blob: 7a1cf7eaa71dc8adccd1a5a5d332edacda8df40f [file] [log] [blame]
Philipp Reisnerb411b362009-09-25 16:07:19 -07001/*
2 drbd_int.h
3
4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
5
6 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
7 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
8 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
9
10 drbd is free software; you can redistribute it and/or modify
11 it under the terms of the GNU General Public License as published by
12 the Free Software Foundation; either version 2, or (at your option)
13 any later version.
14
15 drbd is distributed in the hope that it will be useful,
16 but WITHOUT ANY WARRANTY; without even the implied warranty of
17 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 GNU General Public License for more details.
19
20 You should have received a copy of the GNU General Public License
21 along with drbd; see the file COPYING. If not, write to
22 the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
23
24*/
25
26#ifndef _DRBD_INT_H
27#define _DRBD_INT_H
28
Herbert Xu9534d672016-01-24 21:19:21 +080029#include <crypto/hash.h>
Philipp Reisnerb411b362009-09-25 16:07:19 -070030#include <linux/compiler.h>
31#include <linux/types.h>
Philipp Reisnerb411b362009-09-25 16:07:19 -070032#include <linux/list.h>
33#include <linux/sched.h>
34#include <linux/bitops.h>
35#include <linux/slab.h>
Randy Dunlap132cc532009-10-07 19:26:00 +020036#include <linux/ratelimit.h>
Philipp Reisnerb411b362009-09-25 16:07:19 -070037#include <linux/tcp.h>
38#include <linux/mutex.h>
39#include <linux/major.h>
40#include <linux/blkdev.h>
Tejun Heo66114ca2015-05-22 17:13:32 -040041#include <linux/backing-dev.h>
Philipp Reisnerb411b362009-09-25 16:07:19 -070042#include <linux/genhd.h>
Philipp Reisner062e8792011-02-08 11:09:18 +010043#include <linux/idr.h>
Philipp Reisnerb411b362009-09-25 16:07:19 -070044#include <net/tcp.h>
45#include <linux/lru_cache.h>
Paul Gortmaker70c71602011-05-22 16:47:17 -040046#include <linux/prefetch.h>
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +010047#include <linux/drbd_genl_api.h>
Philipp Reisnerb8907332011-01-27 14:07:51 +010048#include <linux/drbd.h>
Andreas Gruenbacherd9f65222011-09-01 13:18:31 +020049#include "drbd_strings.h"
Philipp Reisnerb8907332011-01-27 14:07:51 +010050#include "drbd_state.h"
Andreas Gruenbachera3603a62011-05-30 11:47:37 +020051#include "drbd_protocol.h"
Philipp Reisnerb411b362009-09-25 16:07:19 -070052
53#ifdef __CHECKER__
54# define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
55# define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
56# define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
57# define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
58#else
59# define __protected_by(x)
60# define __protected_read_by(x)
61# define __protected_write_by(x)
62# define __must_hold(x)
63#endif
64
Philipp Reisnerb411b362009-09-25 16:07:19 -070065/* module parameter, defined in drbd_main.c */
66extern unsigned int minor_count;
Rusty Russell90ab5ee2012-01-13 09:32:20 +103067extern bool disable_sendpage;
68extern bool allow_oos;
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +020069void tl_abort_disk_io(struct drbd_device *device);
Philipp Reisnerb411b362009-09-25 16:07:19 -070070
71#ifdef CONFIG_DRBD_FAULT_INJECTION
72extern int enable_faults;
73extern int fault_rate;
74extern int fault_devs;
75#endif
76
77extern char usermode_helper[];
78
79
Philipp Reisnerb411b362009-09-25 16:07:19 -070080/* This is used to stop/restart our threads.
81 * Cannot use SIGTERM nor SIGKILL, since these
82 * are sent out by init on runlevel changes
83 * I choose SIGHUP for now.
84 */
85#define DRBD_SIGKILL SIGHUP
86
Philipp Reisnerb411b362009-09-25 16:07:19 -070087#define ID_IN_SYNC (4711ULL)
88#define ID_OUT_OF_SYNC (4712ULL)
Philipp Reisnerb411b362009-09-25 16:07:19 -070089#define ID_SYNCER (-1ULL)
Andreas Gruenbacher579b57e2011-01-13 18:40:57 +010090
Philipp Reisner4a23f262011-01-11 17:42:17 +010091#define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
Philipp Reisnerb411b362009-09-25 16:07:19 -070092
Andreas Gruenbacher54761692011-05-30 16:15:21 +020093struct drbd_device;
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +020094struct drbd_connection;
Philipp Reisnerb411b362009-09-25 16:07:19 -070095
Andreas Gruenbacher3b52bef2011-07-06 10:57:39 +020096#define __drbd_printk_device(level, device, fmt, args...) \
97 dev_printk(level, disk_to_dev((device)->vdisk), fmt, ## args)
98#define __drbd_printk_peer_device(level, peer_device, fmt, args...) \
99 dev_printk(level, disk_to_dev((peer_device)->device->vdisk), fmt, ## args)
100#define __drbd_printk_resource(level, resource, fmt, args...) \
101 printk(level "drbd %s: " fmt, (resource)->name, ## args)
102#define __drbd_printk_connection(level, connection, fmt, args...) \
103 printk(level "drbd %s: " fmt, (connection)->resource->name, ## args)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700104
Andreas Gruenbacher3b52bef2011-07-06 10:57:39 +0200105void drbd_printk_with_wrong_object_type(void);
106
107#define __drbd_printk_if_same_type(obj, type, func, level, fmt, args...) \
108 (__builtin_types_compatible_p(typeof(obj), type) || \
109 __builtin_types_compatible_p(typeof(obj), const type)), \
110 func(level, (const type)(obj), fmt, ## args)
111
112#define drbd_printk(level, obj, fmt, args...) \
113 __builtin_choose_expr( \
114 __drbd_printk_if_same_type(obj, struct drbd_device *, \
115 __drbd_printk_device, level, fmt, ## args), \
116 __builtin_choose_expr( \
117 __drbd_printk_if_same_type(obj, struct drbd_resource *, \
118 __drbd_printk_resource, level, fmt, ## args), \
119 __builtin_choose_expr( \
120 __drbd_printk_if_same_type(obj, struct drbd_connection *, \
121 __drbd_printk_connection, level, fmt, ## args), \
122 __builtin_choose_expr( \
123 __drbd_printk_if_same_type(obj, struct drbd_peer_device *, \
124 __drbd_printk_peer_device, level, fmt, ## args), \
125 drbd_printk_with_wrong_object_type()))))
126
127#define drbd_dbg(obj, fmt, args...) \
128 drbd_printk(KERN_DEBUG, obj, fmt, ## args)
129#define drbd_alert(obj, fmt, args...) \
130 drbd_printk(KERN_ALERT, obj, fmt, ## args)
131#define drbd_err(obj, fmt, args...) \
132 drbd_printk(KERN_ERR, obj, fmt, ## args)
133#define drbd_warn(obj, fmt, args...) \
134 drbd_printk(KERN_WARNING, obj, fmt, ## args)
135#define drbd_info(obj, fmt, args...) \
136 drbd_printk(KERN_INFO, obj, fmt, ## args)
137#define drbd_emerg(obj, fmt, args...) \
138 drbd_printk(KERN_EMERG, obj, fmt, ## args)
Andreas Gruenbacherd0180172011-07-03 17:53:52 +0200139
140#define dynamic_drbd_dbg(device, fmt, args...) \
141 dynamic_dev_dbg(disk_to_dev(device->vdisk), fmt, ## args)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700142
Andreas Gruenbacher0b0ba1e2011-06-27 16:23:33 +0200143#define D_ASSERT(device, exp) do { \
144 if (!(exp)) \
145 drbd_err(device, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__); \
146 } while (0)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700147
Andreas Gruenbacher841ce242010-12-15 19:31:20 +0100148/**
149 * expect - Make an assertion
150 *
151 * Unlike the assert macro, this macro returns a boolean result.
152 */
153#define expect(exp) ({ \
154 bool _bool = (exp); \
155 if (!_bool) \
Andreas Gruenbacherd0180172011-07-03 17:53:52 +0200156 drbd_err(device, "ASSERTION %s FAILED in %s\n", \
Andreas Gruenbacher841ce242010-12-15 19:31:20 +0100157 #exp, __func__); \
158 _bool; \
159 })
Philipp Reisnerb411b362009-09-25 16:07:19 -0700160
161/* Defines to control fault insertion */
162enum {
163 DRBD_FAULT_MD_WR = 0, /* meta data write */
164 DRBD_FAULT_MD_RD = 1, /* read */
165 DRBD_FAULT_RS_WR = 2, /* resync */
166 DRBD_FAULT_RS_RD = 3,
167 DRBD_FAULT_DT_WR = 4, /* data */
168 DRBD_FAULT_DT_RD = 5,
169 DRBD_FAULT_DT_RA = 6, /* data read ahead */
170 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */
171 DRBD_FAULT_AL_EE = 8, /* alloc ee */
Philipp Reisner6b4388a2010-04-26 14:11:45 +0200172 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700173
174 DRBD_FAULT_MAX,
175};
176
Philipp Reisnerb411b362009-09-25 16:07:19 -0700177extern unsigned int
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +0200178_drbd_insert_fault(struct drbd_device *device, unsigned int type);
Andreas Gruenbacher0cf9d272010-12-07 10:43:29 +0100179
Philipp Reisnerb411b362009-09-25 16:07:19 -0700180static inline int
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +0200181drbd_insert_fault(struct drbd_device *device, unsigned int type) {
Andreas Gruenbacher0cf9d272010-12-07 10:43:29 +0100182#ifdef CONFIG_DRBD_FAULT_INJECTION
Philipp Reisnerb411b362009-09-25 16:07:19 -0700183 return fault_rate &&
184 (enable_faults & (1<<type)) &&
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +0200185 _drbd_insert_fault(device, type);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700186#else
Andreas Gruenbacher0cf9d272010-12-07 10:43:29 +0100187 return 0;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700188#endif
Andreas Gruenbacher0cf9d272010-12-07 10:43:29 +0100189}
Philipp Reisnerb411b362009-09-25 16:07:19 -0700190
191/* integer division, round _UP_ to the next integer */
192#define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
193/* usual integer division */
194#define div_floor(A, B) ((A)/(B))
195
Philipp Reisnerb411b362009-09-25 16:07:19 -0700196extern struct ratelimit_state drbd_ratelimit_state;
Andreas Gruenbacher05a10ec2011-06-07 22:54:17 +0200197extern struct idr drbd_devices; /* RCU, updates: genl_lock() */
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +0200198extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700199
Andreas Gruenbacherd8763022011-01-26 17:39:41 +0100200extern const char *cmdname(enum drbd_packet cmd);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700201
202/* for sending/receiving the bitmap,
203 * possibly in some encoding scheme */
204struct bm_xfer_ctx {
205 /* "const"
206 * stores total bits and long words
207 * of the bitmap, so we don't need to
208 * call the accessor functions over and again. */
209 unsigned long bm_bits;
210 unsigned long bm_words;
211 /* during xfer, current position within the bitmap */
212 unsigned long bit_offset;
213 unsigned long word_offset;
214
215 /* statistics; index: (h->command == P_BITMAP) */
216 unsigned packets[2];
217 unsigned bytes[2];
218};
219
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +0200220extern void INFO_bm_xfer_stats(struct drbd_device *device,
Philipp Reisnerb411b362009-09-25 16:07:19 -0700221 const char *direction, struct bm_xfer_ctx *c);
222
223static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
224{
225 /* word_offset counts "native long words" (32 or 64 bit),
226 * aligned at 64 bit.
227 * Encoded packet may end at an unaligned bit offset.
228 * In case a fallback clear text packet is transmitted in
229 * between, we adjust this offset back to the last 64bit
230 * aligned "native long word", which makes coding and decoding
231 * the plain text bitmap much more convenient. */
232#if BITS_PER_LONG == 64
233 c->word_offset = c->bit_offset >> 6;
234#elif BITS_PER_LONG == 32
235 c->word_offset = c->bit_offset >> 5;
236 c->word_offset &= ~(1UL);
237#else
238# error "unsupported BITS_PER_LONG"
239#endif
240}
241
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +0200242extern unsigned int drbd_header_size(struct drbd_connection *connection);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700243
Philipp Reisnerb411b362009-09-25 16:07:19 -0700244/**********************************************************************/
245enum drbd_thread_state {
Andreas Gruenbachere77a0a52011-01-25 15:43:39 +0100246 NONE,
247 RUNNING,
248 EXITING,
249 RESTARTING
Philipp Reisnerb411b362009-09-25 16:07:19 -0700250};
251
252struct drbd_thread {
253 spinlock_t t_lock;
254 struct task_struct *task;
255 struct completion stop;
256 enum drbd_thread_state t_state;
257 int (*function) (struct drbd_thread *);
Andreas Gruenbacher2457b6d2011-07-21 13:45:21 +0200258 struct drbd_resource *resource;
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +0200259 struct drbd_connection *connection;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700260 int reset_cpu_mask;
Andreas Gruenbacherc60b0252011-08-10 15:05:02 +0200261 const char *name;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700262};
263
264static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
265{
266 /* THINK testing the t_state seems to be uncritical in all cases
267 * (but thread_{start,stop}), so we can read it *without* the lock.
268 * --lge */
269
270 smp_rmb();
271 return thi->t_state;
272}
273
Philipp Reisnerb411b362009-09-25 16:07:19 -0700274struct drbd_work {
275 struct list_head list;
Andreas Gruenbacher309a8342010-12-21 12:38:39 +0100276 int (*cb)(struct drbd_work *, int cancel);
Andreas Gruenbacher84b8c062011-07-28 15:27:51 +0200277};
278
279struct drbd_device_work {
280 struct drbd_work w;
281 struct drbd_device *device;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700282};
283
Andreas Gruenbacherace652a2011-01-03 17:09:58 +0100284#include "drbd_interval.h"
285
Andreas Gruenbacher54761692011-05-30 16:15:21 +0200286extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *);
Andreas Gruenbacher7be8da02011-02-22 02:15:32 +0100287
Andreas Gruenbacher28bc3b82014-08-14 18:33:30 +0200288extern void lock_all_resources(void);
289extern void unlock_all_resources(void);
290
Philipp Reisnerb411b362009-09-25 16:07:19 -0700291struct drbd_request {
292 struct drbd_work w;
Andreas Gruenbacher84b8c062011-07-28 15:27:51 +0200293 struct drbd_device *device;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700294
295 /* if local IO is not allowed, will be NULL.
296 * if local IO _is_ allowed, holds the locally submitted bio clone,
297 * or, after local IO completion, the ERR_PTR(error).
Andreas Gruenbacherfcefa622011-02-17 16:46:59 +0100298 * see drbd_request_endio(). */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700299 struct bio *private_bio;
300
Andreas Gruenbacherace652a2011-01-03 17:09:58 +0100301 struct drbd_interval i;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700302
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100303 /* epoch: used to check on "completion" whether this req was in
Philipp Reisnerb411b362009-09-25 16:07:19 -0700304 * the current epoch, and we therefore have to close it,
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100305 * causing a p_barrier packet to be send, starting a new epoch.
306 *
307 * This corresponds to "barrier" in struct p_barrier[_ack],
308 * and to "barrier_nr" in struct drbd_epoch (and various
309 * comments/function parameters/local variable names).
