd_device *device, unsigned long still_to_go); |
1401 | extern void __drbd_set_in_sync(struct drbd_device *device, sector_t sector, | |
b411b363 | 1402 | int size, const char *file, const unsigned int line); |
b30ab791 AG |
1403 | #define drbd_set_in_sync(device, sector, size) \ |
1404 | __drbd_set_in_sync(device, sector, size, __FILE__, __LINE__) | |
1405 | extern int __drbd_set_out_of_sync(struct drbd_device *device, sector_t sector, | |
b411b363 | 1406 | int size, const char *file, const unsigned int line); |
b30ab791 AG |
1407 | #define drbd_set_out_of_sync(device, sector, size) \ |
1408 | __drbd_set_out_of_sync(device, sector, size, __FILE__, __LINE__) | |
1409 | extern void drbd_al_shrink(struct drbd_device *device); | |
54761697 | 1410 | extern int drbd_initialize_al(struct drbd_device *, void *); |
b411b363 | 1411 | |
b411b363 | 1412 | /* drbd_nl.c */ |
3b98c0c2 LE |
1413 | /* state info broadcast */ |
1414 | struct sib_info { | |
1415 | enum drbd_state_info_bcast_reason sib_reason; | |
1416 | union { | |
1417 | struct { | |
1418 | char *helper_name; | |
1419 | unsigned helper_exit_code; | |
1420 | }; | |
1421 | struct { | |
1422 | union drbd_state os; | |
1423 | union drbd_state ns; | |
1424 | }; | |
1425 | }; | |
1426 | }; | |
b30ab791 | 1427 | void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib); |
b411b363 PR |
1428 | |
1429 | /* | |
1430 | * inline helper functions | |
1431 | *************************/ | |
1432 | ||
45bb912b LE |
1433 | /* see also page_chain_add and friends in drbd_receiver.c */ |
1434 | static inline struct page *page_chain_next(struct page *page) | |
1435 | { | |
1436 | return (struct page *)page_private(page); | |
1437 | } | |
1438 | #define page_chain_for_each(page) \ | |
1439 | for (; page && ({ prefetch(page_chain_next(page)); 1; }); \ | |
1440 | page = page_chain_next(page)) | |
1441 | #define page_chain_for_each_safe(page, n) \ | |
1442 | for (; page && ({ n = page_chain_next(page); 1; }); page = n) | |
1443 | ||
45bb912b | 1444 | |
045417f7 | 1445 | static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req) |
45bb912b | 1446 | { |
db830c46 | 1447 | struct page *page = peer_req->pages; |
45bb912b LE |
1448 | page_chain_for_each(page) { |
1449 | if (page_count(page) > 1) | |
1450 | return 1; | |
1451 | } | |
1452 | return 0; | |
1453 | } | |
1454 | ||
bf885f8a | 1455 | static inline enum drbd_state_rv |
b30ab791 | 1456 | _drbd_set_state(struct drbd_device *device, union drbd_state ns, |
bf885f8a | 1457 | enum chg_state_flags flags, struct completion *done) |
b411b363 | 1458 | { |
bf885f8a | 1459 | enum drbd_state_rv rv; |
b411b363 PR |
1460 | |
1461 | read_lock(&global_state_lock); | |
b30ab791 | 1462 | rv = __drbd_set_state(device, ns, flags, done); |
b411b363 PR |
1463 | read_unlock(&global_state_lock); |
1464 | ||
1465 | return rv; | |
1466 | } | |
1467 | ||
b30ab791 | 1468 | static inline union drbd_state drbd_read_state(struct drbd_device *device) |
b411b363 | 1469 | { |
78bae59b PR |
1470 | union drbd_state rv; |
1471 | ||
b30ab791 | 1472 | rv.i = device->state.i; |
a6b32bc3 AG |
1473 | rv.susp = first_peer_device(device)->connection->susp; |
1474 | rv.susp_nod = first_peer_device(device)->connection->susp_nod; | |
1475 | rv.susp_fen = first_peer_device(device)->connection->susp_fen; | |
78bae59b PR |
1476 | |
1477 | return rv; | |
b411b363 PR |
1478 | } |
1479 | ||
383606e0 | 1480 | enum drbd_force_detach_flags { |
a2a3c74f LE |
1481 | DRBD_READ_ERROR, |
1482 | DRBD_WRITE_ERROR, | |
383606e0 LE |
1483 | DRBD_META_IO_ERROR, |
1484 | DRBD_FORCE_DETACH, | |
1485 | }; | |
1486 | ||
b411b363 | 1487 | #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__) |
b30ab791 | 1488 | static inline void __drbd_chk_io_error_(struct drbd_device *device, |
a2a3c74f | 1489 | enum drbd_force_detach_flags df, |
383606e0 | 1490 | const char *where) |
b411b363 | 1491 | { |
daeda1cc PR |
1492 | enum drbd_io_error_p ep; |
1493 | ||
1494 | rcu_read_lock(); | |
b30ab791 | 1495 | ep = rcu_dereference(device->ldev->disk_conf)->on_io_error; |
