blob: a30a5ed811bea8e9f4486d004a9159f9a2a3cb9c [file] [log] [blame]
Christoph Böhmwalder93c68cc2022-11-22 14:43:01 +01001/* SPDX-License-Identifier: GPL-2.0-only */
Philipp Reisnerb411b362009-09-25 16:07:19 -07002/*
3 drbd_int.h
4
5 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
6
7 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
8 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
9 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
10
Philipp Reisnerb411b362009-09-25 16:07:19 -070011
12*/
13
14#ifndef _DRBD_INT_H
15#define _DRBD_INT_H
16
Herbert Xu9534d672016-01-24 21:19:21 +080017#include <crypto/hash.h>
Philipp Reisnerb411b362009-09-25 16:07:19 -070018#include <linux/compiler.h>
19#include <linux/types.h>
Philipp Reisnerb411b362009-09-25 16:07:19 -070020#include <linux/list.h>
Ingo Molnar3f07c012017-02-08 18:51:30 +010021#include <linux/sched/signal.h>
Philipp Reisnerb411b362009-09-25 16:07:19 -070022#include <linux/bitops.h>
23#include <linux/slab.h>
Randy Dunlap132cc532009-10-07 19:26:00 +020024#include <linux/ratelimit.h>
Philipp Reisnerb411b362009-09-25 16:07:19 -070025#include <linux/tcp.h>
26#include <linux/mutex.h>
27#include <linux/major.h>
28#include <linux/blkdev.h>
Tejun Heo66114ca2015-05-22 17:13:32 -040029#include <linux/backing-dev.h>
Philipp Reisner062e8792011-02-08 11:09:18 +010030#include <linux/idr.h>
Luis de Bethencourt9d5059c2016-08-02 14:03:47 -070031#include <linux/dynamic_debug.h>
Philipp Reisnerb411b362009-09-25 16:07:19 -070032#include <net/tcp.h>
33#include <linux/lru_cache.h>
Paul Gortmaker70c71602011-05-22 16:47:17 -040034#include <linux/prefetch.h>
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +010035#include <linux/drbd_genl_api.h>
Philipp Reisnerb8907332011-01-27 14:07:51 +010036#include <linux/drbd.h>
Christoph Böhmwalder20f2a342023-01-13 13:35:06 +010037#include <linux/drbd_config.h>
Andreas Gruenbacherd9f65222011-09-01 13:18:31 +020038#include "drbd_strings.h"
Philipp Reisnerb8907332011-01-27 14:07:51 +010039#include "drbd_state.h"
Andreas Gruenbachera3603a62011-05-30 11:47:37 +020040#include "drbd_protocol.h"
Christoph Böhmwalder136160c2022-12-01 12:03:47 +010041#include "drbd_polymorph_printk.h"
Philipp Reisnerb411b362009-09-25 16:07:19 -070042
Roland Kammerer183ece32017-08-29 10:20:46 +020043/* shared module parameters, defined in drbd_main.c */
Philipp Reisnerb411b362009-09-25 16:07:19 -070044#ifdef CONFIG_DRBD_FAULT_INJECTION
Roland Kammerer183ece32017-08-29 10:20:46 +020045extern int drbd_enable_faults;
46extern int drbd_fault_rate;
Philipp Reisnerb411b362009-09-25 16:07:19 -070047#endif
48
Roland Kammerer183ece32017-08-29 10:20:46 +020049extern unsigned int drbd_minor_count;
Greg Kroah-Hartman8ab761e2017-08-29 10:20:45 +020050extern char drbd_usermode_helper[];
Roland Kammerer183ece32017-08-29 10:20:46 +020051extern int drbd_proc_details;
Philipp Reisnerb411b362009-09-25 16:07:19 -070052
53
Philipp Reisnerb411b362009-09-25 16:07:19 -070054/* This is used to stop/restart our threads.
55 * Cannot use SIGTERM nor SIGKILL, since these
56 * are sent out by init on runlevel changes
57 * I choose SIGHUP for now.
58 */
59#define DRBD_SIGKILL SIGHUP
60
Philipp Reisnerb411b362009-09-25 16:07:19 -070061#define ID_IN_SYNC (4711ULL)
62#define ID_OUT_OF_SYNC (4712ULL)
Philipp Reisnerb411b362009-09-25 16:07:19 -070063#define ID_SYNCER (-1ULL)
Andreas Gruenbacher579b57e2011-01-13 18:40:57 +010064
Philipp Reisner4a23f262011-01-11 17:42:17 +010065#define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
Philipp Reisnerb411b362009-09-25 16:07:19 -070066
Andreas Gruenbacher54761692011-05-30 16:15:21 +020067struct drbd_device;
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +020068struct drbd_connection;
Andreas Gruenbacher8164dd62023-03-30 12:27:40 +020069struct drbd_peer_device;
Philipp Reisnerb411b362009-09-25 16:07:19 -070070
Philipp Reisnerb411b362009-09-25 16:07:19 -070071/* Defines to control fault insertion */
72enum {
73 DRBD_FAULT_MD_WR = 0, /* meta data write */
74 DRBD_FAULT_MD_RD = 1, /* read */
75 DRBD_FAULT_RS_WR = 2, /* resync */
76 DRBD_FAULT_RS_RD = 3,
77 DRBD_FAULT_DT_WR = 4, /* data */
78 DRBD_FAULT_DT_RD = 5,
79 DRBD_FAULT_DT_RA = 6, /* data read ahead */
80 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */
81 DRBD_FAULT_AL_EE = 8, /* alloc ee */
Philipp Reisner6b4388a2010-04-26 14:11:45 +020082 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
Philipp Reisnerb411b362009-09-25 16:07:19 -070083
84 DRBD_FAULT_MAX,
85};
86
Philipp Reisnerb411b362009-09-25 16:07:19 -070087extern unsigned int
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +020088_drbd_insert_fault(struct drbd_device *device, unsigned int type);
Andreas Gruenbacher0cf9d272010-12-07 10:43:29 +010089
Philipp Reisnerb411b362009-09-25 16:07:19 -070090static inline int
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +020091drbd_insert_fault(struct drbd_device *device, unsigned int type) {
Andreas Gruenbacher0cf9d272010-12-07 10:43:29 +010092#ifdef CONFIG_DRBD_FAULT_INJECTION
Roland Kammerer183ece32017-08-29 10:20:46 +020093 return drbd_fault_rate &&
94 (drbd_enable_faults & (1<<type)) &&
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +020095 _drbd_insert_fault(device, type);
Philipp Reisnerb411b362009-09-25 16:07:19 -070096#else
Andreas Gruenbacher0cf9d272010-12-07 10:43:29 +010097 return 0;
Philipp Reisnerb411b362009-09-25 16:07:19 -070098#endif
Andreas Gruenbacher0cf9d272010-12-07 10:43:29 +010099}
Philipp Reisnerb411b362009-09-25 16:07:19 -0700100
101/* integer division, round _UP_ to the next integer */
102#define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
103/* usual integer division */
104#define div_floor(A, B) ((A)/(B))
105
Philipp Reisnerb411b362009-09-25 16:07:19 -0700106extern struct ratelimit_state drbd_ratelimit_state;
Andreas Gruenbacher05a10ec2011-06-07 22:54:17 +0200107extern struct idr drbd_devices; /* RCU, updates: genl_lock() */
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +0200108extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700109
Andreas Gruenbacherd8763022011-01-26 17:39:41 +0100110extern const char *cmdname(enum drbd_packet cmd);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700111
112/* for sending/receiving the bitmap,
113 * possibly in some encoding scheme */
114struct bm_xfer_ctx {
115 /* "const"
116 * stores total bits and long words
117 * of the bitmap, so we don't need to
118 * call the accessor functions over and again. */
119 unsigned long bm_bits;
120 unsigned long bm_words;
121 /* during xfer, current position within the bitmap */
122 unsigned long bit_offset;
123 unsigned long word_offset;
124
125 /* statistics; index: (h->command == P_BITMAP) */
126 unsigned packets[2];
127 unsigned bytes[2];
128};
129
Andreas Gruenbacher5e54c2a2023-03-30 12:27:41 +0200130extern void INFO_bm_xfer_stats(struct drbd_peer_device *peer_device,
131 const char *direction, struct bm_xfer_ctx *c);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700132
133static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
134{
135 /* word_offset counts "native long words" (32 or 64 bit),
136 * aligned at 64 bit.
137 * Encoded packet may end at an unaligned bit offset.
138 * In case a fallback clear text packet is transmitted in
139 * between, we adjust this offset back to the last 64bit
140 * aligned "native long word", which makes coding and decoding
141 * the plain text bitmap much more convenient. */
142#if BITS_PER_LONG == 64
143 c->word_offset = c->bit_offset >> 6;
144#elif BITS_PER_LONG == 32
145 c->word_offset = c->bit_offset >> 5;
146 c->word_offset &= ~(1UL);
147#else
148# error "unsupported BITS_PER_LONG"
149#endif
150}
151
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +0200152extern unsigned int drbd_header_size(struct drbd_connection *connection);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700153
Philipp Reisnerb411b362009-09-25 16:07:19 -0700154/**********************************************************************/
155enum drbd_thread_state {
Andreas Gruenbachere77a0a52011-01-25 15:43:39 +0100156 NONE,
157 RUNNING,
158 EXITING,
159 RESTARTING
Philipp Reisnerb411b362009-09-25 16:07:19 -0700160};
161
162struct drbd_thread {
163 spinlock_t t_lock;
164 struct task_struct *task;
165 struct completion stop;
166 enum drbd_thread_state t_state;
167 int (*function) (struct drbd_thread *);
Andreas Gruenbacher2457b6d2011-07-21 13:45:21 +0200168 struct drbd_resource *resource;
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +0200169 struct drbd_connection *connection;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700170 int reset_cpu_mask;
Andreas Gruenbacherc60b0252011-08-10 15:05:02 +0200171 const char *name;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700172};
173
174static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
175{
176 /* THINK testing the t_state seems to be uncritical in all cases
177 * (but thread_{start,stop}), so we can read it *without* the lock.
178 * --lge */
179
180 smp_rmb();
181 return thi->t_state;
182}
183
Philipp Reisnerb411b362009-09-25 16:07:19 -0700184struct drbd_work {
185 struct list_head list;
Andreas Gruenbacher309a8342010-12-21 12:38:39 +0100186 int (*cb)(struct drbd_work *, int cancel);
Andreas Gruenbacher84b8c062011-07-28 15:27:51 +0200187};
188
189struct drbd_device_work {
190 struct drbd_work w;
191 struct drbd_device *device;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700192};
193
Andreas Gruenbacherace652a2011-01-03 17:09:58 +0100194#include "drbd_interval.h"
195
Andreas Gruenbacher54761692011-05-30 16:15:21 +0200196extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *);
Andreas Gruenbacher7be8da02011-02-22 02:15:32 +0100197
Andreas Gruenbacher28bc3b82014-08-14 18:33:30 +0200198extern void lock_all_resources(void);
199extern void unlock_all_resources(void);
200
Philipp Reisnerb411b362009-09-25 16:07:19 -0700201struct drbd_request {
202 struct drbd_work w;
Andreas Gruenbacher84b8c062011-07-28 15:27:51 +0200203 struct drbd_device *device;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700204
205 /* if local IO is not allowed, will be NULL.
206 * if local IO _is_ allowed, holds the locally submitted bio clone,
207 * or, after local IO completion, the ERR_PTR(error).
Andreas Gruenbacherfcefa622011-02-17 16:46:59 +0100208 * see drbd_request_endio(). */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700209 struct bio *private_bio;
210
Andreas Gruenbacherace652a2011-01-03 17:09:58 +0100211 struct drbd_interval i;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700212
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100213 /* epoch: used to check on "completion" whether this req was in
Philipp Reisnerb411b362009-09-25 16:07:19 -0700214 * the current epoch, and we therefore have to close it,
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100215 * causing a p_barrier packet to be send, starting a new epoch.
216 *
217 * This corresponds to "barrier" in struct p_barrier[_ack],
218 * and to "barrier_nr" in struct drbd_epoch (and various
219 * comments/function parameters/local variable names).
Philipp Reisnerb411b362009-09-25 16:07:19 -0700220 */
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100221 unsigned int epoch;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700222
Philipp Reisnerb411b362009-09-25 16:07:19 -0700223 struct list_head tl_requests; /* ring list in the transfer log */
224 struct bio *master_bio; /* master bio pointer */
Lars Ellenberge5f891b2013-11-22 12:32:01 +0100225
Lars Ellenberg844a6ae72013-11-22 12:52:03 +0100226 /* see struct drbd_device */
227 struct list_head req_pending_master_completion;
228 struct list_head req_pending_local;
229
Lars Ellenberge5f891b2013-11-22 12:32:01 +0100230 /* for generic IO accounting */
231 unsigned long start_jif;
232
233 /* for DRBD internal statistics */
234
235 /* Minimal set of time stamps to determine if we wait for activity log
236 * transactions, local disk or peer. 32 bit "jiffies" are good enough,
237 * we don't expect a DRBD request to be stalled for several month.
238 */
239
240 /* before actual request processing */
241 unsigned long in_actlog_jif;
242
243 /* local disk */
244 unsigned long pre_submit_jif;
245
246 /* per connection */
247 unsigned long pre_send_jif;
248 unsigned long acked_jif;
249 unsigned long net_done_jif;
250
251 /* Possibly even more detail to track each phase:
252 * master_completion_jif
253 * how long did it take to complete the master bio
254 * (application visible latency)
255 * allocated_jif
256 * how long the master bio was blocked until we finally allocated
257 * a tracking struct
258 * in_actlog_jif
259 * how long did we wait for activity log transactions
260 *
261 * net_queued_jif
262 * when did we finally queue it for sending
263 * pre_send_jif
264 * when did we start sending it
265 * post_send_jif
266 * how long did we block in the network stack trying to send it
267 * acked_jif
268 * when did we receive (or fake, in protocol A) a remote ACK
269 * net_done_jif
270 * when did we receive final acknowledgement (P_BARRIER_ACK),
271 * or decide, e.g. on connection loss, that we do no longer expect
272 * anything from this peer for this request.