Philipp Reisnerb411b362009-09-25 16:07:19 -0700310 */
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100311 unsigned int epoch;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700312
Philipp Reisnerb411b362009-09-25 16:07:19 -0700313 struct list_head tl_requests; /* ring list in the transfer log */
314 struct bio *master_bio; /* master bio pointer */
Lars Ellenberge5f891b2013-11-22 12:32:01 +0100315
Lars Ellenberg844a6ae2013-11-22 12:52:03 +0100316 /* see struct drbd_device */
317 struct list_head req_pending_master_completion;
318 struct list_head req_pending_local;
319
Lars Ellenberge5f891b2013-11-22 12:32:01 +0100320 /* for generic IO accounting */
321 unsigned long start_jif;
322
323 /* for DRBD internal statistics */
324
325 /* Minimal set of time stamps to determine if we wait for activity log
326 * transactions, local disk or peer. 32 bit "jiffies" are good enough,
327 * we don't expect a DRBD request to be stalled for several month.
328 */
329
330 /* before actual request processing */
331 unsigned long in_actlog_jif;
332
333 /* local disk */
334 unsigned long pre_submit_jif;
335
336 /* per connection */
337 unsigned long pre_send_jif;
338 unsigned long acked_jif;
339 unsigned long net_done_jif;
340
341 /* Possibly even more detail to track each phase:
342 * master_completion_jif
343 * how long did it take to complete the master bio
344 * (application visible latency)
345 * allocated_jif
346 * how long the master bio was blocked until we finally allocated
347 * a tracking struct
348 * in_actlog_jif
349 * how long did we wait for activity log transactions
350 *
351 * net_queued_jif
352 * when did we finally queue it for sending
353 * pre_send_jif
354 * when did we start sending it
355 * post_send_jif
356 * how long did we block in the network stack trying to send it
357 * acked_jif
358 * when did we receive (or fake, in protocol A) a remote ACK
359 * net_done_jif
360 * when did we receive final acknowledgement (P_BARRIER_ACK),
361 * or decide, e.g. on connection loss, that we do no longer expect
362 * anything from this peer for this request.
363 *
364 * pre_submit_jif
365 * post_sub_jif
366 * when did we start submiting to the lower level device,
367 * and how long did we block in that submit function
368 * local_completion_jif
369 * how long did it take the lower level device to complete this request
370 */
371
Lars Ellenbergb4067772012-01-24 16:58:11 +0100372
373 /* once it hits 0, we may complete the master_bio */
374 atomic_t completion_ref;
375 /* once it hits 0, we may destroy this drbd_request object */
376 struct kref kref;
Lars Ellenberga0d856d2012-01-24 17:19:42 +0100377
378 unsigned rq_state; /* see comments above _req_mod() */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700379};
380
Philipp Reisnerb411b362009-09-25 16:07:19 -0700381struct drbd_epoch {
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +0200382 struct drbd_connection *connection;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700383 struct list_head list;
384 unsigned int barrier_nr;
385 atomic_t epoch_size; /* increased on every request added. */
386 atomic_t active; /* increased on every req. added, and dec on every finished. */
387 unsigned long flags;
388};
389
Rashika Kheriade0b2e62013-12-19 15:07:47 +0530390/* Prototype declaration of function defined in drbd_receiver.c */
391int drbdd_init(struct drbd_thread *);
392int drbd_asender(struct drbd_thread *);
393
Philipp Reisnerb411b362009-09-25 16:07:19 -0700394/* drbd_epoch flag bits */
395enum {
Philipp Reisnerb411b362009-09-25 16:07:19 -0700396 DE_HAVE_BARRIER_NUMBER,
Philipp Reisnerb411b362009-09-25 16:07:19 -0700397};
398
399enum epoch_event {
400 EV_PUT,
401 EV_GOT_BARRIER_NR,
Philipp Reisnerb411b362009-09-25 16:07:19 -0700402 EV_BECAME_LAST,
Philipp Reisnerb411b362009-09-25 16:07:19 -0700403 EV_CLEANUP = 32, /* used as flag */
404};
405
Philipp Reisnerb411b362009-09-25 16:07:19 -0700406struct digest_info {
407 int digest_size;
408 void *digest;
409};
410
Andreas Gruenbacherf6ffca92011-02-04 15:30:34 +0100411struct drbd_peer_request {
Andreas Gruenbachera8cd15b2011-08-25 15:49:40 +0200412 struct drbd_work w;
413 struct drbd_peer_device *peer_device;
Philipp Reisner85719572010-07-21 10:20:17 +0200414 struct drbd_epoch *epoch; /* for writes */
Lars Ellenberg45bb9122010-05-14 17:10:48 +0200415 struct page *pages;
416 atomic_t pending_bios;
Andreas Gruenbacher010f6e62011-01-14 20:59:35 +0100417 struct drbd_interval i;
Lars Ellenberg45bb9122010-05-14 17:10:48 +0200418 /* see comments on ee flag bits below */
419 unsigned long flags;
Lars Ellenberg21ae5d72014-05-05 23:42:24 +0200420 unsigned long submit_jif;
Philipp Reisner85719572010-07-21 10:20:17 +0200421 union {
422 u64 block_id;
423 struct digest_info *digest;
424 };
Lars Ellenberg45bb9122010-05-14 17:10:48 +0200425};
426
427/* ee flag bits.
428 * While corresponding bios are in flight, the only modification will be
429 * set_bit WAS_ERROR, which has to be atomic.
430 * If no bios are in flight yet, or all have been completed,
431 * non-atomic modification to ee->flags is ok.
432 */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700433enum {
434 __EE_CALL_AL_COMPLETE_IO,
Philipp Reisnerb411b362009-09-25 16:07:19 -0700435 __EE_MAY_SET_IN_SYNC,
Lars Ellenberg45bb9122010-05-14 17:10:48 +0200436
Lars Ellenberga0fb3c42014-04-28 18:43:23 +0200437 /* is this a TRIM aka REQ_DISCARD? */
438 __EE_IS_TRIM,
439 /* our lower level cannot handle trim,
440 * and we want to fall back to zeroout instead */
441 __EE_IS_TRIM_USE_ZEROOUT,
442
Lars Ellenberg45bb9122010-05-14 17:10:48 +0200443 /* In case a barrier failed,
444 * we need to resubmit without the barrier flag. */
445 __EE_RESUBMITTED,
446
Andreas Gruenbacher6c852be2011-02-04 15:38:52 +0100447 /* we may have several bios per peer request.
Lars Ellenberg45bb9122010-05-14 17:10:48 +0200448 * if any of those fail, we set this flag atomically
449 * from the endio callback */
450 __EE_WAS_ERROR,
Lars Ellenbergc36c3ce2010-08-11 20:42:55 +0200451
452 /* This ee has a pointer to a digest instead of a block id */
453 __EE_HAS_DIGEST,
Andreas Gruenbacher7be8da02011-02-22 02:15:32 +0100454
455 /* Conflicting local requests need to be restarted after this request */
456 __EE_RESTART_REQUESTS,
Philipp Reisner303d1442011-04-13 16:24:47 -0700457
458 /* The peer wants a write ACK for this (wire proto C) */
459 __EE_SEND_WRITE_ACK,
Philipp Reisner302bdea2011-04-21 11:36:49 +0200460
461 /* Is set when net_conf had two_primaries set while creating this peer_req */
462 __EE_IN_INTERVAL_TREE,
Lars Ellenberg21ae5d72014-05-05 23:42:24 +0200463
464 /* for debugfs: */
465 /* has this been submitted, or does it still wait for something else? */
466 __EE_SUBMITTED,
467
468 /* this is/was a write request */
469 __EE_WRITE,
470
471 /* this originates from application on peer
472 * (not some resync or verify or other DRBD internal request) */
473 __EE_APPLICATION,
Philipp Reisnerb411b362009-09-25 16:07:19 -0700474};
475#define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700476#define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
Lars Ellenberga0fb3c42014-04-28 18:43:23 +0200477#define EE_IS_TRIM (1<<__EE_IS_TRIM)
478#define EE_IS_TRIM_USE_ZEROOUT (1<<__EE_IS_TRIM_USE_ZEROOUT)
479#define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
Lars Ellenberg45bb9122010-05-14 17:10:48 +0200480#define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
Lars Ellenbergc36c3ce2010-08-11 20:42:55 +0200481#define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
Andreas Gruenbacher7be8da02011-02-22 02:15:32 +0100482#define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS)
Philipp Reisner303d1442011-04-13 16:24:47 -0700483#define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK)
Philipp Reisner302bdea2011-04-21 11:36:49 +0200484#define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE)
Lars Ellenberg21ae5d72014-05-05 23:42:24 +0200485#define EE_SUBMITTED (1<<__EE_SUBMITTED)
486#define EE_WRITE (1<<__EE_WRITE)
487#define EE_APPLICATION (1<<__EE_APPLICATION)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700488
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +0200489/* flag bits per device */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700490enum {
Philipp Reisnerb411b362009-09-25 16:07:19 -0700491 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
492 MD_DIRTY, /* current uuids and flags not yet on disk */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700493 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700494 CL_ST_CHG_SUCCESS,
495 CL_ST_CHG_FAIL,
496 CRASHED_PRIMARY, /* This node was a crashed primary.
497 * Gets cleared when the state.conn
498 * goes into C_CONNECTED state. */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700499 CONSIDER_RESYNC,
500
Philipp Reisnera8a4e512010-08-25 10:21:04 +0200501 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */
Lars Ellenberg5ab7d2c2014-01-27 15:58:22 +0100502
Philipp Reisnerb411b362009-09-25 16:07:19 -0700503 BITMAP_IO, /* suspend application io;
504 once no more io in flight, start bitmap io */
505 BITMAP_IO_QUEUED, /* Started bitmap IO */
Lars Ellenberga2a3c74f2012-09-22 12:26:57 +0200506 WAS_IO_ERROR, /* Local disk failed, returned IO error */
507 WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */
Lars Ellenberg383606e2012-06-14 14:21:32 +0200508 FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700509 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700510 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
511 * the peer, if it changed there as well. */
Philipp Reisner43a51822010-06-11 11:26:34 +0200512 NEW_CUR_UUID, /* Create new current UUID when thawing IO */
Philipp Reisner07782862010-08-31 12:00:50 +0200513 AL_SUSPENDED, /* Activity logging is currently suspended. */
Philipp Reisner370a43e2011-01-14 16:03:11 +0100514 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */
Philipp Reisnere64a3292011-02-05 17:34:11 +0100515 B_RS_H_DONE, /* Before resync handler done (already executed) */
Philipp Reisner08b165b2011-09-05 16:22:33 +0200516 DISCARD_MY_DATA, /* discard_my_data flag per volume */
Philipp Reisner380207d2011-11-11 12:31:20 +0100517 READ_BALANCE_RR,
Lars Ellenberge334f552014-02-11 09:30:49 +0100518
Lars Ellenbergf4188152014-05-05 23:05:47 +0200519 FLUSH_PENDING, /* if set, device->flush_jif is when we submitted that flush
520 * from drbd_flush_after_epoch() */
521
Lars Ellenberge334f552014-02-11 09:30:49 +0100522 /* cleared only after backing device related structures have been destroyed. */
523 GOING_DISKLESS, /* Disk is being detached, because of io-error, or admin request. */
524
525 /* to be used in drbd_device_post_work() */
526 GO_DISKLESS, /* tell worker to schedule cleanup before detach */
527 DESTROY_DISK, /* tell worker to close backing devices and destroy related structures. */
Lars Ellenbergac0acb92014-02-11 09:47:58 +0100528 MD_SYNC, /* tell worker to call drbd_md_sync() */
529 RS_START, /* tell worker to start resync/OV */
Lars Ellenberge334f552014-02-11 09:30:49 +0100530 RS_PROGRESS, /* tell worker that resync made significant progress */
531 RS_DONE, /* tell worker that resync is done */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700532};
533
Andreas Gruenbacher54761692011-05-30 16:15:21 +0200534struct drbd_bitmap; /* opaque for drbd_device */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700535
Lars Ellenberg20ceb2b2011-01-21 10:56:44 +0100536/* definition of bits in bm_flags to be used in drbd_bm_lock
537 * and drbd_bitmap_io and friends. */
538enum bm_flag {
Lars Ellenberg20ceb2b2011-01-21 10:56:44 +0100539 /* currently locked for bulk operation */
Lars Ellenberg0e8488a2012-04-25 23:06:45 +0200540 BM_LOCKED_MASK = 0xf,
Lars Ellenberg20ceb2b2011-01-21 10:56:44 +0100541
542 /* in detail, that is: */
543 BM_DONT_CLEAR = 0x1,
544 BM_DONT_SET = 0x2,
545 BM_DONT_TEST = 0x4,
546
Lars Ellenberg0e8488a2012-04-25 23:06:45 +0200547 /* so we can mark it locked for bulk operation,
548 * and still allow all non-bulk operations */
549 BM_IS_LOCKED = 0x8,
550
Lars Ellenberg20ceb2b2011-01-21 10:56:44 +0100551 /* (test bit, count bit) allowed (common case) */
Lars Ellenberg0e8488a2012-04-25 23:06:45 +0200552 BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED,
Lars Ellenberg20ceb2b2011-01-21 10:56:44 +0100553
554 /* testing bits, as well as setting new bits allowed, but clearing bits
555 * would be unexpected. Used during bitmap receive. Setting new bits
556 * requires sending of "out-of-sync" information, though. */
Lars Ellenberg0e8488a2012-04-25 23:06:45 +0200557 BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED,
Lars Ellenberg20ceb2b2011-01-21 10:56:44 +0100558
Lars Ellenberg0e8488a2012-04-25 23:06:45 +0200559 /* for drbd_bm_write_copy_pages, everything is allowed,
560 * only concurrent bulk operations are locked out. */
561 BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED,
Lars Ellenberg20ceb2b2011-01-21 10:56:44 +0100562};
563
Philipp Reisnerb411b362009-09-25 16:07:19 -0700564struct drbd_work_queue {
565 struct list_head q;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700566 spinlock_t q_lock; /* to protect the list. */
Lars Ellenberg8c0785a2011-10-19 11:50:57 +0200567 wait_queue_head_t q_wait;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700568};
569
570struct drbd_socket {
Philipp Reisnerb411b362009-09-25 16:07:19 -0700571 struct mutex mutex;
572 struct socket *socket;
573 /* this way we get our
574 * send/receive buffers off the stack */
Andreas Gruenbacher5a87d922011-03-24 21:17:52 +0100575 void *sbuf;
Andreas Gruenbachere6ef8a52011-03-24 18:07:54 +0100576 void *rbuf;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700577};
578
579struct drbd_md {
580 u64 md_offset; /* sector offset to 'super' block */
581
582 u64 la_size_sect; /* last agreed size, unit sectors */
Philipp Reisner9f2247b2012-08-16 14:25:58 +0200583 spinlock_t uuid_lock;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700584 u64 uuid[UI_SIZE];
585 u64 device_uuid;
586 u32 flags;
587 u32 md_size_sect;
588
Lars Ellenbergae8bf312013-03-19 18:16:43 +0100589 s32 al_offset; /* signed relative sector offset to activity log */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700590 s32 bm_offset; /* signed relative sector offset to bitmap */
Lars Ellenberg3a4d4eb2013-03-19 18:16:44 +0100591
592 /* cached value of bdev->disk_conf->meta_dev_idx (see below) */
593 s32 meta_dev_idx;
594
595 /* see al_tr_number_to_on_disk_sector() */
596 u32 al_stripes;
597 u32 al_stripe_size_4k;
598 u32 al_size_4k; /* cached product of the above */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700599};
600
Philipp Reisnerb411b362009-09-25 16:07:19 -0700601struct drbd_backing_dev {
602 struct block_device *backing_bdev;
603 struct block_device *md_bdev;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700604 struct drbd_md md;
Andreas Gruenbacher05008132011-07-07 14:19:42 +0200605 struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700606 sector_t known_size; /* last known size of that backing device */
607};
608
609struct drbd_md_io {
Lars Ellenberge37d2432014-04-01 23:53:30 +0200610 struct page *page;
611 unsigned long start_jif; /* last call to drbd_md_get_buffer */
612 unsigned long submit_jif; /* last _drbd_md_sync_page_io() submit */
613 const char *current_use;
614 atomic_t in_use;
Philipp Reisner0c464422011-06-26 22:26:31 +0200615 unsigned int done;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700616 int error;
617};
618
619struct bm_io_work {
620 struct drbd_work w;
621 char *why;
Lars Ellenberg20ceb2b2011-01-21 10:56:44 +0100622 enum bm_flag flags;
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +0200623 int (*io_fn)(struct drbd_device *device);
624 void (*done)(struct drbd_device *device, int rv);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700625};
626
Philipp Reisner778f2712010-07-06 11:14:00 +0200627struct fifo_buffer {
Philipp Reisner778f2712010-07-06 11:14:00 +0200628 unsigned int head_index;
629 unsigned int size;
Philipp Reisner9958c852011-05-03 16:19:31 +0200630 int total; /* sum of all values */
631 int values[0];
Philipp Reisner778f2712010-07-06 11:14:00 +0200632};
Philipp Reisner9958c852011-05-03 16:19:31 +0200633extern struct fifo_buffer *fifo_alloc(int fifo_size);
Philipp Reisner778f2712010-07-06 11:14:00 +0200634
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +0200635/* flag bits per connection */
Philipp Reisner01a311a2011-02-07 14:30:33 +0100636enum {
637 NET_CONGESTED, /* The data socket is congested */
Lars Ellenberg427c0432012-08-01 12:43:01 +0200638 RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */
Philipp Reisner668700b2015-03-16 16:08:29 +0100639 SEND_PING,
Philipp Reisner2a67d8b2011-02-09 14:10:32 +0100640 GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */
Philipp Reisner4d0fc3f2012-01-20 13:52:27 +0100641 CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */
Philipp Reisnerfc3b10a2011-02-15 11:07:59 +0100642 CONN_WD_ST_CHG_OKAY,
643 CONN_WD_ST_CHG_FAIL,
Philipp Reisner8169e412011-03-15 18:40:27 +0100644 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
Philipp Reisner6936fcb2011-11-10 18:45:36 +0100645 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */
Philipp Reisnera1096a62012-04-06 12:07:34 +0200646 STATE_SENT, /* Do not change state/UUIDs while this is set */
Lars Ellenberg6f3465e2012-07-30 09:08:25 +0200647 CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
648 * pending, from drbd worker context.