daeda1cc PR |
1496 | rcu_read_unlock(); |
1497 | switch (ep) { | |
1498 | case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */ | |
a2a3c74f | 1499 | if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) { |
7383506c | 1500 | if (__ratelimit(&drbd_ratelimit_state)) |
82f59cc6 | 1501 | dev_err(DEV, "Local IO failed in %s.\n", where); |
b30ab791 AG |
1502 | if (device->state.disk > D_INCONSISTENT) |
1503 | _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL); | |
b411b363 PR |
1504 | break; |
1505 | } | |
a2a3c74f | 1506 | /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */ |
b411b363 PR |
1507 | case EP_DETACH: |
1508 | case EP_CALL_HELPER: | |
a2a3c74f LE |
1509 | /* Remember whether we saw a READ or WRITE error. |
1510 | * | |
1511 | * Recovery of the affected area for WRITE failure is covered | |
1512 | * by the activity log. | |
1513 | * READ errors may fall outside that area though. Certain READ | |
1514 | * errors can be "healed" by writing good data to the affected | |
1515 | * blocks, which triggers block re-allocation in lower layers. | |
1516 | * | |
1517 | * If we can not write the bitmap after a READ error, | |
1518 | * we may need to trigger a full sync (see w_go_diskless()). | |
1519 | * | |
1520 | * Force-detach is not really an IO error, but rather a | |
1521 | * desperate measure to try to deal with a completely | |
1522 | * unresponsive lower level IO stack. | |
1523 | * Still it should be treated as a WRITE error. | |
1524 | * | |
1525 | * Meta IO error is always WRITE error: | |
1526 | * we read meta data only once during attach, | |
1527 | * which will fail in case of errors. | |
1528 | */ | |
b30ab791 | 1529 | set_bit(WAS_IO_ERROR, &device->flags); |
a2a3c74f | 1530 | if (df == DRBD_READ_ERROR) |
b30ab791 | 1531 | set_bit(WAS_READ_ERROR, &device->flags); |
a2a3c74f | 1532 | if (df == DRBD_FORCE_DETACH) |
b30ab791 AG |
1533 | set_bit(FORCE_DETACH, &device->flags); |
1534 | if (device->state.disk > D_FAILED) { | |
1535 | _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL); | |
82f59cc6 LE |
1536 | dev_err(DEV, |
1537 | "Local IO failed in %s. Detaching...\n", where); | |
b411b363 PR |
1538 | } |
1539 | break; | |
1540 | } | |
1541 | } | |
1542 | ||
1543 | /** | |
1544 | * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers | |
b30ab791 | 1545 | * @device: DRBD device. |
b411b363 PR |
1546 | * @error: Error code passed to the IO completion callback |
1547 | * @forcedetach: Force detach. I.e. the error happened while accessing the meta data | |
1548 | * | |
1549 | * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED) | |
1550 | */ | |
1551 | #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__) | |
b30ab791 | 1552 | static inline void drbd_chk_io_error_(struct drbd_device *device, |
383606e0 | 1553 | int error, enum drbd_force_detach_flags forcedetach, const char *where) |
b411b363 PR |
1554 | { |
1555 | if (error) { | |
1556 | unsigned long flags; | |
a6b32bc3 | 1557 | spin_lock_irqsave(&first_peer_device(device)->connection->req_lock, flags); |
b30ab791 | 1558 | __drbd_chk_io_error_(device, forcedetach, where); |
a6b32bc3 | 1559 | spin_unlock_irqrestore(&first_peer_device(device)->connection->req_lock, flags); |
b411b363 PR |
1560 | } |
1561 | } | |
1562 | ||
1563 | ||
1564 | /** | |
1565 | * drbd_md_first_sector() - Returns the first sector number of the meta data area | |
1566 | * @bdev: Meta data block device. | |
1567 | * | |
1568 | * BTW, for internal meta data, this happens to be the maximum capacity | |
1569 | * we could agree upon with our peer node. | |
1570 | */ | |
68e41a43 | 1571 | static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev) |
b411b363 | 1572 | { |
68e41a43 | 1573 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
1574 | case DRBD_MD_INDEX_INTERNAL: |
1575 | case DRBD_MD_INDEX_FLEX_INT: | |
1576 | return bdev->md.md_offset + bdev->md.bm_offset; | |
1577 | case DRBD_MD_INDEX_FLEX_EXT: | |
1578 | default: | |