273 *
274 * pre_submit_jif
275 * post_sub_jif
276 * when did we start submiting to the lower level device,
277 * and how long did we block in that submit function
278 * local_completion_jif
279 * how long did it take the lower level device to complete this request
280 */
281
Lars Ellenbergb4067772012-01-24 16:58:11 +0100282
283 /* once it hits 0, we may complete the master_bio */
284 atomic_t completion_ref;
285 /* once it hits 0, we may destroy this drbd_request object */
286 struct kref kref;
Lars Ellenberga0d856d2012-01-24 17:19:42 +0100287
288 unsigned rq_state; /* see comments above _req_mod() */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700289};
290
Philipp Reisnerb411b362009-09-25 16:07:19 -0700291struct drbd_epoch {
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +0200292 struct drbd_connection *connection;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700293 struct list_head list;
294 unsigned int barrier_nr;
295 atomic_t epoch_size; /* increased on every request added. */
296 atomic_t active; /* increased on every req. added, and dec on every finished. */
297 unsigned long flags;
298};
299
Rashika Kheriade0b2e62013-12-19 15:07:47 +0530300/* Prototype declaration of function defined in drbd_receiver.c */
301int drbdd_init(struct drbd_thread *);
302int drbd_asender(struct drbd_thread *);
303
Philipp Reisnerb411b362009-09-25 16:07:19 -0700304/* drbd_epoch flag bits */
305enum {
Philipp Reisnerb411b362009-09-25 16:07:19 -0700306 DE_HAVE_BARRIER_NUMBER,
Philipp Reisnerb411b362009-09-25 16:07:19 -0700307};
308
309enum epoch_event {
310 EV_PUT,
311 EV_GOT_BARRIER_NR,
Philipp Reisnerb411b362009-09-25 16:07:19 -0700312 EV_BECAME_LAST,
Philipp Reisnerb411b362009-09-25 16:07:19 -0700313 EV_CLEANUP = 32, /* used as flag */
314};
315
Philipp Reisnerb411b362009-09-25 16:07:19 -0700316struct digest_info {
317 int digest_size;
318 void *digest;
319};
320
Andreas Gruenbacherf6ffca92011-02-04 15:30:34 +0100321struct drbd_peer_request {
Andreas Gruenbachera8cd15b2011-08-25 15:49:40 +0200322 struct drbd_work w;
323 struct drbd_peer_device *peer_device;
Philipp Reisner85719572010-07-21 10:20:17 +0200324 struct drbd_epoch *epoch; /* for writes */
Lars Ellenberg45bb9122010-05-14 17:10:48 +0200325 struct page *pages;
Christoph Böhmwalderce668b62022-11-09 14:34:53 +0100326 blk_opf_t opf;
Lars Ellenberg45bb9122010-05-14 17:10:48 +0200327 atomic_t pending_bios;
Andreas Gruenbacher010f6e62011-01-14 20:59:35 +0100328 struct drbd_interval i;
Lars Ellenberg45bb9122010-05-14 17:10:48 +0200329 /* see comments on ee flag bits below */
330 unsigned long flags;
Lars Ellenberg21ae5d72014-05-05 23:42:24 +0200331 unsigned long submit_jif;
Philipp Reisner85719572010-07-21 10:20:17 +0200332 union {
333 u64 block_id;
334 struct digest_info *digest;
335 };
Lars Ellenberg45bb9122010-05-14 17:10:48 +0200336};
337
Christoph Böhmwalderce668b62022-11-09 14:34:53 +0100338/* Equivalent to bio_op and req_op. */
339#define peer_req_op(peer_req) \
340 ((peer_req)->opf & REQ_OP_MASK)
341
Lars Ellenberg45bb9122010-05-14 17:10:48 +0200342/* ee flag bits.
343 * While corresponding bios are in flight, the only modification will be
344 * set_bit WAS_ERROR, which has to be atomic.
345 * If no bios are in flight yet, or all have been completed,
346 * non-atomic modification to ee->flags is ok.
347 */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700348enum {
349 __EE_CALL_AL_COMPLETE_IO,
Philipp Reisnerb411b362009-09-25 16:07:19 -0700350 __EE_MAY_SET_IN_SYNC,
Lars Ellenberg45bb9122010-05-14 17:10:48 +0200351
Bart Van Assche93054552018-10-03 13:56:25 -0700352 /* is this a TRIM aka REQ_OP_DISCARD? */
Lars Ellenbergf31e5832018-12-20 17:23:42 +0100353 __EE_TRIM,
354 /* explicit zero-out requested, or
355 * our lower level cannot handle trim,
356 * and we want to fall back to zeroout instead */
357 __EE_ZEROOUT,
Lars Ellenberga0fb3c42014-04-28 18:43:23 +0200358
Lars Ellenberg45bb9122010-05-14 17:10:48 +0200359 /* In case a barrier failed,
360 * we need to resubmit without the barrier flag. */
361 __EE_RESUBMITTED,
362
Andreas Gruenbacher6c852be2011-02-04 15:38:52 +0100363 /* we may have several bios per peer request.
Lars Ellenberg45bb9122010-05-14 17:10:48 +0200364 * if any of those fail, we set this flag atomically
365 * from the endio callback */
366 __EE_WAS_ERROR,
Lars Ellenbergc36c3ce2010-08-11 20:42:55 +0200367
368 /* This ee has a pointer to a digest instead of a block id */
369 __EE_HAS_DIGEST,
Andreas Gruenbacher7be8da02011-02-22 02:15:32 +0100370
371 /* Conflicting local requests need to be restarted after this request */
372 __EE_RESTART_REQUESTS,
Philipp Reisner303d1442011-04-13 16:24:47 -0700373
374 /* The peer wants a write ACK for this (wire proto C) */
375 __EE_SEND_WRITE_ACK,
Philipp Reisner302bdea2011-04-21 11:36:49 +0200376
377 /* Is set when net_conf had two_primaries set while creating this peer_req */
378 __EE_IN_INTERVAL_TREE,
Lars Ellenberg21ae5d72014-05-05 23:42:24 +0200379
380 /* for debugfs: */
381 /* has this been submitted, or does it still wait for something else? */
382 __EE_SUBMITTED,
383
384 /* this is/was a write request */
385 __EE_WRITE,
386
Lars Ellenberg9104d312016-06-14 00:26:31 +0200387 /* this is/was a write same request */
388 __EE_WRITE_SAME,
389
Lars Ellenberg21ae5d72014-05-05 23:42:24 +0200390 /* this originates from application on peer
391 * (not some resync or verify or other DRBD internal request) */
392 __EE_APPLICATION,
Philipp Reisner700ca8c2016-06-14 00:26:13 +0200393
394 /* If it contains only 0 bytes, send back P_RS_DEALLOCATED */
395 __EE_RS_THIN_REQ,
Philipp Reisnerb411b362009-09-25 16:07:19 -0700396};
397#define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700398#define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
Lars Ellenbergf31e5832018-12-20 17:23:42 +0100399#define EE_TRIM (1<<__EE_TRIM)
400#define EE_ZEROOUT (1<<__EE_ZEROOUT)
Lars Ellenberga0fb3c42014-04-28 18:43:23 +0200401#define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
Lars Ellenberg45bb9122010-05-14 17:10:48 +0200402#define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
Lars Ellenbergc36c3ce2010-08-11 20:42:55 +0200403#define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
Andreas Gruenbacher7be8da02011-02-22 02:15:32 +0100404#define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
Philipp Reisner303d1442011-04-13 16:24:47 -0700405#define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK)
Philipp Reisner302bdea2011-04-21 11:36:49 +0200406#define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE)
Lars Ellenberg21ae5d72014-05-05 23:42:24 +0200407#define EE_SUBMITTED (1<<__EE_SUBMITTED)
408#define EE_WRITE (1<<__EE_WRITE)
Lars Ellenberg9104d312016-06-14 00:26:31 +0200409#define EE_WRITE_SAME (1<<__EE_WRITE_SAME)
Lars Ellenberg21ae5d72014-05-05 23:42:24 +0200410#define EE_APPLICATION (1<<__EE_APPLICATION)
Philipp Reisner700ca8c2016-06-14 00:26:13 +0200411#define EE_RS_THIN_REQ (1<<__EE_RS_THIN_REQ)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700412
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +0200413/* flag bits per device */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700414enum {
Philipp Reisnerb411b362009-09-25 16:07:19 -0700415 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
416 MD_DIRTY, /* current uuids and flags not yet on disk */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700417 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700418 CL_ST_CHG_SUCCESS,
419 CL_ST_CHG_FAIL,
420 CRASHED_PRIMARY, /* This node was a crashed primary.
421 * Gets cleared when the state.conn
422 * goes into C_CONNECTED state. */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700423 CONSIDER_RESYNC,
424
Philipp Reisnera8a4e512010-08-25 10:21:04 +0200425 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */
Lars Ellenberg5ab7d2c2014-01-27 15:58:22 +0100426
Philipp Reisnerb411b362009-09-25 16:07:19 -0700427 BITMAP_IO, /* suspend application io;
428 once no more io in flight, start bitmap io */
429 BITMAP_IO_QUEUED, /* Started bitmap IO */
Lars Ellenberga2a3c74f2012-09-22 12:26:57 +0200430 WAS_IO_ERROR, /* Local disk failed, returned IO error */
431 WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */
Lars Ellenberg383606e2012-06-14 14:21:32 +0200432 FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700433 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700434 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
435 * the peer, if it changed there as well. */
Philipp Reisner43a51822010-06-11 11:26:34 +0200436 NEW_CUR_UUID, /* Create new current UUID when thawing IO */
Philipp Reisner07782862010-08-31 12:00:50 +0200437 AL_SUSPENDED, /* Activity logging is currently suspended. */
Philipp Reisner370a43e2011-01-14 16:03:11 +0100438 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */
Philipp Reisnere64a3292011-02-05 17:34:11 +0100439 B_RS_H_DONE, /* Before resync handler done (already executed) */
Philipp Reisner08b165b2011-09-05 16:22:33 +0200440 DISCARD_MY_DATA, /* discard_my_data flag per volume */
Philipp Reisner380207d2011-11-11 12:31:20 +0100441 READ_BALANCE_RR,
Lars Ellenberge334f552014-02-11 09:30:49 +0100442
Lars Ellenbergf4188152014-05-05 23:05:47 +0200443 FLUSH_PENDING, /* if set, device->flush_jif is when we submitted that flush
444 * from drbd_flush_after_epoch() */
445
Lars Ellenberge334f552014-02-11 09:30:49 +0100446 /* cleared only after backing device related structures have been destroyed. */
447 GOING_DISKLESS, /* Disk is being detached, because of io-error, or admin request. */
448
449 /* to be used in drbd_device_post_work() */
450 GO_DISKLESS, /* tell worker to schedule cleanup before detach */
451 DESTROY_DISK, /* tell worker to close backing devices and destroy related structures. */
Lars Ellenbergac0acb92014-02-11 09:47:58 +0100452 MD_SYNC, /* tell worker to call drbd_md_sync() */
453 RS_START, /* tell worker to start resync/OV */
Lars Ellenberge334f552014-02-11 09:30:49 +0100454 RS_PROGRESS, /* tell worker that resync made significant progress */
455 RS_DONE, /* tell worker that resync is done */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700456};
457
Andreas Gruenbacher54761692011-05-30 16:15:21 +0200458struct drbd_bitmap; /* opaque for drbd_device */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700459
Lars Ellenberg20ceb2b22011-01-21 10:56:44 +0100460/* definition of bits in bm_flags to be used in drbd_bm_lock
461 * and drbd_bitmap_io and friends. */
462enum bm_flag {
Lars Ellenberg20ceb2b22011-01-21 10:56:44 +0100463 /* currently locked for bulk operation */
Lars Ellenberg0e8488a2012-04-25 23:06:45 +0200464 BM_LOCKED_MASK = 0xf,
Lars Ellenberg20ceb2b22011-01-21 10:56:44 +0100465
466 /* in detail, that is: */
467 BM_DONT_CLEAR = 0x1,
468 BM_DONT_SET = 0x2,
469 BM_DONT_TEST = 0x4,
470
Lars Ellenberg0e8488a2012-04-25 23:06:45 +0200471 /* so we can mark it locked for bulk operation,
472 * and still allow all non-bulk operations */
473 BM_IS_LOCKED = 0x8,
474
Lars Ellenberg20ceb2b22011-01-21 10:56:44 +0100475 /* (test bit, count bit) allowed (common case) */
Lars Ellenberg0e8488a2012-04-25 23:06:45 +0200476 BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED,
Lars Ellenberg20ceb2b22011-01-21 10:56:44 +0100477
478 /* testing bits, as well as setting new bits allowed, but clearing bits
479 * would be unexpected. Used during bitmap receive. Setting new bits
480 * requires sending of "out-of-sync" information, though. */
Lars Ellenberg0e8488a2012-04-25 23:06:45 +0200481 BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED,
Lars Ellenberg20ceb2b22011-01-21 10:56:44 +0100482
Lars Ellenberg0e8488a2012-04-25 23:06:45 +0200483 /* for drbd_bm_write_copy_pages, everything is allowed,
484 * only concurrent bulk operations are locked out. */
485 BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED,
Lars Ellenberg20ceb2b22011-01-21 10:56:44 +0100486};
487
Philipp Reisnerb411b362009-09-25 16:07:19 -0700488struct drbd_work_queue {
489 struct list_head q;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700490 spinlock_t q_lock; /* to protect the list. */
Lars Ellenberg8c0785a2011-10-19 11:50:57 +0200491 wait_queue_head_t q_wait;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700492};
493
494struct drbd_socket {
Philipp Reisnerb411b362009-09-25 16:07:19 -0700495 struct mutex mutex;
496 struct socket *socket;
497 /* this way we get our
498 * send/receive buffers off the stack */
Andreas Gruenbacher5a87d922011-03-24 21:17:52 +0100499 void *sbuf;
Andreas Gruenbachere6ef8a52011-03-24 18:07:54 +0100500 void *rbuf;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700501};
502
503struct drbd_md {
504 u64 md_offset; /* sector offset to 'super' block */
505
506 u64 la_size_sect; /* last agreed size, unit sectors */
Philipp Reisner9f2247b2012-08-16 14:25:58 +0200507 spinlock_t uuid_lock;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700508 u64 uuid[UI_SIZE];
509 u64 device_uuid;
510 u32 flags;
511 u32 md_size_sect;
512
Lars Ellenbergae8bf312013-03-19 18:16:43 +0100513 s32 al_offset; /* signed relative sector offset to activity log */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700514 s32 bm_offset; /* signed relative sector offset to bitmap */
Lars Ellenberg3a4d4eb2013-03-19 18:16:44 +0100515
516 /* cached value of bdev->disk_conf->meta_dev_idx (see below) */
517 s32 meta_dev_idx;
518
519 /* see al_tr_number_to_on_disk_sector() */
520 u32 al_stripes;
521 u32 al_stripe_size_4k;
522 u32 al_size_4k; /* cached product of the above */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700523};
524
Philipp Reisnerb411b362009-09-25 16:07:19 -0700525struct drbd_backing_dev {
526 struct block_device *backing_bdev;
527 struct block_device *md_bdev;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700528 struct drbd_md md;
Andreas Gruenbacher05008132011-07-07 14:19:42 +0200529 struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700530 sector_t known_size; /* last known size of that backing device */
531};
532
533struct drbd_md_io {
Lars Ellenberge37d2432014-04-01 23:53:30 +0200534 struct page *page;
535 unsigned long start_jif; /* last call to drbd_md_get_buffer */
536 unsigned long submit_jif; /* last _drbd_md_sync_page_io() submit */
537 const char *current_use;
538 atomic_t in_use;
Philipp Reisner0c464422011-06-26 22:26:31 +0200539 unsigned int done;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700540 int error;
541};
542
543struct bm_io_work {
544 struct drbd_work w;
Andreas Gruenbacher8164dd62023-03-30 12:27:40 +0200545 struct drbd_peer_device *peer_device;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700546 char *why;
Lars Ellenberg20ceb2b22011-01-21 10:56:44 +0100547 enum bm_flag flags;
Andreas Gruenbacher8164dd62023-03-30 12:27:40 +0200548 int (*io_fn)(struct drbd_device *device, struct drbd_peer_device *peer_device);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +0200549 void (*done)(struct drbd_device *device, int rv);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700550};
551
Philipp Reisner778f2712010-07-06 11:14:00 +0200552struct fifo_buffer {
Philipp Reisner778f2712010-07-06 11:14:00 +0200553 unsigned int head_index;
554 unsigned int size;
Philipp Reisner9958c852011-05-03 16:19:31 +0200555 int total; /* sum of all values */
Gustavo A. R. Silva1060bfc2020-05-28 09:35:11 -0500556 int values[];
Philipp Reisner778f2712010-07-06 11:14:00 +0200557};
Stephen Kitt6a365872020-01-24 21:03:07 +0100558extern struct fifo_buffer *fifo_alloc(unsigned int fifo_size);
Philipp Reisner778f2712010-07-06 11:14:00 +0200559
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +0200560/* flag bits per connection */
Philipp Reisner01a311a2011-02-07 14:30:33 +0100561enum {
562 NET_CONGESTED, /* The data socket is congested */
Lars Ellenberg427c0432012-08-01 12:43:01 +0200563 RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */
Philipp Reisner668700b2015-03-16 16:08:29 +0100564 SEND_PING,
Philipp Reisner2a67d8b2011-02-09 14:10:32 +0100565 GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */
Philipp Reisner4d0fc3f2012-01-20 13:52:27 +0100566 CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */
Philipp Reisnerfc3b10a2011-02-15 11:07:59 +0100567 CONN_WD_ST_CHG_OKAY,
568 CONN_WD_ST_CHG_FAIL,
Philipp Reisner8169e412011-03-15 18:40:27 +0100569 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
Philipp Reisner6936fcb2011-11-10 18:45:36 +0100570 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */
Philipp Reisnera1096a62012-04-06 12:07:34 +0200571 STATE_SENT, /* Do not change state/UUIDs while this is set */
Lars Ellenberg6f3465e2012-07-30 09:08:25 +0200572 CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
573 * pending, from drbd worker context.