649 * If set, bdi_write_congested() returns true,
650 * so shrink_page_list() would not recurse into,
651 * and potentially deadlock on, this drbd worker.
652 */
Philipp Reisnerb66623e2012-08-08 21:19:09 +0200653 DISCONNECT_SENT,
Lars Ellenberge334f552014-02-11 09:30:49 +0100654
655 DEVICE_WORK_PENDING, /* tell worker that some device has pending work */
Philipp Reisner01a311a2011-02-07 14:30:33 +0100656};
657
Andreas Gruenbachera2972842014-07-31 17:41:33 +0200658enum which_state { NOW, OLD = NOW, NEW };
659
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +0200660struct drbd_resource {
661 char *name;
Lars Ellenberg4d3d5aa2014-05-02 13:19:51 +0200662#ifdef CONFIG_DEBUG_FS
663 struct dentry *debugfs_res;
664 struct dentry *debugfs_res_volumes;
665 struct dentry *debugfs_res_connections;
666 struct dentry *debugfs_res_in_flight_summary;
667#endif
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +0200668 struct kref kref;
Andreas Gruenbacher803ea132011-06-09 01:40:48 +0200669 struct idr devices; /* volume number to device mapping */
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +0200670 struct list_head connections;
671 struct list_head resources;
Andreas Gruenbachereb6bea62011-06-21 16:11:28 +0200672 struct res_opts res_opts;
Andreas Gruenbacher05008132011-07-07 14:19:42 +0200673 struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */
Lars Ellenberg9e276872014-04-28 18:43:22 +0200674 struct mutex adm_mutex; /* mutex to serialize administrative requests */
Andreas Gruenbacher05008132011-07-07 14:19:42 +0200675 spinlock_t req_lock;
Andreas Gruenbacher6bbf53c2011-07-08 01:19:44 +0200676
677 unsigned susp:1; /* IO suspended by user */
678 unsigned susp_nod:1; /* IO suspended because no data */
679 unsigned susp_fen:1; /* IO suspended because fence peer handler runs */
Andreas Gruenbacher625a6ba2011-07-22 14:29:02 +0200680
Philipp Reisnere9526582013-11-22 15:53:41 +0100681 enum write_ordering_e write_ordering;
682
Andreas Gruenbacher625a6ba2011-07-22 14:29:02 +0200683 cpumask_var_t cpu_mask;
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +0200684};
685
Lars Ellenberg944410e2014-05-06 15:02:05 +0200686struct drbd_thread_timing_details
687{
688 unsigned long start_jif;
689 void *cb_addr;
690 const char *caller_fn;
691 unsigned int line;
692 unsigned int cb_nr;
693};
694
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +0200695struct drbd_connection {
696 struct list_head connections;
697 struct drbd_resource *resource;
Lars Ellenberg4d3d5aa2014-05-02 13:19:51 +0200698#ifdef CONFIG_DEBUG_FS
699 struct dentry *debugfs_conn;
700 struct dentry *debugfs_conn_callback_history;
701 struct dentry *debugfs_conn_oldest_requests;
702#endif
Philipp Reisner9dc9fbb2011-04-22 15:23:32 +0200703 struct kref kref;
Andreas Gruenbacherc06ece62011-06-21 17:23:59 +0200704 struct idr peer_devices; /* volume number to peer device mapping */
Philipp Reisner8410da82011-02-11 20:11:10 +0100705 enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */
706 struct mutex cstate_mutex; /* Protects graceful disconnects */
Philipp Reisner28e448b2013-06-25 16:50:06 +0200707 unsigned int connect_cnt; /* Inc each time a connection is established */
Philipp Reisner21114382011-01-19 12:26:59 +0100708
Philipp Reisner062e8792011-02-08 11:09:18 +0100709 unsigned long flags;
Philipp Reisner44ed1672011-04-19 17:10:19 +0200710 struct net_conf *net_conf; /* content protected by rcu */
Philipp Reisner91fd4da2011-04-20 17:47:29 +0200711 wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */
Philipp Reisnere42325a2011-01-19 13:55:45 +0100712
Andreas Gruenbacher089c0752011-06-14 18:28:09 +0200713 struct sockaddr_storage my_addr;
714 int my_addr_len;
715 struct sockaddr_storage peer_addr;
716 int peer_addr_len;
717
Philipp Reisnere42325a2011-01-19 13:55:45 +0100718 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
719 struct drbd_socket meta; /* ping/ack (metadata) packets */
Philipp Reisner31890f42011-01-19 14:12:51 +0100720 int agreed_pro_version; /* actually used protocol version */
Lars Ellenberg20c68fd2014-04-28 18:43:25 +0200721 u32 agreed_features;
Philipp Reisner31890f42011-01-19 14:12:51 +0100722 unsigned long last_received; /* in jiffies, either socket */
723 unsigned int ko_count;
Philipp Reisnere6b3ea82011-01-19 14:02:01 +0100724
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100725 struct list_head transfer_log; /* all requests not yet fully processed */
Philipp Reisner87eeee42011-01-19 14:16:30 +0100726
Herbert Xu9534d672016-01-24 21:19:21 +0800727 struct crypto_shash *cram_hmac_tfm;
728 struct crypto_ahash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */
729 struct crypto_ahash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */
730 struct crypto_ahash *csums_tfm;
731 struct crypto_ahash *verify_tfm;
Philipp Reisnera0638452011-01-19 14:31:32 +0100732 void *int_dig_in;
733 void *int_dig_vv;
734
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100735 /* receiver side */
Philipp Reisner12038a32011-11-09 19:18:00 +0100736 struct drbd_epoch *current_epoch;
737 spinlock_t epoch_lock;
738 unsigned int epochs;
Lars Ellenbergb379c412011-11-17 11:49:46 +0100739 atomic_t current_tle_nr; /* transfer log epoch number */
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100740 unsigned current_tle_writes; /* writes seen within this tl epoch */
Philipp Reisner4b0007c2011-11-09 20:12:34 +0100741
Lars Ellenberg07be15b2012-05-07 11:53:08 +0200742 unsigned long last_reconnect_jif;
Philipp Reisnere6b3ea82011-01-19 14:02:01 +0100743 struct drbd_thread receiver;
744 struct drbd_thread worker;
Philipp Reisner1c03e522015-03-16 15:01:00 +0100745 struct drbd_thread ack_receiver;
Philipp Reisner668700b2015-03-16 16:08:29 +0100746 struct workqueue_struct *ack_sender;
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100747
Lars Ellenberg7753a4c12013-11-22 13:00:12 +0100748 /* cached pointers,
749 * so we can look up the oldest pending requests more quickly.
750 * protected by resource->req_lock */
751 struct drbd_request *req_next; /* DRBD 9: todo.req_next */
752 struct drbd_request *req_ack_pending;
753 struct drbd_request *req_not_net_done;
754
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100755 /* sender side */
Lars Ellenbergd5b27b02011-11-14 15:42:37 +0100756 struct drbd_work_queue sender_work;
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100757
Lars Ellenberg944410e2014-05-06 15:02:05 +0200758#define DRBD_THREAD_DETAILS_HIST 16
759 unsigned int w_cb_nr; /* keeps counting up */
760 unsigned int r_cb_nr; /* keeps counting up */
761 struct drbd_thread_timing_details w_timing_details[DRBD_THREAD_DETAILS_HIST];
762 struct drbd_thread_timing_details r_timing_details[DRBD_THREAD_DETAILS_HIST];
763
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100764 struct {
Lars Ellenberg84d34f22015-02-19 13:54:11 +0100765 unsigned long last_sent_barrier_jif;
766
Lars Ellenbergb6dd1a82011-11-28 15:04:49 +0100767 /* whether this sender thread
768 * has processed a single write yet. */
769 bool seen_any_write_yet;
770
771 /* Which barrier number to send with the next P_BARRIER */
772 int current_epoch_nr;
773
774 /* how many write requests have been sent
775 * with req->epoch == current_epoch_nr.
776 * If none, no P_BARRIER will be sent. */
777 unsigned current_epoch_writes;
778 } send;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700779};
780
Andreas Gruenbachera2972842014-07-31 17:41:33 +0200781static inline bool has_net_conf(struct drbd_connection *connection)
782{
783 bool has_net_conf;
784
785 rcu_read_lock();
786 has_net_conf = rcu_dereference(connection->net_conf);
787 rcu_read_unlock();
788
789 return has_net_conf;
790}
791
Lars Ellenberg944410e2014-05-06 15:02:05 +0200792void __update_timing_details(
793 struct drbd_thread_timing_details *tdp,
794 unsigned int *cb_nr,
795 void *cb,
796 const char *fn, const unsigned int line);
797
798#define update_worker_timing_details(c, cb) \
799 __update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ )
800#define update_receiver_timing_details(c, cb) \
801 __update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ )
802
Lars Ellenberg113fef92013-03-22 18:14:40 -0600803struct submit_worker {
804 struct workqueue_struct *wq;
805 struct work_struct worker;
806
Lars Ellenberg844a6ae2013-11-22 12:52:03 +0100807 /* protected by ..->resource->req_lock */
Lars Ellenberg113fef92013-03-22 18:14:40 -0600808 struct list_head writes;
809};
810
Andreas Gruenbachera6b32bc2011-05-31 14:33:49 +0200811struct drbd_peer_device {
812 struct list_head peer_devices;
813 struct drbd_device *device;
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +0200814 struct drbd_connection *connection;
Philipp Reisner668700b2015-03-16 16:08:29 +0100815 struct work_struct send_acks_work;
Lars Ellenberg4d3d5aa2014-05-02 13:19:51 +0200816#ifdef CONFIG_DEBUG_FS
817 struct dentry *debugfs_peer_dev;
818#endif
Andreas Gruenbachera6b32bc2011-05-31 14:33:49 +0200819};
820
821struct drbd_device {
Andreas Gruenbacherd8628a82011-06-09 01:38:00 +0200822 struct drbd_resource *resource;
Andreas Gruenbachera6b32bc2011-05-31 14:33:49 +0200823 struct list_head peer_devices;
Lars Ellenberg4ce49262014-05-06 00:44:59 +0200824 struct list_head pending_bitmap_io;
Lars Ellenberg4d3d5aa2014-05-02 13:19:51 +0200825
826 unsigned long flush_jif;
827#ifdef CONFIG_DEBUG_FS
828 struct dentry *debugfs_minor;
829 struct dentry *debugfs_vol;
830 struct dentry *debugfs_vol_oldest_requests;
831 struct dentry *debugfs_vol_act_log_extents;
832 struct dentry *debugfs_vol_resync_extents;
833 struct dentry *debugfs_vol_data_gen_id;
Lars Ellenbergf5ec0172015-03-18 17:19:10 +0100834 struct dentry *debugfs_vol_ed_gen_id;
Lars Ellenberg4d3d5aa2014-05-02 13:19:51 +0200835#endif
836
837 unsigned int vnr; /* volume number within the connection */
838 unsigned int minor; /* device minor number */
839
Philipp Reisner81fa2e62011-05-04 15:10:30 +0200840 struct kref kref;
Philipp Reisner21114382011-01-19 12:26:59 +0100841
Philipp Reisnerb411b362009-09-25 16:07:19 -0700842 /* things that are stored as / read from meta data on disk */
843 unsigned long flags;
844
845 /* configured by drbdsetup */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700846 struct drbd_backing_dev *ldev __protected_by(local);
847
848 sector_t p_size; /* partner's disk size */
849 struct request_queue *rq_queue;
850 struct block_device *this_bdev;
851 struct gendisk *vdisk;
852
Lars Ellenberg07be15b2012-05-07 11:53:08 +0200853 unsigned long last_reattach_jif;
Andreas Gruenbacher84b8c062011-07-28 15:27:51 +0200854 struct drbd_work resync_work;
855 struct drbd_work unplug_work;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700856 struct timer_list resync_timer;
857 struct timer_list md_sync_timer;
Philipp Reisner370a43e2011-01-14 16:03:11 +0100858 struct timer_list start_resync_timer;
Philipp Reisner7fde2be2011-03-01 11:08:28 +0100859 struct timer_list request_timer;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700860
861 /* Used after attach while negotiating new disk state. */
862 union drbd_state new_state_tmp;
863
Philipp Reisnerda9fbc22011-03-29 10:52:01 +0200864 union drbd_dev_state state;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700865 wait_queue_head_t misc_wait;
866 wait_queue_head_t state_wait; /* upon each state change. */
867 unsigned int send_cnt;
868 unsigned int recv_cnt;
869 unsigned int read_cnt;
870 unsigned int writ_cnt;
871 unsigned int al_writ_cnt;
872 unsigned int bm_writ_cnt;
873 atomic_t ap_bio_cnt; /* Requests we need to complete */
Lars Ellenbergad3fee72013-12-20 11:22:13 +0100874 atomic_t ap_actlog_cnt; /* Requests waiting for activity log */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700875 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
876 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
Philipp Reisnerd942ae42011-05-31 13:07:24 +0200877 atomic_t unacked_cnt; /* Need to send replies for */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700878 atomic_t local_cnt; /* Waiting for local completion */
Philipp Reisner7dbb4382013-02-28 10:30:19 +0100879 atomic_t suspend_cnt;
Philipp Reisnerb2fb6dbe2011-01-19 13:48:44 +0100880
Andreas Gruenbacherdac13892011-01-21 17:18:39 +0100881 /* Interval tree of pending local requests */
882 struct rb_root read_requests;
Andreas Gruenbacherde696712011-01-20 15:00:24 +0100883 struct rb_root write_requests;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700884
Lars Ellenberg844a6ae2013-11-22 12:52:03 +0100885 /* for statistics and timeouts */
886 /* [0] read, [1] write */
887 struct list_head pending_master_completion[2];
888 struct list_head pending_completion[2];
889
Lars Ellenbergaaaba342014-03-18 12:30:09 +0100890 /* use checksums for *this* resync */
891 bool use_csums;
Lars Ellenberg4b0715f2010-12-14 15:13:04 +0100892 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700893 unsigned long rs_total;
Lars Ellenberg4b0715f2010-12-14 15:13:04 +0100894 /* number of resync blocks that failed in this run */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700895 unsigned long rs_failed;
896 /* Syncer's start time [unit jiffies] */
897 unsigned long rs_start;
898 /* cumulated time in PausedSyncX state [unit jiffies] */
899 unsigned long rs_paused;
Lars Ellenberg1d7734a2010-08-11 21:21:50 +0200900 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700901 unsigned long rs_same_csum;
Lars Ellenberg1d7734a2010-08-11 21:21:50 +0200902#define DRBD_SYNC_MARKS 8
903#define DRBD_SYNC_MARK_STEP (3*HZ)
904 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
905 unsigned long rs_mark_left[DRBD_SYNC_MARKS];
906 /* marks's time [unit jiffies] */
907 unsigned long rs_mark_time[DRBD_SYNC_MARKS];
908 /* current index into rs_mark_{left,time} */
909 int rs_last_mark;
Philipp Reisner328e0f122012-10-19 14:37:47 +0200910 unsigned long rs_last_bcast; /* [unit jiffies] */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700911
912 /* where does the admin want us to start? (sector) */
913 sector_t ov_start_sector;
Lars Ellenberg02b91b52012-06-28 18:26:52 +0200914 sector_t ov_stop_sector;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700915 /* where are we now? (sector) */
916 sector_t ov_position;
917 /* Start sector of out of sync range (to merge printk reporting). */
918 sector_t ov_last_oos_start;
919 /* size of out-of-sync range in sectors. */
920 sector_t ov_last_oos_size;
921 unsigned long ov_left; /* in bits */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700922
Philipp Reisnerb411b362009-09-25 16:07:19 -0700923 struct drbd_bitmap *bitmap;
924 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
925
926 /* Used to track operations of resync... */
927 struct lru_cache *resync;
928 /* Number of locked elements in resync LRU */
929 unsigned int resync_locked;
930 /* resync extent number waiting for application requests */
931 unsigned int resync_wenr;
932
933 int open_cnt;
934 u64 *p_uuid;
Philipp Reisner4b0007c2011-11-09 20:12:34 +0100935
Philipp Reisner85719572010-07-21 10:20:17 +0200936 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
937 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
Andreas Gruenbacher18b75d72011-02-04 15:36:22 +0100938 struct list_head done_ee; /* need to send P_WRITE_ACK */
939 struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700940 struct list_head net_ee; /* zero-copy network send in progress */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700941