1579 | return bdev->md.md_offset; | |
1580 | } | |
1581 | } | |
1582 | ||
1583 | /** | |
1584 | * drbd_md_last_sector() - Return the last sector number of the meta data area | |
1585 | * @bdev: Meta data block device. | |
1586 | */ | |
1587 | static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev) | |
1588 | { | |
68e41a43 | 1589 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
1590 | case DRBD_MD_INDEX_INTERNAL: |
1591 | case DRBD_MD_INDEX_FLEX_INT: | |
ae8bf312 | 1592 | return bdev->md.md_offset + MD_4kB_SECT -1; |
b411b363 PR |
1593 | case DRBD_MD_INDEX_FLEX_EXT: |
1594 | default: | |
ae8bf312 | 1595 | return bdev->md.md_offset + bdev->md.md_size_sect -1; |
b411b363 PR |
1596 | } |
1597 | } | |
1598 | ||
1599 | /* Returns the number of 512 byte sectors of the device */ | |
1600 | static inline sector_t drbd_get_capacity(struct block_device *bdev) | |
1601 | { | |
1602 | /* return bdev ? get_capacity(bdev->bd_disk) : 0; */ | |
77304d2a | 1603 | return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0; |
b411b363 PR |
1604 | } |
1605 | ||
1606 | /** | |
1607 | * drbd_get_max_capacity() - Returns the capacity we announce to out peer | |
1608 | * @bdev: Meta data block device. | |
1609 | * | |
1610 | * returns the capacity we announce to out peer. we clip ourselves at the | |
1611 | * various MAX_SECTORS, because if we don't, current implementation will | |
1612 | * oops sooner or later | |
1613 | */ | |
1614 | static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev) | |
1615 | { | |
1616 | sector_t s; | |
daeda1cc | 1617 | |
68e41a43 | 1618 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
1619 | case DRBD_MD_INDEX_INTERNAL: |
1620 | case DRBD_MD_INDEX_FLEX_INT: | |
1621 | s = drbd_get_capacity(bdev->backing_bdev) | |
1622 | ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX, | |
68e41a43 | 1623 | drbd_md_first_sector(bdev)) |
b411b363 PR |
1624 | : 0; |
1625 | break; | |
1626 | case DRBD_MD_INDEX_FLEX_EXT: | |
1627 | s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX, | |
1628 | drbd_get_capacity(bdev->backing_bdev)); | |
1629 | /* clip at maximum size the meta device can support */ | |
1630 | s = min_t(sector_t, s, | |
1631 | BM_EXT_TO_SECT(bdev->md.md_size_sect | |
1632 | - bdev->md.bm_offset)); | |
1633 | break; | |
1634 | default: | |
1635 | s = min_t(sector_t, DRBD_MAX_SECTORS, | |
1636 | drbd_get_capacity(bdev->backing_bdev)); | |
1637 | } | |
1638 | return s; | |
1639 | } | |
1640 | ||
1641 | /** | |
3a4d4eb3 | 1642 | * drbd_md_ss() - Return the sector number of our meta data super block |
b411b363 PR |
1643 | * @bdev: Meta data block device. |
1644 | */ | |
3a4d4eb3 | 1645 | static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev) |
b411b363 | 1646 | { |
3a4d4eb3 | 1647 | const int meta_dev_idx = bdev->md.meta_dev_idx; |
daeda1cc | 1648 | |
3a4d4eb3 LE |
1649 | if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT) |
1650 | return 0; | |
daeda1cc | 1651 | |
3a4d4eb3 LE |
1652 | /* Since drbd08, internal meta data is always "flexible". |
1653 | * position: last 4k aligned block of 4k size */ | |
1654 | if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL || | |
1655 | meta_dev_idx == DRBD_MD_INDEX_FLEX_INT) | |
ae8bf312 | 1656 | return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8; |
3a4d4eb3 LE |
1657 | |
1658 | /* external, some index; this is the old fixed size layout */ | |
1659 | return MD_128MB_SECT * bdev->md.meta_dev_idx; | |
b411b363 PR |
1660 | } |
1661 | ||
b411b363 PR |
1662 | static inline void |
1663 | drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w) | |
1664 | { | |
1665 | unsigned long flags; | |
1666 | spin_lock_irqsave(&q->q_lock, flags); | |
1667 | list_add(&w->list, &q->q); | |
b411b363 | 1668 | spin_unlock_irqrestore(&q->q_lock, flags); |
8c0785a5 | 1669 | wake_up(&q->q_wait); |
b411b363 PR |
1670 | } |
1671 | ||
1672 | static inline void | |
1673 | drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w) | |
1674 | { | |
1675 | unsigned long flags; | |