Lars Ellenberg6f3465e2012-07-30 09:08:25 +0200574 */
Philipp Reisnerb66623e2012-08-08 21:19:09 +0200575 DISCONNECT_SENT,
Lars Ellenberge334f552014-02-11 09:30:49 +0100576
577 DEVICE_WORK_PENDING, /* tell worker that some device has pending work */
Philipp Reisner01a311a2011-02-07 14:30:33 +0100578};
579
Andreas Gruenbachera2972842014-07-31 17:41:33 +0200580enum which_state { NOW, OLD = NOW, NEW };
581
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +0200582struct drbd_resource {
583 char *name;
Lars Ellenberg4d3d5aa2014-05-02 13:19:51 +0200584#ifdef CONFIG_DEBUG_FS
585 struct dentry *debugfs_res;
586 struct dentry *debugfs_res_volumes;
587 struct dentry *debugfs_res_connections;
588 struct dentry *debugfs_res_in_flight_summary;
589#endif
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +0200590 struct kref kref;
Andreas Gruenbacher803ea132011-06-09 01:40:48 +0200591 struct idr devices; /* volume number to device mapping */
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +0200592 struct list_head connections;
593 struct list_head resources;
Andreas Gruenbachereb6bea62011-06-21 16:11:28 +0200594 struct res_opts res_opts;
Andreas Gruenbacher05008132011-07-07 14:19:42 +0200595 struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */
Lars Ellenberg9e276872014-04-28 18:43:22 +0200596 struct mutex adm_mutex; /* mutex to serialize administrative requests */
Andreas Gruenbacher05008132011-07-07 14:19:42 +0200597 spinlock_t req_lock;
Andreas Gruenbacher6bbf53c2011-07-08 01:19:44 +0200598
599 unsigned susp:1; /* IO suspended by user */
600 unsigned susp_nod:1; /* IO suspended because no data */
601 unsigned susp_fen:1; /* IO suspended because fence peer handler runs */
Andreas Gruenbacher625a6ba2011-07-22 14:29:02 +0200602
Philipp Reisnere9526582013-11-22 15:53:41 +0100603 enum write_ordering_e write_ordering;
604
Andreas Gruenbacher625a6ba2011-07-22 14:29:02 +0200605 cpumask_var_t cpu_mask;
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +0200606};
607
Lars Ellenberg944410e2014-05-06 15:02:05 +0200608struct drbd_thread_timing_details
609{
610 unsigned long start_jif;
611 void *cb_addr;
612 const char *caller_fn;
613 unsigned int line;
614 unsigned int cb_nr;
615};
616
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +0200617struct drbd_connection {
618 struct list_head connections;
619 struct drbd_resource *resource;
Lars Ellenberg4d3d5aa2014-05-02 13:19:51 +0200620#ifdef CONFIG_DEBUG_FS
621 struct dentry *debugfs_conn;
622 struct dentry *debugfs_conn_callback_history;
623 struct dentry *debugfs_conn_oldest_requests;
624#endif
Philipp Reisner9dc9fbb2011-04-22 15:23:32 +0200625 struct kref kref;
Andreas Gruenbacherc06ece62011-06-21 17:23:59 +0200626 struct idr peer_devices; /* volume number to peer device mapping */
Philipp Reisner8410da8f02011-02-11 20:11:10 +0100627 enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
628 struct mutex cstate_mutex; /* Protects graceful disconnects */
Philipp Reisner28e448b2013-06-25 16:50:06 +0200629 unsigned int connect_cnt; /* Inc each time a connection is established */
Philipp Reisner21114382011-01-19 12:26:59 +0100630
Philipp Reisner062e8792011-02-08 11:09:18 +0100631 unsigned long flags;
Philipp Reisner44ed1672011-04-19 17:10:19 +0200632 struct net_conf *net_conf; /* content protected by rcu */
Philipp Reisner91fd4da2011-04-20 17:47:29 +0200633 wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */
Philipp Reisnere42325a2011-01-19 13:55:45 +0100634
Andreas Gruenbacher089c0752011-06-14 18:28:09 +0200635 struct sockaddr_storage my_addr;
636 int my_addr_len;
637 struct sockaddr_storage peer_addr;
638 int peer_addr_len;
639
Philipp Reisnere42325a2011-01-19 13:55:45 +0100640 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
641 struct drbd_socket meta; /* ping/ack (metadata) packets */
Philipp Reisner31890f42011-01-19 14:12:51 +0100642 int agreed_pro_version; /* actually used protocol version */
Lars Ellenberg20c68fd2014-04-28 18:43:25 +0200643 u32 agreed_features;
Philipp Reisner31890f42011-01-19 14:12:51 +0100644 unsigned long last_received; /* in jiffies, either socket */
645 unsigned int ko_count;
Philipp Reisnere6b3ea82011-01-19 14:02:01 +0100646
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100647 struct list_head transfer_log; /* all requests not yet fully processed */
Philipp Reisner87eeee42011-01-19 14:16:30 +0100648
Herbert Xu9534d672016-01-24 21:19:21 +0800649 struct crypto_shash *cram_hmac_tfm;
Kees Cook3d0e6372018-08-06 16:32:16 -0700650 struct crypto_shash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */
651 struct crypto_shash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */
652 struct crypto_shash *csums_tfm;
653 struct crypto_shash *verify_tfm;
Philipp Reisnera0638452011-01-19 14:31:32 +0100654 void *int_dig_in;
655 void *int_dig_vv;
656
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100657 /* receiver side */
Philipp Reisner12038a32011-11-09 19:18:00 +0100658 struct drbd_epoch *current_epoch;
659 spinlock_t epoch_lock;
660 unsigned int epochs;
Lars Ellenbergb379c412011-11-17 11:49:46 +0100661 atomic_t current_tle_nr; /* transfer log epoch number */
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100662 unsigned current_tle_writes; /* writes seen within this tl epoch */
Philipp Reisner4b0007c2011-11-09 20:12:34 +0100663
Lars Ellenberg07be15b2012-05-07 11:53:08 +0200664 unsigned long last_reconnect_jif;
Lars Ellenbergc51a0ef2017-08-29 10:20:32 +0200665 /* empty member on older kernels without blk_start_plug() */
666 struct blk_plug receiver_plug;
Philipp Reisnere6b3ea82011-01-19 14:02:01 +0100667 struct drbd_thread receiver;
668 struct drbd_thread worker;
Philipp Reisner1c03e522015-03-16 15:01:00 +0100669 struct drbd_thread ack_receiver;
Philipp Reisner668700b2015-03-16 16:08:29 +0100670 struct workqueue_struct *ack_sender;
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100671
Lars Ellenberg7753a4c12013-11-22 13:00:12 +0100672 /* cached pointers,
673 * so we can look up the oldest pending requests more quickly.
674 * protected by resource->req_lock */
675 struct drbd_request *req_next; /* DRBD 9: todo.req_next */
676 struct drbd_request *req_ack_pending;
677 struct drbd_request *req_not_net_done;
678
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100679 /* sender side */
Lars Ellenbergd5b27b02011-11-14 15:42:37 +0100680 struct drbd_work_queue sender_work;
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100681
Lars Ellenberg944410e2014-05-06 15:02:05 +0200682#define DRBD_THREAD_DETAILS_HIST 16
683 unsigned int w_cb_nr; /* keeps counting up */
684 unsigned int r_cb_nr; /* keeps counting up */
685 struct drbd_thread_timing_details w_timing_details[DRBD_THREAD_DETAILS_HIST];
686 struct drbd_thread_timing_details r_timing_details[DRBD_THREAD_DETAILS_HIST];
687
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100688 struct {
Lars Ellenberg84d34f22015-02-19 13:54:11 +0100689 unsigned long last_sent_barrier_jif;
690
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100691 /* whether this sender thread
692 * has processed a single write yet. */
693 bool seen_any_write_yet;
694
695 /* Which barrier number to send with the next P_BARRIER */
696 int current_epoch_nr;
697
698 /* how many write requests have been sent
699 * with req->epoch == current_epoch_nr.
700 * If none, no P_BARRIER will be sent. */
701 unsigned current_epoch_writes;
702 } send;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700703};
704
Andreas Gruenbachera2972842014-07-31 17:41:33 +0200705static inline bool has_net_conf(struct drbd_connection *connection)
706{
707 bool has_net_conf;
708
709 rcu_read_lock();
710 has_net_conf = rcu_dereference(connection->net_conf);
711 rcu_read_unlock();
712
713 return has_net_conf;
714}
715
Lars Ellenberg944410e2014-05-06 15:02:05 +0200716void __update_timing_details(
717 struct drbd_thread_timing_details *tdp,
718 unsigned int *cb_nr,
719 void *cb,
720 const char *fn, const unsigned int line);
721
722#define update_worker_timing_details(c, cb) \
723 __update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ )
724#define update_receiver_timing_details(c, cb) \
725 __update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ )
726
Lars Ellenberg113fef92013-03-22 18:14:40 -0600727struct submit_worker {
728 struct workqueue_struct *wq;
729 struct work_struct worker;
730
Lars Ellenberg844a6ae72013-11-22 12:52:03 +0100731 /* protected by ..->resource->req_lock */
Lars Ellenberg113fef92013-03-22 18:14:40 -0600732 struct list_head writes;
733};
734
Andreas Gruenbachera6b32bc32011-05-31 14:33:49 +0200735struct drbd_peer_device {
736 struct list_head peer_devices;
737 struct drbd_device *device;
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +0200738 struct drbd_connection *connection;
Philipp Reisner668700b2015-03-16 16:08:29 +0100739 struct work_struct send_acks_work;
Lars Ellenberg4d3d5aa2014-05-02 13:19:51 +0200740#ifdef CONFIG_DEBUG_FS
741 struct dentry *debugfs_peer_dev;
742#endif
Andreas Gruenbachera6b32bc32011-05-31 14:33:49 +0200743};
744
745struct drbd_device {
Andreas Gruenbacherd8628a82011-06-09 01:38:00 +0200746 struct drbd_resource *resource;
Andreas Gruenbachera6b32bc32011-05-31 14:33:49 +0200747 struct list_head peer_devices;
Lars Ellenberg4ce49262014-05-06 00:44:59 +0200748 struct list_head pending_bitmap_io;
Lars Ellenberg4d3d5aa2014-05-02 13:19:51 +0200749
750 unsigned long flush_jif;
751#ifdef CONFIG_DEBUG_FS
752 struct dentry *debugfs_minor;
753 struct dentry *debugfs_vol;
754 struct dentry *debugfs_vol_oldest_requests;
755 struct dentry *debugfs_vol_act_log_extents;
756 struct dentry *debugfs_vol_resync_extents;
757 struct dentry *debugfs_vol_data_gen_id;
Lars Ellenbergf5ec0172015-03-18 17:19:10 +0100758 struct dentry *debugfs_vol_ed_gen_id;
Lars Ellenberg4d3d5aa2014-05-02 13:19:51 +0200759#endif
760
761 unsigned int vnr; /* volume number within the connection */
762 unsigned int minor; /* device minor number */
763
Philipp Reisner81fa2e62011-05-04 15:10:30 +0200764 struct kref kref;
Philipp Reisner21114382011-01-19 12:26:59 +0100765
Philipp Reisnerb411b362009-09-25 16:07:19 -0700766 /* things that are stored as / read from meta data on disk */
767 unsigned long flags;
768
769 /* configured by drbdsetup */
Christoph Böhmwalder9cf766a2023-01-13 13:35:35 +0100770 struct drbd_backing_dev *ldev;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700771
772 sector_t p_size; /* partner's disk size */
773 struct request_queue *rq_queue;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700774 struct gendisk *vdisk;
775
Lars Ellenberg07be15b2012-05-07 11:53:08 +0200776 unsigned long last_reattach_jif;
Andreas Gruenbacher84b8c062011-07-28 15:27:51 +0200777 struct drbd_work resync_work;
778 struct drbd_work unplug_work;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700779 struct timer_list resync_timer;
780 struct timer_list md_sync_timer;
Philipp Reisner370a43e2011-01-14 16:03:11 +0100781 struct timer_list start_resync_timer;
Philipp Reisner7fde2be2011-03-01 11:08:28 +0100782 struct timer_list request_timer;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700783
784 /* Used after attach while negotiating new disk state. */
785 union drbd_state new_state_tmp;
786
Philipp Reisnerda9fbc22011-03-29 10:52:01 +0200787 union drbd_dev_state state;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700788 wait_queue_head_t misc_wait;
789 wait_queue_head_t state_wait; /* upon each state change. */
790 unsigned int send_cnt;
791 unsigned int recv_cnt;
792 unsigned int read_cnt;
793 unsigned int writ_cnt;
794 unsigned int al_writ_cnt;
795 unsigned int bm_writ_cnt;
796 atomic_t ap_bio_cnt; /* Requests we need to complete */
Lars Ellenbergad3fee72013-12-20 11:22:13 +0100797 atomic_t ap_actlog_cnt; /* Requests waiting for activity log */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700798 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
799 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
Philipp Reisnerd942ae42011-05-31 13:07:24 +0200800 atomic_t unacked_cnt; /* Need to send replies for */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700801 atomic_t local_cnt; /* Waiting for local completion */
Philipp Reisner7dbb4382013-02-28 10:30:19 +0100802 atomic_t suspend_cnt;
Philipp Reisnerb2fb6dbe2011-01-19 13:48:44 +0100803
Andreas Gruenbacherdac13892011-01-21 17:18:39 +0100804 /* Interval tree of pending local requests */
805 struct rb_root read_requests;
Andreas Gruenbacherde696712011-01-20 15:00:24 +0100806 struct rb_root write_requests;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700807
Lars Ellenberg844a6ae72013-11-22 12:52:03 +0100808 /* for statistics and timeouts */
809 /* [0] read, [1] write */
810 struct list_head pending_master_completion[2];
811 struct list_head pending_completion[2];
812
Lars Ellenbergaaaba342014-03-18 12:30:09 +0100813 /* use checksums for *this* resync */
814 bool use_csums;
Lars Ellenberg4b0715f2010-12-14 15:13:04 +0100815 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700816 unsigned long rs_total;
Lars Ellenberg4b0715f2010-12-14 15:13:04 +0100817 /* number of resync blocks that failed in this run */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700818 unsigned long rs_failed;
819 /* Syncer's start time [unit jiffies] */
820 unsigned long rs_start;
821 /* cumulated time in PausedSyncX state [unit jiffies] */
822 unsigned long rs_paused;
Lars Ellenberg1d7734a2010-08-11 21:21:50 +0200823 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700824 unsigned long rs_same_csum;
Lars Ellenberg1d7734a2010-08-11 21:21:50 +0200825#define DRBD_SYNC_MARKS 8
826#define DRBD_SYNC_MARK_STEP (3*HZ)
827 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
828 unsigned long rs_mark_left[DRBD_SYNC_MARKS];
829 /* marks's time [unit jiffies] */
830 unsigned long rs_mark_time[DRBD_SYNC_MARKS];
831 /* current index into rs_mark_{left,time} */
832 int rs_last_mark;
Philipp Reisner328e0f12012-10-19 14:37:47 +0200833 unsigned long rs_last_bcast; /* [unit jiffies] */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700834
835 /* where does the admin want us to start? (sector) */
836 sector_t ov_start_sector;
Lars Ellenberg02b91b52012-06-28 18:26:52 +0200837 sector_t ov_stop_sector;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700838 /* where are we now? (sector) */
839 sector_t ov_position;
840 /* Start sector of out of sync range (to merge printk reporting). */
841 sector_t ov_last_oos_start;
842 /* size of out-of-sync range in sectors. */
843 sector_t ov_last_oos_size;
844 unsigned long ov_left; /* in bits */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700845
Philipp Reisnerb411b362009-09-25 16:07:19 -0700846 struct drbd_bitmap *bitmap;
847 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
848
849 /* Used to track operations of resync... */
850 struct lru_cache *resync;
851 /* Number of locked elements in resync LRU */
852 unsigned int resync_locked;
853 /* resync extent number waiting for application requests */
854 unsigned int resync_wenr;
855
856 int open_cnt;
857 u64 *p_uuid;
Philipp Reisner4b0007c2011-11-09 20:12:34 +0100858
Philipp Reisner85719572010-07-21 10:20:17 +0200859 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
860 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
Andreas Gruenbacher18b75d72011-02-04 15:36:22 +0100861 struct list_head done_ee; /* need to send P_WRITE_ACK */
862 struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700863 struct list_head net_ee; /* zero-copy network send in progress */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700864
865 int next_barrier_nr;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700866 struct list_head resync_reads;
Lars Ellenberg435f0742010-09-06 12:30:25 +0200867 atomic_t pp_in_use; /* allocated from page pool */
868 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700869 wait_queue_head_t ee_wait;
Philipp Reisnercc94c652011-06-26 11:20:27 +0200870 struct drbd_md_io md_io;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700871 spinlock_t al_lock;