942 int next_barrier_nr;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700943 struct list_head resync_reads;
Lars Ellenberg435f0742010-09-06 12:30:25 +0200944 atomic_t pp_in_use; /* allocated from page pool */
945 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700946 wait_queue_head_t ee_wait;
Philipp Reisnercc94c652011-06-26 11:20:27 +0200947 struct drbd_md_io md_io;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700948 spinlock_t al_lock;
949 wait_queue_head_t al_wait;
950 struct lru_cache *act_log; /* activity log */
951 unsigned int al_tr_number;
952 int al_tr_cycle;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700953 wait_queue_head_t seq_wait;
954 atomic_t packet_seq;
955 unsigned int peer_seq;
956 spinlock_t peer_seq_lock;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700957 unsigned long comm_bm_set; /* communicated number of set bits. */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700958 struct bm_io_work bm_io_work;
959 u64 ed_uuid; /* UUID of the exposed data */
Philipp Reisner8410da82011-02-11 20:11:10 +0100960 struct mutex own_state_mutex;
Andreas Gruenbachera6b32bc2011-05-31 14:33:49 +0200961 struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700962 char congestion_reason; /* Why we where congested... */
Lars Ellenberg1d7734a2010-08-11 21:21:50 +0200963 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
964 atomic_t rs_sect_ev; /* for submitted resync data rate, both */
965 int rs_last_sect_ev; /* counter to compare with */
966 int rs_last_events; /* counter of read or write "events" (unit sectors)
967 * on the lower level device when we last looked. */
968 int c_sync_rate; /* current resync rate after syncer throttle magic */
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +0200969 struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */
Philipp Reisner778f2712010-07-06 11:14:00 +0200970 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
Philipp Reisner759fbdf2010-10-26 16:02:27 +0200971 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
Lars Ellenbergdb141b22012-06-25 19:15:58 +0200972 unsigned int peer_max_bio_size;
973 unsigned int local_max_bio_size;
Lars Ellenberg113fef92013-03-22 18:14:40 -0600974
975 /* any requests that would block in drbd_make_request()
976 * are deferred to this single-threaded work queue */
977 struct submit_worker submit;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700978};
979
Lars Ellenberg4ce49262014-05-06 00:44:59 +0200980struct drbd_bm_aio_ctx {
981 struct drbd_device *device;
982 struct list_head list; /* on device->pending_bitmap_io */;
983 unsigned long start_jif;
984 atomic_t in_flight;
985 unsigned int done;
986 unsigned flags;
987#define BM_AIO_COPY_PAGES 1
988#define BM_AIO_WRITE_HINTED 2
989#define BM_AIO_WRITE_ALL_PAGES 4
990#define BM_AIO_READ 8
991 int error;
992 struct kref kref;
993};
994
Lars Ellenberga910b122014-04-28 18:43:21 +0200995struct drbd_config_context {
996 /* assigned from drbd_genlmsghdr */
997 unsigned int minor;
998 /* assigned from request attributes, if present */
999 unsigned int volume;
1000#define VOLUME_UNSPECIFIED (-1U)
1001 /* pointer into the request skb,
1002 * limited lifetime! */
1003 char *resource_name;
1004 struct nlattr *my_addr;
1005 struct nlattr *peer_addr;
1006
1007 /* reply buffer */
1008 struct sk_buff *reply_skb;
1009 /* pointer into reply buffer */
1010 struct drbd_genlmsghdr *reply_dh;
1011 /* resolved from attributes, if possible */
1012 struct drbd_device *device;
1013 struct drbd_resource *resource;
1014 struct drbd_connection *connection;
1015};
1016
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001017static inline struct drbd_device *minor_to_device(unsigned int minor)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001018{
Andreas Gruenbacher05a10ec2011-06-07 22:54:17 +02001019 return (struct drbd_device *)idr_find(&drbd_devices, minor);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001020}
1021
Andreas Gruenbachera6b32bc2011-05-31 14:33:49 +02001022static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device)
1023{
Lars Ellenbergec4a3402014-04-28 18:43:35 +02001024 return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices);
Andreas Gruenbachera6b32bc2011-05-31 14:33:49 +02001025}
1026
Andreas Gruenbachera2972842014-07-31 17:41:33 +02001027static inline struct drbd_peer_device *
1028conn_peer_device(struct drbd_connection *connection, int volume_number)
1029{
1030 return idr_find(&connection->peer_devices, volume_number);
1031}
1032
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +02001033#define for_each_resource(resource, _resources) \
1034 list_for_each_entry(resource, _resources, resources)
1035
1036#define for_each_resource_rcu(resource, _resources) \
1037 list_for_each_entry_rcu(resource, _resources, resources)
1038
1039#define for_each_resource_safe(resource, tmp, _resources) \
1040 list_for_each_entry_safe(resource, tmp, _resources, resources)
1041
1042#define for_each_connection(connection, resource) \
1043 list_for_each_entry(connection, &resource->connections, connections)
1044
1045#define for_each_connection_rcu(connection, resource) \
1046 list_for_each_entry_rcu(connection, &resource->connections, connections)
1047
1048#define for_each_connection_safe(connection, tmp, resource) \
1049 list_for_each_entry_safe(connection, tmp, &resource->connections, connections)
1050
Andreas Gruenbachera6b32bc2011-05-31 14:33:49 +02001051#define for_each_peer_device(peer_device, device) \
1052 list_for_each_entry(peer_device, &device->peer_devices, peer_devices)
1053
1054#define for_each_peer_device_rcu(peer_device, device) \
1055 list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices)
1056
1057#define for_each_peer_device_safe(peer_device, tmp, device) \
1058 list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices)
1059
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001060static inline unsigned int device_to_minor(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001061{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001062 return device->minor;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001063}
1064
Philipp Reisnerb411b362009-09-25 16:07:19 -07001065/*
1066 * function declarations
1067 *************************/
1068
1069/* drbd_main.c */
1070
Philipp Reisnere89b5912010-03-24 17:11:33 +01001071enum dds_flags {
1072 DDSF_FORCED = 1,
1073 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1074};
1075
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001076extern void drbd_init_set_defaults(struct drbd_device *device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001077extern int drbd_thread_start(struct drbd_thread *thi);
1078extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1079#ifdef CONFIG_SMP
Philipp Reisner80822282011-02-08 12:46:30 +01001080extern void drbd_thread_current_set_cpu(struct drbd_thread *thi);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001081#else
1082#define drbd_thread_current_set_cpu(A) ({})
Philipp Reisnerb411b362009-09-25 16:07:19 -07001083#endif
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001084extern void tl_release(struct drbd_connection *, unsigned int barrier_nr,
Philipp Reisnerb411b362009-09-25 16:07:19 -07001085 unsigned int set_size);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001086extern void tl_clear(struct drbd_connection *);
1087extern void drbd_free_sock(struct drbd_connection *connection);
1088extern int drbd_send(struct drbd_connection *connection, struct socket *sock,
Philipp Reisnerbedbd2a2011-02-07 15:08:48 +01001089 void *buf, size_t size, unsigned msg_flags);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001090extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t,
Andreas Gruenbacherfb708e42010-12-15 17:04:36 +01001091 unsigned);
1092
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001093extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd);
1094extern int drbd_send_protocol(struct drbd_connection *connection);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001095extern int drbd_send_uuids(struct drbd_peer_device *);
1096extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *);
1097extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *);
1098extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags);
1099extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s);
1100extern int drbd_send_current_state(struct drbd_peer_device *);
1101extern int drbd_send_sync_param(struct drbd_peer_device *);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001102extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr,
Andreas Gruenbacherd4e67d72011-03-16 01:25:28 +01001103 u32 set_size);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001104extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet,
Andreas Gruenbacherf6ffca92011-02-04 15:30:34 +01001105 struct drbd_peer_request *);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001106extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet,
Andreas Gruenbachera9a99942011-03-16 01:30:14 +01001107 struct p_block_req *rp);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001108extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet,
Andreas Gruenbachera9a99942011-03-16 01:30:14 +01001109 struct p_data *dp, int data_size);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001110extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet,
Philipp Reisnerb411b362009-09-25 16:07:19 -07001111 sector_t sector, int blksize, u64 block_id);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001112extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *);
1113extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet,
Andreas Gruenbacherf6ffca92011-02-04 15:30:34 +01001114 struct drbd_peer_request *);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001115extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req);
1116extern int drbd_send_drequest(struct drbd_peer_device *, int cmd,
Philipp Reisnerb411b362009-09-25 16:07:19 -07001117 sector_t sector, int size, u64 block_id);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001118extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector,
Andreas Gruenbacherd8763022011-01-26 17:39:41 +01001119 int size, void *digest, int digest_size,
1120 enum drbd_packet cmd);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001121extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001122
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001123extern int drbd_send_bitmap(struct drbd_device *device);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001124extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001125extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode);
Lars Ellenberg63a7c8a2015-03-26 20:53:55 +01001126extern void drbd_backing_dev_free(struct drbd_device *device, struct drbd_backing_dev *ldev);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001127extern void drbd_device_cleanup(struct drbd_device *device);
1128void drbd_print_uuids(struct drbd_device *device, const char *text);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001129
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001130extern void conn_md_sync(struct drbd_connection *connection);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001131extern void drbd_md_write(struct drbd_device *device, void *buffer);
1132extern void drbd_md_sync(struct drbd_device *device);
1133extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev);
1134extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1135extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1136extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local);
1137extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local);
1138extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local);
1139extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local);
1140extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local);
1141extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001142extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001143extern void drbd_md_mark_dirty(struct drbd_device *device);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001144extern void drbd_queue_bitmap_io(struct drbd_device *device,
Andreas Gruenbacher54761692011-05-30 16:15:21 +02001145 int (*io_fn)(struct drbd_device *),
1146 void (*done)(struct drbd_device *, int),
Lars Ellenberg20ceb2b2011-01-21 10:56:44 +01001147 char *why, enum bm_flag flags);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001148extern int drbd_bitmap_io(struct drbd_device *device,
Andreas Gruenbacher54761692011-05-30 16:15:21 +02001149 int (*io_fn)(struct drbd_device *),
Lars Ellenberg20ceb2b2011-01-21 10:56:44 +01001150 char *why, enum bm_flag flags);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001151extern int drbd_bitmap_io_from_worker(struct drbd_device *device,
Andreas Gruenbacher54761692011-05-30 16:15:21 +02001152 int (*io_fn)(struct drbd_device *),
Lars Ellenbergedc9f5e2012-09-27 15:18:21 +02001153 char *why, enum bm_flag flags);
Philipp Reisner8fe39aa2013-11-22 13:22:13 +01001154extern int drbd_bmio_set_n_write(struct drbd_device *device) __must_hold(local);
1155extern int drbd_bmio_clear_n_write(struct drbd_device *device) __must_hold(local);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001156
Philipp Reisnerb411b362009-09-25 16:07:19 -07001157/* Meta data layout
Lars Ellenbergae8bf312013-03-19 18:16:43 +01001158 *
1159 * We currently have two possible layouts.
1160 * Offsets in (512 byte) sectors.
1161 * external:
1162 * |----------- md_size_sect ------------------|
1163 * [ 4k superblock ][ activity log ][ Bitmap ]
1164 * | al_offset == 8 |
1165 * | bm_offset = al_offset + X |
1166 * ==> bitmap sectors = md_size_sect - bm_offset
1167 *
1168 * Variants:
1169 * old, indexed fixed size meta data:
1170 *
1171 * internal:
1172 * |----------- md_size_sect ------------------|
1173 * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*]
1174 * | al_offset < 0 |
1175 * | bm_offset = al_offset - Y |
1176 * ==> bitmap sectors = Y = al_offset - bm_offset
1177 *
1178 * [padding*] are zero or up to 7 unused 512 Byte sectors to the
1179 * end of the device, so that the [4k superblock] will be 4k aligned.
1180 *
1181 * The activity log consists of 4k transaction blocks,
1182 * which are written in a ring-buffer, or striped ring-buffer like fashion,
1183 * which are writtensize used to be fixed 32kB,
1184 * but is about to become configurable.
1185 */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001186
Lars Ellenbergae8bf312013-03-19 18:16:43 +01001187/* Our old fixed size meta data layout
1188 * allows up to about 3.8TB, so if you want more,
Lars Ellenberg7ad651b2011-02-21 13:21:03 +01001189 * you need to use the "flexible" meta data format. */
Lars Ellenbergae8bf312013-03-19 18:16:43 +01001190#define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */
1191#define MD_4kB_SECT 8
1192#define MD_32kB_SECT 64
Philipp Reisnerb411b362009-09-25 16:07:19 -07001193
Lars Ellenberg7ad651b2011-02-21 13:21:03 +01001194/* One activity log extent represents 4M of storage */
1195#define AL_EXTENT_SHIFT 22
Philipp Reisnerb411b362009-09-25 16:07:19 -07001196#define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1197
Lars Ellenberg7ad651b2011-02-21 13:21:03 +01001198/* We could make these currently hardcoded constants configurable
1199 * variables at create-md time (or even re-configurable at runtime?).
