Commit | Line | Data |
---|---|---|
b411b363 PR |
1 | /* |
2 | drbd_int.h | |
3 | ||
4 | This file is part of DRBD by Philipp Reisner and Lars Ellenberg. | |
5 | ||
6 | Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. | |
7 | Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. | |
8 | Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. | |
9 | ||
10 | drbd is free software; you can redistribute it and/or modify | |
11 | it under the terms of the GNU General Public License as published by | |
12 | the Free Software Foundation; either version 2, or (at your option) | |
13 | any later version. | |
14 | ||
15 | drbd is distributed in the hope that it will be useful, | |
16 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
18 | GNU General Public License for more details. | |
19 | ||
20 | You should have received a copy of the GNU General Public License | |
21 | along with drbd; see the file COPYING. If not, write to | |
22 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
23 | ||
24 | */ | |
25 | ||
26 | #ifndef _DRBD_INT_H | |
27 | #define _DRBD_INT_H | |
28 | ||
29 | #include <linux/compiler.h> | |
30 | #include <linux/types.h> | |
b411b363 PR |
31 | #include <linux/list.h> |
32 | #include <linux/sched.h> | |
33 | #include <linux/bitops.h> | |
34 | #include <linux/slab.h> | |
35 | #include <linux/crypto.h> | |
132cc538 | 36 | #include <linux/ratelimit.h> |
b411b363 PR |
37 | #include <linux/tcp.h> |
38 | #include <linux/mutex.h> | |
39 | #include <linux/major.h> | |
40 | #include <linux/blkdev.h> | |
41 | #include <linux/genhd.h> | |
062e879c | 42 | #include <linux/idr.h> |
b411b363 PR |
43 | #include <net/tcp.h> |
44 | #include <linux/lru_cache.h> | |
70c71606 | 45 | #include <linux/prefetch.h> |
3b98c0c2 | 46 | #include <linux/drbd_genl_api.h> |
b8907339 | 47 | #include <linux/drbd.h> |
d9f65229 | 48 | #include "drbd_strings.h" |
b8907339 | 49 | #include "drbd_state.h" |
a3603a6e | 50 | #include "drbd_protocol.h" |
b411b363 PR |
51 | |
52 | #ifdef __CHECKER__ | |
53 | # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr"))) | |
54 | # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read"))) | |
55 | # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write"))) | |
56 | # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call"))) | |
57 | #else | |
58 | # define __protected_by(x) | |
59 | # define __protected_read_by(x) | |
60 | # define __protected_write_by(x) | |
61 | # define __must_hold(x) | |
62 | #endif | |
63 | ||
64 | #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0) | |
65 | ||
66 | /* module parameter, defined in drbd_main.c */ | |
67 | extern unsigned int minor_count; | |
90ab5ee9 RR |
68 | extern bool disable_sendpage; |
69 | extern bool allow_oos; | |
b30ab791 | 70 | void tl_abort_disk_io(struct drbd_device *device); |
b411b363 PR |
71 | |
72 | #ifdef CONFIG_DRBD_FAULT_INJECTION | |
73 | extern int enable_faults; | |
74 | extern int fault_rate; | |
75 | extern int fault_devs; | |
76 | #endif | |
77 | ||
78 | extern char usermode_helper[]; | |
79 | ||
80 | ||
b411b363 PR |
81 | /* I don't remember why XCPU ... |
82 | * This is used to wake the asender, | |
83 | * and to interrupt sending the sending task | |
84 | * on disconnect. | |
85 | */ | |
86 | #define DRBD_SIG SIGXCPU | |
87 | ||
88 | /* This is used to stop/restart our threads. | |
89 | * Cannot use SIGTERM nor SIGKILL, since these | |
90 | * are sent out by init on runlevel changes | |
91 | * I choose SIGHUP for now. | |
92 | */ | |
93 | #define DRBD_SIGKILL SIGHUP | |
94 | ||
b411b363 PR |
95 | #define ID_IN_SYNC (4711ULL) |
96 | #define ID_OUT_OF_SYNC (4712ULL) | |
b411b363 | 97 | #define ID_SYNCER (-1ULL) |
579b57ed | 98 | |
4a23f264 | 99 | #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL) |
b411b363 | 100 | |
54761697 | 101 | struct drbd_device; |
bde89a9e | 102 | struct drbd_connection; |
b411b363 | 103 | |
3b52beff AG |
104 | #define __drbd_printk_device(level, device, fmt, args...) \ |
105 | dev_printk(level, disk_to_dev((device)->vdisk), fmt, ## args) | |
106 | #define __drbd_printk_peer_device(level, peer_device, fmt, args...) \ | |
107 | dev_printk(level, disk_to_dev((peer_device)->device->vdisk), fmt, ## args) | |
108 | #define __drbd_printk_resource(level, resource, fmt, args...) \ | |
109 | printk(level "drbd %s: " fmt, (resource)->name, ## args) | |
110 | #define __drbd_printk_connection(level, connection, fmt, args...) \ | |
111 | printk(level "drbd %s: " fmt, (connection)->resource->name, ## args) | |
112 | ||
113 | void drbd_printk_with_wrong_object_type(void); | |
114 | ||
115 | #define __drbd_printk_if_same_type(obj, type, func, level, fmt, args...) \ | |
116 | (__builtin_types_compatible_p(typeof(obj), type) || \ | |
117 | __builtin_types_compatible_p(typeof(obj), const type)), \ | |
118 | func(level, (const type)(obj), fmt, ## args) | |
119 | ||
120 | #define drbd_printk(level, obj, fmt, args...) \ | |
121 | __builtin_choose_expr( \ | |
122 | __drbd_printk_if_same_type(obj, struct drbd_device *, \ | |
123 | __drbd_printk_device, level, fmt, ## args), \ | |
124 | __builtin_choose_expr( \ | |
125 | __drbd_printk_if_same_type(obj, struct drbd_resource *, \ | |
126 | __drbd_printk_resource, level, fmt, ## args), \ | |
127 | __builtin_choose_expr( \ | |
128 | __drbd_printk_if_same_type(obj, struct drbd_connection *, \ | |
129 | __drbd_printk_connection, level, fmt, ## args), \ | |
130 | __builtin_choose_expr( \ | |
131 | __drbd_printk_if_same_type(obj, struct drbd_peer_device *, \ | |
132 | __drbd_printk_peer_device, level, fmt, ## args), \ | |
133 | drbd_printk_with_wrong_object_type())))) | |
134 | ||
135 | #define drbd_dbg(obj, fmt, args...) \ | |
136 | drbd_printk(KERN_DEBUG, obj, fmt, ## args) | |
137 | #define drbd_alert(obj, fmt, args...) \ | |
138 | drbd_printk(KERN_ALERT, obj, fmt, ## args) | |
139 | #define drbd_err(obj, fmt, args...) \ | |
140 | drbd_printk(KERN_ERR, obj, fmt, ## args) | |
141 | #define drbd_warn(obj, fmt, args...) \ | |
142 | drbd_printk(KERN_WARNING, obj, fmt, ## args) | |
143 | #define drbd_info(obj, fmt, args...) \ | |
144 | drbd_printk(KERN_INFO, obj, fmt, ## args) | |
145 | #define drbd_emerg(obj, fmt, args...) \ | |
146 | drbd_printk(KERN_EMERG, obj, fmt, ## args) | |
d0180171 AG |
147 | |
148 | #define dynamic_drbd_dbg(device, fmt, args...) \ | |
149 | dynamic_dev_dbg(disk_to_dev(device->vdisk), fmt, ## args) | |
b411b363 | 150 | |
0b0ba1ef AG |
151 | #define D_ASSERT(device, exp) do { \ |
152 | if (!(exp)) \ | |
153 | drbd_err(device, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__); \ | |
154 | } while (0) | |
b411b363 | 155 | |
841ce241 AG |
156 | /** |
157 | * expect - Make an assertion | |
158 | * | |
159 | * Unlike the assert macro, this macro returns a boolean result. | |
160 | */ | |
161 | #define expect(exp) ({ \ | |
162 | bool _bool = (exp); \ | |
163 | if (!_bool) \ | |
d0180171 | 164 | drbd_err(device, "ASSERTION %s FAILED in %s\n", \ |
841ce241 AG |
165 | #exp, __func__); \ |
166 | _bool; \ | |
167 | }) | |
b411b363 PR |
168 | |
169 | /* Defines to control fault insertion */ | |
170 | enum { | |
171 | DRBD_FAULT_MD_WR = 0, /* meta data write */ | |
172 | DRBD_FAULT_MD_RD = 1, /* read */ | |
173 | DRBD_FAULT_RS_WR = 2, /* resync */ | |
174 | DRBD_FAULT_RS_RD = 3, | |
175 | DRBD_FAULT_DT_WR = 4, /* data */ | |
176 | DRBD_FAULT_DT_RD = 5, | |
177 | DRBD_FAULT_DT_RA = 6, /* data read ahead */ | |
178 | DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */ | |
179 | DRBD_FAULT_AL_EE = 8, /* alloc ee */ | |
6b4388ac | 180 | DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */ |
b411b363 PR |
181 | |
182 | DRBD_FAULT_MAX, | |
183 | }; | |
184 | ||
b411b363 | 185 | extern unsigned int |
b30ab791 | 186 | _drbd_insert_fault(struct drbd_device *device, unsigned int type); |
0cf9d27e | 187 | |
b411b363 | 188 | static inline int |
b30ab791 | 189 | drbd_insert_fault(struct drbd_device *device, unsigned int type) { |
0cf9d27e | 190 | #ifdef CONFIG_DRBD_FAULT_INJECTION |
b411b363 PR |
191 | return fault_rate && |
192 | (enable_faults & (1<<type)) && | |
b30ab791 | 193 | _drbd_insert_fault(device, type); |
b411b363 | 194 | #else |
0cf9d27e | 195 | return 0; |
b411b363 | 196 | #endif |
0cf9d27e | 197 | } |
b411b363 PR |
198 | |
199 | /* integer division, round _UP_ to the next integer */ | |
200 | #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0)) | |
201 | /* usual integer division */ | |
202 | #define div_floor(A, B) ((A)/(B)) | |
203 | ||
b411b363 | 204 | extern struct ratelimit_state drbd_ratelimit_state; |
05a10ec7 | 205 | extern struct idr drbd_devices; /* RCU, updates: genl_lock() */ |
77c556f6 | 206 | extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */ |
b411b363 | 207 | |
d8763023 | 208 | extern const char *cmdname(enum drbd_packet cmd); |
b411b363 PR |
209 | |
210 | /* for sending/receiving the bitmap, | |
211 | * possibly in some encoding scheme */ | |
212 | struct bm_xfer_ctx { | |
213 | /* "const" | |
214 | * stores total bits and long words | |
215 | * of the bitmap, so we don't need to | |
216 | * call the accessor functions over and again. */ | |
217 | unsigned long bm_bits; | |
218 | unsigned long bm_words; | |
219 | /* during xfer, current position within the bitmap */ | |
220 | unsigned long bit_offset; | |
221 | unsigned long word_offset; | |
222 | ||
223 | /* statistics; index: (h->command == P_BITMAP) */ | |
224 | unsigned packets[2]; | |
225 | unsigned bytes[2]; | |
226 | }; | |
227 | ||
b30ab791 | 228 | extern void INFO_bm_xfer_stats(struct drbd_device *device, |
b411b363 PR |
229 | const char *direction, struct bm_xfer_ctx *c); |
230 | ||
231 | static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c) | |
232 | { | |
233 | /* word_offset counts "native long words" (32 or 64 bit), | |
234 | * aligned at 64 bit. | |
235 | * Encoded packet may end at an unaligned bit offset. | |
236 | * In case a fallback clear text packet is transmitted in | |
237 | * between, we adjust this offset back to the last 64bit | |
238 | * aligned "native long word", which makes coding and decoding | |
239 | * the plain text bitmap much more convenient. */ | |
240 | #if BITS_PER_LONG == 64 | |
241 | c->word_offset = c->bit_offset >> 6; | |
242 | #elif BITS_PER_LONG == 32 | |
243 | c->word_offset = c->bit_offset >> 5; | |
244 | c->word_offset &= ~(1UL); | |
245 | #else | |
246 | # error "unsupported BITS_PER_LONG" | |
247 | #endif | |
248 | } | |
249 | ||
bde89a9e | 250 | extern unsigned int drbd_header_size(struct drbd_connection *connection); |
b411b363 | 251 | |
b411b363 PR |
252 | /**********************************************************************/ |
253 | enum drbd_thread_state { | |
e77a0a5c AG |
254 | NONE, |
255 | RUNNING, | |
256 | EXITING, | |
257 | RESTARTING | |
b411b363 PR |
258 | }; |
259 | ||
260 | struct drbd_thread { | |
261 | spinlock_t t_lock; | |
262 | struct task_struct *task; | |
263 | struct completion stop; | |
264 | enum drbd_thread_state t_state; | |
265 | int (*function) (struct drbd_thread *); | |
2457b6d5 | 266 | struct drbd_resource *resource; |
bde89a9e | 267 | struct drbd_connection *connection; |
b411b363 | 268 | int reset_cpu_mask; |
c60b0251 | 269 | const char *name; |
b411b363 PR |
270 | }; |
271 | ||
272 | static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi) | |
273 | { | |
274 | /* THINK testing the t_state seems to be uncritical in all cases | |
275 | * (but thread_{start,stop}), so we can read it *without* the lock. | |
276 | * --lge */ | |
277 | ||
278 | smp_rmb(); | |
279 | return thi->t_state; | |
280 | } | |
281 | ||
b411b363 PR |
282 | struct drbd_work { |
283 | struct list_head list; | |
309a8348 | 284 | int (*cb)(struct drbd_work *, int cancel); |
84b8c06b AG |
285 | }; |
286 | ||
287 | struct drbd_device_work { | |
288 | struct drbd_work w; | |
289 | struct drbd_device *device; | |
b411b363 PR |
290 | }; |
291 | ||
ace652ac AG |
292 | #include "drbd_interval.h" |
293 | ||
54761697 | 294 | extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *); |
7be8da07 | 295 | |
b411b363 PR |
296 | struct drbd_request { |
297 | struct drbd_work w; | |
84b8c06b | 298 | struct drbd_device *device; |
b411b363 PR |
299 | |
300 | /* if local IO is not allowed, will be NULL. | |
301 | * if local IO _is_ allowed, holds the locally submitted bio clone, | |
302 | * or, after local IO completion, the ERR_PTR(error). | |
fcefa62e | 303 | * see drbd_request_endio(). */ |
b411b363 PR |
304 | struct bio *private_bio; |
305 | ||
ace652ac | 306 | struct drbd_interval i; |
b411b363 | 307 | |
b6dd1a89 | 308 | /* epoch: used to check on "completion" whether this req was in |
b411b363 | 309 | * the current epoch, and we therefore have to close it, |
b6dd1a89 LE |
310 | * causing a p_barrier packet to be send, starting a new epoch. |
311 | * | |