872 wait_queue_head_t al_wait;
873 struct lru_cache *act_log; /* activity log */
874 unsigned int al_tr_number;
875 int al_tr_cycle;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700876 wait_queue_head_t seq_wait;
877 atomic_t packet_seq;
878 unsigned int peer_seq;
879 spinlock_t peer_seq_lock;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700880 unsigned long comm_bm_set; /* communicated number of set bits. */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700881 struct bm_io_work bm_io_work;
882 u64 ed_uuid; /* UUID of the exposed data */
Philipp Reisner8410da8f02011-02-11 20:11:10 +0100883 struct mutex own_state_mutex;
Andreas Gruenbachera6b32bc32011-05-31 14:33:49 +0200884 struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700885 char congestion_reason; /* Why we where congested... */
Lars Ellenberg1d7734a2010-08-11 21:21:50 +0200886 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
887 atomic_t rs_sect_ev; /* for submitted resync data rate, both */
888 int rs_last_sect_ev; /* counter to compare with */
889 int rs_last_events; /* counter of read or write "events" (unit sectors)
890 * on the lower level device when we last looked. */
891 int c_sync_rate; /* current resync rate after syncer throttle magic */
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +0200892 struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */
Philipp Reisner778f2712010-07-06 11:14:00 +0200893 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
Philipp Reisner759fbdf2010-10-26 16:02:27 +0200894 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
Lars Ellenbergdb141b22012-06-25 19:15:58 +0200895 unsigned int peer_max_bio_size;
896 unsigned int local_max_bio_size;
Lars Ellenberg113fef92013-03-22 18:14:40 -0600897
898 /* any requests that would block in drbd_make_request()
899 * are deferred to this single-threaded work queue */
900 struct submit_worker submit;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700901};
902
Lars Ellenberg4ce49262014-05-06 00:44:59 +0200903struct drbd_bm_aio_ctx {
904 struct drbd_device *device;
905 struct list_head list; /* on device->pending_bitmap_io */;
906 unsigned long start_jif;
907 atomic_t in_flight;
908 unsigned int done;
909 unsigned flags;
910#define BM_AIO_COPY_PAGES 1
911#define BM_AIO_WRITE_HINTED 2
912#define BM_AIO_WRITE_ALL_PAGES 4
913#define BM_AIO_READ 8
914 int error;
915 struct kref kref;
916};
917
Lars Ellenberga910b122014-04-28 18:43:21 +0200918struct drbd_config_context {
919 /* assigned from drbd_genlmsghdr */
920 unsigned int minor;
921 /* assigned from request attributes, if present */
922 unsigned int volume;
923#define VOLUME_UNSPECIFIED (-1U)
924 /* pointer into the request skb,
925 * limited lifetime! */
926 char *resource_name;
927 struct nlattr *my_addr;
928 struct nlattr *peer_addr;
929
930 /* reply buffer */
931 struct sk_buff *reply_skb;
932 /* pointer into reply buffer */
933 struct drbd_genlmsghdr *reply_dh;
934 /* resolved from attributes, if possible */
935 struct drbd_device *device;
936 struct drbd_resource *resource;
937 struct drbd_connection *connection;
938};
939
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +0200940static inline struct drbd_device *minor_to_device(unsigned int minor)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700941{
Andreas Gruenbacher05a10ec2011-06-07 22:54:17 +0200942 return (struct drbd_device *)idr_find(&drbd_devices, minor);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700943}
944
Andreas Gruenbachera6b32bc32011-05-31 14:33:49 +0200945static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device)
946{
Lars Ellenbergec4a34072014-04-28 18:43:35 +0200947 return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices);
Andreas Gruenbachera6b32bc32011-05-31 14:33:49 +0200948}
949
Andreas Gruenbachera2972842014-07-31 17:41:33 +0200950static inline struct drbd_peer_device *
951conn_peer_device(struct drbd_connection *connection, int volume_number)
952{
953 return idr_find(&connection->peer_devices, volume_number);
954}
955
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +0200956#define for_each_resource(resource, _resources) \
957 list_for_each_entry(resource, _resources, resources)
958
959#define for_each_resource_rcu(resource, _resources) \
960 list_for_each_entry_rcu(resource, _resources, resources)
961
962#define for_each_resource_safe(resource, tmp, _resources) \
963 list_for_each_entry_safe(resource, tmp, _resources, resources)
964
965#define for_each_connection(connection, resource) \
966 list_for_each_entry(connection, &resource->connections, connections)
967
968#define for_each_connection_rcu(connection, resource) \
969 list_for_each_entry_rcu(connection, &resource->connections, connections)
970
971#define for_each_connection_safe(connection, tmp, resource) \
972 list_for_each_entry_safe(connection, tmp, &resource->connections, connections)
973
Andreas Gruenbachera6b32bc32011-05-31 14:33:49 +0200974#define for_each_peer_device(peer_device, device) \
975 list_for_each_entry(peer_device, &device->peer_devices, peer_devices)
976
977#define for_each_peer_device_rcu(peer_device, device) \
978 list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices)
979
980#define for_each_peer_device_safe(peer_device, tmp, device) \
981 list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices)
982
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +0200983static inline unsigned int device_to_minor(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700984{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +0200985 return device->minor;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700986}
987
Philipp Reisnerb411b362009-09-25 16:07:19 -0700988/*
989 * function declarations
990 *************************/
991
992/* drbd_main.c */
993
Philipp Reisnere89b5912010-03-24 17:11:33 +0100994enum dds_flags {
995 DDSF_FORCED = 1,
996 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
997};
998
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +0200999extern void drbd_init_set_defaults(struct drbd_device *device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001000extern int drbd_thread_start(struct drbd_thread *thi);
1001extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1002#ifdef CONFIG_SMP
Philipp Reisner80822282011-02-08 12:46:30 +01001003extern void drbd_thread_current_set_cpu(struct drbd_thread *thi);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001004#else
1005#define drbd_thread_current_set_cpu(A) ({})
Philipp Reisnerb411b362009-09-25 16:07:19 -07001006#endif
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001007extern void tl_release(struct drbd_connection *, unsigned int barrier_nr,
Philipp Reisnerb411b362009-09-25 16:07:19 -07001008 unsigned int set_size);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001009extern void tl_clear(struct drbd_connection *);
1010extern void drbd_free_sock(struct drbd_connection *connection);
1011extern int drbd_send(struct drbd_connection *connection, struct socket *sock,
Philipp Reisnerbedbd2a2011-02-07 15:08:48 +01001012 void *buf, size_t size, unsigned msg_flags);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001013extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t,
Andreas Gruenbacherfb708e42010-12-15 17:04:36 +01001014 unsigned);
1015
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001016extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd);
1017extern int drbd_send_protocol(struct drbd_connection *connection);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001018extern int drbd_send_uuids(struct drbd_peer_device *);
1019extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *);
1020extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *);
1021extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags);
1022extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s);
1023extern int drbd_send_current_state(struct drbd_peer_device *);
1024extern int drbd_send_sync_param(struct drbd_peer_device *);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001025extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr,
Andreas Gruenbacherd4e67d72011-03-16 01:25:28 +01001026 u32 set_size);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001027extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet,
Andreas Gruenbacherf6ffca92011-02-04 15:30:34 +01001028 struct drbd_peer_request *);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001029extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet,
Andreas Gruenbachera9a99942011-03-16 01:30:14 +01001030 struct p_block_req *rp);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001031extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet,
Andreas Gruenbachera9a99942011-03-16 01:30:14 +01001032 struct p_data *dp, int data_size);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001033extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet,
Philipp Reisnerb411b362009-09-25 16:07:19 -07001034 sector_t sector, int blksize, u64 block_id);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001035extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *);
1036extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet,
Andreas Gruenbacherf6ffca92011-02-04 15:30:34 +01001037 struct drbd_peer_request *);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001038extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req);
1039extern int drbd_send_drequest(struct drbd_peer_device *, int cmd,
Philipp Reisnerb411b362009-09-25 16:07:19 -07001040 sector_t sector, int size, u64 block_id);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001041extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector,
Andreas Gruenbacherd8763022011-01-26 17:39:41 +01001042 int size, void *digest, int digest_size,
1043 enum drbd_packet cmd);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001044extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001045
Andreas Gruenbacher8164dd62023-03-30 12:27:40 +02001046extern int drbd_send_bitmap(struct drbd_device *device, struct drbd_peer_device *peer_device);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001047extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001048extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode);
Philipp Reisner700ca8c2016-06-14 00:26:13 +02001049extern int drbd_send_rs_deallocated(struct drbd_peer_device *, struct drbd_peer_request *);
Lars Ellenberg63a7c8a2015-03-26 20:53:55 +01001050extern void drbd_backing_dev_free(struct drbd_device *device, struct drbd_backing_dev *ldev);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001051extern void drbd_device_cleanup(struct drbd_device *device);
Lars Ellenbergc51a0ef2017-08-29 10:20:32 +02001052extern void drbd_print_uuids(struct drbd_device *device, const char *text);
1053extern void drbd_queue_unplug(struct drbd_device *device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001054
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001055extern void conn_md_sync(struct drbd_connection *connection);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001056extern void drbd_md_write(struct drbd_device *device, void *buffer);
1057extern void drbd_md_sync(struct drbd_device *device);
1058extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev);
1059extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1060extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1061extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local);
1062extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local);
1063extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local);
1064extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1065extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local);
1066extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001067extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001068extern void drbd_md_mark_dirty(struct drbd_device *device);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001069extern void drbd_queue_bitmap_io(struct drbd_device *device,
Andreas Gruenbacher8164dd62023-03-30 12:27:40 +02001070 int (*io_fn)(struct drbd_device *, struct drbd_peer_device *),
Andreas Gruenbacher54761692011-05-30 16:15:21 +02001071 void (*done)(struct drbd_device *, int),
Andreas Gruenbacher8164dd62023-03-30 12:27:40 +02001072 char *why, enum bm_flag flags,
1073 struct drbd_peer_device *peer_device);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001074extern int drbd_bitmap_io(struct drbd_device *device,
Andreas Gruenbacher8164dd62023-03-30 12:27:40 +02001075 int (*io_fn)(struct drbd_device *, struct drbd_peer_device *),
1076 char *why, enum bm_flag flags,
1077 struct drbd_peer_device *peer_device);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001078extern int drbd_bitmap_io_from_worker(struct drbd_device *device,
Andreas Gruenbacher8164dd62023-03-30 12:27:40 +02001079 int (*io_fn)(struct drbd_device *, struct drbd_peer_device *),
1080 char *why, enum bm_flag flags,
1081 struct drbd_peer_device *peer_device);
1082extern int drbd_bmio_set_n_write(struct drbd_device *device,
1083 struct drbd_peer_device *peer_device) __must_hold(local);
1084extern int drbd_bmio_clear_n_write(struct drbd_device *device,
1085 struct drbd_peer_device *peer_device) __must_hold(local);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001086
Philipp Reisnerb411b362009-09-25 16:07:19 -07001087/* Meta data layout
Lars Ellenbergae8bf312013-03-19 18:16:43 +01001088 *
1089 * We currently have two possible layouts.
1090 * Offsets in (512 byte) sectors.
1091 * external:
1092 * |----------- md_size_sect ------------------|
1093 * [ 4k superblock ][ activity log ][ Bitmap ]
1094 * | al_offset == 8 |
1095 * | bm_offset = al_offset + X |
1096 * ==> bitmap sectors = md_size_sect - bm_offset
1097 *
1098 * Variants:
1099 * old, indexed fixed size meta data:
1100 *
1101 * internal:
1102 * |----------- md_size_sect ------------------|
1103 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*]
1104 * | al_offset < 0 |
1105 * | bm_offset = al_offset - Y |
1106 * ==> bitmap sectors = Y = al_offset - bm_offset
1107 *
1108 * [padding*] are zero or up to 7 unused 512 Byte sectors to the
1109 * end of the device, so that the [4k superblock] will be 4k aligned.
1110 *
1111 * The activity log consists of 4k transaction blocks,
1112 * which are written in a ring-buffer, or striped ring-buffer like fashion,
1113 * which are writtensize used to be fixed 32kB,
1114 * but is about to become configurable.
1115 */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001116
Lars Ellenbergae8bf312013-03-19 18:16:43 +01001117/* Our old fixed size meta data layout
1118 * allows up to about 3.8TB, so if you want more,
Lars Ellenberg7ad651b2011-02-21 13:21:03 +01001119 * you need to use the "flexible" meta data format. */
Lars Ellenbergae8bf312013-03-19 18:16:43 +01001120#define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */
1121#define MD_4kB_SECT 8
1122#define MD_32kB_SECT 64
Philipp Reisnerb411b362009-09-25 16:07:19 -07001123
Lars Ellenberg7ad651b2011-02-21 13:21:03 +01001124/* One activity log extent represents 4M of storage */
1125#define AL_EXTENT_SHIFT 22
Philipp Reisnerb411b362009-09-25 16:07:19 -07001126#define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1127
Lars Ellenberg7ad651b2011-02-21 13:21:03 +01001128/* We could make these currently hardcoded constants configurable
1129 * variables at create-md time (or even re-configurable at runtime?).
1130 * Which will require some more changes to the DRBD "super block"
1131 * and attach code.
1132 *
1133 * updates per transaction:
1134 * This many changes to the active set can be logged with one transaction.
1135 * This number is arbitrary.
1136 * context per transaction:
1137 * This many context extent numbers are logged with each transaction.
1138 * This number is resulting from the transaction block size (4k), the layout
1139 * of the transaction header, and the number of updates per transaction.