1200 * Which will require some more changes to the DRBD "super block"
1201 * and attach code.
1202 *
1203 * updates per transaction:
1204 * This many changes to the active set can be logged with one transaction.
1205 * This number is arbitrary.
1206 * context per transaction:
1207 * This many context extent numbers are logged with each transaction.
1208 * This number is resulting from the transaction block size (4k), the layout
1209 * of the transaction header, and the number of updates per transaction.
1210 * See drbd_actlog.c:struct al_transaction_on_disk
1211 * */
1212#define AL_UPDATES_PER_TRANSACTION 64 // arbitrary
1213#define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4
1214
Philipp Reisnerb411b362009-09-25 16:07:19 -07001215#if BITS_PER_LONG == 32
1216#define LN2_BPL 5
1217#define cpu_to_lel(A) cpu_to_le32(A)
1218#define lel_to_cpu(A) le32_to_cpu(A)
1219#elif BITS_PER_LONG == 64
1220#define LN2_BPL 6
1221#define cpu_to_lel(A) cpu_to_le64(A)
1222#define lel_to_cpu(A) le64_to_cpu(A)
1223#else
1224#error "LN2 of BITS_PER_LONG unknown!"
1225#endif
1226
1227/* resync bitmap */
1228/* 16MB sized 'bitmap extent' to track syncer usage */
1229struct bm_extent {
1230 int rs_left; /* number of bits set (out of sync) in this extent. */
1231 int rs_failed; /* number of failed resync requests in this extent. */
1232 unsigned long flags;
1233 struct lc_element lce;
1234};
1235
1236#define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1237#define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
Philipp Reisnere3555d82010-11-07 15:56:29 +01001238#define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001239
1240/* drbd_bitmap.c */
1241/*
1242 * We need to store one bit for a block.
1243 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1244 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1245 * Bit 1 ==> local node thinks this block needs to be synced.
1246 */
1247
Philipp Reisner8e26f9c2010-07-06 17:25:54 +02001248#define SLEEP_TIME (HZ/10)
1249
Lars Ellenberg45dfffe2011-02-21 13:21:00 +01001250/* We do bitmap IO in units of 4k blocks.
1251 * We also still have a hardcoded 4k per bit relation. */
1252#define BM_BLOCK_SHIFT 12 /* 4k per bit */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001253#define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
Lars Ellenberg45dfffe2011-02-21 13:21:00 +01001254/* mostly arbitrarily set the represented size of one bitmap extent,
1255 * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap
1256 * at 4k per bit resolution) */
1257#define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001258#define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1259
1260#if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1261#error "HAVE YOU FIXED drbdmeta AS WELL??"
1262#endif
1263
1264/* thus many _storage_ sectors are described by one bit */
1265#define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1266#define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1267#define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1268
1269/* bit to represented kilo byte conversion */
1270#define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1271
1272/* in which _bitmap_ extent (resp. sector) the bit for a certain
1273 * _storage_ sector is located in */
1274#define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
Lars Ellenberg5ab7d2c2014-01-27 15:58:22 +01001275#define BM_BIT_TO_EXT(x) ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
Philipp Reisnerb411b362009-09-25 16:07:19 -07001276
Lars Ellenberg5ab7d2c2014-01-27 15:58:22 +01001277/* first storage sector a bitmap extent corresponds to */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001278#define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
Lars Ellenberg5ab7d2c2014-01-27 15:58:22 +01001279/* how much _storage_ sectors we have per bitmap extent */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001280#define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
Lars Ellenberg5ab7d2c2014-01-27 15:58:22 +01001281/* how many bits are covered by one bitmap extent (resync extent) */
1282#define BM_BITS_PER_EXT (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT))
1283
1284#define BM_BLOCKS_PER_BM_EXT_MASK (BM_BITS_PER_EXT - 1)
1285
Philipp Reisnerb411b362009-09-25 16:07:19 -07001286
1287/* in one sector of the bitmap, we have this many activity_log extents. */
1288#define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
Philipp Reisnerb411b362009-09-25 16:07:19 -07001289
Philipp Reisnerb411b362009-09-25 16:07:19 -07001290/* the extent in "PER_EXTENT" below is an activity log extent
1291 * we need that many (long words/bytes) to store the bitmap
1292 * of one AL_EXTENT_SIZE chunk of storage.
1293 * we can store the bitmap for that many AL_EXTENTS within
1294 * one sector of the _on_disk_ bitmap:
1295 * bit 0 bit 37 bit 38 bit (512*8)-1
1296 * ...|........|........|.. // ..|........|
1297 * sect. 0 `296 `304 ^(512*8*8)-1
1298 *
1299#define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1300#define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1301#define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1302 */
1303
1304#define DRBD_MAX_SECTORS_32 (0xffffffffLU)
Lars Ellenbergae8bf312013-03-19 18:16:43 +01001305/* we have a certain meta data variant that has a fixed on-disk size of 128
1306 * MiB, of which 4k are our "superblock", and 32k are the fixed size activity
1307 * log, leaving this many sectors for the bitmap.
1308 */
1309
1310#define DRBD_MAX_SECTORS_FIXED_BM \
1311 ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9)))
1312#if !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
Philipp Reisnerb411b362009-09-25 16:07:19 -07001313#define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1314#define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1315#else
Lars Ellenbergae8bf312013-03-19 18:16:43 +01001316#define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM
Philipp Reisnerb411b362009-09-25 16:07:19 -07001317/* 16 TB in units of sectors */
1318#if BITS_PER_LONG == 32
1319/* adjust by one page worth of bitmap,
1320 * so we won't wrap around in drbd_bm_find_next_bit.
1321 * you should use 64bit OS for that much storage, anyways. */
1322#define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1323#else
Lars Ellenberg4b0715f2010-12-14 15:13:04 +01001324/* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1325#define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1326/* corresponds to (1UL << 38) bits right now. */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001327#endif
1328#endif
1329
Kirill A. Shutemovea1754a2016-04-01 15:29:48 +03001330/* BIO_MAX_SIZE is 256 * PAGE_SIZE,
1331 * so for typical PAGE_SIZE of 4k, that is (1<<20) Byte.
Lars Ellenberg23361cf2011-03-31 16:36:43 +02001332 * Since we may live in a mixed-platform cluster,
1333 * we limit us to a platform agnostic constant here for now.
1334 * A followup commit may allow even bigger BIO sizes,
1335 * once we thought that through. */
Philipp Reisner98683652012-11-09 14:18:43 +01001336#define DRBD_MAX_BIO_SIZE (1U << 20)
Lars Ellenberg23361cf2011-03-31 16:36:43 +02001337#if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1338#error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE
1339#endif
Lars Ellenbergdb141b22012-06-25 19:15:58 +02001340#define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001341
Philipp Reisner98683652012-11-09 14:18:43 +01001342#define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */
1343#define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001344
Lars Ellenberga0fb3c42014-04-28 18:43:23 +02001345/* For now, don't allow more than one activity log extent worth of data
1346 * to be discarded in one go. We may need to rework drbd_al_begin_io()
1347 * to allow for even larger discard ranges */
1348#define DRBD_MAX_DISCARD_SIZE AL_EXTENT_SIZE
1349#define DRBD_MAX_DISCARD_SECTORS (DRBD_MAX_DISCARD_SIZE >> 9)
1350
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001351extern int drbd_bm_init(struct drbd_device *device);
1352extern int drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits);
1353extern void drbd_bm_cleanup(struct drbd_device *device);
1354extern void drbd_bm_set_all(struct drbd_device *device);
1355extern void drbd_bm_clear_all(struct drbd_device *device);
Lars Ellenberg4b0715f2010-12-14 15:13:04 +01001356/* set/clear/test only a few bits at a time */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001357extern int drbd_bm_set_bits(
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001358 struct drbd_device *device, unsigned long s, unsigned long e);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001359extern int drbd_bm_clear_bits(
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001360 struct drbd_device *device, unsigned long s, unsigned long e);
Lars Ellenberg4b0715f2010-12-14 15:13:04 +01001361extern int drbd_bm_count_bits(
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001362 struct drbd_device *device, const unsigned long s, const unsigned long e);
Lars Ellenberg4b0715f2010-12-14 15:13:04 +01001363/* bm_set_bits variant for use while holding drbd_bm_lock,
1364 * may process the whole bitmap in one go */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001365extern void _drbd_bm_set_bits(struct drbd_device *device,
Philipp Reisnerb411b362009-09-25 16:07:19 -07001366 const unsigned long s, const unsigned long e);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001367extern int drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr);
1368extern int drbd_bm_e_weight(struct drbd_device *device, unsigned long enr);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001369extern int drbd_bm_read(struct drbd_device *device) __must_hold(local);
1370extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr);
1371extern int drbd_bm_write(struct drbd_device *device) __must_hold(local);
1372extern int drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local);
Lars Ellenbergc7a58db2013-12-20 11:39:48 +01001373extern int drbd_bm_write_lazy(struct drbd_device *device, unsigned upper_idx) __must_hold(local);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001374extern int drbd_bm_write_all(struct drbd_device *device) __must_hold(local);
1375extern int drbd_bm_write_copy_pages(struct drbd_device *device) __must_hold(local);
1376extern size_t drbd_bm_words(struct drbd_device *device);
1377extern unsigned long drbd_bm_bits(struct drbd_device *device);
1378extern sector_t drbd_bm_capacity(struct drbd_device *device);
Lars Ellenberg4b0715f2010-12-14 15:13:04 +01001379
1380#define DRBD_END_OF_BITMAP (~(unsigned long)0)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001381extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001382/* bm_find_next variants for use while you hold drbd_bm_lock() */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001383extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo);
1384extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo);
1385extern unsigned long _drbd_bm_total_weight(struct drbd_device *device);
1386extern unsigned long drbd_bm_total_weight(struct drbd_device *device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001387/* for receive_bitmap */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001388extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset,
Philipp Reisnerb411b362009-09-25 16:07:19 -07001389 size_t number, unsigned long *buffer);
Lars Ellenberg19f843a2010-12-15 08:59:11 +01001390/* for _drbd_send_bitmap */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001391extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset,
Philipp Reisnerb411b362009-09-25 16:07:19 -07001392 size_t number, unsigned long *buffer);
1393
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001394extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags);
1395extern void drbd_bm_unlock(struct drbd_device *device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001396/* drbd_main.c */
1397
1398extern struct kmem_cache *drbd_request_cache;
Andreas Gruenbacher6c852be2011-02-04 15:38:52 +01001399extern struct kmem_cache *drbd_ee_cache; /* peer requests */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001400extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1401extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1402extern mempool_t *drbd_request_mempool;
1403extern mempool_t *drbd_ee_mempool;
1404
Lars Ellenberg42818082011-02-23 12:39:46 +01001405/* drbd's page pool, used to buffer data received from the peer,
1406 * or data requested by the peer.
1407 *
1408 * This does not have an emergency reserve.
1409 *
1410 * When allocating from this pool, it first takes pages from the pool.
1411 * Only if the pool is depleted will try to allocate from the system.
1412 *
1413 * The assumption is that pages taken from this pool will be processed,
1414 * and given back, "quickly", and then can be recycled, so we can avoid
1415 * frequent calls to alloc_page(), and still will be able to make progress even
1416 * under memory pressure.
1417 */
1418extern struct page *drbd_pp_pool;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001419extern spinlock_t drbd_pp_lock;
1420extern int drbd_pp_vacant;
1421extern wait_queue_head_t drbd_pp_wait;
1422
Lars Ellenberg42818082011-02-23 12:39:46 +01001423/* We also need a standard (emergency-reserve backed) page pool
1424 * for meta data IO (activity log, bitmap).
1425 * We can keep it global, as long as it is used as "N pages at a time".
1426 * 128 should be plenty, currently we probably can get away with as few as 1.