312 | * This corresponds to "barrier" in struct p_barrier[_ack], | |
313 | * and to "barrier_nr" in struct drbd_epoch (and various | |
314 | * comments/function parameters/local variable names). | |
b411b363 | 315 | */ |
b6dd1a89 | 316 | unsigned int epoch; |
b411b363 | 317 | |
b411b363 PR |
318 | struct list_head tl_requests; /* ring list in the transfer log */ |
319 | struct bio *master_bio; /* master bio pointer */ | |
b411b363 | 320 | unsigned long start_time; |
b411b363 | 321 | |
b406777e LE |
322 | /* once it hits 0, we may complete the master_bio */ |
323 | atomic_t completion_ref; | |
324 | /* once it hits 0, we may destroy this drbd_request object */ | |
325 | struct kref kref; | |
b411b363 | 326 | |
a0d856df | 327 | unsigned rq_state; /* see comments above _req_mod() */ |
b411b363 | 328 | }; |
b411b363 PR |
329 | |
330 | struct drbd_epoch { | |
bde89a9e | 331 | struct drbd_connection *connection; |
b411b363 PR |
332 | struct list_head list; |
333 | unsigned int barrier_nr; | |
334 | atomic_t epoch_size; /* increased on every request added. */ | |
335 | atomic_t active; /* increased on every req. added, and dec on every finished. */ | |
336 | unsigned long flags; | |
337 | }; | |
338 | ||
de0b2e69 RK |
339 | /* Prototype declaration of function defined in drbd_receiver.c */ |
340 | int drbdd_init(struct drbd_thread *); | |
341 | int drbd_asender(struct drbd_thread *); | |
342 | ||
b411b363 PR |
343 | /* drbd_epoch flag bits */ |
344 | enum { | |
b411b363 | 345 | DE_HAVE_BARRIER_NUMBER, |
b411b363 PR |
346 | }; |
347 | ||
348 | enum epoch_event { | |
349 | EV_PUT, | |
350 | EV_GOT_BARRIER_NR, | |
b411b363 | 351 | EV_BECAME_LAST, |
b411b363 PR |
352 | EV_CLEANUP = 32, /* used as flag */ |
353 | }; | |
354 | ||
b411b363 PR |
355 | struct digest_info { |
356 | int digest_size; | |
357 | void *digest; | |
358 | }; | |
359 | ||
f6ffca9f | 360 | struct drbd_peer_request { |
a8cd15ba AG |
361 | struct drbd_work w; |
362 | struct drbd_peer_device *peer_device; | |
85719573 | 363 | struct drbd_epoch *epoch; /* for writes */ |
45bb912b LE |
364 | struct page *pages; |
365 | atomic_t pending_bios; | |
010f6e67 | 366 | struct drbd_interval i; |
45bb912b LE |
367 | /* see comments on ee flag bits below */ |
368 | unsigned long flags; | |
85719573 PR |
369 | union { |
370 | u64 block_id; | |
371 | struct digest_info *digest; | |
372 | }; | |
45bb912b LE |
373 | }; |
374 | ||
375 | /* ee flag bits. | |
376 | * While corresponding bios are in flight, the only modification will be | |
377 | * set_bit WAS_ERROR, which has to be atomic. | |
378 | * If no bios are in flight yet, or all have been completed, | |
379 | * non-atomic modification to ee->flags is ok. | |
380 | */ | |
b411b363 PR |
381 | enum { |
382 | __EE_CALL_AL_COMPLETE_IO, | |
b411b363 | 383 | __EE_MAY_SET_IN_SYNC, |
45bb912b | 384 | |
45bb912b LE |
385 | /* In case a barrier failed, |
386 | * we need to resubmit without the barrier flag. */ | |
387 | __EE_RESUBMITTED, | |
388 | ||
6c852bec | 389 | /* we may have several bios per peer request. |
45bb912b LE |
390 | * if any of those fail, we set this flag atomically |
391 | * from the endio callback */ | |
392 | __EE_WAS_ERROR, | |
c36c3ced LE |
393 | |
394 | /* This ee has a pointer to a digest instead of a block id */ | |
395 | __EE_HAS_DIGEST, | |
7be8da07 AG |
396 | |
397 | /* Conflicting local requests need to be restarted after this request */ | |
398 | __EE_RESTART_REQUESTS, | |
303d1448 PR |
399 | |
400 | /* The peer wants a write ACK for this (wire proto C) */ | |
401 | __EE_SEND_WRITE_ACK, | |
302bdeae PR |
402 | |
403 | /* Is set when net_conf had two_primaries set while creating this peer_req */ | |
404 | __EE_IN_INTERVAL_TREE, | |
b411b363 PR |
405 | }; |
406 | #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO) | |
b411b363 | 407 | #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC) |
45bb912b LE |
408 | #define EE_RESUBMITTED (1<<__EE_RESUBMITTED) |
409 | #define EE_WAS_ERROR (1<<__EE_WAS_ERROR) | |
c36c3ced | 410 | #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST) |
7be8da07 | 411 | #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS) |
303d1448 | 412 | #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK) |
302bdeae | 413 | #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE) |
b411b363 | 414 | |
b30ab791 | 415 | /* flag bits per device */ |
b411b363 | 416 | enum { |
b411b363 PR |
417 | UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */ |
418 | MD_DIRTY, /* current uuids and flags not yet on disk */ | |
b411b363 | 419 | USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */ |
b411b363 PR |
420 | CL_ST_CHG_SUCCESS, |
421 | CL_ST_CHG_FAIL, | |
422 | CRASHED_PRIMARY, /* This node was a crashed primary. | |
423 | * Gets cleared when the state.conn | |
424 | * goes into C_CONNECTED state. */ | |
b411b363 PR |
425 | CONSIDER_RESYNC, |
426 | ||
a8a4e51e | 427 | MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */ |
b411b363 PR |
428 | SUSPEND_IO, /* suspend application io */ |
429 | BITMAP_IO, /* suspend application io; | |
430 | once no more io in flight, start bitmap io */ | |
431 | BITMAP_IO_QUEUED, /* Started bitmap IO */ | |
82f59cc6 | 432 | GO_DISKLESS, /* Disk is being detached, on io-error or admin request. */ |
a2a3c74f LE |
433 | WAS_IO_ERROR, /* Local disk failed, returned IO error */ |
434 | WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */ | |
383606e0 | 435 | FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */ |
b411b363 | 436 | RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */ |
b411b363 PR |
437 | RESIZE_PENDING, /* Size change detected locally, waiting for the response from |
438 | * the peer, if it changed there as well. */ | |
43a5182c | 439 | NEW_CUR_UUID, /* Create new current UUID when thawing IO */ |
0778286a | 440 | AL_SUSPENDED, /* Activity logging is currently suspended. */ |
370a43e7 | 441 | AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */ |
e64a3294 | 442 | B_RS_H_DONE, /* Before resync handler done (already executed) */ |
08b165ba | 443 | DISCARD_MY_DATA, /* discard_my_data flag per volume */ |
380207d0 | 444 | READ_BALANCE_RR, |
b411b363 PR |
445 | }; |
446 | ||
54761697 | 447 | struct drbd_bitmap; /* opaque for drbd_device */ |
b411b363 | 448 | |
20ceb2b2 LE |
449 | /* definition of bits in bm_flags to be used in drbd_bm_lock |
450 | * and drbd_bitmap_io and friends. */ | |
451 | enum bm_flag { | |
452 | /* do we need to kfree, or vfree bm_pages? */ | |
453 | BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */ | |
454 | ||
455 | /* currently locked for bulk operation */ | |
0e8488ad | 456 | BM_LOCKED_MASK = 0xf, |
20ceb2b2 LE |
457 | |
458 | /* in detail, that is: */ | |
459 | BM_DONT_CLEAR = 0x1, | |
460 | BM_DONT_SET = 0x2, | |
461 | BM_DONT_TEST = 0x4, | |
462 | ||
0e8488ad LE |
463 | /* so we can mark it locked for bulk operation, |
464 | * and still allow all non-bulk operations */ | |
465 | BM_IS_LOCKED = 0x8, | |
466 | ||
20ceb2b2 | 467 | /* (test bit, count bit) allowed (common case) */ |
0e8488ad | 468 | BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED, |
20ceb2b2 LE |
469 | |
470 | /* testing bits, as well as setting new bits allowed, but clearing bits | |
471 | * would be unexpected. Used during bitmap receive. Setting new bits | |
472 | * requires sending of "out-of-sync" information, though. */ | |
0e8488ad | 473 | BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED, |
20ceb2b2 | 474 | |
0e8488ad LE |
475 | /* for drbd_bm_write_copy_pages, everything is allowed, |
476 | * only concurrent bulk operations are locked out. */ | |
477 | BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED, | |
20ceb2b2 LE |
478 | }; |
479 | ||
b411b363 PR |
480 | struct drbd_work_queue { |
481 | struct list_head q; | |
b411b363 | 482 | spinlock_t q_lock; /* to protect the list. */ |
8c0785a5 | 483 | wait_queue_head_t q_wait; |
b411b363 PR |
484 | }; |
485 | ||
486 | struct drbd_socket { | |
b411b363 PR |
487 | struct mutex mutex; |
488 | struct socket *socket; | |
489 | /* this way we get our | |
490 | * send/receive buffers off the stack */ | |
5a87d920 | 491 | void *sbuf; |
e6ef8a5c | 492 | void *rbuf; |
b411b363 PR |
493 | }; |
494 | ||
495 | struct drbd_md { | |
496 | u64 md_offset; /* sector offset to 'super' block */ | |
497 | ||
498 | u64 la_size_sect; /* last agreed size, unit sectors */ | |
9f2247bb | 499 | spinlock_t uuid_lock; |
b411b363 PR |
500 | u64 uuid[UI_SIZE]; |
501 | u64 device_uuid; | |
502 | u32 flags; | |
503 | u32 md_size_sect; | |
504 | ||
ae8bf312 | 505 | s32 al_offset; /* signed relative sector offset to activity log */ |
b411b363 | 506 | s32 bm_offset; /* signed relative sector offset to bitmap */ |
3a4d4eb3 LE |
507 | |
508 | /* cached value of bdev->disk_conf->meta_dev_idx (see below) */ | |
509 | s32 meta_dev_idx; | |
510 | ||
511 | /* see al_tr_number_to_on_disk_sector() */ | |
512 | u32 al_stripes; | |
513 | u32 al_stripe_size_4k; | |
514 | u32 al_size_4k; /* cached product of the above */ | |
b411b363 PR |
515 | }; |
516 | ||
b411b363 PR |
517 | struct drbd_backing_dev { |
518 | struct block_device *backing_bdev; | |
519 | struct block_device *md_bdev; | |
b411b363 | 520 | struct drbd_md md; |
0500813f | 521 | struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */ |
b411b363 PR |
522 | sector_t known_size; /* last known size of that backing device */ |
523 | }; | |
524 | ||
525 | struct drbd_md_io { | |
0c464425 | 526 | unsigned int done; |
b411b363 PR |
527 | int error; |
528 | }; | |
529 | ||
530 | struct bm_io_work { | |
531 | struct drbd_work w; | |
532 | char *why; | |
20ceb2b2 | 533 | enum bm_flag flags; |
b30ab791 AG |
534 | int (*io_fn)(struct drbd_device *device); |
535 | void (*done)(struct drbd_device *device, int rv); | |
b411b363 PR |
536 | }; |
537 | ||
538 | enum write_ordering_e { | |
539 | WO_none, | |
540 | WO_drain_io, | |
541 | WO_bdev_flush, | |
b411b363 PR |
542 | }; |
543 | ||
778f271d | 544 | struct fifo_buffer { |
778f271d PR |
545 | unsigned int head_index; |
546 | unsigned int size; | |
9958c857 PR |
547 | int total; /* sum of all values */ |
548 | int values[0]; | |
778f271d | 549 | }; |
9958c857 | 550 | extern struct fifo_buffer *fifo_alloc(int fifo_size); |
778f271d | 551 | |
bde89a9e | 552 | /* flag bits per connection */ |
01a311a5 PR |
553 | enum { |
554 | NET_CONGESTED, /* The data socket is congested */ | |
427c0434 | 555 | RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */ |
e43ef195 | 556 | SEND_PING, /* whether asender should send a ping asap */ |
808e37b8 | 557 | SIGNAL_ASENDER, /* whether asender wants to be interrupted */ |
2a67d8b9 | 558 | GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */ |
4d0fc3fd | 559 | CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */ |
fc3b10a4 PR |
560 | CONN_WD_ST_CHG_OKAY, |
561 | CONN_WD_ST_CHG_FAIL, | |
8169e41b | 562 | CONN_DRY_RUN, /* Expect disconnect after resync handshake. */ |
6936fcb4 | 563 | CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */ |
a1096a6e | 564 | STATE_SENT, /* Do not change state/UUIDs while this is set */ |
6f3465ed LE |
565 | CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC) |
566 | * pending, from drbd worker context. | |
567 | * If set, bdi_write_congested() returns true, | |
568 | * so shrink_page_list() would not recurse into, | |
569 | * and potentially deadlock on, this drbd worker. | |
570 | */ | |
b66623e3 | 571 | DISCONNECT_SENT, |
01a311a5 PR |
572 | }; |
573 | ||
77c556f6 AG |
574 | struct drbd_resource { |
575 | char *name; | |
576 | struct kref kref; | |
803ea134 | 577 | struct idr devices; /* volume number to device mapping */ |
77c556f6 AG |
578 | struct list_head connections; |
579 | struct list_head resources; | |
eb6bea67 | 580 | struct res_opts res_opts; |
0500813f AG |
581 | struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */ |
582 | spinlock_t req_lock; | |
6bbf53ca AG |
583 | |
584 | unsigned susp:1; /* IO suspended by user */ | |
585 | unsigned susp_nod:1; /* IO suspended because no data */ | |
586 | unsigned susp_fen:1; /* IO suspended because fence peer handler runs */ | |
625a6ba2 AG |
587 | |
588 | cpumask_var_t cpu_mask; | |
77c556f6 AG |
589 | }; |
590 | ||
591 | struct drbd_connection { | |
592 | struct list_head connections; | |
593 | struct drbd_resource *resource; | |
9dc9fbb3 | 594 | struct kref kref; |
c06ece6b | 595 | struct idr peer_devices; /* volume number to peer device mapping */ |
8410da8f PR |
596 | enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */ |
597 | struct mutex cstate_mutex; /* Protects graceful disconnects */ | |
28e448bb | 598 | unsigned int connect_cnt; /* Inc each time a connection is established */ |
2111438b | 599 | |
062e879c | 600 | unsigned long flags; |
44ed167d | 601 | struct net_conf *net_conf; /* content protected by rcu */ |