1140 * See drbd_actlog.c:struct al_transaction_on_disk
1141 * */
1142#define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1143#define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1144
Philipp Reisnerb411b362009-09-25 16:07:19 -07001145#if BITS_PER_LONG == 32
1146#define LN2_BPL 5
1147#define cpu_to_lel(A) cpu_to_le32(A)
1148#define lel_to_cpu(A) le32_to_cpu(A)
1149#elif BITS_PER_LONG == 64
1150#define LN2_BPL 6
1151#define cpu_to_lel(A) cpu_to_le64(A)
1152#define lel_to_cpu(A) le64_to_cpu(A)
1153#else
1154#error "LN2 of BITS_PER_LONG unknown!"
1155#endif
1156
1157/* resync bitmap */
1158/* 16MB sized 'bitmap extent' to track syncer usage */
1159struct bm_extent {
1160 int rs_left; /* number of bits set (out of sync) in this extent. */
1161 int rs_failed; /* number of failed resync requests in this extent. */
1162 unsigned long flags;
1163 struct lc_element lce;
1164};
1165
1166#define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1167#define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
Philipp Reisnere3555d82010-11-07 15:56:29 +01001168#define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001169
1170/* drbd_bitmap.c */
1171/*
1172 * We need to store one bit for a block.
1173 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1174 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1175 * Bit 1 ==> local node thinks this block needs to be synced.
1176 */
1177
Philipp Reisner8e26f9c2010-07-06 17:25:54 +02001178#define SLEEP_TIME (HZ/10)
1179
Lars Ellenberg45dfffe2011-02-21 13:21:00 +01001180/* We do bitmap IO in units of 4k blocks.
1181 * We also still have a hardcoded 4k per bit relation. */
1182#define BM_BLOCK_SHIFT 12 /* 4k per bit */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001183#define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
Lars Ellenberg45dfffe2011-02-21 13:21:00 +01001184/* mostly arbitrarily set the represented size of one bitmap extent,
1185 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1186 * at 4k per bit resolution) */
1187#define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001188#define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1189
1190#if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1191#error "HAVE YOU FIXED drbdmeta AS WELL??"
1192#endif
1193
1194/* thus many _storage_ sectors are described by one bit */
1195#define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1196#define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1197#define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1198
1199/* bit to represented kilo byte conversion */
1200#define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1201
1202/* in which _bitmap_ extent (resp. sector) the bit for a certain
1203 * _storage_ sector is located in */
1204#define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
Lars Ellenberg5ab7d2c2014-01-27 15:58:22 +01001205#define BM_BIT_TO_EXT(x) ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
Philipp Reisnerb411b362009-09-25 16:07:19 -07001206
Lars Ellenberg5ab7d2c2014-01-27 15:58:22 +01001207/* first storage sector a bitmap extent corresponds to */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001208#define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
Lars Ellenberg5ab7d2c2014-01-27 15:58:22 +01001209/* how much _storage_ sectors we have per bitmap extent */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001210#define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
Lars Ellenberg5ab7d2c2014-01-27 15:58:22 +01001211/* how many bits are covered by one bitmap extent (resync extent) */
1212#define BM_BITS_PER_EXT (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1213
1214#define BM_BLOCKS_PER_BM_EXT_MASK (BM_BITS_PER_EXT - 1)
1215
Philipp Reisnerb411b362009-09-25 16:07:19 -07001216
1217/* in one sector of the bitmap, we have this many activity_log extents. */
1218#define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
Philipp Reisnerb411b362009-09-25 16:07:19 -07001219
Philipp Reisnerb411b362009-09-25 16:07:19 -07001220/* the extent in "PER_EXTENT" below is an activity log extent
1221 * we need that many (long words/bytes) to store the bitmap
1222 * of one AL_EXTENT_SIZE chunk of storage.
1223 * we can store the bitmap for that many AL_EXTENTS within
1224 * one sector of the _on_disk_ bitmap:
1225 * bit 0 bit 37 bit 38 bit (512*8)-1
1226 * ...|........|........|.. // ..|........|
1227 * sect. 0 `296 `304 ^(512*8*8)-1
1228 *
1229#define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1230#define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1231#define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1232 */
1233
1234#define DRBD_MAX_SECTORS_32 (0xffffffffLU)
Lars Ellenbergae8bf312013-03-19 18:16:43 +01001235/* we have a certain meta data variant that has a fixed on-disk size of 128
1236 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity
1237 * log, leaving this many sectors for the bitmap.
1238 */
1239
1240#define DRBD_MAX_SECTORS_FIXED_BM \
1241 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9)))
Lars Ellenbergae8bf312013-03-19 18:16:43 +01001242#define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM
Philipp Reisnerb411b362009-09-25 16:07:19 -07001243/* 16 TB in units of sectors */
1244#if BITS_PER_LONG == 32
1245/* adjust by one page worth of bitmap,
1246 * so we won't wrap around in drbd_bm_find_next_bit.
1247 * you should use 64bit OS for that much storage, anyways. */
1248#define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1249#else
Lars Ellenberg4b0715f2010-12-14 15:13:04 +01001250/* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1251#define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1252/* corresponds to (1UL << 38) bits right now. */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001253#endif
Philipp Reisnerb411b362009-09-25 16:07:19 -07001254
Ming Lei8bf223c2016-05-30 21:34:35 +08001255/* Estimate max bio size as 256 * PAGE_SIZE,
Kirill A. Shutemovea1754a2016-04-01 15:29:48 +03001256 * so for typical PAGE_SIZE of 4k, that is (1<<20) Byte.
Lars Ellenberg23361cf2011-03-31 16:36:43 +02001257 * Since we may live in a mixed-platform cluster,
1258 * we limit us to a platform agnostic constant here for now.
1259 * A followup commit may allow even bigger BIO sizes,
1260 * once we thought that through. */
Philipp Reisner98683652012-11-09 14:18:43 +01001261#define DRBD_MAX_BIO_SIZE (1U << 20)
Christoph Hellwiga8affc02021-03-11 12:01:37 +01001262#if DRBD_MAX_BIO_SIZE > (BIO_MAX_VECS << PAGE_SHIFT)
Lars Ellenberg23361cf2011-03-31 16:36:43 +02001263#error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1264#endif
Lars Ellenbergdb141b22012-06-25 19:15:58 +02001265#define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001266
Philipp Reisner98683652012-11-09 14:18:43 +01001267#define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */
1268#define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001269
Lars Ellenberg505675f2016-06-14 00:26:23 +02001270/* For now, don't allow more than half of what we can "activate" in one
1271 * activity log transaction to be discarded in one go. We may need to rework
1272 * drbd_al_begin_io() to allow for even larger discard ranges */
Lars Ellenberg9104d312016-06-14 00:26:31 +02001273#define DRBD_MAX_BATCH_BIO_SIZE (AL_UPDATES_PER_TRANSACTION/2*AL_EXTENT_SIZE)
1274#define DRBD_MAX_BBIO_SECTORS (DRBD_MAX_BATCH_BIO_SIZE >> 9)
Lars Ellenberga0fb3c42014-04-28 18:43:23 +02001275
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001276extern int drbd_bm_init(struct drbd_device *device);
1277extern int drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits);
1278extern void drbd_bm_cleanup(struct drbd_device *device);
1279extern void drbd_bm_set_all(struct drbd_device *device);
1280extern void drbd_bm_clear_all(struct drbd_device *device);
Lars Ellenberg4b0715f2010-12-14 15:13:04 +01001281/* set/clear/test only a few bits at a time */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001282extern int drbd_bm_set_bits(
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001283 struct drbd_device *device, unsigned long s, unsigned long e);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001284extern int drbd_bm_clear_bits(
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001285 struct drbd_device *device, unsigned long s, unsigned long e);
Lars Ellenberg4b0715f2010-12-14 15:13:04 +01001286extern int drbd_bm_count_bits(
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001287 struct drbd_device *device, const unsigned long s, const unsigned long e);
Lars Ellenberg4b0715f2010-12-14 15:13:04 +01001288/* bm_set_bits variant for use while holding drbd_bm_lock,
1289 * may process the whole bitmap in one go */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001290extern void _drbd_bm_set_bits(struct drbd_device *device,
Philipp Reisnerb411b362009-09-25 16:07:19 -07001291 const unsigned long s, const unsigned long e);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001292extern int drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr);
1293extern int drbd_bm_e_weight(struct drbd_device *device, unsigned long enr);
Andreas Gruenbacher8164dd62023-03-30 12:27:40 +02001294extern int drbd_bm_read(struct drbd_device *device,
1295 struct drbd_peer_device *peer_device) __must_hold(local);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001296extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr);
Andreas Gruenbacher8164dd62023-03-30 12:27:40 +02001297extern int drbd_bm_write(struct drbd_device *device,
1298 struct drbd_peer_device *peer_device) __must_hold(local);
Lars Ellenberg27ea1d82016-06-14 00:26:38 +02001299extern void drbd_bm_reset_al_hints(struct drbd_device *device) __must_hold(local);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001300extern int drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local);
Lars Ellenbergc7a58db2013-12-20 11:39:48 +01001301extern int drbd_bm_write_lazy(struct drbd_device *device, unsigned upper_idx) __must_hold(local);
Andreas Gruenbacher8164dd62023-03-30 12:27:40 +02001302extern int drbd_bm_write_all(struct drbd_device *device,
1303 struct drbd_peer_device *peer_device) __must_hold(local);
1304extern int drbd_bm_write_copy_pages(struct drbd_device *device,
1305 struct drbd_peer_device *peer_device) __must_hold(local);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001306extern size_t drbd_bm_words(struct drbd_device *device);
1307extern unsigned long drbd_bm_bits(struct drbd_device *device);
1308extern sector_t drbd_bm_capacity(struct drbd_device *device);
Lars Ellenberg4b0715f2010-12-14 15:13:04 +01001309
1310#define DRBD_END_OF_BITMAP (~(unsigned long)0)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001311extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001312/* bm_find_next variants for use while you hold drbd_bm_lock() */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001313extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo);
1314extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo);
1315extern unsigned long _drbd_bm_total_weight(struct drbd_device *device);
1316extern unsigned long drbd_bm_total_weight(struct drbd_device *device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001317/* for receive_bitmap */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001318extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset,
Philipp Reisnerb411b362009-09-25 16:07:19 -07001319 size_t number, unsigned long *buffer);
Lars Ellenberg19f843a2010-12-15 08:59:11 +01001320/* for _drbd_send_bitmap */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001321extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset,
Philipp Reisnerb411b362009-09-25 16:07:19 -07001322 size_t number, unsigned long *buffer);
1323
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001324extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags);
1325extern void drbd_bm_unlock(struct drbd_device *device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001326/* drbd_main.c */
1327
1328extern struct kmem_cache *drbd_request_cache;
Andreas Gruenbacher6c852be2011-02-04 15:38:52 +01001329extern struct kmem_cache *drbd_ee_cache; /* peer requests */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001330extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1331extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
Kent Overstreet0892fac2018-05-20 18:25:48 -04001332extern mempool_t drbd_request_mempool;
1333extern mempool_t drbd_ee_mempool;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001334
Lars Ellenberg42818082011-02-23 12:39:46 +01001335/* drbd's page pool, used to buffer data received from the peer,
1336 * or data requested by the peer.
1337 *
1338 * This does not have an emergency reserve.
1339 *
1340 * When allocating from this pool, it first takes pages from the pool.
1341 * Only if the pool is depleted will try to allocate from the system.
1342 *
1343 * The assumption is that pages taken from this pool will be processed,
1344 * and given back, "quickly", and then can be recycled, so we can avoid
1345 * frequent calls to alloc_page(), and still will be able to make progress even
1346 * under memory pressure.
1347 */
1348extern struct page *drbd_pp_pool;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001349extern spinlock_t drbd_pp_lock;
1350extern int drbd_pp_vacant;
1351extern wait_queue_head_t drbd_pp_wait;
1352
Lars Ellenberg42818082011-02-23 12:39:46 +01001353/* We also need a standard (emergency-reserve backed) page pool
1354 * for meta data IO (activity log, bitmap).
1355 * We can keep it global, as long as it is used as "N pages at a time".
1356 * 128 should be plenty, currently we probably can get away with as few as 1.