1427 */
1428#define DRBD_MIN_POOL_PAGES 128
1429extern mempool_t *drbd_md_io_page_pool;
1430
Lars Ellenberg9476f392011-02-23 17:02:01 +01001431/* We also need to make sure we get a bio
1432 * when we need it for housekeeping purposes */
1433extern struct bio_set *drbd_md_io_bio_set;
1434/* to allocate from that set */
1435extern struct bio *bio_alloc_drbd(gfp_t gfp_mask);
1436
Andreas Gruenbacher28bc3b82014-08-14 18:33:30 +02001437extern struct mutex resources_mutex;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001438
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001439extern int conn_lowest_minor(struct drbd_connection *connection);
Lars Ellenberga910b122014-04-28 18:43:21 +02001440extern enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor);
Andreas Gruenbacher05a10ec2011-06-07 22:54:17 +02001441extern void drbd_destroy_device(struct kref *kref);
Lars Ellenberga910b122014-04-28 18:43:21 +02001442extern void drbd_delete_device(struct drbd_device *device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001443
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +02001444extern struct drbd_resource *drbd_create_resource(const char *name);
1445extern void drbd_free_resource(struct drbd_resource *resource);
1446
Andreas Gruenbachereb6bea62011-06-21 16:11:28 +02001447extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001448extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts);
Andreas Gruenbacher05a10ec2011-06-07 22:54:17 +02001449extern void drbd_destroy_connection(struct kref *kref);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001450extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len,
Andreas Gruenbacher089c0752011-06-14 18:28:09 +02001451 void *peer_addr, int peer_addr_len);
Andreas Gruenbacher4bc76042011-06-13 14:27:45 +02001452extern struct drbd_resource *drbd_find_resource(const char *name);
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +02001453extern void drbd_destroy_resource(struct kref *kref);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001454extern void conn_free_crypto(struct drbd_connection *connection);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001455
1456extern int proc_details;
1457
1458/* drbd_req */
Lars Ellenberg113fef92013-03-22 18:14:40 -06001459extern void do_submit(struct work_struct *ws);
Andreas Gruenbacher54761692011-05-30 16:15:21 +02001460extern void __drbd_make_request(struct drbd_device *, struct bio *, unsigned long);
Jens Axboedece1632015-11-05 10:41:16 -07001461extern blk_qc_t drbd_make_request(struct request_queue *q, struct bio *bio);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001462extern int drbd_read_remote(struct drbd_device *device, struct drbd_request *req);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001463extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1464
1465
1466/* drbd_nl.c */
Andreas Gruenbachera2972842014-07-31 17:41:33 +02001467
1468extern struct mutex notification_mutex;
1469
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001470extern void drbd_suspend_io(struct drbd_device *device);
1471extern void drbd_resume_io(struct drbd_device *device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001472extern char *ppsize(char *buf, unsigned long long size);
Andreas Gruenbacher54761692011-05-30 16:15:21 +02001473extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int);
Philipp Reisnere96c9632013-06-25 16:50:07 +02001474enum determine_dev_size {
Philipp Reisnerd752b262013-06-25 16:50:08 +02001475 DS_ERROR_SHRINK = -3,
1476 DS_ERROR_SPACE_MD = -2,
Philipp Reisnere96c9632013-06-25 16:50:07 +02001477 DS_ERROR = -1,
1478 DS_UNCHANGED = 0,
1479 DS_SHRUNK = 1,
Philipp Reisner57737ad2013-10-23 10:59:17 +02001480 DS_GREW = 2,
1481 DS_GREW_FROM_ZERO = 3,
Philipp Reisnere96c9632013-06-25 16:50:07 +02001482};
Philipp Reisnerd752b262013-06-25 16:50:08 +02001483extern enum determine_dev_size
Andreas Gruenbacher54761692011-05-30 16:15:21 +02001484drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local);
1485extern void resync_after_online_grow(struct drbd_device *);
Philipp Reisner8fe39aa2013-11-22 13:22:13 +01001486extern void drbd_reconsider_max_bio_size(struct drbd_device *device, struct drbd_backing_dev *bdev);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001487extern enum drbd_state_rv drbd_set_role(struct drbd_device *device,
Andreas Gruenbacherbf885f82010-12-08 00:39:32 +01001488 enum drbd_role new_role,
1489 int force);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001490extern bool conn_try_outdate_peer(struct drbd_connection *connection);
1491extern void conn_try_outdate_peer_async(struct drbd_connection *connection);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001492extern int drbd_khelper(struct drbd_device *device, char *cmd);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001493
1494/* drbd_worker.c */
Philipp Reisnerd40e5672014-04-28 18:43:14 +02001495/* bi_end_io handlers */
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001496extern void drbd_md_endio(struct bio *bio);
1497extern void drbd_peer_request_endio(struct bio *bio);
1498extern void drbd_request_endio(struct bio *bio);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001499extern int drbd_worker(struct drbd_thread *thi);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001500enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor);
1501void drbd_resync_after_changed(struct drbd_device *device);
1502extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side);
1503extern void resume_next_sg(struct drbd_device *device);
1504extern void suspend_other_sg(struct drbd_device *device);
1505extern int drbd_resync_finished(struct drbd_device *device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001506/* maybe rather drbd_main.c ? */
Lars Ellenberge37d2432014-04-01 23:53:30 +02001507extern void *drbd_md_get_buffer(struct drbd_device *device, const char *intent);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001508extern void drbd_md_put_buffer(struct drbd_device *device);
1509extern int drbd_md_sync_page_io(struct drbd_device *device,
Philipp Reisnerb411b362009-09-25 16:07:19 -07001510 struct drbd_backing_dev *bdev, sector_t sector, int rw);
Andreas Gruenbacher54761692011-05-30 16:15:21 +02001511extern void drbd_ov_out_of_sync_found(struct drbd_device *, sector_t, int);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001512extern void wait_until_done_or_force_detached(struct drbd_device *device,
Lars Ellenberg44edfb02012-09-27 13:03:45 +02001513 struct drbd_backing_dev *bdev, unsigned int *done);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001514extern void drbd_rs_controller_reset(struct drbd_device *device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001515
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001516static inline void ov_out_of_sync_print(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001517{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001518 if (device->ov_last_oos_size) {
Andreas Gruenbacherd0180172011-07-03 17:53:52 +02001519 drbd_err(device, "Out of sync: start=%llu, size=%lu (sectors)\n",
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001520 (unsigned long long)device->ov_last_oos_start,
1521 (unsigned long)device->ov_last_oos_size);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001522 }
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001523 device->ov_last_oos_size = 0;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001524}
1525
1526
Herbert Xu9534d672016-01-24 21:19:21 +08001527extern void drbd_csum_bio(struct crypto_ahash *, struct bio *, void *);
1528extern void drbd_csum_ee(struct crypto_ahash *, struct drbd_peer_request *, void *);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001529/* worker callbacks */
Andreas Gruenbacher99920dc2011-03-16 15:31:39 +01001530extern int w_e_end_data_req(struct drbd_work *, int);
1531extern int w_e_end_rsdata_req(struct drbd_work *, int);
1532extern int w_e_end_csum_rs_req(struct drbd_work *, int);
1533extern int w_e_end_ov_reply(struct drbd_work *, int);
1534extern int w_e_end_ov_req(struct drbd_work *, int);
1535extern int w_ov_finished(struct drbd_work *, int);
1536extern int w_resync_timer(struct drbd_work *, int);
1537extern int w_send_write_hint(struct drbd_work *, int);
Andreas Gruenbacher99920dc2011-03-16 15:31:39 +01001538extern int w_send_dblock(struct drbd_work *, int);
Andreas Gruenbacher99920dc2011-03-16 15:31:39 +01001539extern int w_send_read_req(struct drbd_work *, int);
Andreas Gruenbacher99920dc2011-03-16 15:31:39 +01001540extern int w_e_reissue(struct drbd_work *, int);
1541extern int w_restart_disk_io(struct drbd_work *, int);
Andreas Gruenbacher8f7bed72010-12-19 23:53:14 +01001542extern int w_send_out_of_sync(struct drbd_work *, int);
Andreas Gruenbacher99920dc2011-03-16 15:31:39 +01001543extern int w_start_resync(struct drbd_work *, int);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001544
1545extern void resync_timer_fn(unsigned long data);
Philipp Reisner370a43e2011-01-14 16:03:11 +01001546extern void start_resync_timer_fn(unsigned long data);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001547
Lars Ellenberga0fb3c42014-04-28 18:43:23 +02001548extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req);
1549
Philipp Reisnerb411b362009-09-25 16:07:19 -07001550/* drbd_receiver.c */
Andreas Gruenbacher753c61912011-07-22 11:14:41 +02001551extern int drbd_receiver(struct drbd_thread *thi);
Philipp Reisner1c03e522015-03-16 15:01:00 +01001552extern int drbd_ack_receiver(struct drbd_thread *thi);
Philipp Reisner668700b2015-03-16 16:08:29 +01001553extern void drbd_send_ping_wf(struct work_struct *ws);
1554extern void drbd_send_acks_wf(struct work_struct *ws);
Lars Ellenberge8299872014-04-28 18:43:19 +02001555extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device);
Lars Ellenbergad3fee72013-12-20 11:22:13 +01001556extern bool drbd_rs_should_slow_down(struct drbd_device *device, sector_t sector,
1557 bool throttle_if_app_is_waiting);
Andreas Gruenbacher54761692011-05-30 16:15:21 +02001558extern int drbd_submit_peer_request(struct drbd_device *,
Andreas Gruenbacherfbe29de2011-02-17 16:38:35 +01001559 struct drbd_peer_request *, const unsigned,
1560 const int);
Andreas Gruenbacher54761692011-05-30 16:15:21 +02001561extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001562extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64,
Andreas Gruenbacher0db55362011-04-06 16:09:15 +02001563 sector_t, unsigned int,
Lars Ellenberga0fb3c42014-04-28 18:43:23 +02001564 bool,
Andreas Gruenbacher0db55362011-04-06 16:09:15 +02001565 gfp_t) __must_hold(local);
Andreas Gruenbacher54761692011-05-30 16:15:21 +02001566extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *,
Andreas Gruenbacher3967deb2011-04-06 16:16:56 +02001567 int);
1568#define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0)
1569#define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1)
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001570extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001571extern void drbd_set_recv_tcq(struct drbd_device *device, int tcq_enabled);
1572extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001573extern int drbd_connected(struct drbd_peer_device *);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001574
Philipp Reisnerb411b362009-09-25 16:07:19 -07001575static inline void drbd_tcp_cork(struct socket *sock)
1576{
Lars Ellenberged439842011-04-23 14:45:14 +02001577 int val = 1;
Philipp Reisnere805b982014-11-10 17:21:15 +01001578 (void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
Lars Ellenberged439842011-04-23 14:45:14 +02001579 (char*)&val, sizeof(val));
Philipp Reisnerb411b362009-09-25 16:07:19 -07001580}
1581
1582static inline void drbd_tcp_uncork(struct socket *sock)
1583{
Lars Ellenberged439842011-04-23 14:45:14 +02001584 int val = 0;
Philipp Reisnere805b982014-11-10 17:21:15 +01001585 (void) kernel_setsockopt(sock, SOL_TCP, TCP_CORK,
Lars Ellenberged439842011-04-23 14:45:14 +02001586 (char*)&val, sizeof(val));
Philipp Reisnerb411b362009-09-25 16:07:19 -07001587}
1588
1589static inline void drbd_tcp_nodelay(struct socket *sock)
1590{
Lars Ellenberged439842011-04-23 14:45:14 +02001591 int val = 1;
Philipp Reisnere805b982014-11-10 17:21:15 +01001592 (void) kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY,
Lars Ellenberged439842011-04-23 14:45:14 +02001593 (char*)&val, sizeof(val));
Philipp Reisnerb411b362009-09-25 16:07:19 -07001594}
1595
1596static inline void drbd_tcp_quickack(struct socket *sock)
1597{
Lars Ellenberged439842011-04-23 14:45:14 +02001598 int val = 2;
Philipp Reisnere805b982014-11-10 17:21:15 +01001599 (void) kernel_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
Lars Ellenberged439842011-04-23 14:45:14 +02001600 (char*)&val, sizeof(val));
Philipp Reisnerb411b362009-09-25 16:07:19 -07001601}
1602
Philipp Reisnerd40e5672014-04-28 18:43:14 +02001603/* sets the number of 512 byte sectors of our virtual device */
1604static inline void drbd_set_my_capacity(struct drbd_device *device,
1605 sector_t size)
1606{
1607 /* set_capacity(device->this_bdev->bd_disk, size); */
1608 set_capacity(device->vdisk, size);
1609 device->this_bdev->bd_inode->i_size = (loff_t)size << 9;
1610}
1611
1612/*
1613 * used to submit our private bio
1614 */
1615static inline void drbd_generic_make_request(struct drbd_device *device,
1616 int fault_type, struct bio *bio)
1617{
1618 __release(local);
1619 if (!bio->bi_bdev) {
Lars Ellenbergf88c5d92014-03-27 14:10:55 +01001620 drbd_err(device, "drbd_generic_make_request: bio->bi_bdev == NULL\n");
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001621 bio->bi_error = -ENODEV;
1622 bio_endio(bio);
Philipp Reisnerd40e5672014-04-28 18:43:14 +02001623 return;
1624 }
1625
1626 if (drbd_insert_fault(device, fault_type))
Christoph Hellwig4246a0b2015-07-20 15:29:37 +02001627 bio_io_error(bio);
Philipp Reisnerd40e5672014-04-28 18:43:14 +02001628 else
1629 generic_make_request(bio);
1630}
1631
Philipp Reisner8fe39aa2013-11-22 13:22:13 +01001632void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev,
1633 enum write_ordering_e wo);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001634
1635/* drbd_proc.c */
1636extern struct proc_dir_entry *drbd_proc;
Emese Revfy7d4e9d02009-12-14 00:59:30 +01001637extern const struct file_operations drbd_proc_fops;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001638extern const char *drbd_conn_str(enum drbd_conns s);
1639extern const char *drbd_role_str(enum drbd_role s);
1640
1641/* drbd_actlog.c */
Lars Ellenberge4d7d6f2014-04-28 18:43:28 +02001642extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001643extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i);
Lars Ellenberg4dd726f2014-02-11 11:15:36 +01001644extern void drbd_al_begin_io_commit(struct drbd_device *device);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001645extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i);
Lars Ellenberg4dd726f2014-02-11 11:15:36 +01001646extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001647extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i);
1648extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector);
1649extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector);
1650extern int drbd_try_rs_begin_io(struct drbd_device *device, sector_t sector);
1651extern void drbd_rs_cancel_all(struct drbd_device *device);
1652extern int drbd_rs_del_all(struct drbd_device *device);
1653extern void drbd_rs_failed_io(struct drbd_device *device,
Philipp Reisnerb411b362009-09-25 16:07:19 -07001654 sector_t sector, int size);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001655extern void drbd_advance_rs_marks(struct drbd_device *device, unsigned long still_to_go);
Lars Ellenberg5ab7d2c2014-01-27 15:58:22 +01001656
1657enum update_sync_bits_mode { RECORD_RS_FAILED, SET_OUT_OF_SYNC, SET_IN_SYNC };
1658extern int __drbd_change_sync(struct drbd_device *device, sector_t sector, int size,
Andreas Gruenbacher179e20b82014-11-10 17:21:09 +01001659 enum update_sync_bits_mode mode);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001660#define drbd_set_in_sync(device, sector, size) \
Andreas Gruenbacher179e20b82014-11-10 17:21:09 +01001661 __drbd_change_sync(device, sector, size, SET_IN_SYNC)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001662#define drbd_set_out_of_sync(device, sector, size) \
Andreas Gruenbacher179e20b82014-11-10 17:21:09 +01001663 __drbd_change_sync(device, sector, size, SET_OUT_OF_SYNC)
Lars Ellenberg5ab7d2c2014-01-27 15:58:22 +01001664#define drbd_rs_failed_io(device, sector, size) \
Andreas Gruenbacher179e20b82014-11-10 17:21:09 +01001665 __drbd_change_sync(device, sector, size, RECORD_RS_FAILED)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001666extern void drbd_al_shrink(struct drbd_device *device);
Lars Ellenberg5f7c0122015-06-08 15:18:45 +02001667extern int drbd_al_initialize(struct drbd_device *, void *);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001668
Philipp Reisnerb411b362009-09-25 16:07:19 -07001669/* drbd_nl.c */
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001670/* state info broadcast */
1671struct sib_info {
1672 enum drbd_state_info_bcast_reason sib_reason;
1673 union {
1674 struct {
1675 char *helper_name;
1676 unsigned helper_exit_code;
1677 };
1678 struct {
1679 union drbd_state os;
1680 union drbd_state ns;
1681 };
1682 };
1683};
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001684void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001685
Andreas Gruenbachera2972842014-07-31 17:41:33 +02001686extern void notify_resource_state(struct sk_buff *,
1687 unsigned int,
1688 struct drbd_resource *,
1689 struct resource_info *,
1690 enum drbd_notification_type);
1691extern void notify_device_state(struct sk_buff *,
1692 unsigned int,
1693 struct drbd_device *,
1694 struct device_info *,
1695 enum drbd_notification_type);
1696extern void notify_connection_state(struct sk_buff *,
1697 unsigned int,
1698 struct drbd_connection *,
1699 struct connection_info *,
1700 enum drbd_notification_type);
1701extern void notify_peer_device_state(struct sk_buff *,
1702 unsigned int,
1703 struct drbd_peer_device *,
1704 struct peer_device_info *,
1705 enum drbd_notification_type);
1706extern void notify_helper(enum drbd_notification_type, struct drbd_device *,
1707 struct drbd_connection *, const char *, int);
1708
Philipp Reisnerb411b362009-09-25 16:07:19 -07001709/*
1710 * inline helper functions
1711 *************************/
1712
Lars Ellenberg45bb9122010-05-14 17:10:48 +02001713/* see also page_chain_add and friends in drbd_receiver.c */
1714static inline struct page *page_chain_next(struct page *page)
1715{
1716 return (struct page *)page_private(page);
1717}
1718#define page_chain_for_each(page) \
1719 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1720 page = page_chain_next(page))
1721#define page_chain_for_each_safe(page, n) \
1722 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1723
Lars Ellenberg45bb9122010-05-14 17:10:48 +02001724
Andreas Gruenbacher045417f2011-04-07 21:34:24 +02001725static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req)
Lars Ellenberg45bb9122010-05-14 17:10:48 +02001726{
Andreas Gruenbacherdb830c42011-02-04 15:57:48 +01001727 struct page *page = peer_req->pages;
Lars Ellenberg45bb9122010-05-14 17:10:48 +02001728 page_chain_for_each(page) {
1729 if (page_count(page) > 1)
1730 return 1;
1731 }
1732 return 0;
1733}
1734
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001735static inline union drbd_state drbd_read_state(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001736{
Andreas Gruenbacher6bbf53c2011-07-08 01:19:44 +02001737 struct drbd_resource *resource = device->resource;
Philipp Reisner78bae592011-03-28 15:40:12 +02001738 union drbd_state rv;
1739
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001740 rv.i = device->state.i;
Andreas Gruenbacher6bbf53c2011-07-08 01:19:44 +02001741 rv.susp = resource->susp;
1742 rv.susp_nod = resource->susp_nod;
1743 rv.susp_fen = resource->susp_fen;
Philipp Reisner78bae592011-03-28 15:40:12 +02001744
1745 return rv;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001746}
1747
Lars Ellenberg383606e2012-06-14 14:21:32 +02001748enum drbd_force_detach_flags {
Lars Ellenberga2a3c74f2012-09-22 12:26:57 +02001749 DRBD_READ_ERROR,
1750 DRBD_WRITE_ERROR,
Lars Ellenberg383606e2012-06-14 14:21:32 +02001751 DRBD_META_IO_ERROR,
1752 DRBD_FORCE_DETACH,
1753};
1754
Philipp Reisnerb411b362009-09-25 16:07:19 -07001755#define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001756static inline void __drbd_chk_io_error_(struct drbd_device *device,
Lars Ellenberga2a3c74f2012-09-22 12:26:57 +02001757 enum drbd_force_detach_flags df,
Lars Ellenberg383606e2012-06-14 14:21:32 +02001758 const char *where)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001759{
Philipp Reisnerdaeda1c2011-05-03 15:00:55 +02001760 enum drbd_io_error_p ep;
1761
1762 rcu_read_lock();
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001763 ep = rcu_dereference(device->ldev->disk_conf)->on_io_error;
Philipp Reisnerdaeda1c2011-05-03 15:00:55 +02001764 rcu_read_unlock();
1765 switch (ep) {
1766 case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */
Lars Ellenberga2a3c74f2012-09-22 12:26:57 +02001767 if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) {
Lars Ellenberg73835062010-05-27 11:51:56 +02001768 if (__ratelimit(&drbd_ratelimit_state))
Andreas Gruenbacherd0180172011-07-03 17:53:52 +02001769 drbd_err(device, "Local IO failed in %s.\n", where);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001770 if (device->state.disk > D_INCONSISTENT)
1771 _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001772 break;
1773 }
Lars Ellenberga2a3c74f2012-09-22 12:26:57 +02001774 /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001775 case EP_DETACH:
1776 case EP_CALL_HELPER:
Lars Ellenberga2a3c74f2012-09-22 12:26:57 +02001777 /* Remember whether we saw a READ or WRITE error.