91fd4dad | 602 | wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */ |
e42325a5 | 603 | |
089c075d AG |
604 | struct sockaddr_storage my_addr; |
605 | int my_addr_len; | |
606 | struct sockaddr_storage peer_addr; | |
607 | int peer_addr_len; | |
608 | ||
e42325a5 PR |
609 | struct drbd_socket data; /* data/barrier/cstate/parameter packets */ |
610 | struct drbd_socket meta; /* ping/ack (metadata) packets */ | |
31890f4a PR |
611 | int agreed_pro_version; /* actually used protocol version */ |
612 | unsigned long last_received; /* in jiffies, either socket */ | |
613 | unsigned int ko_count; | |
e6b3ea83 | 614 | |
b6dd1a89 | 615 | struct list_head transfer_log; /* all requests not yet fully processed */ |
87eeee41 | 616 | |
a0638456 | 617 | struct crypto_hash *cram_hmac_tfm; |
bde89a9e | 618 | struct crypto_hash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */ |
036b17ea | 619 | struct crypto_hash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */ |
f399002e LE |
620 | struct crypto_hash *csums_tfm; |
621 | struct crypto_hash *verify_tfm; | |
a0638456 PR |
622 | void *int_dig_in; |
623 | void *int_dig_vv; | |
624 | ||
b6dd1a89 | 625 | /* receiver side */ |
12038a3a PR |
626 | struct drbd_epoch *current_epoch; |
627 | spinlock_t epoch_lock; | |
628 | unsigned int epochs; | |
4b0007c0 | 629 | enum write_ordering_e write_ordering; |
b379c41e | 630 | atomic_t current_tle_nr; /* transfer log epoch number */ |
b6dd1a89 | 631 | unsigned current_tle_writes; /* writes seen within this tl epoch */ |
4b0007c0 | 632 | |
07be15b1 | 633 | unsigned long last_reconnect_jif; |
e6b3ea83 PR |
634 | struct drbd_thread receiver; |
635 | struct drbd_thread worker; | |
636 | struct drbd_thread asender; | |
b6dd1a89 LE |
637 | |
638 | /* sender side */ | |
d5b27b01 | 639 | struct drbd_work_queue sender_work; |
b6dd1a89 LE |
640 | |
641 | struct { | |
642 | /* whether this sender thread | |
643 | * has processed a single write yet. */ | |
644 | bool seen_any_write_yet; | |
645 | ||
646 | /* Which barrier number to send with the next P_BARRIER */ | |
647 | int current_epoch_nr; | |
648 | ||
649 | /* how many write requests have been sent | |
650 | * with req->epoch == current_epoch_nr. | |
651 | * If none, no P_BARRIER will be sent. */ | |
652 | unsigned current_epoch_writes; | |
653 | } send; | |
778f271d PR |
654 | }; |
655 | ||
113fef9e LE |
656 | struct submit_worker { |
657 | struct workqueue_struct *wq; | |
658 | struct work_struct worker; | |
659 | ||
660 | spinlock_t lock; | |
661 | struct list_head writes; | |
662 | }; | |
663 | ||
a6b32bc3 AG |
664 | struct drbd_peer_device { |
665 | struct list_head peer_devices; | |
666 | struct drbd_device *device; | |
bde89a9e | 667 | struct drbd_connection *connection; |
a6b32bc3 AG |
668 | }; |
669 | ||
670 | struct drbd_device { | |
d8628a86 | 671 | struct drbd_resource *resource; |
a6b32bc3 | 672 | struct list_head peer_devices; |
2111438b | 673 | int vnr; /* volume number within the connection */ |
81fa2e67 | 674 | struct kref kref; |
2111438b | 675 | |
b411b363 PR |
676 | /* things that are stored as / read from meta data on disk */ |
677 | unsigned long flags; | |
678 | ||
679 | /* configured by drbdsetup */ | |
b411b363 PR |
680 | struct drbd_backing_dev *ldev __protected_by(local); |
681 | ||
682 | sector_t p_size; /* partner's disk size */ | |
683 | struct request_queue *rq_queue; | |
684 | struct block_device *this_bdev; | |
685 | struct gendisk *vdisk; | |
686 | ||
07be15b1 | 687 | unsigned long last_reattach_jif; |
84b8c06b AG |
688 | struct drbd_work resync_work; |
689 | struct drbd_work unplug_work; | |
690 | struct drbd_work go_diskless; | |
691 | struct drbd_work md_sync_work; | |
692 | struct drbd_work start_resync_work; | |
b411b363 PR |
693 | struct timer_list resync_timer; |
694 | struct timer_list md_sync_timer; | |
370a43e7 | 695 | struct timer_list start_resync_timer; |
7fde2be9 | 696 | struct timer_list request_timer; |
ee15b038 LE |
697 | #ifdef DRBD_DEBUG_MD_SYNC |
698 | struct { | |
699 | unsigned int line; | |
700 | const char* func; | |
701 | } last_md_mark_dirty; | |
702 | #endif | |
b411b363 PR |
703 | |
704 | /* Used after attach while negotiating new disk state. */ | |
705 | union drbd_state new_state_tmp; | |
706 | ||
da9fbc27 | 707 | union drbd_dev_state state; |
b411b363 PR |
708 | wait_queue_head_t misc_wait; |
709 | wait_queue_head_t state_wait; /* upon each state change. */ | |
710 | unsigned int send_cnt; | |
711 | unsigned int recv_cnt; | |
712 | unsigned int read_cnt; | |
713 | unsigned int writ_cnt; | |
714 | unsigned int al_writ_cnt; | |
715 | unsigned int bm_writ_cnt; | |
716 | atomic_t ap_bio_cnt; /* Requests we need to complete */ | |
717 | atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */ | |
718 | atomic_t rs_pending_cnt; /* RS request/data packets on the wire */ | |
d942ae44 | 719 | atomic_t unacked_cnt; /* Need to send replies for */ |
b411b363 | 720 | atomic_t local_cnt; /* Waiting for local completion */ |
b2fb6dbe | 721 | |
dac1389c AG |
722 | /* Interval tree of pending local requests */ |
723 | struct rb_root read_requests; | |
de696716 | 724 | struct rb_root write_requests; |
b411b363 | 725 | |
4b0715f0 | 726 | /* blocks to resync in this run [unit BM_BLOCK_SIZE] */ |
b411b363 | 727 | unsigned long rs_total; |
4b0715f0 | 728 | /* number of resync blocks that failed in this run */ |
b411b363 PR |
729 | unsigned long rs_failed; |
730 | /* Syncer's start time [unit jiffies] */ | |
731 | unsigned long rs_start; | |
732 | /* cumulated time in PausedSyncX state [unit jiffies] */ | |
733 | unsigned long rs_paused; | |
1d7734a0 LE |
734 | /* skipped because csum was equal [unit BM_BLOCK_SIZE] */ |
735 | unsigned long rs_same_csum; | |
736 | #define DRBD_SYNC_MARKS 8 | |
737 | #define DRBD_SYNC_MARK_STEP (3*HZ) | |
b411b363 | 738 | /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */ |
1d7734a0 | 739 | unsigned long rs_mark_left[DRBD_SYNC_MARKS]; |
b411b363 | 740 | /* marks's time [unit jiffies] */ |
1d7734a0 LE |
741 | unsigned long rs_mark_time[DRBD_SYNC_MARKS]; |
742 | /* current index into rs_mark_{left,time} */ | |
743 | int rs_last_mark; | |
328e0f12 | 744 | unsigned long rs_last_bcast; /* [unit jiffies] */ |
b411b363 PR |
745 | |
746 | /* where does the admin want us to start? (sector) */ | |
747 | sector_t ov_start_sector; | |
02b91b55 | 748 | sector_t ov_stop_sector; |
b411b363 PR |
749 | /* where are we now? (sector) */ |
750 | sector_t ov_position; | |
751 | /* Start sector of out of sync range (to merge printk reporting). */ | |
752 | sector_t ov_last_oos_start; | |
753 | /* size of out-of-sync range in sectors. */ | |
754 | sector_t ov_last_oos_size; | |
755 | unsigned long ov_left; /* in bits */ | |
b411b363 | 756 | |
b411b363 PR |
757 | struct drbd_bitmap *bitmap; |
758 | unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */ | |
759 | ||
760 | /* Used to track operations of resync... */ | |
761 | struct lru_cache *resync; | |
762 | /* Number of locked elements in resync LRU */ | |
763 | unsigned int resync_locked; | |
764 | /* resync extent number waiting for application requests */ | |
765 | unsigned int resync_wenr; | |
766 | ||
767 | int open_cnt; | |
768 | u64 *p_uuid; | |
4b0007c0 | 769 | |
85719573 PR |
770 | struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */ |
771 | struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */ | |
18b75d75 AG |
772 | struct list_head done_ee; /* need to send P_WRITE_ACK */ |
773 | struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */ | |
b411b363 | 774 | struct list_head net_ee; /* zero-copy network send in progress */ |
b411b363 PR |
775 | |
776 | int next_barrier_nr; | |
b411b363 | 777 | struct list_head resync_reads; |
435f0740 LE |
778 | atomic_t pp_in_use; /* allocated from page pool */ |
779 | atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */ | |
b411b363 PR |
780 | wait_queue_head_t ee_wait; |
781 | struct page *md_io_page; /* one page buffer for md_io */ | |
cc94c650 | 782 | struct drbd_md_io md_io; |
e1711731 | 783 | atomic_t md_io_in_use; /* protects the md_io, md_io_page and md_io_tmpp */ |
b411b363 PR |
784 | spinlock_t al_lock; |
785 | wait_queue_head_t al_wait; | |
786 | struct lru_cache *act_log; /* activity log */ | |
787 | unsigned int al_tr_number; | |
788 | int al_tr_cycle; | |
b411b363 PR |
789 | wait_queue_head_t seq_wait; |
790 | atomic_t packet_seq; | |
791 | unsigned int peer_seq; | |
792 | spinlock_t peer_seq_lock; | |
793 | unsigned int minor; | |
794 | unsigned long comm_bm_set; /* communicated number of set bits. */ | |
b411b363 PR |
795 | struct bm_io_work bm_io_work; |
796 | u64 ed_uuid; /* UUID of the exposed data */ | |
8410da8f | 797 | struct mutex own_state_mutex; |
a6b32bc3 | 798 | struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */ |
b411b363 | 799 | char congestion_reason; /* Why we where congested... */ |
1d7734a0 LE |
800 | atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */ |
801 | atomic_t rs_sect_ev; /* for submitted resync data rate, both */ | |
802 | int rs_last_sect_ev; /* counter to compare with */ | |
803 | int rs_last_events; /* counter of read or write "events" (unit sectors) | |
804 | * on the lower level device when we last looked. */ | |
805 | int c_sync_rate; /* current resync rate after syncer throttle magic */ | |
bde89a9e | 806 | struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */ |
778f271d | 807 | int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */ |
759fbdfb | 808 | atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */ |
db141b2f LE |
809 | unsigned int peer_max_bio_size; |
810 | unsigned int local_max_bio_size; | |
113fef9e LE |
811 | |
812 | /* any requests that would block in drbd_make_request() | |
813 | * are deferred to this single-threaded work queue */ | |
814 | struct submit_worker submit; | |
b411b363 PR |
815 | }; |
816 | ||
b30ab791 | 817 | static inline struct drbd_device *minor_to_device(unsigned int minor) |
b411b363 | 818 | { |
05a10ec7 | 819 | return (struct drbd_device *)idr_find(&drbd_devices, minor); |
b411b363 PR |
820 | } |
821 | ||
a6b32bc3 AG |
822 | static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device) |
823 | { | |
824 | return list_first_entry(&device->peer_devices, struct drbd_peer_device, peer_devices); | |
825 | } | |
826 | ||
77c556f6 AG |
827 | #define for_each_resource(resource, _resources) \ |
828 | list_for_each_entry(resource, _resources, resources) | |
829 | ||
830 | #define for_each_resource_rcu(resource, _resources) \ | |
831 | list_for_each_entry_rcu(resource, _resources, resources) | |
832 | ||
833 | #define for_each_resource_safe(resource, tmp, _resources) \ | |
834 | list_for_each_entry_safe(resource, tmp, _resources, resources) | |
835 | ||
836 | #define for_each_connection(connection, resource) \ | |
837 | list_for_each_entry(connection, &resource->connections, connections) | |
838 | ||
839 | #define for_each_connection_rcu(connection, resource) \ | |
840 | list_for_each_entry_rcu(connection, &resource->connections, connections) | |
841 | ||
842 | #define for_each_connection_safe(connection, tmp, resource) \ | |
843 | list_for_each_entry_safe(connection, tmp, &resource->connections, connections) | |
844 | ||
a6b32bc3 AG |
845 | #define for_each_peer_device(peer_device, device) \ |
846 | list_for_each_entry(peer_device, &device->peer_devices, peer_devices) | |
847 | ||
848 | #define for_each_peer_device_rcu(peer_device, device) \ | |
849 | list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices) | |
850 | ||
851 | #define for_each_peer_device_safe(peer_device, tmp, device) \ | |
852 | list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices) | |
853 | ||
b30ab791 | 854 | static inline unsigned int device_to_minor(struct drbd_device *device) |
b411b363 | 855 | { |
b30ab791 | 856 | return device->minor; |
b411b363 PR |
857 | } |
858 | ||
b411b363 PR |
859 | /* |
860 | * function declarations | |
861 | *************************/ | |
862 | ||
863 | /* drbd_main.c */ | |
864 | ||
e89b591c PR |
865 | enum dds_flags { |
866 | DDSF_FORCED = 1, | |
867 | DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */ | |
868 | }; | |
869 | ||
b30ab791 | 870 | extern void drbd_init_set_defaults(struct drbd_device *device); |
b411b363 PR |
871 | extern int drbd_thread_start(struct drbd_thread *thi); |
872 | extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait); | |
873 | #ifdef CONFIG_SMP | |
80822284 | 874 | extern void drbd_thread_current_set_cpu(struct drbd_thread *thi); |
b411b363 PR |
875 | #else |
876 | #define drbd_thread_current_set_cpu(A) ({}) | |
b411b363 | 877 | #endif |
bde89a9e | 878 | extern void tl_release(struct drbd_connection *, unsigned int barrier_nr, |
b411b363 | 879 | unsigned int set_size); |
bde89a9e AG |
880 | extern void tl_clear(struct drbd_connection *); |