1357 */
1358#define DRBD_MIN_POOL_PAGES 128
Kent Overstreet0892fac2018-05-20 18:25:48 -04001359extern mempool_t drbd_md_io_page_pool;
Lars Ellenberg42818082011-02-23 12:39:46 +01001360
Lars Ellenberg9476f392011-02-23 17:02:01 +01001361/* We also need to make sure we get a bio
1362 * when we need it for housekeeping purposes */
Kent Overstreet0892fac2018-05-20 18:25:48 -04001363extern struct bio_set drbd_md_io_bio_set;
Lars Ellenberg9476f392011-02-23 17:02:01 +01001364
NeilBrown8cb0def2017-06-18 14:38:58 +10001365/* And a bio_set for cloning */
Kent Overstreet0892fac2018-05-20 18:25:48 -04001366extern struct bio_set drbd_io_bio_set;
NeilBrown8cb0def2017-06-18 14:38:58 +10001367
Andreas Gruenbacher28bc3b82014-08-14 18:33:30 +02001368extern struct mutex resources_mutex;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001369
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001370extern int conn_lowest_minor(struct drbd_connection *connection);
Lars Ellenberga910b122014-04-28 18:43:21 +02001371extern enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor);
Andreas Gruenbacher05a10ec2011-06-07 22:54:17 +02001372extern void drbd_destroy_device(struct kref *kref);
Lars Ellenberga910b122014-04-28 18:43:21 +02001373extern void drbd_delete_device(struct drbd_device *device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001374
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +02001375extern struct drbd_resource *drbd_create_resource(const char *name);
1376extern void drbd_free_resource(struct drbd_resource *resource);
1377
Andreas Gruenbachereb6bea62011-06-21 16:11:28 +02001378extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001379extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts);
Andreas Gruenbacher05a10ec2011-06-07 22:54:17 +02001380extern void drbd_destroy_connection(struct kref *kref);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001381extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len,
Andreas Gruenbacher089c0752011-06-14 18:28:09 +02001382 void *peer_addr, int peer_addr_len);
Andreas Gruenbacher4bc76042011-06-13 14:27:45 +02001383extern struct drbd_resource *drbd_find_resource(const char *name);
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +02001384extern void drbd_destroy_resource(struct kref *kref);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001385extern void conn_free_crypto(struct drbd_connection *connection);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001386
Philipp Reisnerb411b362009-09-25 16:07:19 -07001387/* drbd_req */
Lars Ellenberg113fef92013-03-22 18:14:40 -06001388extern void do_submit(struct work_struct *ws);
Guoqing Jiang370276b2021-01-21 15:21:50 +01001389extern void __drbd_make_request(struct drbd_device *, struct bio *);
Christoph Hellwig3e087732021-10-12 13:12:24 +02001390void drbd_submit_bio(struct bio *bio);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001391extern int drbd_read_remote(struct drbd_device *device, struct drbd_request *req);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001392extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1393
1394
1395/* drbd_nl.c */
Andreas Gruenbachera2972842014-07-31 17:41:33 +02001396
1397extern struct mutex notification_mutex;
1398
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001399extern void drbd_suspend_io(struct drbd_device *device);
1400extern void drbd_resume_io(struct drbd_device *device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001401extern char *ppsize(char *buf, unsigned long long size);
Andreas Gruenbacher54761692011-05-30 16:15:21 +02001402extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int);
Philipp Reisnere96c9632013-06-25 16:50:07 +02001403enum determine_dev_size {
Philipp Reisnerd752b262013-06-25 16:50:08 +02001404 DS_ERROR_SHRINK = -3,
1405 DS_ERROR_SPACE_MD = -2,
Philipp Reisnere96c9632013-06-25 16:50:07 +02001406 DS_ERROR = -1,
1407 DS_UNCHANGED = 0,
1408 DS_SHRUNK = 1,
Philipp Reisner57737ad2013-10-23 10:59:17 +02001409 DS_GREW = 2,
1410 DS_GREW_FROM_ZERO = 3,
Philipp Reisnere96c9632013-06-25 16:50:07 +02001411};
Philipp Reisnerd752b262013-06-25 16:50:08 +02001412extern enum determine_dev_size
Andreas Gruenbacher54761692011-05-30 16:15:21 +02001413drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local);
1414extern void resync_after_online_grow(struct drbd_device *);
Lars Ellenberg9104d312016-06-14 00:26:31 +02001415extern void drbd_reconsider_queue_parameters(struct drbd_device *device,
1416 struct drbd_backing_dev *bdev, struct o_qlim *o);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001417extern enum drbd_state_rv drbd_set_role(struct drbd_device *device,
Andreas Gruenbacherbf885f82010-12-08 00:39:32 +01001418 enum drbd_role new_role,
1419 int force);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001420extern bool conn_try_outdate_peer(struct drbd_connection *connection);
1421extern void conn_try_outdate_peer_async(struct drbd_connection *connection);
Fabian Frederick7e5fec32016-06-14 00:26:35 +02001422extern enum drbd_peer_state conn_khelper(struct drbd_connection *connection, char *cmd);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001423extern int drbd_khelper(struct drbd_device *device, char *cmd);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001424
1425/* drbd_worker.c */
Philipp Reisnerd40e5672014-04-28 18:43:14 +02001426/* bi_end_io handlers */
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001427extern void drbd_md_endio(struct bio *bio);
1428extern void drbd_peer_request_endio(struct bio *bio);
1429extern void drbd_request_endio(struct bio *bio);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001430extern int drbd_worker(struct drbd_thread *thi);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001431enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor);
1432void drbd_resync_after_changed(struct drbd_device *device);
1433extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side);
1434extern void resume_next_sg(struct drbd_device *device);
1435extern void suspend_other_sg(struct drbd_device *device);
Christoph Böhmwalder0d11f3c2023-03-30 12:27:44 +02001436extern int drbd_resync_finished(struct drbd_peer_device *peer_device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001437/* maybe rather drbd_main.c ? */
Lars Ellenberge37d2432014-04-01 23:53:30 +02001438extern void *drbd_md_get_buffer(struct drbd_device *device, const char *intent);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001439extern void drbd_md_put_buffer(struct drbd_device *device);
1440extern int drbd_md_sync_page_io(struct drbd_device *device,
Bart Van Assche99451722022-07-14 11:06:39 -07001441 struct drbd_backing_dev *bdev, sector_t sector, enum req_op op);
Christoph Böhmwalder0d11f3c2023-03-30 12:27:44 +02001442extern void drbd_ov_out_of_sync_found(struct drbd_peer_device *peer_device,
1443 sector_t sector, int size);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001444extern void wait_until_done_or_force_detached(struct drbd_device *device,
Lars Ellenberg44edfb02012-09-27 13:03:45 +02001445 struct drbd_backing_dev *bdev, unsigned int *done);
Christoph Böhmwalder0d11f3c2023-03-30 12:27:44 +02001446extern void drbd_rs_controller_reset(struct drbd_peer_device *peer_device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001447
Christoph Böhmwalder0d11f3c2023-03-30 12:27:44 +02001448static inline void ov_out_of_sync_print(struct drbd_peer_device *peer_device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001449{
Christoph Böhmwalder0d11f3c2023-03-30 12:27:44 +02001450 struct drbd_device *device = peer_device->device;
1451
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001452 if (device->ov_last_oos_size) {
Christoph Böhmwalder0d11f3c2023-03-30 12:27:44 +02001453 drbd_err(peer_device, "Out of sync: start=%llu, size=%lu (sectors)\n",
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001454 (unsigned long long)device->ov_last_oos_start,
1455 (unsigned long)device->ov_last_oos_size);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001456 }
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001457 device->ov_last_oos_size = 0;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001458}
1459
1460
Kees Cook3d0e6372018-08-06 16:32:16 -07001461extern void drbd_csum_bio(struct crypto_shash *, struct bio *, void *);
1462extern void drbd_csum_ee(struct crypto_shash *, struct drbd_peer_request *,
1463 void *);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001464/* worker callbacks */
Andreas Gruenbacher99920dc2011-03-16 15:31:39 +01001465extern int w_e_end_data_req(struct drbd_work *, int);
1466extern int w_e_end_rsdata_req(struct drbd_work *, int);
1467extern int w_e_end_csum_rs_req(struct drbd_work *, int);
1468extern int w_e_end_ov_reply(struct drbd_work *, int);
1469extern int w_e_end_ov_req(struct drbd_work *, int);
1470extern int w_ov_finished(struct drbd_work *, int);
1471extern int w_resync_timer(struct drbd_work *, int);
1472extern int w_send_write_hint(struct drbd_work *, int);
Andreas Gruenbacher99920dc2011-03-16 15:31:39 +01001473extern int w_send_dblock(struct drbd_work *, int);
Andreas Gruenbacher99920dc2011-03-16 15:31:39 +01001474extern int w_send_read_req(struct drbd_work *, int);
Andreas Gruenbacher99920dc2011-03-16 15:31:39 +01001475extern int w_e_reissue(struct drbd_work *, int);
1476extern int w_restart_disk_io(struct drbd_work *, int);
Andreas Gruenbacher8f7bed72010-12-19 23:53:14 +01001477extern int w_send_out_of_sync(struct drbd_work *, int);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001478
Kees Cook2bccef32017-10-17 20:33:01 -07001479extern void resync_timer_fn(struct timer_list *t);
1480extern void start_resync_timer_fn(struct timer_list *t);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001481
Lars Ellenberga0fb3c42014-04-28 18:43:23 +02001482extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req);
1483
Philipp Reisnerb411b362009-09-25 16:07:19 -07001484/* drbd_receiver.c */
Lars Ellenbergf31e5832018-12-20 17:23:42 +01001485extern int drbd_issue_discard_or_zero_out(struct drbd_device *device,
1486 sector_t start, unsigned int nr_sectors, int flags);
Andreas Gruenbacher753c61912011-07-22 11:14:41 +02001487extern int drbd_receiver(struct drbd_thread *thi);
Philipp Reisner1c03e522015-03-16 15:01:00 +01001488extern int drbd_ack_receiver(struct drbd_thread *thi);
Philipp Reisner668700b2015-03-16 16:08:29 +01001489extern void drbd_send_ping_wf(struct work_struct *ws);
1490extern void drbd_send_acks_wf(struct work_struct *ws);
Lars Ellenberge8299872014-04-28 18:43:19 +02001491extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device);
Christoph Böhmwalder0d11f3c2023-03-30 12:27:44 +02001492extern bool drbd_rs_should_slow_down(struct drbd_peer_device *peer_device, sector_t sector,
Lars Ellenbergad3fee72013-12-20 11:22:13 +01001493 bool throttle_if_app_is_waiting);
Christoph Böhmwalderce668b62022-11-09 14:34:53 +01001494extern int drbd_submit_peer_request(struct drbd_peer_request *peer_req);
Andreas Gruenbacher54761692011-05-30 16:15:21 +02001495extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001496extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64,
Andreas Gruenbacher0db55362011-04-06 16:09:15 +02001497 sector_t, unsigned int,
Lars Ellenberg9104d312016-06-14 00:26:31 +02001498 unsigned int,
Andreas Gruenbacher0db55362011-04-06 16:09:15 +02001499 gfp_t) __must_hold(local);
Andreas Gruenbacher54761692011-05-30 16:15:21 +02001500extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *,
Andreas Gruenbacher3967deb2011-04-06 16:16:56 +02001501 int);
1502#define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1503#define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001504extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001505extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
1506extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001507extern int drbd_connected(struct drbd_peer_device *);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001508
Philipp Reisnerd40e5672014-04-28 18:43:14 +02001509/* sets the number of 512 byte sectors of our virtual device */
Lars Ellenbergd5412e82018-12-20 17:23:30 +01001510void drbd_set_my_capacity(struct drbd_device *device, sector_t size);
Philipp Reisnerd40e5672014-04-28 18:43:14 +02001511
1512/*
1513 * used to submit our private bio
1514 */
Christoph Hellwiged00aab2020-07-01 10:59:44 +02001515static inline void drbd_submit_bio_noacct(struct drbd_device *device,
Philipp Reisnerd40e5672014-04-28 18:43:14 +02001516 int fault_type, struct bio *bio)
1517{
1518 __release(local);
Christoph Hellwig309dca302021-01-24 11:02:34 +01001519 if (!bio->bi_bdev) {
1520 drbd_err(device, "drbd_submit_bio_noacct: bio->bi_bdev == NULL\n");
Christoph Hellwig4e4cbee2017-06-03 09:38:06 +02001521 bio->bi_status = BLK_STS_IOERR;
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001522 bio_endio(bio);
Philipp Reisnerd40e5672014-04-28 18:43:14 +02001523 return;
1524 }
1525
1526 if (drbd_insert_fault(device, fault_type))
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001527 bio_io_error(bio);
Philipp Reisnerd40e5672014-04-28 18:43:14 +02001528 else
Christoph Hellwiged00aab2020-07-01 10:59:44 +02001529 submit_bio_noacct(bio);
Philipp Reisnerd40e5672014-04-28 18:43:14 +02001530}
1531
Philipp Reisner8fe39aa2013-11-22 13:22:13 +01001532void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev,
1533 enum write_ordering_e wo);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001534
1535/* drbd_proc.c */
1536extern struct proc_dir_entry *drbd_proc;
Christoph Hellwig004fd112018-04-11 16:46:11 +02001537int drbd_seq_show(struct seq_file *seq, void *v);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001538
1539/* drbd_actlog.c */
Lars Ellenberge4d7d6f2014-04-28 18:43:28 +02001540extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001541extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i);
Lars Ellenberg4dd726f2014-02-11 11:15:36 +01001542extern void drbd_al_begin_io_commit(struct drbd_device *device);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001543extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i);
Lars Ellenberg4dd726f2014-02-11 11:15:36 +01001544extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001545extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i);
1546extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector);
1547extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector);
Christoph Böhmwalder0d11f3c2023-03-30 12:27:44 +02001548extern int drbd_try_rs_begin_io(struct drbd_peer_device *peer_device, sector_t sector);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001549extern void drbd_rs_cancel_all(struct drbd_device *device);
1550extern int drbd_rs_del_all(struct drbd_device *device);
Christoph Böhmwalder0d11f3c2023-03-30 12:27:44 +02001551extern void drbd_rs_failed_io(struct drbd_peer_device *peer_device,
Philipp Reisnerb411b362009-09-25 16:07:19 -07001552 sector_t sector, int size);
Christoph Böhmwalder0d11f3c2023-03-30 12:27:44 +02001553extern void drbd_advance_rs_marks(struct drbd_peer_device *peer_device, unsigned long still_to_go);
Lars Ellenberg5ab7d2c2014-01-27 15:58:22 +01001554
1555enum update_sync_bits_mode { RECORD_RS_FAILED, SET_OUT_OF_SYNC, SET_IN_SYNC };
Christoph Böhmwalder0d11f3c2023-03-30 12:27:44 +02001556extern int __drbd_change_sync(struct drbd_peer_device *peer_device, sector_t sector, int size,
Andreas Gruenbacher179e20b82014-11-10 17:21:09 +01001557 enum update_sync_bits_mode mode);
Christoph Böhmwalder0d11f3c2023-03-30 12:27:44 +02001558#define drbd_set_in_sync(peer_device, sector, size) \
1559 __drbd_change_sync(peer_device, sector, size, SET_IN_SYNC)
1560#define drbd_set_out_of_sync(peer_device, sector, size) \
1561 __drbd_change_sync(peer_device, sector, size, SET_OUT_OF_SYNC)
1562#define drbd_rs_failed_io(peer_device, sector, size) \
1563 __drbd_change_sync(peer_device, sector, size, RECORD_RS_FAILED)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001564extern void drbd_al_shrink(struct drbd_device *device);
Lars Ellenberg5f7c0122015-06-08 15:18:45 +02001565extern int drbd_al_initialize(struct drbd_device *, void *);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001566
Philipp Reisnerb411b362009-09-25 16:07:19 -07001567/* drbd_nl.c */
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001568/* state info broadcast */
1569struct sib_info {
1570 enum drbd_state_info_bcast_reason sib_reason;
1571 union {
1572 struct {
1573 char *helper_name;
1574 unsigned helper_exit_code;
1575 };
1576 struct {
1577 union drbd_state os;
1578 union drbd_state ns;
1579 };
1580 };
1581};
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001582void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001583
Lv Yunlongaadb22b2022-04-06 21:04:43 +02001584extern int notify_resource_state(struct sk_buff *,
Andreas Gruenbachera2972842014-07-31 17:41:33 +02001585 unsigned int,
1586 struct drbd_resource *,
1587 struct resource_info *,
1588 enum drbd_notification_type);
Lv Yunlongaadb22b2022-04-06 21:04:43 +02001589extern int notify_device_state(struct sk_buff *,
Andreas Gruenbachera2972842014-07-31 17:41:33 +02001590 unsigned int,
1591 struct drbd_device *,
1592 struct device_info *,
1593 enum drbd_notification_type);
Lv Yunlongaadb22b2022-04-06 21:04:43 +02001594extern int notify_connection_state(struct sk_buff *,
Andreas Gruenbachera2972842014-07-31 17:41:33 +02001595 unsigned int,
1596 struct drbd_connection *,
1597 struct connection_info *,
1598 enum drbd_notification_type);
Lv Yunlongaadb22b2022-04-06 21:04:43 +02001599extern int notify_peer_device_state(struct sk_buff *,
Andreas Gruenbachera2972842014-07-31 17:41:33 +02001600 unsigned int,
1601 struct drbd_peer_device *,
1602 struct peer_device_info *,
1603 enum drbd_notification_type);
1604extern void notify_helper(enum drbd_notification_type, struct drbd_device *,
1605 struct drbd_connection *, const char *, int);
1606
Philipp Reisnerb411b362009-09-25 16:07:19 -07001607/*
1608 * inline helper functions
1609 *************************/
1610
Lars Ellenberg45bb9122010-05-14 17:10:48 +02001611/* see also page_chain_add and friends in drbd_receiver.c */
1612static inline struct page *page_chain_next(struct page *page)
1613{
1614 return (struct page *)page_private(page);
1615}
1616#define page_chain_for_each(page) \
1617 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1618 page = page_chain_next(page))
1619#define page_chain_for_each_safe(page, n) \
1620 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1621
Lars Ellenberg45bb9122010-05-14 17:10:48 +02001622
Andreas Gruenbacher045417f2011-04-07 21:34:24 +02001623static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req)
Lars Ellenberg45bb9122010-05-14 17:10:48 +02001624{
Andreas Gruenbacherdb830c42011-02-04 15:57:48 +01001625 struct page *page = peer_req->pages;
Lars Ellenberg45bb9122010-05-14 17:10:48 +02001626 page_chain_for_each(page) {
1627 if (page_count(page) > 1)
1628 return 1;
1629 }
1630 return 0;
1631}
1632
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001633static inline union drbd_state drbd_read_state(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001634{
Andreas Gruenbacher6bbf53c2011-07-08 01:19:44 +02001635 struct drbd_resource *resource = device->resource;
Philipp Reisner78bae592011-03-28 15:40:12 +02001636 union drbd_state rv;
1637
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001638 rv.i = device->state.i;
Andreas Gruenbacher6bbf53c2011-07-08 01:19:44 +02001639 rv.susp = resource->susp;
1640 rv.susp_nod = resource->susp_nod;
1641 rv.susp_fen = resource->susp_fen;
Philipp Reisner78bae592011-03-28 15:40:12 +02001642
1643 return rv;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001644}
1645
Lars Ellenberg383606e2012-06-14 14:21:32 +02001646enum drbd_force_detach_flags {
Lars Ellenberga2a3c74f2012-09-22 12:26:57 +02001647 DRBD_READ_ERROR,
1648 DRBD_WRITE_ERROR,
Lars Ellenberg383606e2012-06-14 14:21:32 +02001649 DRBD_META_IO_ERROR,
1650 DRBD_FORCE_DETACH,
1651};
1652
Philipp Reisnerb411b362009-09-25 16:07:19 -07001653#define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001654static inline void __drbd_chk_io_error_(struct drbd_device *device,
Lars Ellenberga2a3c74f2012-09-22 12:26:57 +02001655 enum drbd_force_detach_flags df,
Lars Ellenberg383606e2012-06-14 14:21:32 +02001656 const char *where)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001657{
Philipp Reisnerdaeda1c2011-05-03 15:00:55 +02001658 enum drbd_io_error_p ep;
1659
1660 rcu_read_lock();
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001661 ep = rcu_dereference(device->ldev->disk_conf)->on_io_error;
Philipp Reisnerdaeda1c2011-05-03 15:00:55 +02001662 rcu_read_unlock();
1663 switch (ep) {
1664 case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */
Lars Ellenberga2a3c74f2012-09-22 12:26:57 +02001665 if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) {
Christoph Böhmwaldere3fa02d2022-12-01 12:03:49 +01001666 if (drbd_ratelimit())
Andreas Gruenbacherd0180172011-07-03 17:53:52 +02001667 drbd_err(device, "Local IO failed in %s.\n", where);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001668 if (device->state.disk > D_INCONSISTENT)
1669 _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001670 break;
1671 }
Gustavo A. R. Silvadf561f662020-08-23 17:36:59 -05001672 fallthrough; /* for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001673 case EP_DETACH:
1674 case EP_CALL_HELPER:
Lars Ellenberga2a3c74f2012-09-22 12:26:57 +02001675 /* Remember whether we saw a READ or WRITE error.