1676 | spin_lock_irqsave(&q->q_lock, flags); | |
1677 | list_add_tail(&w->list, &q->q); | |
b411b363 | 1678 | spin_unlock_irqrestore(&q->q_lock, flags); |
8c0785a5 | 1679 | wake_up(&q->q_wait); |
b411b363 PR |
1680 | } |
1681 | ||
bde89a9e | 1682 | static inline void wake_asender(struct drbd_connection *connection) |
b411b363 | 1683 | { |
bde89a9e AG |
1684 | if (test_bit(SIGNAL_ASENDER, &connection->flags)) |
1685 | force_sig(DRBD_SIG, connection->asender.task); | |
b411b363 PR |
1686 | } |
1687 | ||
bde89a9e | 1688 | static inline void request_ping(struct drbd_connection *connection) |
b411b363 | 1689 | { |
bde89a9e AG |
1690 | set_bit(SEND_PING, &connection->flags); |
1691 | wake_asender(connection); | |
b411b363 PR |
1692 | } |
1693 | ||
bde89a9e | 1694 | extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *); |
54761697 | 1695 | extern void *drbd_prepare_command(struct drbd_device *, struct drbd_socket *); |
bde89a9e | 1696 | extern int conn_send_command(struct drbd_connection *, struct drbd_socket *, |
dba58587 AG |
1697 | enum drbd_packet, unsigned int, void *, |
1698 | unsigned int); | |
54761697 | 1699 | extern int drbd_send_command(struct drbd_device *, struct drbd_socket *, |
dba58587 AG |
1700 | enum drbd_packet, unsigned int, void *, |
1701 | unsigned int); | |
b411b363 | 1702 | |
bde89a9e AG |
1703 | extern int drbd_send_ping(struct drbd_connection *connection); |
1704 | extern int drbd_send_ping_ack(struct drbd_connection *connection); | |
54761697 | 1705 | extern int drbd_send_state_req(struct drbd_device *, union drbd_state, union drbd_state); |
bde89a9e | 1706 | extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state); |
b411b363 PR |
1707 | |
1708 | static inline void drbd_thread_stop(struct drbd_thread *thi) | |
1709 | { | |
81e84650 | 1710 | _drbd_thread_stop(thi, false, true); |
b411b363 PR |
1711 | } |
1712 | ||
1713 | static inline void drbd_thread_stop_nowait(struct drbd_thread *thi) | |
1714 | { | |
81e84650 | 1715 | _drbd_thread_stop(thi, false, false); |
b411b363 PR |
1716 | } |
1717 | ||
1718 | static inline void drbd_thread_restart_nowait(struct drbd_thread *thi) | |
1719 | { | |
81e84650 | 1720 | _drbd_thread_stop(thi, true, false); |
b411b363 PR |
1721 | } |
1722 | ||
1723 | /* counts how many answer packets packets we expect from our peer, | |
1724 | * for either explicit application requests, | |
1725 | * or implicit barrier packets as necessary. | |
1726 | * increased: | |
1727 | * w_send_barrier | |
8554df1c | 1728 | * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ); |
b411b363 PR |
1729 | * it is much easier and equally valid to count what we queue for the |
1730 | * worker, even before it actually was queued or send. | |
1731 | * (drbd_make_request_common; recovery path on read io-error) | |
1732 | * decreased: | |
1733 | * got_BarrierAck (respective tl_clear, tl_clear_barrier) | |
8554df1c | 1734 | * _req_mod(req, DATA_RECEIVED) |
b411b363 | 1735 | * [from receive_DataReply] |
8554df1c | 1736 | * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED) |
b411b363 PR |
1737 | * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)] |
1738 | * for some reason it is NOT decreased in got_NegAck, | |
1739 | * but in the resulting cleanup code from report_params. | |
1740 | * we should try to remember the reason for that... | |
8554df1c AG |
1741 | * _req_mod(req, SEND_FAILED or SEND_CANCELED) |
1742 | * _req_mod(req, CONNECTION_LOST_WHILE_PENDING) | |
b411b363 PR |
1743 | * [from tl_clear_barrier] |
1744 | */ | |
b30ab791 | 1745 | static inline void inc_ap_pending(struct drbd_device *device) |
b411b363 | 1746 | { |
b30ab791 | 1747 | atomic_inc(&device->ap_pending_cnt); |
b411b363 PR |
1748 | } |
1749 | ||
49559d87 | 1750 | #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \ |
b30ab791 | 1751 | if (atomic_read(&device->which) < 0) \ |
b411b363 | 1752 | dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \ |
49559d87 | 1753 | func, line, \ |
b30ab791 | 1754 | atomic_read(&device->which)) |
