aboutsummaryrefslogtreecommitdiffstats
path: root/include
diff options
context:
space:
mode:
Diffstat (limited to 'include')
-rw-r--r--include/asm-generic/cacheflush.h1
-rw-r--r--include/linux/backing-dev.h13
-rw-r--r--include/linux/bio.h20
-rw-r--r--include/linux/blkdev.h56
-rw-r--r--include/linux/cgroup_subsys.h6
-rw-r--r--include/linux/connector.h2
-rw-r--r--include/linux/drbd.h343
-rw-r--r--include/linux/drbd_limits.h137
-rw-r--r--include/linux/drbd_nl.h137
-rw-r--r--include/linux/drbd_tag_magic.h83
-rw-r--r--include/linux/fs.h5
-rw-r--r--include/linux/genhd.h1
-rw-r--r--include/linux/iocontext.h14
-rw-r--r--include/linux/lru_cache.h294
-rw-r--r--include/linux/writeback.h1
15 files changed, 1085 insertions, 28 deletions
diff --git a/include/asm-generic/cacheflush.h b/include/asm-generic/cacheflush.h
index ba4ec39a1131..57b5c3c82e86 100644
--- a/include/asm-generic/cacheflush.h
+++ b/include/asm-generic/cacheflush.h
@@ -13,6 +13,7 @@
13#define flush_cache_dup_mm(mm) do { } while (0) 13#define flush_cache_dup_mm(mm) do { } while (0)
14#define flush_cache_range(vma, start, end) do { } while (0) 14#define flush_cache_range(vma, start, end) do { } while (0)
15#define flush_cache_page(vma, vmaddr, pfn) do { } while (0) 15#define flush_cache_page(vma, vmaddr, pfn) do { } while (0)
16#define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 0
16#define flush_dcache_page(page) do { } while (0) 17#define flush_dcache_page(page) do { } while (0)
17#define flush_dcache_mmap_lock(mapping) do { } while (0) 18#define flush_dcache_mmap_lock(mapping) do { } while (0)
18#define flush_dcache_mmap_unlock(mapping) do { } while (0) 19#define flush_dcache_mmap_unlock(mapping) do { } while (0)
diff --git a/include/linux/backing-dev.h b/include/linux/backing-dev.h
index b449e738533a..fcbc26af00e4 100644
--- a/include/linux/backing-dev.h
+++ b/include/linux/backing-dev.h
@@ -331,4 +331,17 @@ static inline int bdi_sched_wait(void *word)
331 return 0; 331 return 0;
332} 332}
333 333
334static inline void blk_run_backing_dev(struct backing_dev_info *bdi,
335 struct page *page)
336{
337 if (bdi && bdi->unplug_io_fn)
338 bdi->unplug_io_fn(bdi, page);
339}
340
341static inline void blk_run_address_space(struct address_space *mapping)
342{
343 if (mapping)
344 blk_run_backing_dev(mapping->backing_dev_info, NULL);
345}
346
334#endif /* _LINUX_BACKING_DEV_H */ 347#endif /* _LINUX_BACKING_DEV_H */
diff --git a/include/linux/bio.h b/include/linux/bio.h
index 5be93f18d842..7fc5606e6ea5 100644
--- a/include/linux/bio.h
+++ b/include/linux/bio.h
@@ -391,6 +391,18 @@ extern struct bio *bio_copy_kern(struct request_queue *, void *, unsigned int,
391 gfp_t, int); 391 gfp_t, int);
392extern void bio_set_pages_dirty(struct bio *bio); 392extern void bio_set_pages_dirty(struct bio *bio);
393extern void bio_check_pages_dirty(struct bio *bio); 393extern void bio_check_pages_dirty(struct bio *bio);
394
395#ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
396# error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
397#endif
398#if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
399extern void bio_flush_dcache_pages(struct bio *bi);
400#else
401static inline void bio_flush_dcache_pages(struct bio *bi)
402{
403}
404#endif
405
394extern struct bio *bio_copy_user(struct request_queue *, struct rq_map_data *, 406extern struct bio *bio_copy_user(struct request_queue *, struct rq_map_data *,
395 unsigned long, unsigned int, int, gfp_t); 407 unsigned long, unsigned int, int, gfp_t);
396extern struct bio *bio_copy_user_iov(struct request_queue *, 408extern struct bio *bio_copy_user_iov(struct request_queue *,
@@ -450,11 +462,8 @@ extern struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly;
450/* 462/*
451 * remember never ever reenable interrupts between a bvec_kmap_irq and 463 * remember never ever reenable interrupts between a bvec_kmap_irq and
452 * bvec_kunmap_irq! 464 * bvec_kunmap_irq!
453 *
454 * This function MUST be inlined - it plays with the CPU interrupt flags.
455 */ 465 */
456static __always_inline char *bvec_kmap_irq(struct bio_vec *bvec, 466static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags)
457 unsigned long *flags)
458{ 467{
459 unsigned long addr; 468 unsigned long addr;
460 469
@@ -470,8 +479,7 @@ static __always_inline char *bvec_kmap_irq(struct bio_vec *bvec,
470 return (char *) addr + bvec->bv_offset; 479 return (char *) addr + bvec->bv_offset;
471} 480}
472 481
473static __always_inline void bvec_kunmap_irq(char *buffer, 482static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags)
474 unsigned long *flags)
475{ 483{
476 unsigned long ptr = (unsigned long) buffer & PAGE_MASK; 484 unsigned long ptr = (unsigned long) buffer & PAGE_MASK;
477 485
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 221cecd86bd3..784a919aa0d0 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -312,13 +312,17 @@ struct queue_limits {
312 unsigned int io_min; 312 unsigned int io_min;
313 unsigned int io_opt; 313 unsigned int io_opt;
314 unsigned int max_discard_sectors; 314 unsigned int max_discard_sectors;
315 unsigned int discard_granularity;
316 unsigned int discard_alignment;
315 317
316 unsigned short logical_block_size; 318 unsigned short logical_block_size;
317 unsigned short max_hw_segments; 319 unsigned short max_hw_segments;
318 unsigned short max_phys_segments; 320 unsigned short max_phys_segments;
319 321
320 unsigned char misaligned; 322 unsigned char misaligned;
323 unsigned char discard_misaligned;
321 unsigned char no_cluster; 324 unsigned char no_cluster;
325 signed char discard_zeroes_data;
322}; 326};
323 327
324struct request_queue 328struct request_queue
@@ -749,6 +753,17 @@ struct req_iterator {
749#define rq_iter_last(rq, _iter) \ 753#define rq_iter_last(rq, _iter) \
750 (_iter.bio->bi_next == NULL && _iter.i == _iter.bio->bi_vcnt-1) 754 (_iter.bio->bi_next == NULL && _iter.i == _iter.bio->bi_vcnt-1)
751 755
756#ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
757# error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
758#endif
759#if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
760extern void rq_flush_dcache_pages(struct request *rq);
761#else
762static inline void rq_flush_dcache_pages(struct request *rq)
763{
764}
765#endif
766
752extern int blk_register_queue(struct gendisk *disk); 767extern int blk_register_queue(struct gendisk *disk);
753extern void blk_unregister_queue(struct gendisk *disk); 768extern void blk_unregister_queue(struct gendisk *disk);
754extern void register_disk(struct gendisk *dev); 769extern void register_disk(struct gendisk *dev);
@@ -823,19 +838,6 @@ static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
823 return bdev->bd_disk->queue; 838 return bdev->bd_disk->queue;
824} 839}
825 840
826static inline void blk_run_backing_dev(struct backing_dev_info *bdi,
827 struct page *page)
828{
829 if (bdi && bdi->unplug_io_fn)