881 | extern void drbd_free_sock(struct drbd_connection *connection); | |
882 | extern int drbd_send(struct drbd_connection *connection, struct socket *sock, | |
bedbd2a5 | 883 | void *buf, size_t size, unsigned msg_flags); |
bde89a9e | 884 | extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t, |
fb708e40 AG |
885 | unsigned); |
886 | ||
bde89a9e AG |
887 | extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd); |
888 | extern int drbd_send_protocol(struct drbd_connection *connection); | |
69a22773 AG |
889 | extern int drbd_send_uuids(struct drbd_peer_device *); |
890 | extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *); | |
891 | extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *); | |
892 | extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags); | |
893 | extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s); | |
894 | extern int drbd_send_current_state(struct drbd_peer_device *); | |
895 | extern int drbd_send_sync_param(struct drbd_peer_device *); | |
bde89a9e | 896 | extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr, |
d4e67d7c | 897 | u32 set_size); |
69a22773 | 898 | extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet, |
f6ffca9f | 899 | struct drbd_peer_request *); |
69a22773 | 900 | extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet, |
a9a9994d | 901 | struct p_block_req *rp); |
69a22773 | 902 | extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet, |
a9a9994d | 903 | struct p_data *dp, int data_size); |
69a22773 | 904 | extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet, |
b411b363 | 905 | sector_t sector, int blksize, u64 block_id); |
69a22773 AG |
906 | extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *); |
907 | extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet, | |
f6ffca9f | 908 | struct drbd_peer_request *); |
69a22773 AG |
909 | extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req); |
910 | extern int drbd_send_drequest(struct drbd_peer_device *, int cmd, | |
b411b363 | 911 | sector_t sector, int size, u64 block_id); |
69a22773 | 912 | extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector, |
d8763023 AG |
913 | int size, void *digest, int digest_size, |
914 | enum drbd_packet cmd); | |
69a22773 | 915 | extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size); |
b411b363 | 916 | |
b30ab791 | 917 | extern int drbd_send_bitmap(struct drbd_device *device); |
69a22773 | 918 | extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode); |
bde89a9e | 919 | extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode); |
b411b363 | 920 | extern void drbd_free_bc(struct drbd_backing_dev *ldev); |
b30ab791 AG |
921 | extern void drbd_device_cleanup(struct drbd_device *device); |
922 | void drbd_print_uuids(struct drbd_device *device, const char *text); | |
b411b363 | 923 | |
bde89a9e | 924 | extern void conn_md_sync(struct drbd_connection *connection); |
b30ab791 AG |
925 | extern void drbd_md_write(struct drbd_device *device, void *buffer); |
926 | extern void drbd_md_sync(struct drbd_device *device); | |
927 | extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev); | |
928 | extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); | |
929 | extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); | |
930 | extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local); | |
931 | extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local); | |
932 | extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local); | |
933 | extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); | |
934 | extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local); | |
935 | extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local); | |
b411b363 | 936 | extern int drbd_md_test_flag(struct drbd_backing_dev *, int); |
ee15b038 | 937 | #ifndef DRBD_DEBUG_MD_SYNC |
b30ab791 | 938 | extern void drbd_md_mark_dirty(struct drbd_device *device); |
ee15b038 LE |
939 | #else |
940 | #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ ) | |
b30ab791 | 941 | extern void drbd_md_mark_dirty_(struct drbd_device *device, |
ee15b038 LE |
942 | unsigned int line, const char *func); |
943 | #endif | |
b30ab791 | 944 | extern void drbd_queue_bitmap_io(struct drbd_device *device, |
54761697 AG |
945 | int (*io_fn)(struct drbd_device *), |
946 | void (*done)(struct drbd_device *, int), | |
20ceb2b2 | 947 | char *why, enum bm_flag flags); |
b30ab791 | 948 | extern int drbd_bitmap_io(struct drbd_device *device, |
54761697 | 949 | int (*io_fn)(struct drbd_device *), |
20ceb2b2 | 950 | char *why, enum bm_flag flags); |
b30ab791 | 951 | extern int drbd_bitmap_io_from_worker(struct drbd_device *device, |
54761697 | 952 | int (*io_fn)(struct drbd_device *), |
edc9f5eb | 953 | char *why, enum bm_flag flags); |
b30ab791 AG |
954 | extern int drbd_bmio_set_n_write(struct drbd_device *device); |
955 | extern int drbd_bmio_clear_n_write(struct drbd_device *device); | |
956 | extern void drbd_ldev_destroy(struct drbd_device *device); | |
b411b363 | 957 | |
b411b363 | 958 | /* Meta data layout |
ae8bf312 LE |
959 | * |
960 | * We currently have two possible layouts. | |
961 | * Offsets in (512 byte) sectors. | |
962 | * external: | |
963 | * |----------- md_size_sect ------------------| | |
964 | * [ 4k superblock ][ activity log ][ Bitmap ] | |
965 | * | al_offset == 8 | | |
966 | * | bm_offset = al_offset + X | | |
967 | * ==> bitmap sectors = md_size_sect - bm_offset | |
968 | * | |
969 | * Variants: | |
970 | * old, indexed fixed size meta data: | |
971 | * | |
972 | * internal: | |
973 | * |----------- md_size_sect ------------------| | |
974 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*] | |
975 | * | al_offset < 0 | | |
976 | * | bm_offset = al_offset - Y | | |
977 | * ==> bitmap sectors = Y = al_offset - bm_offset | |
978 | * | |
979 | * [padding*] are zero or up to 7 unused 512 Byte sectors to the | |
980 | * end of the device, so that the [4k superblock] will be 4k aligned. | |
981 | * | |
982 | * The activity log consists of 4k transaction blocks, | |
983 | * which are written in a ring-buffer, or striped ring-buffer like fashion, | |
984 | * which are writtensize used to be fixed 32kB, | |
985 | * but is about to become configurable. | |
986 | */ | |
b411b363 | 987 | |
ae8bf312 LE |
988 | /* Our old fixed size meta data layout |
989 | * allows up to about 3.8TB, so if you want more, | |
7ad651b5 | 990 | * you need to use the "flexible" meta data format. */ |
ae8bf312 LE |
991 | #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */ |
992 | #define MD_4kB_SECT 8 | |
993 | #define MD_32kB_SECT 64 | |
7ad651b5 LE |
994 | |
995 | /* One activity log extent represents 4M of storage */ | |
996 | #define AL_EXTENT_SHIFT 22 | |
b411b363 PR |
997 | #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT) |
998 | ||
7ad651b5 LE |
999 | /* We could make these currently hardcoded constants configurable |
1000 | * variables at create-md time (or even re-configurable at runtime?). | |
1001 | * Which will require some more changes to the DRBD "super block" | |
1002 | * and attach code. | |
1003 | * | |
1004 | * updates per transaction: | |
1005 | * This many changes to the active set can be logged with one transaction. | |
1006 | * This number is arbitrary. | |
1007 | * context per transaction: | |
1008 | * This many context extent numbers are logged with each transaction. | |
1009 | * This number is resulting from the transaction block size (4k), the layout | |
1010 | * of the transaction header, and the number of updates per transaction. | |
1011 | * See drbd_actlog.c:struct al_transaction_on_disk | |
1012 | * */ | |
1013 | #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary | |
1014 | #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4 | |
1015 | ||
b411b363 PR |
1016 | #if BITS_PER_LONG == 32 |
1017 | #define LN2_BPL 5 | |
1018 | #define cpu_to_lel(A) cpu_to_le32(A) | |
1019 | #define lel_to_cpu(A) le32_to_cpu(A) | |
1020 | #elif BITS_PER_LONG == 64 | |
1021 | #define LN2_BPL 6 | |
1022 | #define cpu_to_lel(A) cpu_to_le64(A) | |
1023 | #define lel_to_cpu(A) le64_to_cpu(A) | |
1024 | #else | |
1025 | #error "LN2 of BITS_PER_LONG unknown!" | |
1026 | #endif | |
1027 | ||
1028 | /* resync bitmap */ | |
1029 | /* 16MB sized 'bitmap extent' to track syncer usage */ | |
1030 | struct bm_extent { | |
1031 | int rs_left; /* number of bits set (out of sync) in this extent. */ | |
1032 | int rs_failed; /* number of failed resync requests in this extent. */ | |
1033 | unsigned long flags; | |
1034 | struct lc_element lce; | |
1035 | }; | |
1036 | ||
1037 | #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */ | |
1038 | #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */ | |
e3555d85 | 1039 | #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */ |
b411b363 PR |
1040 | |
1041 | /* drbd_bitmap.c */ | |
1042 | /* | |
1043 | * We need to store one bit for a block. | |
1044 | * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap. | |
1045 | * Bit 0 ==> local node thinks this block is binary identical on both nodes | |
1046 | * Bit 1 ==> local node thinks this block needs to be synced. | |
1047 | */ | |
1048 | ||
8e26f9cc PR |
1049 | #define SLEEP_TIME (HZ/10) |
1050 | ||
45dfffeb LE |
1051 | /* We do bitmap IO in units of 4k blocks. |
1052 | * We also still have a hardcoded 4k per bit relation. */ | |
1053 | #define BM_BLOCK_SHIFT 12 /* 4k per bit */ | |
b411b363 | 1054 | #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT) |
45dfffeb LE |
1055 | /* mostly arbitrarily set the represented size of one bitmap extent, |
1056 | * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap | |
1057 | * at 4k per bit resolution) */ | |
1058 | #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */ | |
b411b363 PR |
1059 | #define BM_EXT_SIZE (1<<BM_EXT_SHIFT) |
1060 | ||
1061 | #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12) | |
1062 | #error "HAVE YOU FIXED drbdmeta AS WELL??" | |
1063 | #endif | |
1064 | ||
1065 | /* thus many _storage_ sectors are described by one bit */ | |
1066 | #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9)) | |
1067 | #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9)) | |
1068 | #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1) | |
1069 | ||
1070 | /* bit to represented kilo byte conversion */ | |
1071 | #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10)) | |
1072 | ||
1073 | /* in which _bitmap_ extent (resp. sector) the bit for a certain | |
1074 | * _storage_ sector is located in */ | |
1075 | #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9)) | |
1076 | ||
1077 | /* how much _storage_ sectors we have per bitmap sector */ | |
1078 | #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9)) | |
1079 | #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1) | |
1080 | ||
1081 | /* in one sector of the bitmap, we have this many activity_log extents. */ | |
1082 | #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT)) | |
b411b363 PR |
1083 | |
1084 | #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT) | |
1085 | #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1) | |
1086 | ||
1087 | /* the extent in "PER_EXTENT" below is an activity log extent | |
1088 | * we need that many (long words/bytes) to store the bitmap | |
1089 | * of one AL_EXTENT_SIZE chunk of storage. | |
1090 | * we can store the bitmap for that many AL_EXTENTS within | |
1091 | * one sector of the _on_disk_ bitmap: | |
1092 | * bit 0 bit 37 bit 38 bit (512*8)-1 | |
1093 | * ...|........|........|.. // ..|........| | |
1094 | * sect. 0 `296 `304 ^(512*8*8)-1 | |
1095 | * | |
1096 | #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG ) | |
1097 | #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128 | |
1098 | #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4 | |
1099 | */ | |
1100 | ||
1101 | #define DRBD_MAX_SECTORS_32 (0xffffffffLU) | |
ae8bf312 LE |
1102 | /* we have a certain meta data variant that has a fixed on-disk size of 128 |
1103 | * MiB, of which 4k are our "superblock", and 32k are the fixed size activity | |
1104 | * log, leaving this many sectors for the bitmap. | |
1105 | */ | |
1106 | ||
1107 | #define DRBD_MAX_SECTORS_FIXED_BM \ | |
1108 | ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9))) | |
1109 | #if !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32 | |
b411b363 PR |
1110 | #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32 |
1111 | #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32 | |
1112 | #else | |
ae8bf312 | 1113 | #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM |
b411b363 PR |
1114 | /* 16 TB in units of sectors */ |
1115 | #if BITS_PER_LONG == 32 | |
1116 | /* adjust by one page worth of bitmap, | |
1117 | * so we won't wrap around in drbd_bm_find_next_bit. | |