1778 *
1779 * Recovery of the affected area for WRITE failure is covered
1780 * by the activity log.
1781 * READ errors may fall outside that area though. Certain READ
1782 * errors can be "healed" by writing good data to the affected
1783 * blocks, which triggers block re-allocation in lower layers.
1784 *
1785 * If we can not write the bitmap after a READ error,
1786 * we may need to trigger a full sync (see w_go_diskless()).
1787 *
1788 * Force-detach is not really an IO error, but rather a
1789 * desperate measure to try to deal with a completely
1790 * unresponsive lower level IO stack.
1791 * Still it should be treated as a WRITE error.
1792 *
1793 * Meta IO error is always WRITE error:
1794 * we read meta data only once during attach,
1795 * which will fail in case of errors.
1796 */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001797 set_bit(WAS_IO_ERROR, &device->flags);
Lars Ellenberga2a3c74f2012-09-22 12:26:57 +02001798 if (df == DRBD_READ_ERROR)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001799 set_bit(WAS_READ_ERROR, &device->flags);
Lars Ellenberga2a3c74f2012-09-22 12:26:57 +02001800 if (df == DRBD_FORCE_DETACH)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001801 set_bit(FORCE_DETACH, &device->flags);
1802 if (device->state.disk > D_FAILED) {
1803 _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL);
Andreas Gruenbacherd0180172011-07-03 17:53:52 +02001804 drbd_err(device,
Lars Ellenberg82f59cc2010-10-16 12:13:47 +02001805 "Local IO failed in %s. Detaching...\n", where);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001806 }
1807 break;
1808 }
1809}
1810
1811/**
1812 * 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 +02001813 * @device: DRBD device.
Philipp Reisnerb411b362009-09-25 16:07:19 -07001814 * @error: Error code passed to the IO completion callback
1815 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1816 *
1817 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1818 */
1819#define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001820static inline void drbd_chk_io_error_(struct drbd_device *device,
Lars Ellenberg383606e2012-06-14 14:21:32 +02001821 int error, enum drbd_force_detach_flags forcedetach, const char *where)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001822{
1823 if (error) {
1824 unsigned long flags;
Andreas Gruenbacher05008132011-07-07 14:19:42 +02001825 spin_lock_irqsave(&device->resource->req_lock, flags);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02001826 __drbd_chk_io_error_(device, forcedetach, where);
Andreas Gruenbacher05008132011-07-07 14:19:42 +02001827 spin_unlock_irqrestore(&device->resource->req_lock, flags);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001828 }
1829}
1830
1831
1832/**
1833 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1834 * @bdev: Meta data block device.
1835 *
1836 * BTW, for internal meta data, this happens to be the maximum capacity
1837 * we could agree upon with our peer node.
1838 */
Lars Ellenberg68e41a42013-03-19 18:16:45 +01001839static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001840{
Lars Ellenberg68e41a42013-03-19 18:16:45 +01001841 switch (bdev->md.meta_dev_idx) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07001842 case DRBD_MD_INDEX_INTERNAL:
1843 case DRBD_MD_INDEX_FLEX_INT:
1844 return bdev->md.md_offset + bdev->md.bm_offset;
1845 case DRBD_MD_INDEX_FLEX_EXT:
1846 default:
1847 return bdev->md.md_offset;
1848 }
1849}
1850
1851/**
1852 * drbd_md_last_sector() - Return the last sector number of the meta data area
1853 * @bdev: Meta data block device.
1854 */
1855static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1856{
Lars Ellenberg68e41a42013-03-19 18:16:45 +01001857 switch (bdev->md.meta_dev_idx) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07001858 case DRBD_MD_INDEX_INTERNAL:
1859 case DRBD_MD_INDEX_FLEX_INT:
Lars Ellenbergae8bf312013-03-19 18:16:43 +01001860 return bdev->md.md_offset + MD_4kB_SECT -1;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001861 case DRBD_MD_INDEX_FLEX_EXT:
1862 default:
Lars Ellenbergae8bf312013-03-19 18:16:43 +01001863 return bdev->md.md_offset + bdev->md.md_size_sect -1;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001864 }
1865}
1866
1867/* Returns the number of 512 byte sectors of the device */
1868static inline sector_t drbd_get_capacity(struct block_device *bdev)
1869{
1870 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
Mike Snitzer77304d22010-11-08 14:39:12 +01001871 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001872}
1873
1874/**
1875 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1876 * @bdev: Meta data block device.
1877 *
1878 * returns the capacity we announce to out peer. we clip ourselves at the
1879 * various MAX_SECTORS, because if we don't, current implementation will
1880 * oops sooner or later
1881 */
1882static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1883{
1884 sector_t s;
Philipp Reisnerdaeda1c2011-05-03 15:00:55 +02001885
Lars Ellenberg68e41a42013-03-19 18:16:45 +01001886 switch (bdev->md.meta_dev_idx) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07001887 case DRBD_MD_INDEX_INTERNAL:
1888 case DRBD_MD_INDEX_FLEX_INT:
1889 s = drbd_get_capacity(bdev->backing_bdev)
1890 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
Lars Ellenberg68e41a42013-03-19 18:16:45 +01001891 drbd_md_first_sector(bdev))
Philipp Reisnerb411b362009-09-25 16:07:19 -07001892 : 0;
1893 break;
1894 case DRBD_MD_INDEX_FLEX_EXT:
1895 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1896 drbd_get_capacity(bdev->backing_bdev));
1897 /* clip at maximum size the meta device can support */
1898 s = min_t(sector_t, s,
1899 BM_EXT_TO_SECT(bdev->md.md_size_sect
1900 - bdev->md.bm_offset));
1901 break;
1902 default:
1903 s = min_t(sector_t, DRBD_MAX_SECTORS,
1904 drbd_get_capacity(bdev->backing_bdev));
1905 }
1906 return s;
1907}
1908
1909/**
Lars Ellenberg3a4d4eb2013-03-19 18:16:44 +01001910 * drbd_md_ss() - Return the sector number of our meta data super block
Philipp Reisnerb411b362009-09-25 16:07:19 -07001911 * @bdev: Meta data block device.
1912 */
Lars Ellenberg3a4d4eb2013-03-19 18:16:44 +01001913static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001914{
Lars Ellenberg3a4d4eb2013-03-19 18:16:44 +01001915 const int meta_dev_idx = bdev->md.meta_dev_idx;
Philipp Reisnerdaeda1c2011-05-03 15:00:55 +02001916
Lars Ellenberg3a4d4eb2013-03-19 18:16:44 +01001917 if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001918 return 0;
Lars Ellenberg3a4d4eb2013-03-19 18:16:44 +01001919
1920 /* Since drbd08, internal meta data is always "flexible".
1921 * position: last 4k aligned block of 4k size */
1922 if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL ||
1923 meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)
1924 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8;
1925
1926 /* external, some index; this is the old fixed size layout */
1927 return MD_128MB_SECT * bdev->md.meta_dev_idx;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001928}
1929
1930static inline void
Philipp Reisnerb411b362009-09-25 16:07:19 -07001931drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1932{
1933 unsigned long flags;
1934 spin_lock_irqsave(&q->q_lock, flags);
1935 list_add_tail(&w->list, &q->q);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001936 spin_unlock_irqrestore(&q->q_lock, flags);
Lars Ellenberg8c0785a2011-10-19 11:50:57 +02001937 wake_up(&q->q_wait);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001938}
1939
Lars Ellenberge334f552014-02-11 09:30:49 +01001940static inline void
Lars Ellenberg15e26f62014-04-28 11:43:21 +02001941drbd_queue_work_if_unqueued(struct drbd_work_queue *q, struct drbd_work *w)
1942{
1943 unsigned long flags;
1944 spin_lock_irqsave(&q->q_lock, flags);
1945 if (list_empty_careful(&w->list))
1946 list_add_tail(&w->list, &q->q);
1947 spin_unlock_irqrestore(&q->q_lock, flags);
1948 wake_up(&q->q_wait);
1949}
1950
1951static inline void
Lars Ellenberge334f552014-02-11 09:30:49 +01001952drbd_device_post_work(struct drbd_device *device, int work_bit)
1953{
1954 if (!test_and_set_bit(work_bit, &device->flags)) {
1955 struct drbd_connection *connection =
1956 first_peer_device(device)->connection;
1957 struct drbd_work_queue *q = &connection->sender_work;
1958 if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags))
1959 wake_up(&q->q_wait);
1960 }
1961}
1962
Andreas Gruenbacherb5043c52011-07-28 15:56:02 +02001963extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue);
1964
Philipp Reisner668700b2015-03-16 16:08:29 +01001965/* To get the ack_receiver out of the blocking network stack,
1966 * so it can change its sk_rcvtimeo from idle- to ping-timeout,
1967 * and send a ping, we need to send a signal.
1968 * Which signal we send is irrelevant. */
1969static inline void wake_ack_receiver(struct drbd_connection *connection)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001970{
Philipp Reisner668700b2015-03-16 16:08:29 +01001971 struct task_struct *task = connection->ack_receiver.task;
1972 if (task && get_t_state(&connection->ack_receiver) == RUNNING)
1973 force_sig(SIGXCPU, task);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001974}
1975
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001976static inline void request_ping(struct drbd_connection *connection)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001977{
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001978 set_bit(SEND_PING, &connection->flags);
Philipp Reisner668700b2015-03-16 16:08:29 +01001979 wake_ack_receiver(connection);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001980}
1981
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001982extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001983extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001984extern int conn_send_command(struct drbd_connection *, struct drbd_socket *,
Andreas Gruenbacherdba58582011-03-29 16:55:40 +02001985 enum drbd_packet, unsigned int, void *,
1986 unsigned int);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001987extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *,
Andreas Gruenbacherdba58582011-03-29 16:55:40 +02001988 enum drbd_packet, unsigned int, void *,
1989 unsigned int);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001990
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001991extern int drbd_send_ping(struct drbd_connection *connection);
1992extern int drbd_send_ping_ack(struct drbd_connection *connection);
Andreas Gruenbacher69a22772011-08-09 00:47:13 +02001993extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state);
Andreas Gruenbacherbde89a92011-05-30 16:32:41 +02001994extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001995
1996static inline void drbd_thread_stop(struct drbd_thread *thi)
1997{
Andreas Gruenbacher81e84652010-12-09 15:03:57 +01001998 _drbd_thread_stop(thi, false, true);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001999}
2000
2001static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
2002{
Andreas Gruenbacher81e84652010-12-09 15:03:57 +01002003 _drbd_thread_stop(thi, false, false);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002004}
2005
2006static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
2007{
Andreas Gruenbacher81e84652010-12-09 15:03:57 +01002008 _drbd_thread_stop(thi, true, false);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002009}
2010
2011/* counts how many answer packets packets we expect from our peer,
2012 * for either explicit application requests,
2013 * or implicit barrier packets as necessary.
2014 * increased:
2015 * w_send_barrier
Andreas Gruenbacher8554df12011-01-25 15:37:43 +01002016 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002017 * it is much easier and equally valid to count what we queue for the
2018 * worker, even before it actually was queued or send.
2019 * (drbd_make_request_common; recovery path on read io-error)
2020 * decreased:
2021 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
Andreas Gruenbacher8554df12011-01-25 15:37:43 +01002022 * _req_mod(req, DATA_RECEIVED)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002023 * [from receive_DataReply]
Andreas Gruenbacher8554df12011-01-25 15:37:43 +01002024 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002025 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
2026 * for some reason it is NOT decreased in got_NegAck,
2027 * but in the resulting cleanup code from report_params.
2028 * we should try to remember the reason for that...
Andreas Gruenbacher8554df12011-01-25 15:37:43 +01002029 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
2030 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002031 * [from tl_clear_barrier]
2032 */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002033static inline void inc_ap_pending(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002034{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002035 atomic_inc(&device->ap_pending_cnt);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002036}
2037
Philipp Reisner49559d82011-02-21 14:19:44 +01002038#define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002039 if (atomic_read(&device->which) < 0) \
Andreas Gruenbacherd0180172011-07-03 17:53:52 +02002040 drbd_err(device, "in %s:%d: " #which " = %d < 0 !\n", \
Philipp Reisner49559d82011-02-21 14:19:44 +01002041 func, line, \
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002042 atomic_read(&device->which))
Philipp Reisnerb411b362009-09-25 16:07:19 -07002043
Joe Perches659b2e32014-03-25 12:35:05 -07002044#define dec_ap_pending(device) _dec_ap_pending(device, __func__, __LINE__)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002045static inline void _dec_ap_pending(struct drbd_device *device, const char *func, int line)
Philipp Reisner49559d82011-02-21 14:19:44 +01002046{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002047 if (atomic_dec_and_test(&device->ap_pending_cnt))
2048 wake_up(&device->misc_wait);
Philipp Reisner49559d82011-02-21 14:19:44 +01002049 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line);
2050}
Philipp Reisnerb411b362009-09-25 16:07:19 -07002051
2052/* counts how many resync-related answers we still expect from the peer
2053 * increase decrease
2054 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
Lucas De Marchi25985ed2011-03-30 22:57:33 -03002055 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002056 * (or P_NEG_ACK with ID_SYNCER)
2057 */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002058static inline void inc_rs_pending(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002059{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002060 atomic_inc(&device->rs_pending_cnt);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002061}
2062
Joe Perches659b2e32014-03-25 12:35:05 -07002063#define dec_rs_pending(device) _dec_rs_pending(device, __func__, __LINE__)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002064static inline void _dec_rs_pending(struct drbd_device *device, const char *func, int line)
Philipp Reisner49559d82011-02-21 14:19:44 +01002065{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002066 atomic_dec(&device->rs_pending_cnt);
Philipp Reisner49559d82011-02-21 14:19:44 +01002067 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line);
2068}
Philipp Reisnerb411b362009-09-25 16:07:19 -07002069
2070/* counts how many answers we still need to send to the peer.