830 bdi->unplug_io_fn(bdi, page);
831}
832
833static inline void blk_run_address_space(struct address_space *mapping)
834{
835 if (mapping)
836 blk_run_backing_dev(mapping->backing_dev_info, NULL);
837}
838
839/* 841/*
840 * blk_rq_pos() : the current sector 842 * blk_rq_pos() : the current sector
841 * blk_rq_bytes() : bytes left in the entire request 843 * blk_rq_bytes() : bytes left in the entire request
@@ -1134,6 +1136,34 @@ static inline int bdev_alignment_offset(struct block_device *bdev)
1134 return q->limits.alignment_offset; 1136 return q->limits.alignment_offset;
1135} 1137}
1136 1138
1139static inline int queue_discard_alignment(struct request_queue *q)
1140{
1141 if (q->limits.discard_misaligned)
1142 return -1;
1143
1144 return q->limits.discard_alignment;
1145}
1146
1147static inline int queue_sector_discard_alignment(struct request_queue *q,
1148 sector_t sector)
1149{
1150 return ((sector << 9) - q->limits.discard_alignment)
1151 & (q->limits.discard_granularity - 1);
1152}
1153
1154static inline unsigned int queue_discard_zeroes_data(struct request_queue *q)
1155{
1156 if (q->limits.discard_zeroes_data == 1)
1157 return 1;
1158
1159 return 0;
1160}
1161
1162static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev)
1163{
1164 return queue_discard_zeroes_data(bdev_get_queue(bdev));
1165}
1166
1137static inline int queue_dma_alignment(struct request_queue *q) 1167static inline int queue_dma_alignment(struct request_queue *q)
1138{ 1168{
1139 return q ? q->dma_alignment : 511; 1169 return q ? q->dma_alignment : 511;
diff --git a/include/linux/cgroup_subsys.h b/include/linux/cgroup_subsys.h
index 9c8d31bacf46..ccefff02b6cb 100644
--- a/include/linux/cgroup_subsys.h
+++ b/include/linux/cgroup_subsys.h
@@ -60,3 +60,9 @@ SUBSYS(net_cls)
60#endif 60#endif
61 61
62/* */ 62/* */
63
64#ifdef CONFIG_BLK_CGROUP
65SUBSYS(blkio)
66#endif
67
68/* */
diff --git a/include/linux/connector.h b/include/linux/connector.h
index 3a14615fd35c..72ba63eb83c5 100644
--- a/include/linux/connector.h
+++ b/include/linux/connector.h
@@ -43,6 +43,8 @@
43#define CN_DST_VAL 0x1 43#define CN_DST_VAL 0x1
44#define CN_IDX_DM 0x7 /* Device Mapper */ 44#define CN_IDX_DM 0x7 /* Device Mapper */
45#define CN_VAL_DM_USERSPACE_LOG 0x1 45#define CN_VAL_DM_USERSPACE_LOG 0x1
46#define CN_IDX_DRBD 0x8
47#define CN_VAL_DRBD 0x1
46 48
47#define CN_NETLINK_USERS 8 49#define CN_NETLINK_USERS 8
48 50
diff --git a/include/linux/drbd.h b/include/linux/drbd.h
new file mode 100644
index 000000000000..e84f4733cb55
--- /dev/null
+++ b/include/linux/drbd.h
@@ -0,0 +1,343 @@
1/*
2 drbd.h
3 Kernel module for 2.6.x Kernels
4
5 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
6
7 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
8 Copyright (C) 2001-2008, Philipp Reisner <philipp.reisner@linbit.com>.
9 Copyright (C) 2001-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
10
11 drbd is free software; you can redistribute it and/or modify
12 it under the terms of the GNU General Public License as published by
13 the Free Software Foundation; either version 2, or (at your option)
14 any later version.
15
16 drbd is distributed in the hope that it will be useful,
17 but WITHOUT ANY WARRANTY; without even the implied warranty of
18 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19 GNU General Public License for more details.
20
21 You should have received a copy of the GNU General Public License
22 along with drbd; see the file COPYING. If not, write to
23 the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
24
25*/
26#ifndef DRBD_H
27#define DRBD_H
28#include <linux/connector.h>
29#include <asm/types.h>
30
31#ifdef __KERNEL__
32#include <linux/types.h>
33#include <asm/byteorder.h>
34#else
35#include <sys/types.h>
36#include <sys/wait.h>
37#include <limits.h>
38
39/* Altough the Linux source code makes a difference between
40 generic endianness and the bitfields' endianness, there is no
41 architecture as of Linux-2.6.24-rc4 where the bitfileds' endianness
42 does not match the generic endianness. */
43
44#if __BYTE_ORDER == __LITTLE_ENDIAN
45#define __LITTLE_ENDIAN_BITFIELD
46#elif __BYTE_ORDER == __BIG_ENDIAN
47#define __BIG_ENDIAN_BITFIELD
48#else
49# error "sorry, weird endianness on this box"
50#endif
51
52#endif
53
54
55extern const char *drbd_buildtag(void);
56#define REL_VERSION "8.3.6"
57#define API_VERSION 88
58#define PRO_VERSION_MIN 86
59#define PRO_VERSION_MAX 91
60
61
62enum drbd_io_error_p {
63 EP_PASS_ON, /* FIXME should the better be named "Ignore"? */
64 EP_CALL_HELPER,
65 EP_DETACH
66};
67
68enum drbd_fencing_p {
69 FP_DONT_CARE,
70 FP_RESOURCE,
71 FP_STONITH
72};
73
74enum drbd_disconnect_p {
75 DP_RECONNECT,
76 DP_DROP_NET_CONF,
77 DP_FREEZE_IO
78};
79
80enum drbd_after_sb_p {
81 ASB_DISCONNECT,
82 ASB_DISCARD_YOUNGER_PRI,
83 ASB_DISCARD_OLDER_PRI,
84 ASB_DISCARD_ZERO_CHG,
85 ASB_DISCARD_LEAST_CHG,
86 ASB_DISCARD_LOCAL,
87 ASB_DISCARD_REMOTE,
88 ASB_CONSENSUS,
89 ASB_DISCARD_SECONDARY,
90 ASB_CALL_HELPER,
91 ASB_VIOLENTLY
92};
93
94/* KEEP the order, do not delete or insert. Only append. */
95enum drbd_ret_codes {
96 ERR_CODE_BASE = 100,
97 NO_ERROR = 101,
98 ERR_LOCAL_ADDR = 102,
99 ERR_PEER_ADDR = 103,
100 ERR_OPEN_DISK = 104,
101 ERR_OPEN_MD_DISK = 105,
102 ERR_DISK_NOT_BDEV = 107,
103 ERR_MD_NOT_BDEV = 108,
104 ERR_DISK_TO_SMALL = 111,
105 ERR_MD_DISK_TO_SMALL = 112,
106 ERR_BDCLAIM_DISK = 114,
107 ERR_BDCLAIM_MD_DISK = 115,
108 ERR_MD_IDX_INVALID = 116,
109 ERR_IO_MD_DISK = 118,
110 ERR_MD_INVALID = 119,
111 ERR_AUTH_ALG = 120,
112 ERR_AUTH_ALG_ND = 121,
113 ERR_NOMEM = 122,
114 ERR_DISCARD = 123,
115 ERR_DISK_CONFIGURED = 124,
116 ERR_NET_CONFIGURED = 125,
117 ERR_MANDATORY_TAG = 126,
118 ERR_MINOR_INVALID = 127,
119 ERR_INTR = 129, /* EINTR */
120 ERR_RESIZE_RESYNC = 130,
121 ERR_NO_PRIMARY = 131,
122 ERR_SYNC_AFTER = 132,
123 ERR_SYNC_AFTER_CYCLE = 133,
124 ERR_PAUSE_IS_SET = 134,
125 ERR_PAUSE_IS_CLEAR = 135,
126 ERR_PACKET_NR = 137,
127 ERR_NO_DISK = 138,
128 ERR_NOT_PROTO_C = 139,
129 ERR_NOMEM_BITMAP = 140,
130 ERR_INTEGRITY_ALG = 141, /* DRBD 8.2 only */
131 ERR_INTEGRITY_ALG_ND = 142, /* DRBD 8.2 only */
132 ERR_CPU_MASK_PARSE = 143, /* DRBD 8.2 only */
133 ERR_CSUMS_ALG = 144, /* DRBD 8.2 only */
134 ERR_CSUMS_ALG_ND = 145, /* DRBD 8.2 only */
135 ERR_VERIFY_ALG = 146, /* DRBD 8.2 only */
136 ERR_VERIFY_ALG_ND = 147, /* DRBD 8.2 only */
137 ERR_CSUMS_RESYNC_RUNNING= 148, /* DRBD 8.2 only */
138 ERR_VERIFY_RUNNING = 149, /* DRBD 8.2 only */
139 ERR_DATA_NOT_CURRENT = 150,
140 ERR_CONNECTED = 151, /* DRBD 8.3 only */
141 ERR_PERM = 152,
142
143 /* insert new ones above this line */
144 AFTER_LAST_ERR_CODE
145};
146
147#define DRBD_PROT_A 1
148#define DRBD_PROT_B 2
149#define DRBD_PROT_C 3
150
151enum drbd_role {
152 R_UNKNOWN = 0,
153 R_PRIMARY = 1, /* role */
154 R_SECONDARY = 2, /* role */
155 R_MASK = 3,
156};
157
158/* The order of these constants is important.