1118 | * you should use 64bit OS for that much storage, anyways. */ | |
1119 | #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff) | |
1120 | #else | |
4b0715f0 LE |
1121 | /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */ |
1122 | #define DRBD_MAX_SECTORS_FLEX (1UL << 51) | |
1123 | /* corresponds to (1UL << 38) bits right now. */ | |
b411b363 PR |
1124 | #endif |
1125 | #endif | |
1126 | ||
23361cf3 LE |
1127 | /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE, |
1128 | * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte. | |
1129 | * Since we may live in a mixed-platform cluster, | |
1130 | * we limit us to a platform agnostic constant here for now. | |
1131 | * A followup commit may allow even bigger BIO sizes, | |
1132 | * once we thought that through. */ | |
98683650 | 1133 | #define DRBD_MAX_BIO_SIZE (1U << 20) |
23361cf3 LE |
1134 | #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE |
1135 | #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE | |
1136 | #endif | |
db141b2f | 1137 | #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */ |
b411b363 | 1138 | |
98683650 PR |
1139 | #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */ |
1140 | #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */ | |
b411b363 | 1141 | |
b30ab791 AG |
1142 | extern int drbd_bm_init(struct drbd_device *device); |
1143 | extern int drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits); | |
1144 | extern void drbd_bm_cleanup(struct drbd_device *device); | |
1145 | extern void drbd_bm_set_all(struct drbd_device *device); | |
1146 | extern void drbd_bm_clear_all(struct drbd_device *device); | |
4b0715f0 | 1147 | /* set/clear/test only a few bits at a time */ |
b411b363 | 1148 | extern int drbd_bm_set_bits( |
b30ab791 | 1149 | struct drbd_device *device, unsigned long s, unsigned long e); |
b411b363 | 1150 | extern int drbd_bm_clear_bits( |
b30ab791 | 1151 | struct drbd_device *device, unsigned long s, unsigned long e); |
4b0715f0 | 1152 | extern int drbd_bm_count_bits( |
b30ab791 | 1153 | struct drbd_device *device, const unsigned long s, const unsigned long e); |
4b0715f0 LE |
1154 | /* bm_set_bits variant for use while holding drbd_bm_lock, |
1155 | * may process the whole bitmap in one go */ | |
b30ab791 | 1156 | extern void _drbd_bm_set_bits(struct drbd_device *device, |
b411b363 | 1157 | const unsigned long s, const unsigned long e); |
b30ab791 AG |
1158 | extern int drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr); |
1159 | extern int drbd_bm_e_weight(struct drbd_device *device, unsigned long enr); | |
1160 | extern int drbd_bm_write_page(struct drbd_device *device, unsigned int idx) __must_hold(local); | |
1161 | extern int drbd_bm_read(struct drbd_device *device) __must_hold(local); | |
1162 | extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr); | |
1163 | extern int drbd_bm_write(struct drbd_device *device) __must_hold(local); | |
1164 | extern int drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local); | |
1165 | extern int drbd_bm_write_all(struct drbd_device *device) __must_hold(local); | |
1166 | extern int drbd_bm_write_copy_pages(struct drbd_device *device) __must_hold(local); | |
1167 | extern size_t drbd_bm_words(struct drbd_device *device); | |
1168 | extern unsigned long drbd_bm_bits(struct drbd_device *device); | |
1169 | extern sector_t drbd_bm_capacity(struct drbd_device *device); | |
4b0715f0 LE |
1170 | |
1171 | #define DRBD_END_OF_BITMAP (~(unsigned long)0) | |
b30ab791 | 1172 | extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); |
b411b363 | 1173 | /* bm_find_next variants for use while you hold drbd_bm_lock() */ |
b30ab791 AG |
1174 | extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); |
1175 | extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo); | |
1176 | extern unsigned long _drbd_bm_total_weight(struct drbd_device *device); | |
1177 | extern unsigned long drbd_bm_total_weight(struct drbd_device *device); | |
1178 | extern int drbd_bm_rs_done(struct drbd_device *device); | |
b411b363 | 1179 | /* for receive_bitmap */ |
b30ab791 | 1180 | extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset, |
b411b363 | 1181 | size_t number, unsigned long *buffer); |
19f843aa | 1182 | /* for _drbd_send_bitmap */ |
b30ab791 | 1183 | extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset, |
b411b363 PR |
1184 | size_t number, unsigned long *buffer); |
1185 | ||
b30ab791 AG |
1186 | extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags); |
1187 | extern void drbd_bm_unlock(struct drbd_device *device); | |
b411b363 PR |
1188 | /* drbd_main.c */ |
1189 | ||
1190 | extern struct kmem_cache *drbd_request_cache; | |
6c852bec | 1191 | extern struct kmem_cache *drbd_ee_cache; /* peer requests */ |
b411b363 PR |
1192 | extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */ |
1193 | extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */ | |
1194 | extern mempool_t *drbd_request_mempool; | |
1195 | extern mempool_t *drbd_ee_mempool; | |
1196 | ||
4281808f LE |
1197 | /* drbd's page pool, used to buffer data received from the peer, |
1198 | * or data requested by the peer. | |
1199 | * | |
1200 | * This does not have an emergency reserve. | |
1201 | * | |
1202 | * When allocating from this pool, it first takes pages from the pool. | |
1203 | * Only if the pool is depleted will try to allocate from the system. | |
1204 | * | |
1205 | * The assumption is that pages taken from this pool will be processed, | |
1206 | * and given back, "quickly", and then can be recycled, so we can avoid | |
1207 | * frequent calls to alloc_page(), and still will be able to make progress even | |
1208 | * under memory pressure. | |
1209 | */ | |
1210 | extern struct page *drbd_pp_pool; | |
b411b363 PR |
1211 | extern spinlock_t drbd_pp_lock; |
1212 | extern int drbd_pp_vacant; | |
1213 | extern wait_queue_head_t drbd_pp_wait; | |
1214 | ||
4281808f LE |
1215 | /* We also need a standard (emergency-reserve backed) page pool |
1216 | * for meta data IO (activity log, bitmap). | |
1217 | * We can keep it global, as long as it is used as "N pages at a time". | |
1218 | * 128 should be plenty, currently we probably can get away with as few as 1. | |
1219 | */ | |
1220 | #define DRBD_MIN_POOL_PAGES 128 | |
1221 | extern mempool_t *drbd_md_io_page_pool; | |
1222 | ||
9476f39d LE |
1223 | /* We also need to make sure we get a bio |
1224 | * when we need it for housekeeping purposes */ | |
1225 | extern struct bio_set *drbd_md_io_bio_set; | |
1226 | /* to allocate from that set */ | |
1227 | extern struct bio *bio_alloc_drbd(gfp_t gfp_mask); | |
1228 | ||
b411b363 PR |
1229 | extern rwlock_t global_state_lock; |
1230 | ||
bde89a9e | 1231 | extern int conn_lowest_minor(struct drbd_connection *connection); |
59515a2e | 1232 | enum drbd_ret_code drbd_create_device(struct drbd_resource *resource, unsigned int minor, int vnr); |
05a10ec7 | 1233 | extern void drbd_destroy_device(struct kref *kref); |
f82795d6 | 1234 | extern void drbd_delete_device(struct drbd_device *mdev); |
b411b363 | 1235 | |
77c556f6 AG |
1236 | extern struct drbd_resource *drbd_create_resource(const char *name); |
1237 | extern void drbd_free_resource(struct drbd_resource *resource); | |
1238 | ||
eb6bea67 | 1239 | extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts); |
bde89a9e | 1240 | extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts); |
05a10ec7 | 1241 | extern void drbd_destroy_connection(struct kref *kref); |
bde89a9e | 1242 | extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len, |
089c075d | 1243 | void *peer_addr, int peer_addr_len); |
4bc76048 | 1244 | extern struct drbd_resource *drbd_find_resource(const char *name); |
77c556f6 | 1245 | extern void drbd_destroy_resource(struct kref *kref); |
bde89a9e | 1246 | extern void conn_free_crypto(struct drbd_connection *connection); |
b411b363 PR |
1247 | |
1248 | extern int proc_details; | |
1249 | ||
1250 | /* drbd_req */ | |
113fef9e | 1251 | extern void do_submit(struct work_struct *ws); |
54761697 | 1252 | extern void __drbd_make_request(struct drbd_device *, struct bio *, unsigned long); |
5a7bbad2 | 1253 | extern void drbd_make_request(struct request_queue *q, struct bio *bio); |
b30ab791 | 1254 | extern int drbd_read_remote(struct drbd_device *device, struct drbd_request *req); |
b411b363 PR |
1255 | extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec); |
1256 | extern int is_valid_ar_handle(struct drbd_request *, sector_t); | |
1257 | ||
1258 | ||
1259 | /* drbd_nl.c */ | |
8432b314 | 1260 | extern int drbd_msg_put_info(const char *info); |
b30ab791 AG |
1261 | extern void drbd_suspend_io(struct drbd_device *device); |
1262 | extern void drbd_resume_io(struct drbd_device *device); | |
b411b363 | 1263 | extern char *ppsize(char *buf, unsigned long long size); |
54761697 | 1264 | extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int); |
e96c9633 | 1265 | enum determine_dev_size { |
d752b269 PR |
1266 | DS_ERROR_SHRINK = -3, |
1267 | DS_ERROR_SPACE_MD = -2, | |
e96c9633 PR |
1268 | DS_ERROR = -1, |
1269 | DS_UNCHANGED = 0, | |
1270 | DS_SHRUNK = 1, | |
57737adc PR |
1271 | DS_GREW = 2, |
1272 | DS_GREW_FROM_ZERO = 3, | |
e96c9633 | 1273 | }; |
d752b269 | 1274 | extern enum determine_dev_size |
54761697 AG |
1275 | drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local); |
1276 | extern void resync_after_online_grow(struct drbd_device *); | |
b30ab791 AG |
1277 | extern void drbd_reconsider_max_bio_size(struct drbd_device *device); |
1278 | extern enum drbd_state_rv drbd_set_role(struct drbd_device *device, | |
bf885f8a AG |
1279 | enum drbd_role new_role, |
1280 | int force); | |
bde89a9e AG |
1281 | extern bool conn_try_outdate_peer(struct drbd_connection *connection); |
1282 | extern void conn_try_outdate_peer_async(struct drbd_connection *connection); | |
b30ab791 | 1283 | extern int drbd_khelper(struct drbd_device *device, char *cmd); |
b411b363 PR |
1284 | |
1285 | /* drbd_worker.c */ | |
d40e5671 PR |
1286 | /* bi_end_io handlers */ |
1287 | extern void drbd_md_io_complete(struct bio *bio, int error); | |
1288 | extern void drbd_peer_request_endio(struct bio *bio, int error); | |
1289 | extern void drbd_request_endio(struct bio *bio, int error); | |
b411b363 | 1290 | extern int drbd_worker(struct drbd_thread *thi); |
b30ab791 AG |
1291 | enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor); |
1292 | void drbd_resync_after_changed(struct drbd_device *device); | |
1293 | extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side); | |
1294 | extern void resume_next_sg(struct drbd_device *device); | |
1295 | extern void suspend_other_sg(struct drbd_device *device); | |
1296 | extern int drbd_resync_finished(struct drbd_device *device); | |
b411b363 | 1297 | /* maybe rather drbd_main.c ? */ |
b30ab791 AG |
1298 | extern void *drbd_md_get_buffer(struct drbd_device *device); |
1299 | extern void drbd_md_put_buffer(struct drbd_device *device); | |
1300 | extern int drbd_md_sync_page_io(struct drbd_device *device, | |
b411b363 | 1301 | struct drbd_backing_dev *bdev, sector_t sector, int rw); |
54761697 | 1302 | extern void drbd_ov_out_of_sync_found(struct drbd_device *, sector_t, int); |
b30ab791 | 1303 | extern void wait_until_done_or_force_detached(struct drbd_device *device, |
44edfb0d | 1304 | struct drbd_backing_dev *bdev, unsigned int *done); |
b30ab791 | 1305 | extern void drbd_rs_controller_reset(struct drbd_device *device); |
b411b363 | 1306 | |
b30ab791 | 1307 | static inline void ov_out_of_sync_print(struct drbd_device *device) |
b411b363 | 1308 | { |
b30ab791 | 1309 | if (device->ov_last_oos_size) { |
d0180171 | 1310 | drbd_err(device, "Out of sync: start=%llu, size=%lu (sectors)\n", |
b30ab791 AG |
1311 | (unsigned long long)device->ov_last_oos_start, |
1312 | (unsigned long)device->ov_last_oos_size); | |
b411b363 | 1313 | } |
b30ab791 | 1314 | device->ov_last_oos_size = 0; |
b411b363 PR |
1315 | } |
1316 | ||
1317 | ||
79a3c8d3 AG |
1318 | extern void drbd_csum_bio(struct crypto_hash *, struct bio *, void *); |
1319 | extern void drbd_csum_ee(struct crypto_hash *, struct drbd_peer_request *, void *); | |
b411b363 | 1320 | /* worker callbacks */ |
99920dc5 AG |
1321 | extern int w_e_end_data_req(struct drbd_work *, int); |
1322 | extern int w_e_end_rsdata_req(struct drbd_work *, int); | |
1323 | extern int w_e_end_csum_rs_req(struct drbd_work *, int); | |
1324 | extern int w_e_end_ov_reply(struct drbd_work *, int); | |
1325 | extern int w_e_end_ov_req(struct drbd_work *, int); | |
1326 | extern int w_ov_finished(struct drbd_work *, int); | |
1327 | extern int w_resync_timer(struct drbd_work *, int); | |
1328 | extern int w_send_write_hint(struct drbd_work *, int); | |
99920dc5 | 1329 | extern int w_send_dblock(struct drbd_work *, int); |