2071 * increased on
2072 * receive_Data unless protocol A;
2073 * we need to send a P_RECV_ACK (proto B)
2074 * or P_WRITE_ACK (proto C)
2075 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2076 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2077 * receive_Barrier_* we need to send a P_BARRIER_ACK
2078 */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002079static inline void inc_unacked(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002080{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002081 atomic_inc(&device->unacked_cnt);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002082}
2083
Joe Perches659b2e32014-03-25 12:35:05 -07002084#define dec_unacked(device) _dec_unacked(device, __func__, __LINE__)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002085static inline void _dec_unacked(struct drbd_device *device, const char *func, int line)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002086{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002087 atomic_dec(&device->unacked_cnt);
Philipp Reisner49559d82011-02-21 14:19:44 +01002088 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002089}
2090
Joe Perches659b2e32014-03-25 12:35:05 -07002091#define sub_unacked(device, n) _sub_unacked(device, n, __func__, __LINE__)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002092static inline void _sub_unacked(struct drbd_device *device, int n, const char *func, int line)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002093{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002094 atomic_sub(n, &device->unacked_cnt);
Philipp Reisner49559d82011-02-21 14:19:44 +01002095 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002096}
2097
Lars Ellenberg5ab7d2c2014-01-27 15:58:22 +01002098static inline bool is_sync_state(enum drbd_conns connection_state)
2099{
2100 return
2101 (connection_state == C_SYNC_SOURCE
2102 || connection_state == C_SYNC_TARGET
2103 || connection_state == C_PAUSED_SYNC_S
2104 || connection_state == C_PAUSED_SYNC_T);
2105}
2106
Philipp Reisnerb411b362009-09-25 16:07:19 -07002107/**
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002108 * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev
Andreas Gruenbacherd1b80852014-09-11 14:29:09 +02002109 * @_device: DRBD device.
2110 * @_min_state: Minimum device state required for success.
Philipp Reisnerb411b362009-09-25 16:07:19 -07002111 *
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002112 * You have to call put_ldev() when finished working with device->ldev.
Philipp Reisnerb411b362009-09-25 16:07:19 -07002113 */
Andreas Gruenbacherd1b80852014-09-11 14:29:09 +02002114#define get_ldev_if_state(_device, _min_state) \
2115 (_get_ldev_if_state((_device), (_min_state)) ? \
2116 ({ __acquire(x); true; }) : false)
2117#define get_ldev(_device) get_ldev_if_state(_device, D_INCONSISTENT)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002118
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002119static inline void put_ldev(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002120{
Andreas Gruenbacher11f8b2b2014-09-11 14:29:05 +02002121 enum drbd_disk_state disk_state = device->state.disk;
Lars Ellenbergba3c6fb2014-02-11 08:57:18 +01002122 /* We must check the state *before* the atomic_dec becomes visible,
2123 * or we have a theoretical race where someone hitting zero,
2124 * while state still D_FAILED, will then see D_DISKLESS in the
2125 * condition below and calling into destroy, where he must not, yet. */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002126 int i = atomic_dec_return(&device->local_cnt);
Lars Ellenberg9a0d9d02011-05-02 11:51:31 +02002127
2128 /* This may be called from some endio handler,
2129 * so we must not sleep here. */
2130
Philipp Reisnerb411b362009-09-25 16:07:19 -07002131 __release(local);
Andreas Gruenbacher0b0ba1e2011-06-27 16:23:33 +02002132 D_ASSERT(device, i >= 0);
Lars Ellenberge9e6f3e2010-09-14 20:26:27 +02002133 if (i == 0) {
Andreas Gruenbacher11f8b2b2014-09-11 14:29:05 +02002134 if (disk_state == D_DISKLESS)
Lars Ellenberg82f59cc2010-10-16 12:13:47 +02002135 /* even internal references gone, safe to destroy */
Lars Ellenberge334f552014-02-11 09:30:49 +01002136 drbd_device_post_work(device, DESTROY_DISK);
Andreas Gruenbacher11f8b2b2014-09-11 14:29:05 +02002137 if (disk_state == D_FAILED)
Lars Ellenberg82f59cc2010-10-16 12:13:47 +02002138 /* all application IO references gone. */
Lars Ellenberge334f552014-02-11 09:30:49 +01002139 if (!test_and_set_bit(GOING_DISKLESS, &device->flags))
2140 drbd_device_post_work(device, GO_DISKLESS);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002141 wake_up(&device->misc_wait);
Lars Ellenberge9e6f3e2010-09-14 20:26:27 +02002142 }
Philipp Reisnerb411b362009-09-25 16:07:19 -07002143}
2144
2145#ifndef __CHECKER__
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002146static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002147{
2148 int io_allowed;
2149
Lars Ellenberg82f59cc2010-10-16 12:13:47 +02002150 /* never get a reference while D_DISKLESS */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002151 if (device->state.disk == D_DISKLESS)
Lars Ellenberg82f59cc2010-10-16 12:13:47 +02002152 return 0;
2153
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002154 atomic_inc(&device->local_cnt);
2155 io_allowed = (device->state.disk >= mins);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002156 if (!io_allowed)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002157 put_ldev(device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002158 return io_allowed;
2159}
2160#else
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002161extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002162#endif
2163
Philipp Reisnerb411b362009-09-25 16:07:19 -07002164/* this throttles on-the-fly application requests
2165 * according to max_buffers settings;
2166 * maybe re-implement using semaphores? */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002167static inline int drbd_get_max_buffers(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002168{
Philipp Reisner44ed1672011-04-19 17:10:19 +02002169 struct net_conf *nc;
2170 int mxb;
2171
2172 rcu_read_lock();
Andreas Gruenbachera6b32bc2011-05-31 14:33:49 +02002173 nc = rcu_dereference(first_peer_device(device)->connection->net_conf);
Philipp Reisner44ed1672011-04-19 17:10:19 +02002174 mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */
2175 rcu_read_unlock();
2176
Philipp Reisnerb411b362009-09-25 16:07:19 -07002177 return mxb;
2178}
2179
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002180static inline int drbd_state_is_stable(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002181{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002182 union drbd_dev_state s = device->state;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002183
2184 /* DO NOT add a default clause, we want the compiler to warn us
2185 * for any newly introduced state we may have forgotten to add here */
2186
2187 switch ((enum drbd_conns)s.conn) {
2188 /* new io only accepted when there is no connection, ... */
2189 case C_STANDALONE:
2190 case C_WF_CONNECTION:
2191 /* ... or there is a well established connection. */
2192 case C_CONNECTED:
2193 case C_SYNC_SOURCE:
2194 case C_SYNC_TARGET:
2195 case C_VERIFY_S:
2196 case C_VERIFY_T:
2197 case C_PAUSED_SYNC_S:
2198 case C_PAUSED_SYNC_T:
Philipp Reisner67531712010-10-27 12:21:30 +02002199 case C_AHEAD:
2200 case C_BEHIND:
Philipp Reisner37190942010-11-10 12:08:37 +01002201 /* transitional states, IO allowed */
Philipp Reisnerb411b362009-09-25 16:07:19 -07002202 case C_DISCONNECTING:
2203 case C_UNCONNECTED:
2204 case C_TIMEOUT:
2205 case C_BROKEN_PIPE:
2206 case C_NETWORK_FAILURE:
2207 case C_PROTOCOL_ERROR:
2208 case C_TEAR_DOWN:
2209 case C_WF_REPORT_PARAMS:
2210 case C_STARTING_SYNC_S:
2211 case C_STARTING_SYNC_T:
Philipp Reisner37190942010-11-10 12:08:37 +01002212 break;
2213
2214 /* Allow IO in BM exchange states with new protocols */
Philipp Reisnerb411b362009-09-25 16:07:19 -07002215 case C_WF_BITMAP_S:
Andreas Gruenbachera6b32bc2011-05-31 14:33:49 +02002216 if (first_peer_device(device)->connection->agreed_pro_version < 96)
Philipp Reisner37190942010-11-10 12:08:37 +01002217 return 0;
2218 break;
2219
2220 /* no new io accepted in these states */
Philipp Reisnerb411b362009-09-25 16:07:19 -07002221 case C_WF_BITMAP_T:
2222 case C_WF_SYNC_UUID:
2223 case C_MASK:
2224 /* not "stable" */
2225 return 0;
2226 }
2227
2228 switch ((enum drbd_disk_state)s.disk) {
2229 case D_DISKLESS:
2230 case D_INCONSISTENT:
2231 case D_OUTDATED:
2232 case D_CONSISTENT:
2233 case D_UP_TO_DATE:
Philipp Reisner5ca1de02011-06-28 17:01:19 +02002234 case D_FAILED:
Philipp Reisnerb411b362009-09-25 16:07:19 -07002235 /* disk state is stable as well. */
2236 break;
2237
Philipp Reisnerd942ae42011-05-31 13:07:24 +02002238 /* no new io accepted during transitional states */
Philipp Reisnerb411b362009-09-25 16:07:19 -07002239 case D_ATTACHING:
Philipp Reisnerb411b362009-09-25 16:07:19 -07002240 case D_NEGOTIATING:
2241 case D_UNKNOWN:
2242 case D_MASK:
2243 /* not "stable" */
2244 return 0;
2245 }
2246
2247 return 1;
2248}
2249
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002250static inline int drbd_suspended(struct drbd_device *device)
Philipp Reisnerfb22c402010-09-08 23:20:21 +02002251{
Andreas Gruenbacher6bbf53c2011-07-08 01:19:44 +02002252 struct drbd_resource *resource = device->resource;
Philipp Reisner8e0af252011-03-28 16:18:39 +02002253
Andreas Gruenbacher6bbf53c2011-07-08 01:19:44 +02002254 return resource->susp || resource->susp_fen || resource->susp_nod;
Philipp Reisnerfb22c402010-09-08 23:20:21 +02002255}
2256
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002257static inline bool may_inc_ap_bio(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002258{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002259 int mxb = drbd_get_max_buffers(device);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002260
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002261 if (drbd_suspended(device))
Andreas Gruenbacher1b881ef2010-12-13 18:03:38 +01002262 return false;
Philipp Reisner7dbb4382013-02-28 10:30:19 +01002263 if (atomic_read(&device->suspend_cnt))
Andreas Gruenbacher1b881ef2010-12-13 18:03:38 +01002264 return false;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002265
2266 /* to avoid potential deadlock or bitmap corruption,
2267 * in various places, we only allow new application io
2268 * to start during "stable" states. */
2269
2270 /* no new io accepted when attaching or detaching the disk */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002271 if (!drbd_state_is_stable(device))
Andreas Gruenbacher1b881ef2010-12-13 18:03:38 +01002272 return false;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002273
2274 /* since some older kernels don't have atomic_add_unless,
2275 * and we are within the spinlock anyways, we have this workaround. */
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002276 if (atomic_read(&device->ap_bio_cnt) > mxb)
Andreas Gruenbacher1b881ef2010-12-13 18:03:38 +01002277 return false;
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002278 if (test_bit(BITMAP_IO, &device->flags))
Andreas Gruenbacher1b881ef2010-12-13 18:03:38 +01002279 return false;
2280 return true;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002281}
2282
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002283static inline bool inc_ap_bio_cond(struct drbd_device *device)
Philipp Reisner8869d682010-11-17 18:24:19 +01002284{
Andreas Gruenbacher1b881ef2010-12-13 18:03:38 +01002285 bool rv = false;
Philipp Reisner8869d682010-11-17 18:24:19 +01002286
Andreas Gruenbacher05008132011-07-07 14:19:42 +02002287 spin_lock_irq(&device->resource->req_lock);
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002288 rv = may_inc_ap_bio(device);
Philipp Reisner8869d682010-11-17 18:24:19 +01002289 if (rv)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002290 atomic_inc(&device->ap_bio_cnt);
Andreas Gruenbacher05008132011-07-07 14:19:42 +02002291 spin_unlock_irq(&device->resource->req_lock);
Philipp Reisner8869d682010-11-17 18:24:19 +01002292
2293 return rv;
2294}
2295
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002296static inline void inc_ap_bio(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002297{
Philipp Reisnerb411b362009-09-25 16:07:19 -07002298 /* we wait here
2299 * as long as the device is suspended
2300 * until the bitmap is no longer on the fly during connection
Philipp Reisnerd942ae42011-05-31 13:07:24 +02002301 * handshake as long as we would exceed the max_buffer limit.
Philipp Reisnerb411b362009-09-25 16:07:19 -07002302 *
2303 * to avoid races with the reconnect code,
2304 * we need to atomic_inc within the spinlock. */
2305
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002306 wait_event(device->misc_wait, inc_ap_bio_cond(device));
Philipp Reisnerb411b362009-09-25 16:07:19 -07002307}
2308
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002309static inline void dec_ap_bio(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002310{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002311 int mxb = drbd_get_max_buffers(device);
2312 int ap_bio = atomic_dec_return(&device->ap_bio_cnt);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002313
Andreas Gruenbacher0b0ba1e2011-06-27 16:23:33 +02002314 D_ASSERT(device, ap_bio >= 0);
Lars Ellenberg7ee1fb92012-06-19 10:27:58 +02002315
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002316 if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) {
2317 if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags))
Andreas Gruenbacher84b8c062011-07-28 15:27:51 +02002318 drbd_queue_work(&first_peer_device(device)->
2319 connection->sender_work,
2320 &device->bm_io_work.w);
Lars Ellenberg7ee1fb92012-06-19 10:27:58 +02002321 }
2322
Philipp Reisnerb411b362009-09-25 16:07:19 -07002323 /* this currently does wake_up for every dec_ap_bio!
2324 * maybe rather introduce some type of hysteresis?
2325 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2326 if (ap_bio < mxb)
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002327 wake_up(&device->misc_wait);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002328}
2329
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002330static inline bool verify_can_do_stop_sector(struct drbd_device *device)
Lars Ellenberg58ffa582012-07-26 14:09:49 +02002331{
Andreas Gruenbachera6b32bc2011-05-31 14:33:49 +02002332 return first_peer_device(device)->connection->agreed_pro_version >= 97 &&
2333 first_peer_device(device)->connection->agreed_pro_version != 100;
Lars Ellenberg58ffa582012-07-26 14:09:49 +02002334}
2335
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002336static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002337{
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002338 int changed = device->ed_uuid != val;
2339 device->ed_uuid = val;
Lars Ellenberg62b0da32011-01-20 13:25:21 +01002340 return changed;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002341}
2342
Andreas Gruenbacherb30ab792011-07-03 13:26:43 +02002343static inline int drbd_queue_order_type(struct drbd_device *device)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002344{
2345 /* sorry, we currently have no working implementation
2346 * of distributed TCQ stuff */
2347#ifndef QUEUE_ORDERED_NONE
2348#define QUEUE_ORDERED_NONE 0
2349#endif
2350 return QUEUE_ORDERED_NONE;
2351}
2352
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +02002353static inline struct drbd_connection *first_connection(struct drbd_resource *resource)
2354{
Lars Ellenbergec4a3402014-04-28 18:43:35 +02002355 return list_first_entry_or_null(&resource->connections,
Andreas Gruenbacher77c556f2011-06-08 22:17:38 +02002356 struct drbd_connection, connections);
2357}
2358
Philipp Reisnerb411b362009-09-25 16:07:19 -07002359#endif