159 * The lower ones (<C_WF_REPORT_PARAMS) indicate
160 * that there is no socket!
161 * >=C_WF_REPORT_PARAMS ==> There is a socket
162 */
163enum drbd_conns {
164 C_STANDALONE,
165 C_DISCONNECTING, /* Temporal state on the way to StandAlone. */
166 C_UNCONNECTED, /* >= C_UNCONNECTED -> inc_net() succeeds */
167
168 /* These temporal states are all used on the way
169 * from >= C_CONNECTED to Unconnected.
170 * The 'disconnect reason' states
171 * I do not allow to change beween them. */
172 C_TIMEOUT,
173 C_BROKEN_PIPE,
174 C_NETWORK_FAILURE,
175 C_PROTOCOL_ERROR,
176 C_TEAR_DOWN,
177
178 C_WF_CONNECTION,
179 C_WF_REPORT_PARAMS, /* we have a socket */
180 C_CONNECTED, /* we have introduced each other */
181 C_STARTING_SYNC_S, /* starting full sync by admin request. */
182 C_STARTING_SYNC_T, /* stariing full sync by admin request. */
183 C_WF_BITMAP_S,
184 C_WF_BITMAP_T,
185 C_WF_SYNC_UUID,
186
187 /* All SyncStates are tested with this comparison
188 * xx >= C_SYNC_SOURCE && xx <= C_PAUSED_SYNC_T */
189 C_SYNC_SOURCE,
190 C_SYNC_TARGET,
191 C_VERIFY_S,
192 C_VERIFY_T,
193 C_PAUSED_SYNC_S,
194 C_PAUSED_SYNC_T,
195 C_MASK = 31
196};
197
198enum drbd_disk_state {
199 D_DISKLESS,
200 D_ATTACHING, /* In the process of reading the meta-data */
201 D_FAILED, /* Becomes D_DISKLESS as soon as we told it the peer */
202 /* when >= D_FAILED it is legal to access mdev->bc */
203 D_NEGOTIATING, /* Late attaching state, we need to talk to the peer */
204 D_INCONSISTENT,
205 D_OUTDATED,
206 D_UNKNOWN, /* Only used for the peer, never for myself */
207 D_CONSISTENT, /* Might be D_OUTDATED, might be D_UP_TO_DATE ... */
208 D_UP_TO_DATE, /* Only this disk state allows applications' IO ! */
209 D_MASK = 15
210};
211
212union drbd_state {
213/* According to gcc's docs is the ...
214 * The order of allocation of bit-fields within a unit (C90 6.5.2.1, C99 6.7.2.1).
215 * Determined by ABI.
216 * pointed out by Maxim Uvarov q<muvarov@ru.mvista.com>
217 * even though we transmit as "cpu_to_be32(state)",
218 * the offsets of the bitfields still need to be swapped
219 * on different endianess.
220 */
221 struct {
222#if defined(__LITTLE_ENDIAN_BITFIELD)
223 unsigned role:2 ; /* 3/4 primary/secondary/unknown */
224 unsigned peer:2 ; /* 3/4 primary/secondary/unknown */
225 unsigned conn:5 ; /* 17/32 cstates */
226 unsigned disk:4 ; /* 8/16 from D_DISKLESS to D_UP_TO_DATE */
227 unsigned pdsk:4 ; /* 8/16 from D_DISKLESS to D_UP_TO_DATE */
228 unsigned susp:1 ; /* 2/2 IO suspended no/yes */
229 unsigned aftr_isp:1 ; /* isp .. imposed sync pause */
230 unsigned peer_isp:1 ;
231 unsigned user_isp:1 ;
232 unsigned _pad:11; /* 0 unused */
233#elif defined(__BIG_ENDIAN_BITFIELD)
234 unsigned _pad:11; /* 0 unused */
235 unsigned user_isp:1 ;
236 unsigned peer_isp:1 ;
237 unsigned aftr_isp:1 ; /* isp .. imposed sync pause */
238 unsigned susp:1 ; /* 2/2 IO suspended no/yes */
239 unsigned pdsk:4 ; /* 8/16 from D_DISKLESS to D_UP_TO_DATE */
240 unsigned disk:4 ; /* 8/16 from D_DISKLESS to D_UP_TO_DATE */
241 unsigned conn:5 ; /* 17/32 cstates */
242 unsigned peer:2 ; /* 3/4 primary/secondary/unknown */
243 unsigned role:2 ; /* 3/4 primary/secondary/unknown */
244#else
245# error "this endianess is not supported"
246#endif
247 };
248 unsigned int i;
249};
250
251enum drbd_state_ret_codes {
252 SS_CW_NO_NEED = 4,
253 SS_CW_SUCCESS = 3,
254 SS_NOTHING_TO_DO = 2,
255 SS_SUCCESS = 1,
256 SS_UNKNOWN_ERROR = 0, /* Used to sleep longer in _drbd_request_state */
257 SS_TWO_PRIMARIES = -1,
258 SS_NO_UP_TO_DATE_DISK = -2,
259 SS_NO_LOCAL_DISK = -4,
260 SS_NO_REMOTE_DISK = -5,
261 SS_CONNECTED_OUTDATES = -6,
262 SS_PRIMARY_NOP = -7,
263 SS_RESYNC_RUNNING = -8,
264 SS_ALREADY_STANDALONE = -9,
265 SS_CW_FAILED_BY_PEER = -10,
266 SS_IS_DISKLESS = -11,
267 SS_DEVICE_IN_USE = -12,
268 SS_NO_NET_CONFIG = -13,
269 SS_NO_VERIFY_ALG = -14, /* drbd-8.2 only */
270 SS_NEED_CONNECTION = -15, /* drbd-8.2 only */
271 SS_LOWER_THAN_OUTDATED = -16,
272 SS_NOT_SUPPORTED = -17, /* drbd-8.2 only */
273 SS_IN_TRANSIENT_STATE = -18, /* Retry after the next state change */
274 SS_CONCURRENT_ST_CHG = -19, /* Concurrent cluster side state change! */
275 SS_AFTER_LAST_ERROR = -20, /* Keep this at bottom */
276};
277
278/* from drbd_strings.c */
279extern const char *drbd_conn_str(enum drbd_conns);
280extern const char *drbd_role_str(enum drbd_role);
281extern const char *drbd_disk_str(enum drbd_disk_state);
282extern const char *drbd_set_st_err_str(enum drbd_state_ret_codes);
283
284#define SHARED_SECRET_MAX 64
285
286#define MDF_CONSISTENT (1 << 0)
287#define MDF_PRIMARY_IND (1 << 1)