99920dc5 | 1330 | extern int w_send_read_req(struct drbd_work *, int); |
99920dc5 AG |
1331 | extern int w_e_reissue(struct drbd_work *, int); |
1332 | extern int w_restart_disk_io(struct drbd_work *, int); | |
8f7bed77 | 1333 | extern int w_send_out_of_sync(struct drbd_work *, int); |
99920dc5 | 1334 | extern int w_start_resync(struct drbd_work *, int); |
b411b363 PR |
1335 | |
1336 | extern void resync_timer_fn(unsigned long data); | |
370a43e7 | 1337 | extern void start_resync_timer_fn(unsigned long data); |
b411b363 PR |
1338 | |
1339 | /* drbd_receiver.c */ | |
753c6191 AG |
1340 | extern int drbd_receiver(struct drbd_thread *thi); |
1341 | extern int drbd_asender(struct drbd_thread *thi); | |
e8299874 LE |
1342 | extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device); |
1343 | extern bool drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector); | |
54761697 | 1344 | extern int drbd_submit_peer_request(struct drbd_device *, |
fbe29dec AG |
1345 | struct drbd_peer_request *, const unsigned, |
1346 | const int); | |
54761697 | 1347 | extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *); |
69a22773 | 1348 | extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64, |
0db55363 AG |
1349 | sector_t, unsigned int, |
1350 | gfp_t) __must_hold(local); | |
54761697 | 1351 | extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *, |
3967deb1 AG |
1352 | int); |
1353 | #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0) | |
1354 | #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1) | |
69a22773 | 1355 | extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool); |
b30ab791 AG |
1356 | extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled); |
1357 | extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed); | |
69a22773 | 1358 | extern int drbd_connected(struct drbd_peer_device *); |
b411b363 | 1359 | |
ed439848 LE |
1360 | /* Yes, there is kernel_setsockopt, but only since 2.6.18. |
1361 | * So we have our own copy of it here. */ | |
b411b363 | 1362 | static inline int drbd_setsockopt(struct socket *sock, int level, int optname, |
ed439848 | 1363 | char *optval, int optlen) |
b411b363 | 1364 | { |
ed439848 LE |
1365 | mm_segment_t oldfs = get_fs(); |
1366 | char __user *uoptval; | |
b411b363 | 1367 | int err; |
ed439848 LE |
1368 | |
1369 | uoptval = (char __user __force *)optval; | |
1370 | ||
1371 | set_fs(KERNEL_DS); | |
b411b363 | 1372 | if (level == SOL_SOCKET) |
ed439848 | 1373 | err = sock_setsockopt(sock, level, optname, uoptval, optlen); |
b411b363 | 1374 | else |
ed439848 | 1375 | err = sock->ops->setsockopt(sock, level, optname, uoptval, |
b411b363 | 1376 | optlen); |
ed439848 | 1377 | set_fs(oldfs); |
b411b363 PR |
1378 | return err; |
1379 | } | |
1380 | ||
1381 | static inline void drbd_tcp_cork(struct socket *sock) | |
1382 | { | |
ed439848 | 1383 | int val = 1; |
b411b363 | 1384 | (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK, |
ed439848 | 1385 | (char*)&val, sizeof(val)); |
b411b363 PR |
1386 | } |
1387 | ||
1388 | static inline void drbd_tcp_uncork(struct socket *sock) | |
1389 | { | |
ed439848 | 1390 | int val = 0; |
b411b363 | 1391 | (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK, |
ed439848 | 1392 | (char*)&val, sizeof(val)); |
b411b363 PR |
1393 | } |
1394 | ||
1395 | static inline void drbd_tcp_nodelay(struct socket *sock) | |
1396 | { | |
ed439848 | 1397 | int val = 1; |
b411b363 | 1398 | (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY, |
ed439848 | 1399 | (char*)&val, sizeof(val)); |
b411b363 PR |
1400 | } |
1401 | ||
1402 | static inline void drbd_tcp_quickack(struct socket *sock) | |
1403 | { | |
ed439848 | 1404 | int val = 2; |
b411b363 | 1405 | (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK, |
ed439848 | 1406 | (char*)&val, sizeof(val)); |
b411b363 PR |
1407 | } |
1408 | ||
d40e5671 PR |
1409 | /* sets the number of 512 byte sectors of our virtual device */ |
1410 | static inline void drbd_set_my_capacity(struct drbd_device *device, | |
1411 | sector_t size) | |
1412 | { | |
1413 | /* set_capacity(device->this_bdev->bd_disk, size); */ | |
1414 | set_capacity(device->vdisk, size); | |
1415 | device->this_bdev->bd_inode->i_size = (loff_t)size << 9; | |
1416 | } | |
1417 | ||
1418 | /* | |
1419 | * used to submit our private bio | |
1420 | */ | |
1421 | static inline void drbd_generic_make_request(struct drbd_device *device, | |
1422 | int fault_type, struct bio *bio) | |
1423 | { | |
1424 | __release(local); | |
1425 | if (!bio->bi_bdev) { | |
1426 | printk(KERN_ERR "drbd%d: drbd_generic_make_request: " | |
1427 | "bio->bi_bdev == NULL\n", | |
1428 | device_to_minor(device)); | |
1429 | dump_stack(); | |
1430 | bio_endio(bio, -ENODEV); | |
1431 | return; | |
1432 | } | |
1433 | ||
1434 | if (drbd_insert_fault(device, fault_type)) | |
1435 | bio_endio(bio, -EIO); | |
1436 | else | |
1437 | generic_make_request(bio); | |
1438 | } | |
1439 | ||
bde89a9e | 1440 | void drbd_bump_write_ordering(struct drbd_connection *connection, enum write_ordering_e wo); |
b411b363 PR |
1441 | |
1442 | /* drbd_proc.c */ | |
1443 | extern struct proc_dir_entry *drbd_proc; | |
7d4e9d09 | 1444 | extern const struct file_operations drbd_proc_fops; |
b411b363 PR |
1445 | extern const char *drbd_conn_str(enum drbd_conns s); |
1446 | extern const char *drbd_role_str(enum drbd_role s); | |
1447 | ||
1448 | /* drbd_actlog.c */ | |
b30ab791 AG |
1449 | extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i); |
1450 | extern void drbd_al_begin_io_commit(struct drbd_device *device, bool delegate); | |
1451 | extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i); | |
1452 | extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i, bool delegate); | |
1453 | extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i); | |
1454 | extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector); | |
1455 | extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector); | |
1456 | extern int drbd_try_rs_begin_io(struct drbd_device *device, sector_t sector); | |
1457 | extern void drbd_rs_cancel_all(struct drbd_device *device); | |
1458 | extern int drbd_rs_del_all(struct drbd_device *device); | |
1459 | extern void drbd_rs_failed_io(struct drbd_device *device, | |
b411b363 | 1460 | sector_t sector, int size); |
b30ab791 AG |
1461 | extern void drbd_advance_rs_marks(struct drbd_device *device, unsigned long still_to_go); |
1462 | extern void __drbd_set_in_sync(struct drbd_device *device, sector_t sector, | |
b411b363 | 1463 | int size, const char *file, const unsigned int line); |
b30ab791 AG |
1464 | #define drbd_set_in_sync(device, sector, size) \ |
1465 | __drbd_set_in_sync(device, sector, size, __FILE__, __LINE__) | |
1466 | extern int __drbd_set_out_of_sync(struct drbd_device *device, sector_t sector, | |
b411b363 | 1467 | int size, const char *file, const unsigned int line); |
b30ab791 AG |
1468 | #define drbd_set_out_of_sync(device, sector, size) \ |
1469 | __drbd_set_out_of_sync(device, sector, size, __FILE__, __LINE__) | |
1470 | extern void drbd_al_shrink(struct drbd_device *device); | |
54761697 | 1471 | extern int drbd_initialize_al(struct drbd_device *, void *); |
b411b363 | 1472 | |
b411b363 | 1473 | /* drbd_nl.c */ |
3b98c0c2 LE |
1474 | /* state info broadcast */ |
1475 | struct sib_info { | |
1476 | enum drbd_state_info_bcast_reason sib_reason; | |
1477 | union { | |
1478 | struct { | |
1479 | char *helper_name; | |
1480 | unsigned helper_exit_code; | |
1481 | }; | |
1482 | struct { | |
1483 | union drbd_state os; | |
1484 | union drbd_state ns; | |
1485 | }; | |
1486 | }; | |
1487 | }; | |
b30ab791 | 1488 | void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib); |
b411b363 PR |
1489 | |
1490 | /* | |
1491 | * inline helper functions | |
1492 | *************************/ | |
1493 | ||
45bb912b LE |
1494 | /* see also page_chain_add and friends in drbd_receiver.c */ |
1495 | static inline struct page *page_chain_next(struct page *page) | |
1496 | { | |
1497 | return (struct page *)page_private(page); | |
1498 | } | |
1499 | #define page_chain_for_each(page) \ | |
1500 | for (; page && ({ prefetch(page_chain_next(page)); 1; }); \ | |
1501 | page = page_chain_next(page)) | |
1502 | #define page_chain_for_each_safe(page, n) \ | |
1503 | for (; page && ({ n = page_chain_next(page); 1; }); page = n) | |
1504 | ||
45bb912b | 1505 | |
045417f7 | 1506 | static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req) |
45bb912b | 1507 | { |
db830c46 | 1508 | struct page *page = peer_req->pages; |
45bb912b LE |
1509 | page_chain_for_each(page) { |
1510 | if (page_count(page) > 1) | |
1511 | return 1; | |
1512 | } | |
1513 | return 0; | |
1514 | } | |
1515 | ||
bf885f8a | 1516 | static inline enum drbd_state_rv |
b30ab791 | 1517 | _drbd_set_state(struct drbd_device *device, union drbd_state ns, |
bf885f8a | 1518 | enum chg_state_flags flags, struct completion *done) |
b411b363 | 1519 | { |
bf885f8a | 1520 | enum drbd_state_rv rv; |
b411b363 PR |
1521 | |
1522 | read_lock(&global_state_lock); | |
b30ab791 | 1523 | rv = __drbd_set_state(device, ns, flags, done); |
b411b363 PR |
1524 | read_unlock(&global_state_lock); |
1525 | ||
1526 | return rv; | |
1527 | } | |
1528 | ||
b30ab791 | 1529 | static inline union drbd_state drbd_read_state(struct drbd_device *device) |
b411b363 | 1530 | { |
6bbf53ca | 1531 | struct drbd_resource *resource = device->resource; |
78bae59b PR |
1532 | union drbd_state rv; |
1533 | ||
b30ab791 | 1534 | rv.i = device->state.i; |
6bbf53ca AG |
1535 | rv.susp = resource->susp; |
1536 | rv.susp_nod = resource->susp_nod; | |
1537 | rv.susp_fen = resource->susp_fen; | |
78bae59b PR |
1538 | |
1539 | return rv; | |
b411b363 PR |
1540 | } |
1541 | ||
383606e0 | 1542 | enum drbd_force_detach_flags { |
a2a3c74f LE |
1543 | DRBD_READ_ERROR, |
1544 | DRBD_WRITE_ERROR, | |
383606e0 LE |
1545 | DRBD_META_IO_ERROR, |
1546 | DRBD_FORCE_DETACH, | |
1547 | }; | |
1548 | ||
b411b363 | 1549 | #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__) |
b30ab791 | 1550 | static inline void __drbd_chk_io_error_(struct drbd_device *device, |
a2a3c74f | 1551 | enum drbd_force_detach_flags df, |
383606e0 | 1552 | const char *where) |
b411b363 | 1553 | { |
daeda1cc PR |
1554 | enum drbd_io_error_p ep; |
1555 | ||
1556 | rcu_read_lock(); | |
b30ab791 | 1557 | ep = rcu_dereference(device->ldev->disk_conf)->on_io_error; |
daeda1cc PR |
1558 | rcu_read_unlock(); |
1559 | switch (ep) { | |
1560 | case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */ | |
a2a3c74f | 1561 | if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) { |
7383506c | 1562 | if (__ratelimit(&drbd_ratelimit_state)) |
d0180171 | 1563 | drbd_err(device, "Local IO failed in %s.\n", where); |
b30ab791 AG |
1564 | if (device->state.disk > D_INCONSISTENT) |
1565 | _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL); | |
b411b363 PR |
1566 | break; |
1567 | } | |
a2a3c74f | 1568 | /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */ |
b411b363 PR |
1569 | case EP_DETACH: |
1570 | case EP_CALL_HELPER: | |
a2a3c74f LE |
1571 | /* Remember whether we saw a READ or WRITE error. |
1572 | * | |
1573 | * Recovery of the affected area for WRITE failure is covered | |
1574 | * by the activity log. | |
1575 | * READ errors may fall outside that area though. Certain READ | |
1576 | * errors can be "healed" by writing good data to the affected | |
1577 | * blocks, which triggers block re-allocation in lower layers. | |
1578 | * | |
1579 | * If we can not write the bitmap after a READ error, | |
1580 | * we may need to trigger a full sync (see w_go_diskless()). | |
1581 | * | |
1582 | * Force-detach is not really an IO error, but rather a | |
1583 | * desperate measure to try to deal with a completely | |
1584 | * unresponsive lower level IO stack. | |
1585 | * Still it should be treated as a WRITE error. | |
1586 | * | |
1587 | * Meta IO error is always WRITE error: | |
1588 | * we read meta data only once during attach, | |
1589 | * which will fail in case of errors. | |
1590 | */ | |
b30ab791 | 1591 | set_bit(WAS_IO_ERROR, &device->flags); |
a2a3c74f | 1592 | if (df == DRBD_READ_ERROR) |
b30ab791 | 1593 | set_bit(WAS_READ_ERROR, &device->flags); |
a2a3c74f | 1594 | if (df == DRBD_FORCE_DETACH) |
b30ab791 AG |
1595 | set_bit(FORCE_DETACH, &device->flags); |
1596 | if (device->state.disk > D_FAILED) { | |
1597 | _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL); | |
d0180171 | 1598 | drbd_err(device, |
82f59cc6 | 1599 | "Local IO failed in %s. Detaching...\n", where); |
b411b363 PR |
1600 | } |
1601 | break; | |
1602 | } | |
1603 | } | |
1604 | ||
1605 | /** | |
1606 | * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers | |
b30ab791 | 1607 | * @device: DRBD device. |
b411b363 PR |
1608 | * @error: Error code passed to the IO completion callback |