b411b363 | 1755 | |
b30ab791 AG |
1756 | #define dec_ap_pending(device) _dec_ap_pending(device, __FUNCTION__, __LINE__) |
1757 | static inline void _dec_ap_pending(struct drbd_device *device, const char *func, int line) | |
49559d87 | 1758 | { |
b30ab791 AG |
1759 | if (atomic_dec_and_test(&device->ap_pending_cnt)) |
1760 | wake_up(&device->misc_wait); | |
49559d87 PR |
1761 | ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line); |
1762 | } | |
b411b363 PR |
1763 | |
1764 | /* counts how many resync-related answers we still expect from the peer | |
1765 | * increase decrease | |
1766 | * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY) | |
25985edc | 1767 | * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER) |
b411b363 PR |
1768 | * (or P_NEG_ACK with ID_SYNCER) |
1769 | */ | |
b30ab791 | 1770 | static inline void inc_rs_pending(struct drbd_device *device) |
b411b363 | 1771 | { |
b30ab791 | 1772 | atomic_inc(&device->rs_pending_cnt); |
b411b363 PR |
1773 | } |
1774 | ||
b30ab791 AG |
1775 | #define dec_rs_pending(device) _dec_rs_pending(device, __FUNCTION__, __LINE__) |
1776 | static inline void _dec_rs_pending(struct drbd_device *device, const char *func, int line) | |
49559d87 | 1777 | { |
b30ab791 | 1778 | atomic_dec(&device->rs_pending_cnt); |
49559d87 PR |
1779 | ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line); |
1780 | } | |
b411b363 PR |
1781 | |
1782 | /* counts how many answers we still need to send to the peer. | |
1783 | * increased on | |
1784 | * receive_Data unless protocol A; | |
1785 | * we need to send a P_RECV_ACK (proto B) | |
1786 | * or P_WRITE_ACK (proto C) | |
1787 | * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK | |
1788 | * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA | |
1789 | * receive_Barrier_* we need to send a P_BARRIER_ACK | |
1790 | */ | |
b30ab791 | 1791 | static inline void inc_unacked(struct drbd_device *device) |
b411b363 | 1792 | { |
b30ab791 | 1793 | atomic_inc(&device->unacked_cnt); |
b411b363 PR |
1794 | } |
1795 | ||
b30ab791 AG |
1796 | #define dec_unacked(device) _dec_unacked(device, __FUNCTION__, __LINE__) |
1797 | static inline void _dec_unacked(struct drbd_device *device, const char *func, int line) | |
b411b363 | 1798 | { |
b30ab791 | 1799 | atomic_dec(&device->unacked_cnt); |
49559d87 | 1800 | ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line); |
b411b363 PR |
1801 | } |
1802 | ||
b30ab791 AG |
1803 | #define sub_unacked(device, n) _sub_unacked(device, n, __FUNCTION__, __LINE__) |
1804 | static inline void _sub_unacked(struct drbd_device *device, int n, const char *func, int line) | |
b411b363 | 1805 | { |
b30ab791 | 1806 | atomic_sub(n, &device->unacked_cnt); |
49559d87 | 1807 | ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line); |
b411b363 PR |
1808 | } |
1809 | ||
1810 | /** | |
b30ab791 | 1811 | * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev |
b411b363 PR |
1812 | * @M: DRBD device. |
1813 | * | |
b30ab791 | 1814 | * You have to call put_ldev() when finished working with device->ldev. |
b411b363 PR |
1815 | */ |
1816 | #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT)) | |
1817 | #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS)) | |
1818 | ||
b30ab791 | 1819 | static inline void put_ldev(struct drbd_device *device) |
b411b363 | 1820 | { |
b30ab791 | 1821 | int i = atomic_dec_return(&device->local_cnt); |
9a0d9d03 LE |
1822 | |
1823 | /* This may be called from some endio handler, | |
1824 | * so we must not sleep here. */ | |
1825 | ||
b411b363 | 1826 | __release(local); |
1d7734a0 | 1827 | D_ASSERT(i >= 0); |
e9e6f3ec | 1828 | if (i == 0) { |
b30ab791 | 1829 | if (device->state.disk == D_DISKLESS) |
82f59cc6 | 1830 | /* even internal references gone, safe to destroy */ |
b30ab791 AG |
1831 | drbd_ldev_destroy(device); |
1832 | if (device->state.disk == D_FAILED) { | |
82f59cc6 | 1833 | /* all application IO references gone. */ |
b30ab791 | 1834 | if (!test_and_set_bit(GO_DISKLESS, &device->flags)) |