288#define MDF_CONNECTED_IND (1 << 2)
289#define MDF_FULL_SYNC (1 << 3)
290#define MDF_WAS_UP_TO_DATE (1 << 4)
291#define MDF_PEER_OUT_DATED (1 << 5)
292#define MDF_CRASHED_PRIMARY (1 << 6)
293
294enum drbd_uuid_index {
295 UI_CURRENT,
296 UI_BITMAP,
297 UI_HISTORY_START,
298 UI_HISTORY_END,
299 UI_SIZE, /* nl-packet: number of dirty bits */
300 UI_FLAGS, /* nl-packet: flags */
301 UI_EXTENDED_SIZE /* Everything. */
302};
303
304enum drbd_timeout_flag {
305 UT_DEFAULT = 0,
306 UT_DEGRADED = 1,
307 UT_PEER_OUTDATED = 2,
308};
309
310#define UUID_JUST_CREATED ((__u64)4)
311
312#define DRBD_MAGIC 0x83740267
313#define BE_DRBD_MAGIC __constant_cpu_to_be32(DRBD_MAGIC)
314
315/* these are of type "int" */
316#define DRBD_MD_INDEX_INTERNAL -1
317#define DRBD_MD_INDEX_FLEX_EXT -2
318#define DRBD_MD_INDEX_FLEX_INT -3
319
320/* Start of the new netlink/connector stuff */
321
322#define DRBD_NL_CREATE_DEVICE 0x01
323#define DRBD_NL_SET_DEFAULTS 0x02
324
325
326/* For searching a vacant cn_idx value */
327#define CN_IDX_STEP 6977
328
329struct drbd_nl_cfg_req {
330 int packet_type;
331 unsigned int drbd_minor;
332 int flags;
333 unsigned short tag_list[];
334};
335
336struct drbd_nl_cfg_reply {
337 int packet_type;
338 unsigned int minor;
339 int ret_code; /* enum ret_code or set_st_err_t */
340 unsigned short tag_list[]; /* only used with get_* calls */
341};
342
343#endif
diff --git a/include/linux/drbd_limits.h b/include/linux/drbd_limits.h
new file mode 100644
index 000000000000..51f47a586ad8
--- /dev/null
+++ b/include/linux/drbd_limits.h
@@ -0,0 +1,137 @@
1/*
2 drbd_limits.h
3 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
4*/
5
6/*
7 * Our current limitations.
8 * Some of them are hard limits,
9 * some of them are arbitrary range limits, that make it easier to provide
10 * feedback about nonsense settings for certain configurable values.
11 */
12
13#ifndef DRBD_LIMITS_H
14#define DRBD_LIMITS_H 1
15
16#define DEBUG_RANGE_CHECK 0
17
18#define DRBD_MINOR_COUNT_MIN 1
19#define DRBD_MINOR_COUNT_MAX 255
20
21#define DRBD_DIALOG_REFRESH_MIN 0
22#define DRBD_DIALOG_REFRESH_MAX 600
23
24/* valid port number */
25#define DRBD_PORT_MIN 1
26#define DRBD_PORT_MAX 0xffff
27
28/* startup { */
29 /* if you want more than 3.4 days, disable */
30#define DRBD_WFC_TIMEOUT_MIN 0
31#define DRBD_WFC_TIMEOUT_MAX 300000
32#define DRBD_WFC_TIMEOUT_DEF 0
33
34#define DRBD_DEGR_WFC_TIMEOUT_MIN 0
35#define DRBD_DEGR_WFC_TIMEOUT_MAX 300000
36#define DRBD_DEGR_WFC_TIMEOUT_DEF 0
37
38#define DRBD_OUTDATED_WFC_TIMEOUT_MIN 0
39#define DRBD_OUTDATED_WFC_TIMEOUT_MAX 300000
40#define DRBD_OUTDATED_WFC_TIMEOUT_DEF 0
41/* }*/
42
43/* net { */
44 /* timeout, unit centi seconds
45 * more than one minute timeout is not usefull */
46#define DRBD_TIMEOUT_MIN 1
47#define DRBD_TIMEOUT_MAX 600
48#define DRBD_TIMEOUT_DEF 60 /* 6 seconds */
49
50 /* active connection retries when C_WF_CONNECTION */
51#define DRBD_CONNECT_INT_MIN 1
52#define DRBD_CONNECT_INT_MAX 120
53#define DRBD_CONNECT_INT_DEF 10 /* seconds */
54
55 /* keep-alive probes when idle */
56#define DRBD_PING_INT_MIN 1
57#define DRBD_PING_INT_MAX 120
58#define DRBD_PING_INT_DEF 10
59
60 /* timeout for the ping packets.*/
61#define DRBD_PING_TIMEO_MIN 1
62#define DRBD_PING_TIMEO_MAX 100
63#define DRBD_PING_TIMEO_DEF 5
64
65 /* max number of write requests between write barriers */
66#define DRBD_MAX_EPOCH_SIZE_MIN 1
67#define DRBD_MAX_EPOCH_SIZE_MAX 20000
68#define DRBD_MAX_EPOCH_SIZE_DEF 2048
69
70 /* I don't think that a tcp send buffer of more than 10M is usefull */
71#define DRBD_SNDBUF_SIZE_MIN 0
72#define DRBD_SNDBUF_SIZE_MAX (10<<20)
73#define DRBD_SNDBUF_SIZE_DEF 0
74
75#define DRBD_RCVBUF_SIZE_MIN 0
76#define DRBD_RCVBUF_SIZE_MAX (10<<20)
77#define DRBD_RCVBUF_SIZE_DEF 0
78
79 /* @4k PageSize -> 128kB - 512MB */
80#define DRBD_MAX_BUFFERS_MIN 32
81#define DRBD_MAX_BUFFERS_MAX 131072
82#define DRBD_MAX_BUFFERS_DEF 2048
83
84 /* @4k PageSize -> 4kB - 512MB */
85#define DRBD_UNPLUG_WATERMARK_MIN 1
86#define DRBD_UNPLUG_WATERMARK_MAX 131072
87#define DRBD_UNPLUG_WATERMARK_DEF (DRBD_MAX_BUFFERS_DEF/16)
88
89 /* 0 is disabled.
90 * 200 should be more than enough even for very short timeouts */
91#define DRBD_KO_COUNT_MIN 0
92#define DRBD_KO_COUNT_MAX 200
93#define DRBD_KO_COUNT_DEF 0
94/* } */
95
96/* syncer { */
97 /* FIXME allow rate to be zero? */
98#define DRBD_RATE_MIN 1
99/* channel bonding 10 GbE, or other hardware */
100#define DRBD_RATE_MAX (4 << 20)
101#define DRBD_RATE_DEF 250 /* kb/second */
102
103 /* less than 7 would hit performance unneccessarily.