1609 | * @forcedetach: Force detach. I.e. the error happened while accessing the meta data | |
1610 | * | |
1611 | * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED) | |
1612 | */ | |
1613 | #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__) | |
b30ab791 | 1614 | static inline void drbd_chk_io_error_(struct drbd_device *device, |
383606e0 | 1615 | int error, enum drbd_force_detach_flags forcedetach, const char *where) |
b411b363 PR |
1616 | { |
1617 | if (error) { | |
1618 | unsigned long flags; | |
0500813f | 1619 | spin_lock_irqsave(&device->resource->req_lock, flags); |
b30ab791 | 1620 | __drbd_chk_io_error_(device, forcedetach, where); |
0500813f | 1621 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
b411b363 PR |
1622 | } |
1623 | } | |
1624 | ||
1625 | ||
1626 | /** | |
1627 | * drbd_md_first_sector() - Returns the first sector number of the meta data area | |
1628 | * @bdev: Meta data block device. | |
1629 | * | |
1630 | * BTW, for internal meta data, this happens to be the maximum capacity | |
1631 | * we could agree upon with our peer node. | |
1632 | */ | |
68e41a43 | 1633 | static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev) |
b411b363 | 1634 | { |
68e41a43 | 1635 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
1636 | case DRBD_MD_INDEX_INTERNAL: |
1637 | case DRBD_MD_INDEX_FLEX_INT: | |
1638 | return bdev->md.md_offset + bdev->md.bm_offset; | |
1639 | case DRBD_MD_INDEX_FLEX_EXT: | |
1640 | default: | |
1641 | return bdev->md.md_offset; | |
1642 | } | |
1643 | } | |
1644 | ||
1645 | /** | |
1646 | * drbd_md_last_sector() - Return the last sector number of the meta data area | |
1647 | * @bdev: Meta data block device. | |
1648 | */ | |
1649 | static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev) | |
1650 | { | |
68e41a43 | 1651 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
1652 | case DRBD_MD_INDEX_INTERNAL: |
1653 | case DRBD_MD_INDEX_FLEX_INT: | |
ae8bf312 | 1654 | return bdev->md.md_offset + MD_4kB_SECT -1; |
b411b363 PR |
1655 | case DRBD_MD_INDEX_FLEX_EXT: |
1656 | default: | |
ae8bf312 | 1657 | return bdev->md.md_offset + bdev->md.md_size_sect -1; |
b411b363 PR |
1658 | } |
1659 | } | |
1660 | ||
1661 | /* Returns the number of 512 byte sectors of the device */ | |
1662 | static inline sector_t drbd_get_capacity(struct block_device *bdev) | |
1663 | { | |
1664 | /* return bdev ? get_capacity(bdev->bd_disk) : 0; */ | |
77304d2a | 1665 | return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0; |
b411b363 PR |
1666 | } |
1667 | ||
1668 | /** | |
1669 | * drbd_get_max_capacity() - Returns the capacity we announce to out peer | |
1670 | * @bdev: Meta data block device. | |
1671 | * | |
1672 | * returns the capacity we announce to out peer. we clip ourselves at the | |
1673 | * various MAX_SECTORS, because if we don't, current implementation will | |
1674 | * oops sooner or later | |
1675 | */ | |
1676 | static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev) | |
1677 | { | |
1678 | sector_t s; | |
daeda1cc | 1679 | |
68e41a43 | 1680 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
1681 | case DRBD_MD_INDEX_INTERNAL: |
1682 | case DRBD_MD_INDEX_FLEX_INT: | |
1683 | s = drbd_get_capacity(bdev->backing_bdev) | |
1684 | ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX, | |
68e41a43 | 1685 | drbd_md_first_sector(bdev)) |
b411b363 PR |
1686 | : 0; |
1687 | break; | |
1688 | case DRBD_MD_INDEX_FLEX_EXT: | |
1689 | s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX, | |
1690 | drbd_get_capacity(bdev->backing_bdev)); | |
1691 | /* clip at maximum size the meta device can support */ | |
1692 | s = min_t(sector_t, s, | |
1693 | BM_EXT_TO_SECT(bdev->md.md_size_sect | |
1694 | - bdev->md.bm_offset)); | |
1695 | break; | |
1696 | default: | |
1697 | s = min_t(sector_t, DRBD_MAX_SECTORS, | |
1698 | drbd_get_capacity(bdev->backing_bdev)); | |
1699 | } | |
1700 | return s; | |
1701 | } | |
1702 | ||
1703 | /** | |
3a4d4eb3 | 1704 | * drbd_md_ss() - Return the sector number of our meta data super block |
b411b363 PR |
1705 | * @bdev: Meta data block device. |
1706 | */ | |
3a4d4eb3 | 1707 | static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev) |
b411b363 | 1708 | { |
3a4d4eb3 | 1709 | const int meta_dev_idx = bdev->md.meta_dev_idx; |
daeda1cc | 1710 | |
3a4d4eb3 LE |
1711 | if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT) |
1712 | return 0; | |
daeda1cc | 1713 | |
3a4d4eb3 LE |
1714 | /* Since drbd08, internal meta data is always "flexible". |
1715 | * position: last 4k aligned block of 4k size */ | |
1716 | if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL || | |
1717 | meta_dev_idx == DRBD_MD_INDEX_FLEX_INT) | |
ae8bf312 | 1718 | return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8; |
3a4d4eb3 LE |
1719 | |
1720 | /* external, some index; this is the old fixed size layout */ | |
1721 | return MD_128MB_SECT * bdev->md.meta_dev_idx; | |
b411b363 PR |
1722 | } |
1723 | ||
b411b363 PR |
1724 | static inline void |
1725 | drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w) | |
1726 | { | |
1727 | unsigned long flags; | |
1728 | spin_lock_irqsave(&q->q_lock, flags); | |
1729 | list_add(&w->list, &q->q); | |
b411b363 | 1730 | spin_unlock_irqrestore(&q->q_lock, flags); |
8c0785a5 | 1731 | wake_up(&q->q_wait); |
b411b363 PR |
1732 | } |
1733 | ||
1734 | static inline void | |
1735 | drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w) | |
1736 | { | |
1737 | unsigned long flags; | |
1738 | spin_lock_irqsave(&q->q_lock, flags); | |
1739 | list_add_tail(&w->list, &q->q); | |
b411b363 | 1740 | spin_unlock_irqrestore(&q->q_lock, flags); |
8c0785a5 | 1741 | wake_up(&q->q_wait); |
b411b363 PR |
1742 | } |
1743 | ||
b5043c5e AG |
1744 | extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue); |
1745 | ||
bde89a9e | 1746 | static inline void wake_asender(struct drbd_connection *connection) |
b411b363 | 1747 | { |
bde89a9e AG |
1748 | if (test_bit(SIGNAL_ASENDER, &connection->flags)) |
1749 | force_sig(DRBD_SIG, connection->asender.task); | |
b411b363 PR |
1750 | } |
1751 | ||
bde89a9e | 1752 | static inline void request_ping(struct drbd_connection *connection) |
b411b363 | 1753 | { |
bde89a9e AG |
1754 | set_bit(SEND_PING, &connection->flags); |
1755 | wake_asender(connection); | |
b411b363 PR |
1756 | } |
1757 | ||
bde89a9e | 1758 | extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *); |
69a22773 | 1759 | extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *); |
bde89a9e | 1760 | extern int conn_send_command(struct drbd_connection *, struct drbd_socket *, |
dba58587 AG |
1761 | enum drbd_packet, unsigned int, void *, |
1762 | unsigned int); | |
69a22773 | 1763 | extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *, |
dba58587 AG |
1764 | enum drbd_packet, unsigned int, void *, |
1765 | unsigned int); | |
b411b363 | 1766 | |
bde89a9e AG |
1767 | extern int drbd_send_ping(struct drbd_connection *connection); |
1768 | extern int drbd_send_ping_ack(struct drbd_connection *connection); | |
69a22773 | 1769 | extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state); |
bde89a9e | 1770 | extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state); |
b411b363 PR |
1771 | |
1772 | static inline void drbd_thread_stop(struct drbd_thread *thi) | |
1773 | { | |
81e84650 | 1774 | _drbd_thread_stop(thi, false, true); |
b411b363 PR |
1775 | } |
1776 | ||
1777 | static inline void drbd_thread_stop_nowait(struct drbd_thread *thi) | |
1778 | { | |
81e84650 | 1779 | _drbd_thread_stop(thi, false, false); |
b411b363 PR |
1780 | } |
1781 | ||
1782 | static inline void drbd_thread_restart_nowait(struct drbd_thread *thi) | |
1783 | { | |
81e84650 | 1784 | _drbd_thread_stop(thi, true, false); |
b411b363 PR |
1785 | } |
1786 | ||
1787 | /* counts how many answer packets packets we expect from our peer, | |
1788 | * for either explicit application requests, | |
1789 | * or implicit barrier packets as necessary. | |
1790 | * increased: | |
1791 | * w_send_barrier | |
8554df1c | 1792 | * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ); |
b411b363 PR |
1793 | * it is much easier and equally valid to count what we queue for the |
1794 | * worker, even before it actually was queued or send. | |
1795 | * (drbd_make_request_common; recovery path on read io-error) | |
1796 | * decreased: | |
1797 | * got_BarrierAck (respective tl_clear, tl_clear_barrier) | |
8554df1c | 1798 | * _req_mod(req, DATA_RECEIVED) |
b411b363 | 1799 | * [from receive_DataReply] |
8554df1c | 1800 | * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED) |
b411b363 PR |
1801 | * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)] |
1802 | * for some reason it is NOT decreased in got_NegAck, | |
1803 | * but in the resulting cleanup code from report_params. | |
1804 | * we should try to remember the reason for that... | |
8554df1c AG |
1805 | * _req_mod(req, SEND_FAILED or SEND_CANCELED) |
1806 | * _req_mod(req, CONNECTION_LOST_WHILE_PENDING) | |
b411b363 PR |
1807 | * [from tl_clear_barrier] |
1808 | */ | |
b30ab791 | 1809 | static inline void inc_ap_pending(struct drbd_device *device) |
b411b363 | 1810 | { |
b30ab791 | 1811 | atomic_inc(&device->ap_pending_cnt); |
b411b363 PR |
1812 | } |
1813 | ||
49559d87 | 1814 | #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \ |
b30ab791 | 1815 | if (atomic_read(&device->which) < 0) \ |
d0180171 | 1816 | drbd_err(device, "in %s:%d: " #which " = %d < 0 !\n", \ |
49559d87 | 1817 | func, line, \ |
b30ab791 | 1818 | atomic_read(&device->which)) |
b411b363 | 1819 | |
b30ab791 AG |
1820 | #define dec_ap_pending(device) _dec_ap_pending(device, __FUNCTION__, __LINE__) |
1821 | static inline void _dec_ap_pending(struct drbd_device *device, const char *func, int line) | |
49559d87 | 1822 | { |
b30ab791 AG |
1823 | if (atomic_dec_and_test(&device->ap_pending_cnt)) |
1824 | wake_up(&device->misc_wait); | |
49559d87 PR |
1825 | ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line); |
1826 | } | |
b411b363 PR |
1827 | |
1828 | /* counts how many resync-related answers we still expect from the peer | |
1829 | * increase decrease | |
1830 | * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY) | |
25985edc | 1831 | * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER) |
b411b363 PR |
1832 | * (or P_NEG_ACK with ID_SYNCER) |
1833 | */ | |
b30ab791 | 1834 | static inline void inc_rs_pending(struct drbd_device *device) |
b411b363 | 1835 | { |
b30ab791 | 1836 | atomic_inc(&device->rs_pending_cnt); |
b411b363 PR |
1837 | } |
1838 | ||
b30ab791 AG |
1839 | #define dec_rs_pending(device) _dec_rs_pending(device, __FUNCTION__, __LINE__) |
1840 | static inline void _dec_rs_pending(struct drbd_device *device, const char *func, int line) | |
49559d87 | 1841 | { |
b30ab791 | 1842 | atomic_dec(&device->rs_pending_cnt); |
49559d87 PR |
1843 | ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line); |
1844 | } | |
b411b363 PR |
1845 | |
1846 | /* counts how many answers we still need to send to the peer. | |
1847 | * increased on | |
1848 | * receive_Data unless protocol A; | |
1849 | * we need to send a P_RECV_ACK (proto B) | |
1850 | * or P_WRITE_ACK (proto C) | |
1851 | * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK | |
1852 | * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA | |
1853 | * receive_Barrier_* we need to send a P_BARRIER_ACK | |
1854 | */ | |
b30ab791 | 1855 | static inline void inc_unacked(struct drbd_device *device) |
b411b363 | 1856 | { |
b30ab791 | 1857 | atomic_inc(&device->unacked_cnt); |
b411b363 PR |
1858 | } |
1859 | ||
b30ab791 AG |
1860 | #define dec_unacked(device) _dec_unacked(device, __FUNCTION__, __LINE__) |
1861 | static inline void _dec_unacked(struct drbd_device *device, const char *func, int line) | |
b411b363 | 1862 | { |
b30ab791 | 1863 | atomic_dec(&device->unacked_cnt); |
49559d87 | 1864 | ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line); |
b411b363 PR |
1865 | } |
1866 | ||
b30ab791 AG |
1867 | #define sub_unacked(device, n) _sub_unacked(device, n, __FUNCTION__, __LINE__) |
1868 | static inline void _sub_unacked(struct drbd_device *device, int n, const char *func, int line) | |
b411b363 | 1869 | { |
b30ab791 | 1870 | atomic_sub(n, &device->unacked_cnt); |
49559d87 | 1871 | ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line); |
b411b363 PR |
1872 | } |
1873 | ||
1874 | /** | |
b30ab791 | 1875 | * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev |
b411b363 PR |
1876 | * @M: DRBD device. |
1877 | * | |
b30ab791 | 1878 | * You have to call put_ldev() when finished working with device->ldev. |
b411b363 PR |
1879 | */ |
1880 | #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT)) | |
1881 | #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS)) | |
1882 | ||