a6b32bc3 | 1835 | drbd_queue_work(&first_peer_device(device)->connection->sender_work, &device->go_diskless); |
9114d795 | 1836 | } |
b30ab791 | 1837 | wake_up(&device->misc_wait); |
e9e6f3ec | 1838 | } |
b411b363 PR |
1839 | } |
1840 | ||
1841 | #ifndef __CHECKER__ | |
b30ab791 | 1842 | static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins) |
b411b363 PR |
1843 | { |
1844 | int io_allowed; | |
1845 | ||
82f59cc6 | 1846 | /* never get a reference while D_DISKLESS */ |
b30ab791 | 1847 | if (device->state.disk == D_DISKLESS) |
82f59cc6 LE |
1848 | return 0; |
1849 | ||
b30ab791 AG |
1850 | atomic_inc(&device->local_cnt); |
1851 | io_allowed = (device->state.disk >= mins); | |
b411b363 | 1852 | if (!io_allowed) |
b30ab791 | 1853 | put_ldev(device); |
b411b363 PR |
1854 | return io_allowed; |
1855 | } | |
1856 | #else | |
b30ab791 | 1857 | extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins); |
b411b363 PR |
1858 | #endif |
1859 | ||
1860 | /* you must have an "get_ldev" reference */ | |
b30ab791 | 1861 | static inline void drbd_get_syncer_progress(struct drbd_device *device, |
b411b363 PR |
1862 | unsigned long *bits_left, unsigned int *per_mil_done) |
1863 | { | |
4b0715f0 LE |
1864 | /* this is to break it at compile time when we change that, in case we |
1865 | * want to support more than (1<<32) bits on a 32bit arch. */ | |
b30ab791 | 1866 | typecheck(unsigned long, device->rs_total); |
b411b363 PR |
1867 | |
1868 | /* note: both rs_total and rs_left are in bits, i.e. in | |
1869 | * units of BM_BLOCK_SIZE. | |
1870 | * for the percentage, we don't care. */ | |
1871 | ||
b30ab791 AG |
1872 | if (device->state.conn == C_VERIFY_S || device->state.conn == C_VERIFY_T) |
1873 | *bits_left = device->ov_left; | |
439d5953 | 1874 | else |
b30ab791 | 1875 | *bits_left = drbd_bm_total_weight(device) - device->rs_failed; |
b411b363 PR |
1876 | /* >> 10 to prevent overflow, |
1877 | * +1 to prevent division by zero */ | |
b30ab791 | 1878 | if (*bits_left > device->rs_total) { |
b411b363 PR |
1879 | /* doh. maybe a logic bug somewhere. |
1880 | * may also be just a race condition | |
1881 | * between this and a disconnect during sync. | |
1882 | * for now, just prevent in-kernel buffer overflow. | |
1883 | */ | |
1884 | smp_rmb(); | |
1885 | dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n", | |
b30ab791 AG |
1886 | drbd_conn_str(device->state.conn), |
1887 | *bits_left, device->rs_total, device->rs_failed); | |
b411b363 PR |
1888 | *per_mil_done = 0; |
1889 | } else { | |
4b0715f0 LE |
1890 | /* Make sure the division happens in long context. |
1891 | * We allow up to one petabyte storage right now, | |
1892 | * at a granularity of 4k per bit that is 2**38 bits. | |
1893 | * After shift right and multiplication by 1000, | |
1894 | * this should still fit easily into a 32bit long, | |
1895 | * so we don't need a 64bit division on 32bit arch. | |
1896 | * Note: currently we don't support such large bitmaps on 32bit | |
1897 | * arch anyways, but no harm done to be prepared for it here. | |
1898 | */ | |
b30ab791 | 1899 | unsigned int shift = device->rs_total > UINT_MAX ? 16 : 10; |
4b0715f0 | 1900 | unsigned long left = *bits_left >> shift; |
b30ab791 | 1901 | unsigned long total = 1UL + (device->rs_total >> shift); |
4b0715f0 | 1902 | unsigned long tmp = 1000UL - left * 1000UL/total; |
b411b363 PR |
1903 | *per_mil_done = tmp; |
1904 | } | |
1905 | } | |
1906 | ||
1907 | ||
1908 | /* this throttles on-the-fly application requests | |
1909 | * according to max_buffers settings; | |
1910 | * maybe re-implement using semaphores? */ | |
b30ab791 | 1911 | static inline int drbd_get_max_buffers(struct drbd_device *device) |
b411b363 | 1912 | { |
44ed167d PR |
1913 | struct net_conf *nc; |
1914 | int mxb; | |
1915 | ||
1916 | rcu_read_lock(); | |
a6b32bc3 | 1917 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d PR |
1918 | mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */ |
1919 | rcu_read_unlock(); | |
1920 | ||