104 * 3833 is the largest prime that still does fit
105 * into 64 sectors of activity log */
106#define DRBD_AL_EXTENTS_MIN 7
107#define DRBD_AL_EXTENTS_MAX 3833
108#define DRBD_AL_EXTENTS_DEF 127
109
110#define DRBD_AFTER_MIN -1
111#define DRBD_AFTER_MAX 255
112#define DRBD_AFTER_DEF -1
113
114/* } */
115
116/* drbdsetup XY resize -d Z
117 * you are free to reduce the device size to nothing, if you want to.
118 * the upper limit with 64bit kernel, enough ram and flexible meta data
119 * is 16 TB, currently. */
120/* DRBD_MAX_SECTORS */
121#define DRBD_DISK_SIZE_SECT_MIN 0
122#define DRBD_DISK_SIZE_SECT_MAX (16 * (2LLU << 30))
123#define DRBD_DISK_SIZE_SECT_DEF 0 /* = disabled = no user size... */
124
125#define DRBD_ON_IO_ERROR_DEF EP_PASS_ON
126#define DRBD_FENCING_DEF FP_DONT_CARE
127#define DRBD_AFTER_SB_0P_DEF ASB_DISCONNECT
128#define DRBD_AFTER_SB_1P_DEF ASB_DISCONNECT
129#define DRBD_AFTER_SB_2P_DEF ASB_DISCONNECT
130#define DRBD_RR_CONFLICT_DEF ASB_DISCONNECT
131
132#define DRBD_MAX_BIO_BVECS_MIN 0
133#define DRBD_MAX_BIO_BVECS_MAX 128
134#define DRBD_MAX_BIO_BVECS_DEF 0
135
136#undef RANGE
137#endif
diff --git a/include/linux/drbd_nl.h b/include/linux/drbd_nl.h
new file mode 100644
index 000000000000..db5721ad50d1
--- /dev/null
+++ b/include/linux/drbd_nl.h
@@ -0,0 +1,137 @@
1/*
2 PAKET( name,
3 TYPE ( pn, pr, member )
4 ...
5 )
6
7 You may never reissue one of the pn arguments
8*/
9
10#if !defined(NL_PACKET) || !defined(NL_STRING) || !defined(NL_INTEGER) || !defined(NL_BIT) || !defined(NL_INT64)
11#error "The macros NL_PACKET, NL_STRING, NL_INTEGER, NL_INT64 and NL_BIT needs to be defined"
12#endif
13
14NL_PACKET(primary, 1,
15 NL_BIT( 1, T_MAY_IGNORE, overwrite_peer)
16)
17
18NL_PACKET(secondary, 2, )
19
20NL_PACKET(disk_conf, 3,
21 NL_INT64( 2, T_MAY_IGNORE, disk_size)
22 NL_STRING( 3, T_MANDATORY, backing_dev, 128)
23 NL_STRING( 4, T_MANDATORY, meta_dev, 128)
24 NL_INTEGER( 5, T_MANDATORY, meta_dev_idx)
25 NL_INTEGER( 6, T_MAY_IGNORE, on_io_error)
26 NL_INTEGER( 7, T_MAY_IGNORE, fencing)
27 NL_BIT( 37, T_MAY_IGNORE, use_bmbv)
28 NL_BIT( 53, T_MAY_IGNORE, no_disk_flush)
29 NL_BIT( 54, T_MAY_IGNORE, no_md_flush)
30 /* 55 max_bio_size was available in 8.2.6rc2 */
31 NL_INTEGER( 56, T_MAY_IGNORE, max_bio_bvecs)
32 NL_BIT( 57, T_MAY_IGNORE, no_disk_barrier)
33 NL_BIT( 58, T_MAY_IGNORE, no_disk_drain)
34)
35
36NL_PACKET(detach, 4, )
37
38NL_PACKET(net_conf, 5,
39 NL_STRING( 8, T_MANDATORY, my_addr, 128)
40 NL_STRING( 9, T_MANDATORY, peer_addr, 128)
41 NL_STRING( 10, T_MAY_IGNORE, shared_secret, SHARED_SECRET_MAX)
42 NL_STRING( 11, T_MAY_IGNORE, cram_hmac_alg, SHARED_SECRET_MAX)
43 NL_STRING( 44, T_MAY_IGNORE, integrity_alg, SHARED_SECRET_MAX)
44 NL_INTEGER( 14, T_MAY_IGNORE, timeout)
45 NL_INTEGER( 15, T_MANDATORY, wire_protocol)
46 NL_INTEGER( 16, T_MAY_IGNORE, try_connect_int)
47 NL_INTEGER( 17, T_MAY_IGNORE, ping_int)
48 NL_INTEGER( 18, T_MAY_IGNORE, max_epoch_size)
49 NL_INTEGER( 19, T_MAY_IGNORE, max_buffers)
50 NL_INTEGER( 20, T_MAY_IGNORE, unplug_watermark)
51 NL_INTEGER( 21, T_MAY_IGNORE, sndbuf_size)
52 NL_INTEGER( 22, T_MAY_IGNORE, ko_count)
53 NL_INTEGER( 24, T_MAY_IGNORE, after_sb_0p)
54 NL_INTEGER( 25, T_MAY_IGNORE, after_sb_1p)
55 NL_INTEGER( 26, T_MAY_IGNORE, after_sb_2p)
56 NL_INTEGER( 39, T_MAY_IGNORE, rr_conflict)
57 NL_INTEGER( 40, T_MAY_IGNORE, ping_timeo)
58 NL_INTEGER( 67, T_MAY_IGNORE, rcvbuf_size)
59 /* 59 addr_family was available in GIT, never released */
60 NL_BIT( 60, T_MANDATORY, mind_af)
61 NL_BIT( 27, T_MAY_IGNORE, want_lose)
62 NL_BIT( 28, T_MAY_IGNORE, two_primaries)
63 NL_BIT( 41, T_MAY_IGNORE, always_asbp)
64 NL_BIT( 61, T_MAY_IGNORE, no_cork)
65 NL_BIT( 62, T_MANDATORY, auto_sndbuf_size)
66)
67
68NL_PACKET(disconnect, 6, )
69
70NL_PACKET(resize, 7,
71 NL_INT64( 29, T_MAY_IGNORE, resize_size)
72)
73
74NL_PACKET(syncer_conf, 8,
75 NL_INTEGER( 30, T_MAY_IGNORE, rate)
76 NL_INTEGER( 31, T_MAY_IGNORE, after)
77 NL_INTEGER( 32, T_MAY_IGNORE, al_extents)
78 NL_STRING( 52, T_MAY_IGNORE, verify_alg, SHARED_SECRET_MAX)
79 NL_STRING( 51, T_MAY_IGNORE, cpu_mask, 32)
80 NL_STRING( 64, T_MAY_IGNORE, csums_alg, SHARED_SECRET_MAX)
81 NL_BIT( 65, T_MAY_IGNORE, use_rle)
82)
83
84NL_PACKET(invalidate, 9, )
85NL_PACKET(invalidate_peer, 10, )
86NL_PACKET(pause_sync, 11, )
87NL_PACKET(resume_sync, 12, )
88NL_PACKET(suspend_io, 13, )
89NL_PACKET(resume_io, 14, )
90NL_PACKET(outdate, 15, )
91NL_PACKET(get_config, 16, )
92NL_PACKET(get_state, 17,
93 NL_INTEGER( 33, T_MAY_IGNORE, state_i)
94)
95
96NL_PACKET(get_uuids, 18,
97 NL_STRING( 34, T_MAY_IGNORE, uuids, (UI_SIZE*sizeof(__u64)))
98 NL_INTEGER( 35, T_MAY_IGNORE, uuids_flags)
99)
100
101NL_PACKET(get_timeout_flag, 19,
102 NL_BIT( 36, T_MAY_IGNORE, use_degraded)