b30ab791 | 1883 | static inline void put_ldev(struct drbd_device *device) |
b411b363 | 1884 | { |
b30ab791 | 1885 | int i = atomic_dec_return(&device->local_cnt); |
9a0d9d03 LE |
1886 | |
1887 | /* This may be called from some endio handler, | |
1888 | * so we must not sleep here. */ | |
1889 | ||
b411b363 | 1890 | __release(local); |
0b0ba1ef | 1891 | D_ASSERT(device, i >= 0); |
e9e6f3ec | 1892 | if (i == 0) { |
b30ab791 | 1893 | if (device->state.disk == D_DISKLESS) |
82f59cc6 | 1894 | /* even internal references gone, safe to destroy */ |
b30ab791 AG |
1895 | drbd_ldev_destroy(device); |
1896 | if (device->state.disk == D_FAILED) { | |
82f59cc6 | 1897 | /* all application IO references gone. */ |
b30ab791 | 1898 | if (!test_and_set_bit(GO_DISKLESS, &device->flags)) |
84b8c06b AG |
1899 | drbd_queue_work(&first_peer_device(device)->connection->sender_work, |
1900 | &device->go_diskless); | |
9114d795 | 1901 | } |
b30ab791 | 1902 | wake_up(&device->misc_wait); |
e9e6f3ec | 1903 | } |
b411b363 PR |
1904 | } |
1905 | ||
1906 | #ifndef __CHECKER__ | |
b30ab791 | 1907 | static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins) |
b411b363 PR |
1908 | { |
1909 | int io_allowed; | |
1910 | ||
82f59cc6 | 1911 | /* never get a reference while D_DISKLESS */ |
b30ab791 | 1912 | if (device->state.disk == D_DISKLESS) |
82f59cc6 LE |
1913 | return 0; |
1914 | ||
b30ab791 AG |
1915 | atomic_inc(&device->local_cnt); |
1916 | io_allowed = (device->state.disk >= mins); | |
b411b363 | 1917 | if (!io_allowed) |
b30ab791 | 1918 | put_ldev(device); |
b411b363 PR |
1919 | return io_allowed; |
1920 | } | |
1921 | #else | |
b30ab791 | 1922 | extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins); |
b411b363 PR |
1923 | #endif |
1924 | ||
1925 | /* you must have an "get_ldev" reference */ | |
b30ab791 | 1926 | static inline void drbd_get_syncer_progress(struct drbd_device *device, |
b411b363 PR |
1927 | unsigned long *bits_left, unsigned int *per_mil_done) |
1928 | { | |
4b0715f0 LE |
1929 | /* this is to break it at compile time when we change that, in case we |
1930 | * want to support more than (1<<32) bits on a 32bit arch. */ | |
b30ab791 | 1931 | typecheck(unsigned long, device->rs_total); |
b411b363 PR |
1932 | |
1933 | /* note: both rs_total and rs_left are in bits, i.e. in | |
1934 | * units of BM_BLOCK_SIZE. | |
1935 | * for the percentage, we don't care. */ | |
1936 | ||
b30ab791 AG |
1937 | if (device->state.conn == C_VERIFY_S || device->state.conn == C_VERIFY_T) |
1938 | *bits_left = device->ov_left; | |
439d5953 | 1939 | else |
b30ab791 | 1940 | *bits_left = drbd_bm_total_weight(device) - device->rs_failed; |
b411b363 PR |
1941 | /* >> 10 to prevent overflow, |
1942 | * +1 to prevent division by zero */ | |
b30ab791 | 1943 | if (*bits_left > device->rs_total) { |
b411b363 PR |
1944 | /* doh. maybe a logic bug somewhere. |
1945 | * may also be just a race condition | |
1946 | * between this and a disconnect during sync. | |
1947 | * for now, just prevent in-kernel buffer overflow. | |
1948 | */ | |
1949 | smp_rmb(); | |
d0180171 | 1950 | drbd_warn(device, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n", |
b30ab791 AG |
1951 | drbd_conn_str(device->state.conn), |
1952 | *bits_left, device->rs_total, device->rs_failed); | |
b411b363 PR |
1953 | *per_mil_done = 0; |
1954 | } else { | |
4b0715f0 LE |
1955 | /* Make sure the division happens in long context. |
1956 | * We allow up to one petabyte storage right now, | |
1957 | * at a granularity of 4k per bit that is 2**38 bits. | |
1958 | * After shift right and multiplication by 1000, | |
1959 | * this should still fit easily into a 32bit long, | |
1960 | * so we don't need a 64bit division on 32bit arch. | |
1961 | * Note: currently we don't support such large bitmaps on 32bit | |
1962 | * arch anyways, but no harm done to be prepared for it here. | |
1963 | */ | |
b30ab791 | 1964 | unsigned int shift = device->rs_total > UINT_MAX ? 16 : 10; |
4b0715f0 | 1965 | unsigned long left = *bits_left >> shift; |
b30ab791 | 1966 | unsigned long total = 1UL + (device->rs_total >> shift); |
4b0715f0 | 1967 | unsigned long tmp = 1000UL - left * 1000UL/total; |
b411b363 PR |
1968 | *per_mil_done = tmp; |
1969 | } | |
1970 | } | |
1971 | ||
1972 | ||
1973 | /* this throttles on-the-fly application requests | |
1974 | * according to max_buffers settings; | |
1975 | * maybe re-implement using semaphores? */ | |
b30ab791 | 1976 | static inline int drbd_get_max_buffers(struct drbd_device *device) |
b411b363 | 1977 | { |
44ed167d PR |
1978 | struct net_conf *nc; |
1979 | int mxb; | |
1980 | ||
1981 | rcu_read_lock(); | |
a6b32bc3 | 1982 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d PR |
1983 | mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */ |
1984 | rcu_read_unlock(); | |
1985 | ||
b411b363 PR |
1986 | return mxb; |
1987 | } | |
1988 | ||
b30ab791 | 1989 | static inline int drbd_state_is_stable(struct drbd_device *device) |
b411b363 | 1990 | { |
b30ab791 | 1991 | union drbd_dev_state s = device->state; |
b411b363 PR |
1992 | |
1993 | /* DO NOT add a default clause, we want the compiler to warn us | |
1994 | * for any newly introduced state we may have forgotten to add here */ | |
1995 | ||
1996 | switch ((enum drbd_conns)s.conn) { | |
1997 | /* new io only accepted when there is no connection, ... */ | |
1998 | case C_STANDALONE: | |
1999 | case C_WF_CONNECTION: | |
2000 | /* ... or there is a well established connection. */ | |
2001 | case C_CONNECTED: | |
2002 | case C_SYNC_SOURCE: | |
2003 | case C_SYNC_TARGET: | |
2004 | case C_VERIFY_S: | |
2005 | case C_VERIFY_T: | |
2006 | case C_PAUSED_SYNC_S: | |
2007 | case C_PAUSED_SYNC_T: | |
67531718 PR |
2008 | case C_AHEAD: |
2009 | case C_BEHIND: | |
3719094e | 2010 | /* transitional states, IO allowed */ |
b411b363 PR |
2011 | case C_DISCONNECTING: |
2012 | case C_UNCONNECTED: | |
2013 | case C_TIMEOUT: | |
2014 | case C_BROKEN_PIPE: | |
2015 | case C_NETWORK_FAILURE: | |
2016 | case C_PROTOCOL_ERROR: | |
2017 | case C_TEAR_DOWN: | |
2018 | case C_WF_REPORT_PARAMS: | |
2019 | case C_STARTING_SYNC_S: | |
2020 | case C_STARTING_SYNC_T: | |
3719094e PR |
2021 | break; |
2022 | ||
2023 | /* Allow IO in BM exchange states with new protocols */ | |
b411b363 | 2024 | case C_WF_BITMAP_S: |
a6b32bc3 | 2025 | if (first_peer_device(device)->connection->agreed_pro_version < 96) |
3719094e PR |
2026 | return 0; |
2027 | break; | |
2028 | ||
2029 | /* no new io accepted in these states */ | |
b411b363 PR |
2030 | case C_WF_BITMAP_T: |
2031 | case C_WF_SYNC_UUID: | |
2032 | case C_MASK: | |
2033 | /* not "stable" */ | |
2034 | return 0; | |
2035 | } | |
2036 | ||
2037 | switch ((enum drbd_disk_state)s.disk) { | |
2038 | case D_DISKLESS: | |
2039 | case D_INCONSISTENT: | |
2040 | case D_OUTDATED: | |
2041 | case D_CONSISTENT: | |
2042 | case D_UP_TO_DATE: | |
5ca1de03 | 2043 | case D_FAILED: |
b411b363 PR |
2044 | /* disk state is stable as well. */ |
2045 | break; | |
2046 | ||
d942ae44 | 2047 | /* no new io accepted during transitional states */ |
b411b363 | 2048 | case D_ATTACHING: |
b411b363 PR |
2049 | case D_NEGOTIATING: |
2050 | case D_UNKNOWN: | |
2051 | case D_MASK: | |
2052 | /* not "stable" */ | |
2053 | return 0; | |
2054 | } | |
2055 | ||
2056 | return 1; | |
2057 | } | |
2058 | ||
b30ab791 | 2059 | static inline int drbd_suspended(struct drbd_device *device) |
fb22c402 | 2060 | { |
6bbf53ca | 2061 | struct drbd_resource *resource = device->resource; |
8e0af25f | 2062 | |
6bbf53ca | 2063 | return resource->susp || resource->susp_fen || resource->susp_nod; |
fb22c402 PR |
2064 | } |
2065 | ||
b30ab791 | 2066 | static inline bool may_inc_ap_bio(struct drbd_device *device) |
b411b363 | 2067 | { |
b30ab791 | 2068 | int mxb = drbd_get_max_buffers(device); |
b411b363 | 2069 | |
b30ab791 | 2070 | if (drbd_suspended(device)) |
1b881ef7 | 2071 | return false; |
b30ab791 | 2072 | if (test_bit(SUSPEND_IO, &device->flags)) |
1b881ef7 | 2073 | return false; |
b411b363 PR |
2074 | |
2075 | /* to avoid potential deadlock or bitmap corruption, | |
2076 | * in various places, we only allow new application io | |
2077 | * to start during "stable" states. */ | |
2078 | ||
2079 | /* no new io accepted when attaching or detaching the disk */ | |
b30ab791 | 2080 | if (!drbd_state_is_stable(device)) |
1b881ef7 | 2081 | return false; |
b411b363 PR |
2082 | |
2083 | /* since some older kernels don't have atomic_add_unless, | |
2084 | * and we are within the spinlock anyways, we have this workaround. */ | |
b30ab791 | 2085 | if (atomic_read(&device->ap_bio_cnt) > mxb) |
1b881ef7 | 2086 | return false; |
b30ab791 | 2087 | if (test_bit(BITMAP_IO, &device->flags)) |
1b881ef7 AG |
2088 | return false; |
2089 | return true; | |
b411b363 PR |
2090 | } |
2091 | ||
b30ab791 | 2092 | static inline bool inc_ap_bio_cond(struct drbd_device *device) |
b411b363 | 2093 | { |
1b881ef7 | 2094 | bool rv = false; |
8869d683 | 2095 | |
0500813f | 2096 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 2097 | rv = may_inc_ap_bio(device); |
8869d683 | 2098 | if (rv) |
b30ab791 | 2099 | atomic_inc(&device->ap_bio_cnt); |
0500813f | 2100 | spin_unlock_irq(&device->resource->req_lock); |
8869d683 PR |
2101 | |
2102 | return rv; | |
2103 | } | |
b411b363 | 2104 | |
b30ab791 | 2105 | static inline void inc_ap_bio(struct drbd_device *device) |
8869d683 | 2106 | { |
b411b363 PR |
2107 | /* we wait here |
2108 | * as long as the device is suspended | |
2109 | * until the bitmap is no longer on the fly during connection | |
d942ae44 | 2110 | * handshake as long as we would exceed the max_buffer limit. |
b411b363 PR |
2111 | * |
2112 | * to avoid races with the reconnect code, | |
2113 | * we need to atomic_inc within the spinlock. */ | |
2114 | ||
b30ab791 | 2115 | wait_event(device->misc_wait, inc_ap_bio_cond(device)); |
b411b363 PR |
2116 | } |
2117 | ||
b30ab791 | 2118 | static inline void dec_ap_bio(struct drbd_device *device) |
b411b363 | 2119 | { |
b30ab791 AG |
2120 | int mxb = drbd_get_max_buffers(device); |
2121 | int ap_bio = atomic_dec_return(&device->ap_bio_cnt); | |
b411b363 | 2122 | |
0b0ba1ef | 2123 | D_ASSERT(device, ap_bio >= 0); |
7ee1fb93 | 2124 | |
b30ab791 AG |
2125 | if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) { |
2126 | if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags)) | |
84b8c06b AG |
2127 | drbd_queue_work(&first_peer_device(device)-> |
2128 | connection->sender_work, | |
2129 | &device->bm_io_work.w); | |
7ee1fb93 LE |
2130 | } |
2131 | ||
b411b363 PR |
2132 | /* this currently does wake_up for every dec_ap_bio! |
2133 | * maybe rather introduce some type of hysteresis? | |
2134 | * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */ | |
2135 | if (ap_bio < mxb) | |
b30ab791 | 2136 | wake_up(&device->misc_wait); |
b411b363 PR |
2137 | } |
2138 | ||
b30ab791 | 2139 | static inline bool verify_can_do_stop_sector(struct drbd_device *device) |
58ffa580 | 2140 | { |
a6b32bc3 AG |
2141 | return first_peer_device(device)->connection->agreed_pro_version >= 97 && |
2142 | first_peer_device(device)->connection->agreed_pro_version != 100; | |
58ffa580 LE |
2143 | } |
2144 | ||
b30ab791 | 2145 | static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val) |
b411b363 | 2146 | { |
b30ab791 AG |
2147 | int changed = device->ed_uuid != val; |
2148 | device->ed_uuid = val; | |
62b0da3a | 2149 | return changed; |
b411b363 PR |
2150 | } |
2151 | ||
b30ab791 | 2152 | static inline int drbd_queue_order_type(struct drbd_device *device) |
b411b363 PR |
2153 | { |
2154 | /* sorry, we currently have no working implementation | |
2155 | * of distributed TCQ stuff */ | |
2156 | #ifndef QUEUE_ORDERED_NONE | |
2157 | #define QUEUE_ORDERED_NONE 0 | |
2158 | #endif | |
2159 | return QUEUE_ORDERED_NONE; | |
2160 | } | |
2161 | ||
b30ab791 | 2162 | static inline void drbd_md_flush(struct drbd_device *device) |
b411b363 PR |
2163 | { |
2164 | int r; | |
2165 | ||
b30ab791 | 2166 | if (device->ldev == NULL) { |
d0180171 | 2167 | drbd_warn(device, "device->ldev == NULL in drbd_md_flush\n"); |
fd0017c1 PR |
2168 | return; |
2169 | } | |
2170 | ||
b30ab791 | 2171 | if (test_bit(MD_NO_FUA, &device->flags)) |
b411b363 PR |
2172 | return; |
2173 | ||
b30ab791 | 2174 | r = blkdev_issue_flush(device->ldev->md_bdev, GFP_NOIO, NULL); |
b411b363 | 2175 | if (r) { |
b30ab791 | 2176 | set_bit(MD_NO_FUA, &device->flags); |
d0180171 | 2177 | drbd_err(device, "meta data flush failed with status %d, disabling md-flushes\n", r); |
b411b363 PR |
2178 | } |
2179 | } | |
2180 | ||
77c556f6 AG |
2181 | static inline struct drbd_connection *first_connection(struct drbd_resource *resource) |
2182 | { | |
2183 | return list_first_entry(&resource->connections, | |
2184 | struct drbd_connection, connections); | |
2185 | } | |
2186 | ||
b411b363 | 2187 | #endif |