b411b363 PR |
1921 | return mxb; |
1922 | } | |
1923 | ||
b30ab791 | 1924 | static inline int drbd_state_is_stable(struct drbd_device *device) |
b411b363 | 1925 | { |
b30ab791 | 1926 | union drbd_dev_state s = device->state; |
b411b363 PR |
1927 | |
1928 | /* DO NOT add a default clause, we want the compiler to warn us | |
1929 | * for any newly introduced state we may have forgotten to add here */ | |
1930 | ||
1931 | switch ((enum drbd_conns)s.conn) { | |
1932 | /* new io only accepted when there is no connection, ... */ | |
1933 | case C_STANDALONE: | |
1934 | case C_WF_CONNECTION: | |
1935 | /* ... or there is a well established connection. */ | |
1936 | case C_CONNECTED: | |
1937 | case C_SYNC_SOURCE: | |
1938 | case C_SYNC_TARGET: | |
1939 | case C_VERIFY_S: | |
1940 | case C_VERIFY_T: | |
1941 | case C_PAUSED_SYNC_S: | |
1942 | case C_PAUSED_SYNC_T: | |
67531718 PR |
1943 | case C_AHEAD: |
1944 | case C_BEHIND: | |
3719094e | 1945 | /* transitional states, IO allowed */ |
b411b363 PR |
1946 | case C_DISCONNECTING: |
1947 | case C_UNCONNECTED: | |
1948 | case C_TIMEOUT: | |
1949 | case C_BROKEN_PIPE: | |
1950 | case C_NETWORK_FAILURE: | |
1951 | case C_PROTOCOL_ERROR: | |
1952 | case C_TEAR_DOWN: | |
1953 | case C_WF_REPORT_PARAMS: | |
1954 | case C_STARTING_SYNC_S: | |
1955 | case C_STARTING_SYNC_T: | |
3719094e PR |
1956 | break; |
1957 | ||
1958 | /* Allow IO in BM exchange states with new protocols */ | |
b411b363 | 1959 | case C_WF_BITMAP_S: |
a6b32bc3 | 1960 | if (first_peer_device(device)->connection->agreed_pro_version < 96) |
3719094e PR |
1961 | return 0; |
1962 | break; | |
1963 | ||
1964 | /* no new io accepted in these states */ | |
b411b363 PR |
1965 | case C_WF_BITMAP_T: |
1966 | case C_WF_SYNC_UUID: | |
1967 | case C_MASK: | |
1968 | /* not "stable" */ | |
1969 | return 0; | |
1970 | } | |
1971 | ||
1972 | switch ((enum drbd_disk_state)s.disk) { | |
1973 | case D_DISKLESS: | |
1974 | case D_INCONSISTENT: | |
1975 | case D_OUTDATED: | |
1976 | case D_CONSISTENT: | |
1977 | case D_UP_TO_DATE: | |
5ca1de03 | 1978 | case D_FAILED: |
b411b363 PR |
1979 | /* disk state is stable as well. */ |
1980 | break; | |
1981 | ||
d942ae44 | 1982 | /* no new io accepted during transitional states */ |
b411b363 | 1983 | case D_ATTACHING: |
b411b363 PR |
1984 | case D_NEGOTIATING: |
1985 | case D_UNKNOWN: | |
1986 | case D_MASK: | |
1987 | /* not "stable" */ | |
1988 | return 0; | |
1989 | } | |
1990 | ||
1991 | return 1; | |
1992 | } | |
1993 | ||
b30ab791 | 1994 | static inline int drbd_suspended(struct drbd_device *device) |
fb22c402 | 1995 | { |
a6b32bc3 | 1996 | struct drbd_connection *connection = first_peer_device(device)->connection; |
8e0af25f | 1997 | |
bde89a9e | 1998 | return connection->susp || connection->susp_fen || connection->susp_nod; |
fb22c402 PR |
1999 | } |
2000 | ||
b30ab791 | 2001 | static inline bool may_inc_ap_bio(struct drbd_device *device) |
b411b363 | 2002 | { |
b30ab791 | 2003 | int mxb = drbd_get_max_buffers(device); |
b411b363 | 2004 | |
b30ab791 | 2005 | if (drbd_suspended(device)) |
1b881ef7 | 2006 | return false; |
b30ab791 | 2007 | if (test_bit(SUSPEND_IO, &device->flags)) |
1b881ef7 | 2008 | return false; |
b411b363 PR |
2009 | |
2010 | /* to avoid potential deadlock or bitmap corruption, | |
2011 | * in various places, we only allow new application io | |
2012 | * to start during "stable" states. */ | |
2013 | ||
2014 | /* no new io accepted when attaching or detaching the disk */ | |
b30ab791 | 2015 | if (!drbd_state_is_stable(device)) |
1b881ef7 | 2016 | return false; |
b411b363 PR |
2017 | |
2018 | /* since some older kernels don't have atomic_add_unless, | |
2019 | * and we are within the spinlock anyways, we have this workaround. */ | |
b30ab791 | 2020 | if (atomic_read(&device->ap_bio_cnt) > mxb) |
1b881ef7 | 2021 | return false; |
b30ab791 | 2022 | if (test_bit(BITMAP_IO, &device->flags)) |
1b881ef7 AG |
2023 | return false; |
2024 | return true; | |
b411b363 PR |
2025 | } |
2026 | ||