103)
104
105NL_PACKET(call_helper, 20,
106 NL_STRING( 38, T_MAY_IGNORE, helper, 32)
107)
108
109/* Tag nr 42 already allocated in drbd-8.1 development. */
110
111NL_PACKET(sync_progress, 23,
112 NL_INTEGER( 43, T_MAY_IGNORE, sync_progress)
113)
114
115NL_PACKET(dump_ee, 24,
116 NL_STRING( 45, T_MAY_IGNORE, dump_ee_reason, 32)
117 NL_STRING( 46, T_MAY_IGNORE, seen_digest, SHARED_SECRET_MAX)
118 NL_STRING( 47, T_MAY_IGNORE, calc_digest, SHARED_SECRET_MAX)
119 NL_INT64( 48, T_MAY_IGNORE, ee_sector)
120 NL_INT64( 49, T_MAY_IGNORE, ee_block_id)
121 NL_STRING( 50, T_MAY_IGNORE, ee_data, 32 << 10)
122)
123
124NL_PACKET(start_ov, 25,
125 NL_INT64( 66, T_MAY_IGNORE, start_sector)
126)
127
128NL_PACKET(new_c_uuid, 26,
129 NL_BIT( 63, T_MANDATORY, clear_bm)
130)
131
132#undef NL_PACKET
133#undef NL_INTEGER
134#undef NL_INT64
135#undef NL_BIT
136#undef NL_STRING
137
diff --git a/include/linux/drbd_tag_magic.h b/include/linux/drbd_tag_magic.h
new file mode 100644
index 000000000000..fcdff8410e99
--- /dev/null
+++ b/include/linux/drbd_tag_magic.h
@@ -0,0 +1,83 @@
1#ifndef DRBD_TAG_MAGIC_H
2#define DRBD_TAG_MAGIC_H
3
4#define TT_END 0
5#define TT_REMOVED 0xE000
6
7/* declare packet_type enums */
8enum packet_types {
9#define NL_PACKET(name, number, fields) P_ ## name = number,
10#define NL_INTEGER(pn, pr, member)
11#define NL_INT64(pn, pr, member)
12#define NL_BIT(pn, pr, member)
13#define NL_STRING(pn, pr, member, len)
14#include "drbd_nl.h"
15 P_nl_after_last_packet,
16};
17
18/* These struct are used to deduce the size of the tag lists: */
19#define NL_PACKET(name, number, fields) \
20 struct name ## _tag_len_struct { fields };
21#define NL_INTEGER(pn, pr, member) \
22 int member; int tag_and_len ## member;
23#define NL_INT64(pn, pr, member) \
24 __u64 member; int tag_and_len ## member;
25#define NL_BIT(pn, pr, member) \
26 unsigned char member:1; int tag_and_len ## member;
27#define NL_STRING(pn, pr, member, len) \
28 unsigned char member[len]; int member ## _len; \
29 int tag_and_len ## member;
30#include "linux/drbd_nl.h"
31
32/* declate tag-list-sizes */
33static const int tag_list_sizes[] = {
34#define NL_PACKET(name, number, fields) 2 fields ,
35#define NL_INTEGER(pn, pr, member) + 4 + 4
36#define NL_INT64(pn, pr, member) + 4 + 8
37#define NL_BIT(pn, pr, member) + 4 + 1
38#define NL_STRING(pn, pr, member, len) + 4 + (len)
39#include "drbd_nl.h"
40};
41
42/* The two highest bits are used for the tag type */
43#define TT_MASK 0xC000
44#define TT_INTEGER 0x0000
45#define TT_INT64 0x4000
46#define TT_BIT 0x8000
47#define TT_STRING 0xC000
48/* The next bit indicates if processing of the tag is mandatory */
49#define T_MANDATORY 0x2000
50#define T_MAY_IGNORE 0x0000
51#define TN_MASK 0x1fff
52/* The remaining 13 bits are used to enumerate the tags */
53
54#define tag_type(T) ((T) & TT_MASK)
55#define tag_number(T) ((T) & TN_MASK)
56
57/* declare tag enums */
58#define NL_PACKET(name, number, fields) fields
59enum drbd_tags {
60#define NL_INTEGER(pn, pr, member) T_ ## member = pn | TT_INTEGER | pr ,
61#define NL_INT64(pn, pr, member) T_ ## member = pn | TT_INT64 | pr ,
62#define NL_BIT(pn, pr, member) T_ ## member = pn | TT_BIT | pr ,
63#define NL_STRING(pn, pr, member, len) T_ ## member = pn | TT_STRING | pr ,
64#include "drbd_nl.h"
65};
66
67struct tag {
68 const char *name;
69 int type_n_flags;
70 int max_len;
71};
72
73/* declare tag names */
74#define NL_PACKET(name, number, fields) fields
75static const struct tag tag_descriptions[] = {
76#define NL_INTEGER(pn, pr, member) [ pn ] = { #member, TT_INTEGER | pr, sizeof(int) },
77#define NL_INT64(pn, pr, member) [ pn ] = { #member, TT_INT64 | pr, sizeof(__u64) },
78#define NL_BIT(pn, pr, member) [ pn ] = { #member, TT_BIT | pr, sizeof(int) },
79#define NL_STRING(pn, pr, member, len) [ pn ] = { #member, TT_STRING | pr, (len) },
80#include "drbd_nl.h"
81};
82
83#endif
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 2620a8c63571..891f7d642e5c 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -129,7 +129,7 @@ struct inodes_stat_t {
129 * WRITE_SYNC Like WRITE_SYNC_PLUG, but also unplugs the device 129 * WRITE_SYNC Like WRITE_SYNC_PLUG, but also unplugs the device
130 * immediately after submission. The write equivalent 130 * immediately after submission. The write equivalent
131 * of READ_SYNC. 131 * of READ_SYNC.
132 * WRITE_ODIRECT Special case write for O_DIRECT only. 132 * WRITE_ODIRECT_PLUG Special case write for O_DIRECT only.
133 * SWRITE_SYNC 133 * SWRITE_SYNC
134 * SWRITE_SYNC_PLUG Like WRITE_SYNC/WRITE_SYNC_PLUG, but locks the buffer. 134 * SWRITE_SYNC_PLUG Like WRITE_SYNC/WRITE_SYNC_PLUG, but locks the buffer.
135 * See SWRITE. 135 * See SWRITE.