1676 *
1677 * Recovery of the affected area for WRITE failure is covered
1678 * by the activity log.
1679 * READ errors may fall outside that area though. Certain READ
1680 * errors can be "healed" by writing good data to the affected
1681 * blocks, which triggers block re-allocation in lower layers.
1682 *
1683 * If we can not write the bitmap after a READ error,
1684 * we may need to trigger a full sync (see w_go_diskless()).
1685 *
1686 * Force-detach is not really an IO error, but rather a
1687 * desperate measure to try to deal with a completely
1688 * unresponsive lower level IO stack.
1689 * Still it should be treated as a WRITE error.
1690 *
1691 * Meta IO error is always WRITE error:
1692 * we read meta data only once during attach,
1693 * which will fail in case of errors.
1694 */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001695 set_bit(WAS_IO_ERROR, &device->flags);
Lars Ellenberga2a3c74f2012-09-22 12:26:57 +02001696 if (df == DRBD_READ_ERROR)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001697 set_bit(WAS_READ_ERROR, &device->flags);
Lars Ellenberga2a3c74f2012-09-22 12:26:57 +02001698 if (df == DRBD_FORCE_DETACH)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001699 set_bit(FORCE_DETACH, &device->flags);
1700 if (device->state.disk > D_FAILED) {
1701 _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL);
Andreas Gruenbacherd0180172011-07-03 17:53:52 +02001702 drbd_err(device,
Lars Ellenberg82f59cc2010-10-16 12:13:47 +02001703 "Local IO failed in %s. Detaching...\n", where);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001704 }
1705 break;
1706 }
1707}
1708
1709/**
1710 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001711 * @device: DRBD device.
Philipp Reisnerb411b362009-09-25 16:07:19 -07001712 * @error: Error code passed to the IO completion callback
1713 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1714 *
1715 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1716 */
1717#define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001718static inline void drbd_chk_io_error_(struct drbd_device *device,
Lars Ellenberg383606e2012-06-14 14:21:32 +02001719 int error, enum drbd_force_detach_flags forcedetach, const char *where)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001720{
1721 if (error) {
1722 unsigned long flags;
Andreas Gruenbacher05008132011-07-07 14:19:42 +02001723 spin_lock_irqsave(&device->resource->req_lock, flags);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001724 __drbd_chk_io_error_(device, forcedetach, where);
Andreas Gruenbacher05008132011-07-07 14:19:42 +02001725 spin_unlock_irqrestore(&device->resource->req_lock, flags);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001726 }
1727}
1728
1729
1730/**
1731 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1732 * @bdev: Meta data block device.
1733 *
1734 * BTW, for internal meta data, this happens to be the maximum capacity
1735 * we could agree upon with our peer node.
1736 */
Lars Ellenberg68e41a42013-03-19 18:16:45 +01001737static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001738{
Lars Ellenberg68e41a42013-03-19 18:16:45 +01001739 switch (bdev->md.meta_dev_idx) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07001740 case DRBD_MD_INDEX_INTERNAL:
1741 case DRBD_MD_INDEX_FLEX_INT:
1742 return bdev->md.md_offset + bdev->md.bm_offset;
1743 case DRBD_MD_INDEX_FLEX_EXT:
1744 default:
1745 return bdev->md.md_offset;
1746 }
1747}
1748
1749/**
1750 * drbd_md_last_sector() - Return the last sector number of the meta data area
1751 * @bdev: Meta data block device.
1752 */
1753static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1754{
Lars Ellenberg68e41a42013-03-19 18:16:45 +01001755 switch (bdev->md.meta_dev_idx) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07001756 case DRBD_MD_INDEX_INTERNAL:
1757 case DRBD_MD_INDEX_FLEX_INT:
Lars Ellenbergae8bf312013-03-19 18:16:43 +01001758 return bdev->md.md_offset + MD_4kB_SECT -1;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001759 case DRBD_MD_INDEX_FLEX_EXT:
1760 default:
Lars Ellenbergae8bf312013-03-19 18:16:43 +01001761 return bdev->md.md_offset + bdev->md.md_size_sect -1;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001762 }
1763}
1764
1765/* Returns the number of 512 byte sectors of the device */
1766static inline sector_t drbd_get_capacity(struct block_device *bdev)
1767{
Christoph Hellwigda7b3922021-10-18 12:11:04 +02001768 return bdev ? bdev_nr_sectors(bdev) : 0;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001769}
1770
1771/**
1772 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1773 * @bdev: Meta data block device.
1774 *
1775 * returns the capacity we announce to out peer. we clip ourselves at the
1776 * various MAX_SECTORS, because if we don't, current implementation will
1777 * oops sooner or later
1778 */
1779static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1780{
1781 sector_t s;
Philipp Reisnerdaeda1c2011-05-03 15:00:55 +02001782
Lars Ellenberg68e41a42013-03-19 18:16:45 +01001783 switch (bdev->md.meta_dev_idx) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07001784 case DRBD_MD_INDEX_INTERNAL:
1785 case DRBD_MD_INDEX_FLEX_INT:
1786 s = drbd_get_capacity(bdev->backing_bdev)
1787 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
Lars Ellenberg68e41a42013-03-19 18:16:45 +01001788 drbd_md_first_sector(bdev))
Philipp Reisnerb411b362009-09-25 16:07:19 -07001789 : 0;
1790 break;
1791 case DRBD_MD_INDEX_FLEX_EXT:
1792 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1793 drbd_get_capacity(bdev->backing_bdev));
1794 /* clip at maximum size the meta device can support */
1795 s = min_t(sector_t, s,
1796 BM_EXT_TO_SECT(bdev->md.md_size_sect
1797 - bdev->md.bm_offset));
1798 break;
1799 default:
1800 s = min_t(sector_t, DRBD_MAX_SECTORS,
1801 drbd_get_capacity(bdev->backing_bdev));
1802 }
1803 return s;
1804}
1805
1806/**
Lars Ellenberg3a4d4eb2013-03-19 18:16:44 +01001807 * drbd_md_ss() - Return the sector number of our meta data super block
Philipp Reisnerb411b362009-09-25 16:07:19 -07001808 * @bdev: Meta data block device.
1809 */
Lars Ellenberg3a4d4eb2013-03-19 18:16:44 +01001810static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001811{
Lars Ellenberg3a4d4eb2013-03-19 18:16:44 +01001812 const int meta_dev_idx = bdev->md.meta_dev_idx;
Philipp Reisnerdaeda1c2011-05-03 15:00:55 +02001813
Lars Ellenberg3a4d4eb2013-03-19 18:16:44 +01001814 if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001815 return 0;
Lars Ellenberg3a4d4eb2013-03-19 18:16:44 +01001816
1817 /* Since drbd08, internal meta data is always "flexible".
1818 * position: last 4k aligned block of 4k size */
1819 if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL ||
1820 meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)
1821 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8;
1822
1823 /* external, some index; this is the old fixed size layout */
1824 return MD_128MB_SECT * bdev->md.meta_dev_idx;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001825}
1826
1827static inline void
Philipp Reisnerb411b362009-09-25 16:07:19 -07001828drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1829{
1830 unsigned long flags;
1831 spin_lock_irqsave(&q->q_lock, flags);
1832 list_add_tail(&w->list, &q->q);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001833 spin_unlock_irqrestore(&q->q_lock, flags);
Lars Ellenberg8c0785a2011-10-19 11:50:57 +02001834 wake_up(&q->q_wait);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001835}
1836
Lars Ellenberge334f552014-02-11 09:30:49 +01001837static inline void
Lars Ellenberg15e26f6a2014-04-28 11:43:21 +02001838drbd_queue_work_if_unqueued(struct drbd_work_queue *q, struct drbd_work *w)
1839{
1840 unsigned long flags;
1841 spin_lock_irqsave(&q->q_lock, flags);
1842 if (list_empty_careful(&w->list))
1843 list_add_tail(&w->list, &q->q);
1844 spin_unlock_irqrestore(&q->q_lock, flags);
1845 wake_up(&q->q_wait);
1846}
1847
1848static inline void
Lars Ellenberge334f552014-02-11 09:30:49 +01001849drbd_device_post_work(struct drbd_device *device, int work_bit)
1850{
1851 if (!test_and_set_bit(work_bit, &device->flags)) {
1852 struct drbd_connection *connection =
1853 first_peer_device(device)->connection;
1854 struct drbd_work_queue *q = &connection->sender_work;
1855 if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags))
1856 wake_up(&q->q_wait);
1857 }
1858}
1859
Andreas Gruenbacherb5043c52011-07-28 15:56:02 +02001860extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue);
1861
Philipp Reisner668700b2015-03-16 16:08:29 +01001862/* To get the ack_receiver out of the blocking network stack,
1863 * so it can change its sk_rcvtimeo from idle- to ping-timeout,
1864 * and send a ping, we need to send a signal.
1865 * Which signal we send is irrelevant. */
1866static inline void wake_ack_receiver(struct drbd_connection *connection)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001867{
Philipp Reisner668700b2015-03-16 16:08:29 +01001868 struct task_struct *task = connection->ack_receiver.task;
1869 if (task && get_t_state(&connection->ack_receiver) == RUNNING)
Eric W. Biedermanfee10992019-02-05 17:25:02 -06001870 send_sig(SIGXCPU, task, 1);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001871}
1872
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001873static inline void request_ping(struct drbd_connection *connection)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001874{
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001875 set_bit(SEND_PING, &connection->flags);
Philipp Reisner668700b2015-03-16 16:08:29 +01001876 wake_ack_receiver(connection);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001877}
1878
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001879extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001880extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001881extern int conn_send_command(struct drbd_connection *, struct drbd_socket *,
Andreas Gruenbacherdba58582011-03-29 16:55:40 +02001882 enum drbd_packet, unsigned int, void *,
1883 unsigned int);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001884extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *,
Andreas Gruenbacherdba58582011-03-29 16:55:40 +02001885 enum drbd_packet, unsigned int, void *,
1886 unsigned int);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001887
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001888extern int drbd_send_ping(struct drbd_connection *connection);
1889extern int drbd_send_ping_ack(struct drbd_connection *connection);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001890extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001891extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001892
1893static inline void drbd_thread_stop(struct drbd_thread *thi)
1894{
Andreas Gruenbacher81e84652010-12-09 15:03:57 +01001895 _drbd_thread_stop(thi, false, true);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001896}
1897
1898static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
1899{
Andreas Gruenbacher81e84652010-12-09 15:03:57 +01001900 _drbd_thread_stop(thi, false, false);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001901}
1902
1903static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
1904{
Andreas Gruenbacher81e84652010-12-09 15:03:57 +01001905 _drbd_thread_stop(thi, true, false);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001906}
1907
1908/* counts how many answer packets packets we expect from our peer,
1909 * for either explicit application requests,
1910 * or implicit barrier packets as necessary.
1911 * increased:
1912 * w_send_barrier
Andreas Gruenbacher8554df12011-01-25 15:37:43 +01001913 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001914 * it is much easier and equally valid to count what we queue for the
1915 * worker, even before it actually was queued or send.
1916 * (drbd_make_request_common; recovery path on read io-error)
1917 * decreased:
1918 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
Andreas Gruenbacher8554df12011-01-25 15:37:43 +01001919 * _req_mod(req, DATA_RECEIVED)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001920 * [from receive_DataReply]
Andreas Gruenbacher8554df12011-01-25 15:37:43 +01001921 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001922 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
1923 * for some reason it is NOT decreased in got_NegAck,
1924 * but in the resulting cleanup code from report_params.
1925 * we should try to remember the reason for that...
Andreas Gruenbacher8554df12011-01-25 15:37:43 +01001926 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
1927 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001928 * [from tl_clear_barrier]
1929 */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001930static inline void inc_ap_pending(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001931{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001932 atomic_inc(&device->ap_pending_cnt);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001933}
1934
Andreas Gruenbacher33f7d312023-03-30 12:27:39 +02001935#define dec_ap_pending(device) ((void)expect((device), __dec_ap_pending(device) >= 0))
1936static inline int __dec_ap_pending(struct drbd_device *device)
Philipp Reisner49559d82011-02-21 14:19:44 +01001937{
Andreas Gruenbacher33f7d312023-03-30 12:27:39 +02001938 int ap_pending_cnt = atomic_dec_return(&device->ap_pending_cnt);
1939
1940 if (ap_pending_cnt == 0)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001941 wake_up(&device->misc_wait);
Andreas Gruenbacher33f7d312023-03-30 12:27:39 +02001942 return ap_pending_cnt;
Philipp Reisner49559d82011-02-21 14:19:44 +01001943}
Philipp Reisnerb411b362009-09-25 16:07:19 -07001944
1945/* counts how many resync-related answers we still expect from the peer
1946 * increase decrease
1947 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
Lucas De Marchi25985ed2011-03-30 22:57:33 -03001948 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001949 * (or P_NEG_ACK with ID_SYNCER)
1950 */
Christoph Böhmwalder0d11f3c2023-03-30 12:27:44 +02001951static inline void inc_rs_pending(struct drbd_peer_device *peer_device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001952{
Christoph Böhmwalder0d11f3c2023-03-30 12:27:44 +02001953 atomic_inc(&peer_device->device->rs_pending_cnt);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001954}
1955
Christoph Böhmwalder0d11f3c2023-03-30 12:27:44 +02001956#define dec_rs_pending(peer_device) \
1957 ((void)expect((peer_device), __dec_rs_pending(peer_device) >= 0))
1958static inline int __dec_rs_pending(struct drbd_peer_device *peer_device)
Philipp Reisner49559d82011-02-21 14:19:44 +01001959{
Christoph Böhmwalder0d11f3c2023-03-30 12:27:44 +02001960 return atomic_dec_return(&peer_device->device->rs_pending_cnt);
Philipp Reisner49559d82011-02-21 14:19:44 +01001961}
Philipp Reisnerb411b362009-09-25 16:07:19 -07001962
1963/* counts how many answers we still need to send to the peer.