b30ab791 | 2027 | static inline bool inc_ap_bio_cond(struct drbd_device *device) |
b411b363 | 2028 | { |
1b881ef7 | 2029 | bool rv = false; |
8869d683 | 2030 | |
a6b32bc3 | 2031 | spin_lock_irq(&first_peer_device(device)->connection->req_lock); |
b30ab791 | 2032 | rv = may_inc_ap_bio(device); |
8869d683 | 2033 | if (rv) |
b30ab791 | 2034 | atomic_inc(&device->ap_bio_cnt); |
a6b32bc3 | 2035 | spin_unlock_irq(&first_peer_device(device)->connection->req_lock); |
8869d683 PR |
2036 | |
2037 | return rv; | |
2038 | } | |
b411b363 | 2039 | |
b30ab791 | 2040 | static inline void inc_ap_bio(struct drbd_device *device) |
8869d683 | 2041 | { |
b411b363 PR |
2042 | /* we wait here |
2043 | * as long as the device is suspended | |
2044 | * until the bitmap is no longer on the fly during connection | |
d942ae44 | 2045 | * handshake as long as we would exceed the max_buffer limit. |
b411b363 PR |
2046 | * |
2047 | * to avoid races with the reconnect code, | |
2048 | * we need to atomic_inc within the spinlock. */ | |
2049 | ||
b30ab791 | 2050 | wait_event(device->misc_wait, inc_ap_bio_cond(device)); |
b411b363 PR |
2051 | } |
2052 | ||
b30ab791 | 2053 | static inline void dec_ap_bio(struct drbd_device *device) |
b411b363 | 2054 | { |
b30ab791 AG |
2055 | int mxb = drbd_get_max_buffers(device); |
2056 | int ap_bio = atomic_dec_return(&device->ap_bio_cnt); | |
b411b363 PR |
2057 | |
2058 | D_ASSERT(ap_bio >= 0); | |
7ee1fb93 | 2059 | |
b30ab791 AG |
2060 | if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) { |
2061 | if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags)) | |
a6b32bc3 | 2062 | drbd_queue_work(&first_peer_device(device)->connection->sender_work, &device->bm_io_work.w); |
7ee1fb93 LE |
2063 | } |
2064 | ||
b411b363 PR |
2065 | /* this currently does wake_up for every dec_ap_bio! |
2066 | * maybe rather introduce some type of hysteresis? | |
2067 | * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */ | |
2068 | if (ap_bio < mxb) | |
b30ab791 | 2069 | wake_up(&device->misc_wait); |
b411b363 PR |
2070 | } |
2071 | ||
b30ab791 | 2072 | static inline bool verify_can_do_stop_sector(struct drbd_device *device) |
58ffa580 | 2073 | { |
a6b32bc3 AG |
2074 | return first_peer_device(device)->connection->agreed_pro_version >= 97 && |
2075 | first_peer_device(device)->connection->agreed_pro_version != 100; | |
58ffa580 LE |
2076 | } |
2077 | ||
b30ab791 | 2078 | static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val) |
b411b363 | 2079 | { |
b30ab791 AG |
2080 | int changed = device->ed_uuid != val; |
2081 | device->ed_uuid = val; | |
62b0da3a | 2082 | return changed; |
b411b363 PR |
2083 | } |
2084 | ||
b30ab791 | 2085 | static inline int drbd_queue_order_type(struct drbd_device *device) |
b411b363 PR |
2086 | { |
2087 | /* sorry, we currently have no working implementation | |
2088 | * of distributed TCQ stuff */ | |
2089 | #ifndef QUEUE_ORDERED_NONE | |
2090 | #define QUEUE_ORDERED_NONE 0 | |
2091 | #endif | |
2092 | return QUEUE_ORDERED_NONE; | |
2093 | } | |
2094 | ||
b30ab791 | 2095 | static inline void drbd_md_flush(struct drbd_device *device) |
b411b363 PR |
2096 | { |
2097 | int r; | |
2098 | ||
b30ab791 AG |
2099 | if (device->ldev == NULL) { |
2100 | dev_warn(DEV, "device->ldev == NULL in drbd_md_flush\n"); | |
fd0017c1 PR |
2101 | return; |
2102 | } | |
2103 | ||
b30ab791 | 2104 | if (test_bit(MD_NO_FUA, &device->flags)) |
b411b363 PR |
2105 | return; |
2106 | ||
b30ab791 | 2107 | r = blkdev_issue_flush(device->ldev->md_bdev, GFP_NOIO, NULL); |
b411b363 | 2108 | if (r) { |
b30ab791 | 2109 | set_bit(MD_NO_FUA, &device->flags); |
b411b363 PR |
2110 | dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r); |
2111 | } | |
2112 | } | |
2113 | ||
77c556f6 AG |
2114 | static inline struct drbd_connection *first_connection(struct drbd_resource *resource) |
2115 | { | |
2116 | return list_first_entry(&resource->connections, | |
2117 | struct drbd_connection, connections); | |
2118 | } | |
2119 | ||
b411b363 | 2120 | #endif |