@@ -151,7 +151,7 @@ struct inodes_stat_t {
151#define READ_META (READ | (1 << BIO_RW_META)) 151#define READ_META (READ | (1 << BIO_RW_META))
152#define WRITE_SYNC_PLUG (WRITE | (1 << BIO_RW_SYNCIO) | (1 << BIO_RW_NOIDLE)) 152#define WRITE_SYNC_PLUG (WRITE | (1 << BIO_RW_SYNCIO) | (1 << BIO_RW_NOIDLE))
153#define WRITE_SYNC (WRITE_SYNC_PLUG | (1 << BIO_RW_UNPLUG)) 153#define WRITE_SYNC (WRITE_SYNC_PLUG | (1 << BIO_RW_UNPLUG))
154#define WRITE_ODIRECT (WRITE | (1 << BIO_RW_SYNCIO) | (1 << BIO_RW_UNPLUG)) 154#define WRITE_ODIRECT_PLUG (WRITE | (1 << BIO_RW_SYNCIO))
155#define SWRITE_SYNC_PLUG \ 155#define SWRITE_SYNC_PLUG \
156 (SWRITE | (1 << BIO_RW_SYNCIO) | (1 << BIO_RW_NOIDLE)) 156 (SWRITE | (1 << BIO_RW_SYNCIO) | (1 << BIO_RW_NOIDLE))
157#define SWRITE_SYNC (SWRITE_SYNC_PLUG | (1 << BIO_RW_UNPLUG)) 157#define SWRITE_SYNC (SWRITE_SYNC_PLUG | (1 << BIO_RW_UNPLUG))
@@ -304,6 +304,7 @@ struct inodes_stat_t {
304#define BLKIOOPT _IO(0x12,121) 304#define BLKIOOPT _IO(0x12,121)
305#define BLKALIGNOFF _IO(0x12,122) 305#define BLKALIGNOFF _IO(0x12,122)
306#define BLKPBSZGET _IO(0x12,123) 306#define BLKPBSZGET _IO(0x12,123)
307#define BLKDISCARDZEROES _IO(0x12,124)
307 308
308#define BMAP_IOCTL 1 /* obsolete - kept for compatibility */ 309#define BMAP_IOCTL 1 /* obsolete - kept for compatibility */
309#define FIBMAP _IO(0x00,1) /* bmap access */ 310#define FIBMAP _IO(0x00,1) /* bmap access */
diff --git a/include/linux/genhd.h b/include/linux/genhd.h
index 297df45ffd0a..c6c0c41af35f 100644
--- a/include/linux/genhd.h
+++ b/include/linux/genhd.h
@@ -91,6 +91,7 @@ struct hd_struct {
91 sector_t start_sect; 91 sector_t start_sect;
92 sector_t nr_sects; 92 sector_t nr_sects;
93 sector_t alignment_offset; 93 sector_t alignment_offset;
94 unsigned int discard_alignment;
94 struct device __dev; 95 struct device __dev;
95 struct kobject *holder_dir; 96 struct kobject *holder_dir;
96 int policy, partno; 97 int policy, partno;
diff --git a/include/linux/iocontext.h b/include/linux/iocontext.h
index 4da4a75c3f1e..a63235996309 100644
--- a/include/linux/iocontext.h
+++ b/include/linux/iocontext.h
@@ -40,16 +40,11 @@ struct cfq_io_context {
40 struct io_context *ioc; 40 struct io_context *ioc;
41 41
42 unsigned long last_end_request; 42 unsigned long last_end_request;
43 sector_t last_request_pos;
44 43
45 unsigned long ttime_total; 44 unsigned long ttime_total;
46 unsigned long ttime_samples; 45 unsigned long ttime_samples;
47 unsigned long ttime_mean; 46 unsigned long ttime_mean;
48 47
49 unsigned int seek_samples;
50 u64 seek_total;
51 sector_t seek_mean;
52
53 struct list_head queue_list; 48 struct list_head queue_list;
54 struct hlist_node cic_list; 49 struct hlist_node cic_list;
55 50
@@ -73,6 +68,10 @@ struct io_context {
73 unsigned short ioprio; 68 unsigned short ioprio;
74 unsigned short ioprio_changed; 69 unsigned short ioprio_changed;
75 70
71#ifdef CONFIG_BLK_CGROUP
72 unsigned short cgroup_changed;
73#endif
74
76 /* 75 /*
77 * For request batching 76 * For request batching
78 */ 77 */
@@ -99,14 +98,15 @@ static inline struct io_context *ioc_task_link(struct io_context *ioc)
99 return NULL; 98 return NULL;
100} 99}
101 100
101struct task_struct;
102#ifdef CONFIG_BLOCK 102#ifdef CONFIG_BLOCK
103int put_io_context(struct io_context *ioc); 103int put_io_context(struct io_context *ioc);
104void exit_io_context(void); 104void exit_io_context(struct task_struct *task);
105struct io_context *get_io_context(gfp_t gfp_flags, int node); 105struct io_context *get_io_context(gfp_t gfp_flags, int node);
106struct io_context *alloc_io_context(gfp_t gfp_flags, int node); 106struct io_context *alloc_io_context(gfp_t gfp_flags, int node);
107void copy_io_context(struct io_context **pdst, struct io_context **psrc); 107void copy_io_context(struct io_context **pdst, struct io_context **psrc);
108#else 108#else
109static inline void exit_io_context(void) 109static inline void exit_io_context(struct task_struct *task)
110{ 110{
111} 111}
112 112
diff --git a/include/linux/lru_cache.h b/include/linux/lru_cache.h
new file mode 100644
index 000000000000..3a2b2d9b0472
--- /dev/null
+++ b/include/linux/lru_cache.h
@@ -0,0 +1,294 @@
1/*
2 lru_cache.c
3
4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
5
6 Copyright (C) 2003-2008, LINBIT Information Technologies GmbH.
7 Copyright (C) 2003-2008, Philipp Reisner <philipp.reisner@linbit.com>.
8 Copyright (C) 2003-2008, 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 LRU_CACHE_H
27#define LRU_CACHE_H
28
29#include <linux/list.h>
30#include <linux/slab.h>
31#include <linux/bitops.h>
32#include <linux/string.h> /* for memset */
33#include <linux/seq_file.h>
34
35/*
36This header file (and its .c file; kernel-doc of functions see there)
37 define a helper framework to easily keep track of index:label associations,
38 and changes to an "active set" of objects, as well as pending transactions,
39 to persistently record those changes.
40
41 We use an LRU policy if it is necessary to "cool down" a region currently in
42 the active set before we can "heat" a previously unused region.
43
44 Because of this later property, it is called "lru_cache".
45 As it actually Tracks Objects in an Active SeT, we could also call it
46 toast (incidentally that is what may happen to the data on the
47 backend storage uppon next resync, if we don't get it right).
48
49What for?
50
51We replicate IO (more or less synchronously) to local and remote disk.
52
53For crash recovery after replication node failure,
54 we need to resync all regions that have been target of in-flight WRITE IO
55 (in use, or "hot", regions), as we don't know wether or not those WRITEs have
56 made it to stable storage.
57
58 To avoid a "full resync", we need to persistently track these regions.
59
60 This is known as "write intent log", and can be implemented as on-disk
61 (coarse or fine grained) bitmap, or other meta data.
62
63 To avoid the overhead of frequent extra writes to this meta data area,
64 usually the condition is softened to regions that _may_ have been target of
65 in-flight WRITE IO, e.g. by only lazily clearing the on-disk write-intent
66 bitmap, trading frequency of meta data transactions against amount of
67 (possibly unneccessary) resync traffic.
68
69 If we set a hard limit on the area that may be "hot" at any given time, we
70 limit the amount of resync traffic needed for crash recovery.
71
72For recovery after replication link failure,
73 we need to resync all blocks that have been changed on the other replica
74 in the mean time, or, if both replica have been changed independently [*],
75 all blocks that have been changed on either replica in the mean time.
76 [*] usually as a result of a cluster split-brain and insufficient protection.
77 but there are valid use cases to do this on purpose.
78
79 Tracking those blocks can be implemented as "dirty bitmap".
80 Having it fine-grained reduces the amount of resync traffic.
81 It should also be persistent, to allow for reboots (or crashes)
82 while the replication link is down.
83
84There are various possible implementations for persistently storing
85write intent log information, three of which are mentioned here.
86
87"Chunk dirtying"
88 The on-disk "dirty bitmap" may be re-used as "write-intent" bitmap as well.
89 To reduce the frequency of bitmap updates for write-intent log purposes,
90 one could dirty "chunks" (of some size) at a time of the (fine grained)
91 on-disk bitmap, while keeping the in-memory "dirty" bitmap as clean as
92 possible, flushing it to disk again when a previously "hot" (and on-disk
93 dirtied as full chunk) area "cools down" again (no IO in flight anymore,
94 and none expected in the near future either).