1964 * increased on
1965 * receive_Data unless protocol A;
1966 * we need to send a P_RECV_ACK (proto B)
1967 * or P_WRITE_ACK (proto C)
1968 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
1969 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
1970 * receive_Barrier_* we need to send a P_BARRIER_ACK
1971 */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001972static inline void inc_unacked(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001973{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001974 atomic_inc(&device->unacked_cnt);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001975}
1976
Andreas Gruenbacher33f7d312023-03-30 12:27:39 +02001977#define dec_unacked(device) ((void)expect(device, __dec_unacked(device) >= 0))
1978static inline int __dec_unacked(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001979{
Andreas Gruenbacher33f7d312023-03-30 12:27:39 +02001980 return atomic_dec_return(&device->unacked_cnt);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001981}
1982
Andreas Gruenbacher33f7d312023-03-30 12:27:39 +02001983#define sub_unacked(device, n) ((void)expect(device, __sub_unacked(device) >= 0))
1984static inline int __sub_unacked(struct drbd_device *device, int n)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001985{
Andreas Gruenbacher33f7d312023-03-30 12:27:39 +02001986 return atomic_sub_return(n, &device->unacked_cnt);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001987}
1988
Lars Ellenberg5052fee2016-06-14 00:26:24 +02001989static inline bool is_sync_target_state(enum drbd_conns connection_state)
1990{
1991 return connection_state == C_SYNC_TARGET ||
1992 connection_state == C_PAUSED_SYNC_T;
1993}
1994
1995static inline bool is_sync_source_state(enum drbd_conns connection_state)
1996{
1997 return connection_state == C_SYNC_SOURCE ||
1998 connection_state == C_PAUSED_SYNC_S;
1999}
2000
Lars Ellenberg5ab7d2c2014-01-27 15:58:22 +01002001static inline bool is_sync_state(enum drbd_conns connection_state)
2002{
Lars Ellenberg5052fee2016-06-14 00:26:24 +02002003 return is_sync_source_state(connection_state) ||
2004 is_sync_target_state(connection_state);
Lars Ellenberg5ab7d2c2014-01-27 15:58:22 +01002005}
2006
Philipp Reisnerb411b362009-09-25 16:07:19 -07002007/**
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002008 * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev
Andreas Gruenbacherd1b80852014-09-11 14:29:09 +02002009 * @_device: DRBD device.
2010 * @_min_state: Minimum device state required for success.
Philipp Reisnerb411b362009-09-25 16:07:19 -07002011 *
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002012 * You have to call put_ldev() when finished working with device->ldev.
Philipp Reisnerb411b362009-09-25 16:07:19 -07002013 */
Andreas Gruenbacherd1b80852014-09-11 14:29:09 +02002014#define get_ldev_if_state(_device, _min_state) \
2015 (_get_ldev_if_state((_device), (_min_state)) ? \
2016 ({ __acquire(x); true; }) : false)
2017#define get_ldev(_device) get_ldev_if_state(_device, D_INCONSISTENT)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002018
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002019static inline void put_ldev(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002020{
Andreas Gruenbacher11f8b2b2014-09-11 14:29:05 +02002021 enum drbd_disk_state disk_state = device->state.disk;
Lars Ellenbergba3c6fb2014-02-11 08:57:18 +01002022 /* We must check the state *before* the atomic_dec becomes visible,
2023 * or we have a theoretical race where someone hitting zero,
2024 * while state still D_FAILED, will then see D_DISKLESS in the
2025 * condition below and calling into destroy, where he must not, yet. */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002026 int i = atomic_dec_return(&device->local_cnt);
Lars Ellenberg9a0d9d032011-05-02 11:51:31 +02002027
2028 /* This may be called from some endio handler,
2029 * so we must not sleep here. */
2030
Philipp Reisnerb411b362009-09-25 16:07:19 -07002031 __release(local);
Andreas Gruenbacher0b0ba1e2011-06-27 16:23:33 +02002032 D_ASSERT(device, i >= 0);
Lars Ellenberge9e6f3e2010-09-14 20:26:27 +02002033 if (i == 0) {
Andreas Gruenbacher11f8b2b2014-09-11 14:29:05 +02002034 if (disk_state == D_DISKLESS)
Lars Ellenberg82f59cc2010-10-16 12:13:47 +02002035 /* even internal references gone, safe to destroy */
Lars Ellenberge334f552014-02-11 09:30:49 +01002036 drbd_device_post_work(device, DESTROY_DISK);
Andreas Gruenbacher11f8b2b2014-09-11 14:29:05 +02002037 if (disk_state == D_FAILED)
Lars Ellenberg82f59cc2010-10-16 12:13:47 +02002038 /* all application IO references gone. */
Lars Ellenberge334f552014-02-11 09:30:49 +01002039 if (!test_and_set_bit(GOING_DISKLESS, &device->flags))
2040 drbd_device_post_work(device, GO_DISKLESS);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002041 wake_up(&device->misc_wait);
Lars Ellenberge9e6f3e2010-09-14 20:26:27 +02002042 }
Philipp Reisnerb411b362009-09-25 16:07:19 -07002043}
2044
2045#ifndef __CHECKER__
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002046static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002047{
2048 int io_allowed;
2049
Lars Ellenberg82f59cc2010-10-16 12:13:47 +02002050 /* never get a reference while D_DISKLESS */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002051 if (device->state.disk == D_DISKLESS)
Lars Ellenberg82f59cc2010-10-16 12:13:47 +02002052 return 0;
2053
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002054 atomic_inc(&device->local_cnt);
2055 io_allowed = (device->state.disk >= mins);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002056 if (!io_allowed)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002057 put_ldev(device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002058 return io_allowed;
2059}
2060#else
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002061extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002062#endif
2063
Philipp Reisnerb411b362009-09-25 16:07:19 -07002064/* this throttles on-the-fly application requests
2065 * according to max_buffers settings;
2066 * maybe re-implement using semaphores? */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002067static inline int drbd_get_max_buffers(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002068{
Philipp Reisner44ed1672011-04-19 17:10:19 +02002069 struct net_conf *nc;
2070 int mxb;
2071
2072 rcu_read_lock();
Andreas Gruenbachera6b32bc32011-05-31 14:33:49 +02002073 nc = rcu_dereference(first_peer_device(device)->connection->net_conf);
Philipp Reisner44ed1672011-04-19 17:10:19 +02002074 mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */
2075 rcu_read_unlock();
2076
Philipp Reisnerb411b362009-09-25 16:07:19 -07002077 return mxb;
2078}
2079
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002080static inline int drbd_state_is_stable(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002081{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002082 union drbd_dev_state s = device->state;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002083
2084 /* DO NOT add a default clause, we want the compiler to warn us
2085 * for any newly introduced state we may have forgotten to add here */
2086
2087 switch ((enum drbd_conns)s.conn) {
2088 /* new io only accepted when there is no connection, ... */
2089 case C_STANDALONE:
2090 case C_WF_CONNECTION:
2091 /* ... or there is a well established connection. */
2092 case C_CONNECTED:
2093 case C_SYNC_SOURCE:
2094 case C_SYNC_TARGET:
2095 case C_VERIFY_S:
2096 case C_VERIFY_T:
2097 case C_PAUSED_SYNC_S:
2098 case C_PAUSED_SYNC_T:
Philipp Reisner67531712010-10-27 12:21:30 +02002099 case C_AHEAD:
2100 case C_BEHIND:
Philipp Reisner37190942010-11-10 12:08:37 +01002101 /* transitional states, IO allowed */
Philipp Reisnerb411b362009-09-25 16:07:19 -07002102 case C_DISCONNECTING:
2103 case C_UNCONNECTED:
2104 case C_TIMEOUT:
2105 case C_BROKEN_PIPE:
2106 case C_NETWORK_FAILURE:
2107 case C_PROTOCOL_ERROR:
2108 case C_TEAR_DOWN:
2109 case C_WF_REPORT_PARAMS:
2110 case C_STARTING_SYNC_S:
2111 case C_STARTING_SYNC_T:
Philipp Reisner37190942010-11-10 12:08:37 +01002112 break;
2113
2114 /* Allow IO in BM exchange states with new protocols */
Philipp Reisnerb411b362009-09-25 16:07:19 -07002115 case C_WF_BITMAP_S:
Andreas Gruenbachera6b32bc32011-05-31 14:33:49 +02002116 if (first_peer_device(device)->connection->agreed_pro_version < 96)
Philipp Reisner37190942010-11-10 12:08:37 +01002117 return 0;
2118 break;
2119
2120 /* no new io accepted in these states */
Philipp Reisnerb411b362009-09-25 16:07:19 -07002121 case C_WF_BITMAP_T:
2122 case C_WF_SYNC_UUID:
2123 case C_MASK:
2124 /* not "stable" */
2125 return 0;
2126 }
2127
2128 switch ((enum drbd_disk_state)s.disk) {
2129 case D_DISKLESS:
2130 case D_INCONSISTENT:
2131 case D_OUTDATED:
2132 case D_CONSISTENT:
2133 case D_UP_TO_DATE:
Philipp Reisner5ca1de02011-06-28 17:01:19 +02002134 case D_FAILED:
Philipp Reisnerb411b362009-09-25 16:07:19 -07002135 /* disk state is stable as well. */
2136 break;
2137
Philipp Reisnerd942ae42011-05-31 13:07:24 +02002138 /* no new io accepted during transitional states */
Philipp Reisnerb411b362009-09-25 16:07:19 -07002139 case D_ATTACHING:
Philipp Reisnerb411b362009-09-25 16:07:19 -07002140 case D_NEGOTIATING:
2141 case D_UNKNOWN:
2142 case D_MASK:
2143 /* not "stable" */
2144 return 0;
2145 }
2146
2147 return 1;
2148}
2149
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002150static inline int drbd_suspended(struct drbd_device *device)
Philipp Reisnerfb22c402010-09-08 23:20:21 +02002151{
Andreas Gruenbacher6bbf53c2011-07-08 01:19:44 +02002152 struct drbd_resource *resource = device->resource;
Philipp Reisner8e0af252011-03-28 16:18:39 +02002153
Andreas Gruenbacher6bbf53c2011-07-08 01:19:44 +02002154 return resource->susp || resource->susp_fen || resource->susp_nod;
Philipp Reisnerfb22c402010-09-08 23:20:21 +02002155}
2156
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002157static inline bool may_inc_ap_bio(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002158{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002159 int mxb = drbd_get_max_buffers(device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002160
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002161 if (drbd_suspended(device))
Andreas Gruenbacher1b881ef2010-12-13 18:03:38 +01002162 return false;
Philipp Reisner7dbb4382013-02-28 10:30:19 +01002163 if (atomic_read(&device->suspend_cnt))
Andreas Gruenbacher1b881ef2010-12-13 18:03:38 +01002164 return false;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002165
2166 /* to avoid potential deadlock or bitmap corruption,
2167 * in various places, we only allow new application io
2168 * to start during "stable" states. */
2169
2170 /* no new io accepted when attaching or detaching the disk */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002171 if (!drbd_state_is_stable(device))
Andreas Gruenbacher1b881ef2010-12-13 18:03:38 +01002172 return false;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002173
2174 /* since some older kernels don't have atomic_add_unless,
2175 * and we are within the spinlock anyways, we have this workaround. */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002176 if (atomic_read(&device->ap_bio_cnt) > mxb)
Andreas Gruenbacher1b881ef2010-12-13 18:03:38 +01002177 return false;
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002178 if (test_bit(BITMAP_IO, &device->flags))
Andreas Gruenbacher1b881ef2010-12-13 18:03:38 +01002179 return false;
2180 return true;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002181}
2182
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002183static inline bool inc_ap_bio_cond(struct drbd_device *device)
Philipp Reisner8869d682010-11-17 18:24:19 +01002184{
Andreas Gruenbacher1b881ef2010-12-13 18:03:38 +01002185 bool rv = false;
Philipp Reisner8869d682010-11-17 18:24:19 +01002186
Andreas Gruenbacher05008132011-07-07 14:19:42 +02002187 spin_lock_irq(&device->resource->req_lock);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002188 rv = may_inc_ap_bio(device);
Philipp Reisner8869d682010-11-17 18:24:19 +01002189 if (rv)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002190 atomic_inc(&device->ap_bio_cnt);
Andreas Gruenbacher05008132011-07-07 14:19:42 +02002191 spin_unlock_irq(&device->resource->req_lock);
Philipp Reisner8869d682010-11-17 18:24:19 +01002192
2193 return rv;
2194}
2195
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002196static inline void inc_ap_bio(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002197{
Philipp Reisnerb411b362009-09-25 16:07:19 -07002198 /* we wait here
2199 * as long as the device is suspended
2200 * until the bitmap is no longer on the fly during connection
Philipp Reisnerd942ae42011-05-31 13:07:24 +02002201 * handshake as long as we would exceed the max_buffer limit.
Philipp Reisnerb411b362009-09-25 16:07:19 -07002202 *
2203 * to avoid races with the reconnect code,
2204 * we need to atomic_inc within the spinlock. */
2205
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002206 wait_event(device->misc_wait, inc_ap_bio_cond(device));
Philipp Reisnerb411b362009-09-25 16:07:19 -07002207}
2208
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002209static inline void dec_ap_bio(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002210{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002211 int mxb = drbd_get_max_buffers(device);
2212 int ap_bio = atomic_dec_return(&device->ap_bio_cnt);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002213
Andreas Gruenbacher0b0ba1e2011-06-27 16:23:33 +02002214 D_ASSERT(device, ap_bio >= 0);
Lars Ellenberg7ee1fb92012-06-19 10:27:58 +02002215
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002216 if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) {
2217 if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags))
Andreas Gruenbacher84b8c062011-07-28 15:27:51 +02002218 drbd_queue_work(&first_peer_device(device)->
2219 connection->sender_work,
2220 &device->bm_io_work.w);
Lars Ellenberg7ee1fb92012-06-19 10:27:58 +02002221 }
2222
Philipp Reisnerb411b362009-09-25 16:07:19 -07002223 /* this currently does wake_up for every dec_ap_bio!
2224 * maybe rather introduce some type of hysteresis?
2225 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2226 if (ap_bio < mxb)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002227 wake_up(&device->misc_wait);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002228}
2229
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002230static inline bool verify_can_do_stop_sector(struct drbd_device *device)
Lars Ellenberg58ffa582012-07-26 14:09:49 +02002231{
Andreas Gruenbachera6b32bc32011-05-31 14:33:49 +02002232 return first_peer_device(device)->connection->agreed_pro_version >= 97 &&
2233 first_peer_device(device)->connection->agreed_pro_version != 100;
Lars Ellenberg58ffa582012-07-26 14:09:49 +02002234}
2235
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002236static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002237{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002238 int changed = device->ed_uuid != val;
2239 device->ed_uuid = val;
Lars Ellenberg62b0da32011-01-20 13:25:21 +01002240 return changed;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002241}
2242
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002243static inline int drbd_queue_order_type(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002244{
2245 /* sorry, we currently have no working implementation
2246 * of distributed TCQ stuff */
2247#ifndef QUEUE_ORDERED_NONE
2248#define QUEUE_ORDERED_NONE 0
2249#endif
2250 return QUEUE_ORDERED_NONE;
2251}
2252
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +02002253static inline struct drbd_connection *first_connection(struct drbd_resource *resource)
2254{
Lars Ellenbergec4a34072014-04-28 18:43:35 +02002255 return list_first_entry_or_null(&resource->connections,
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +02002256 struct drbd_connection, connections);
2257}
2258
Philipp Reisnerb411b362009-09-25 16:07:19 -07002259#endif