95
96"Explicit (coarse) write intent bitmap"
97 An other implementation could chose a (probably coarse) explicit bitmap,
98 for write-intent log purposes, additionally to the fine grained dirty bitmap.
99
100"Activity log"
101 Yet an other implementation may keep track of the hot regions, by starting
102 with an empty set, and writing down a journal of region numbers that have
103 become "hot", or have "cooled down" again.
104
105 To be able to use a ring buffer for this journal of changes to the active
106 set, we not only record the actual changes to that set, but also record the
107 not changing members of the set in a round robin fashion. To do so, we use a
108 fixed (but configurable) number of slots which we can identify by index, and
109 associate region numbers (labels) with these indices.
110 For each transaction recording a change to the active set, we record the
111 change itself (index: -old_label, +new_label), and which index is associated
112 with which label (index: current_label) within a certain sliding window that
113 is moved further over the available indices with each such transaction.
114
115 Thus, for crash recovery, if the ringbuffer is sufficiently large, we can
116 accurately reconstruct the active set.
117
118 Sufficiently large depends only on maximum number of active objects, and the
119 size of the sliding window recording "index: current_label" associations within
120 each transaction.
121
122 This is what we call the "activity log".
123
124 Currently we need one activity log transaction per single label change, which
125 does not give much benefit over the "dirty chunks of bitmap" approach, other
126 than potentially less seeks.
127
128 We plan to change the transaction format to support multiple changes per
129 transaction, which then would reduce several (disjoint, "random") updates to
130 the bitmap into one transaction to the activity log ring buffer.
131*/
132
133/* this defines an element in a tracked set
134 * .colision is for hash table lookup.
135 * When we process a new IO request, we know its sector, thus can deduce the
136 * region number (label) easily. To do the label -> object lookup without a
137 * full list walk, we use a simple hash table.
138 *
139 * .list is on one of three lists:
140 * in_use: currently in use (refcnt > 0, lc_number != LC_FREE)
141 * lru: unused but ready to be reused or recycled
142 * (ts_refcnt == 0, lc_number != LC_FREE),
143 * free: unused but ready to be recycled
144 * (ts_refcnt == 0, lc_number == LC_FREE),
145 *
146 * an element is said to be "in the active set",
147 * if either on "in_use" or "lru", i.e. lc_number != LC_FREE.
148 *
149 * DRBD currently (May 2009) only uses 61 elements on the resync lru_cache
150 * (total memory usage 2 pages), and up to 3833 elements on the act_log
151 * lru_cache, totalling ~215 kB for 64bit architechture, ~53 pages.
152 *
153 * We usually do not actually free these objects again, but only "recycle"
154 * them, as the change "index: -old_label, +LC_FREE" would need a transaction
155 * as well. Which also means that using a kmem_cache to allocate the objects
156 * from wastes some resources.
157 * But it avoids high order page allocations in kmalloc.
158 */
159struct lc_element {
160 struct hlist_node colision;
161 struct list_head list; /* LRU list or free list */
162 unsigned refcnt;
163 /* back "pointer" into ts_cache->element[index],
164 * for paranoia, and for "ts_element_to_index" */
165 unsigned lc_index;
166 /* if we want to track a larger set of objects,
167 * it needs to become arch independend u64 */
168 unsigned lc_number;
169
170 /* special label when on free list */
171#define LC_FREE (~0U)
172};
173
174struct lru_cache {
175 /* the least recently used item is kept at lru->prev */
176 struct list_head lru;
177 struct list_head free;
178 struct list_head in_use;
179
180 /* the pre-created kmem cache to allocate the objects from */
181 struct kmem_cache *lc_cache;
182
183 /* size of tracked objects, used to memset(,0,) them in lc_reset */
184 size_t element_size;
185 /* offset of struct lc_element member in the tracked object */
186 size_t element_off;
187
188 /* number of elements (indices) */
189 unsigned int nr_elements;
190 /* Arbitrary limit on maximum tracked objects. Practical limit is much
191 * lower due to allocation failures, probably. For typical use cases,
192 * nr_elements should be a few thousand at most.
193 * This also limits the maximum value of ts_element.ts_index, allowing the
194 * 8 high bits of .ts_index to be overloaded with flags in the future. */
195#define LC_MAX_ACTIVE (1<<24)
196
197 /* statistics */
198 unsigned used; /* number of lelements currently on in_use list */
199 unsigned long hits, misses, starving, dirty, changed;
200
201 /* see below: flag-bits for lru_cache */
202 unsigned long flags;
203
204 /* when changing the label of an index element */
205 unsigned int new_number;
206
207 /* for paranoia when changing the label of an index element */
208 struct lc_element *changing_element;
209
210 void *lc_private;
211 const char *name;
212
213 /* nr_elements there */
214 struct hlist_head *lc_slot;
215 struct lc_element **lc_element;
216};
217
218
219/* flag-bits for lru_cache */
220enum {
221 /* debugging aid, to catch concurrent access early.
222 * user needs to guarantee exclusive access by proper locking! */
223 __LC_PARANOIA,
224 /* if we need to change the set, but currently there is a changing
225 * transaction pending, we are "dirty", and must deferr further
226 * changing requests */
227 __LC_DIRTY,
228 /* if we need to change the set, but currently there is no free nor
229 * unused element available, we are "starving", and must not give out
230 * further references, to guarantee that eventually some refcnt will
231 * drop to zero and we will be able to make progress again, changing
232 * the set, writing the transaction.
233 * if the statistics say we are frequently starving,
234 * nr_elements is too small. */
235 __LC_STARVING,
236};
237#define LC_PARANOIA (1<<__LC_PARANOIA)
238#define LC_DIRTY (1<<__LC_DIRTY)
239#define LC_STARVING (1<<__LC_STARVING)
240
241extern struct lru_cache *lc_create(const char *name, struct kmem_cache *cache,
242 unsigned e_count, size_t e_size, size_t e_off);
243extern void lc_reset(struct lru_cache *lc);
244extern void lc_destroy(struct lru_cache *lc);
245extern void lc_set(struct lru_cache *lc, unsigned int enr, int index);
246extern void lc_del(struct lru_cache *lc, struct lc_element *element);
247
248extern struct lc_element *lc_try_get(struct lru_cache *lc, unsigned int enr);
249extern struct lc_element *lc_find(struct lru_cache *lc, unsigned int enr);
250extern struct lc_element *lc_get(struct lru_cache *lc, unsigned int enr);
251extern unsigned int lc_put(struct lru_cache *lc, struct lc_element *e);
252extern void lc_changed(struct lru_cache *lc, struct lc_element *e);
253
254struct seq_file;
255extern size_t lc_seq_printf_stats(struct seq_file *seq, struct lru_cache *lc);
256
257extern void lc_seq_dump_details(struct seq_file *seq, struct lru_cache *lc, char *utext,
258 void (*detail) (struct seq_file *, struct lc_element *));
259
260/**
261 * lc_try_lock - can be used to stop lc_get() from changing the tracked set
262 * @lc: the lru cache to operate on
263 *
264 * Note that the reference counts and order on the active and lru lists may
265 * still change. Returns true if we aquired the lock.
266 */
267static inline int lc_try_lock(struct lru_cache *lc)
268{
269 return !test_and_set_bit(__LC_DIRTY, &lc->flags);
270}
271
272/**
273 * lc_unlock - unlock @lc, allow lc_get() to change the set again
274 * @lc: the lru cache to operate on
275 */
276static inline void lc_unlock(struct lru_cache *lc)
277{
278 clear_bit(__LC_DIRTY, &lc->flags);
279 smp_mb__after_clear_bit();
280}
281
282static inline int lc_is_used(struct lru_cache *lc, unsigned int enr)
283{
284 struct lc_element *e = lc_find(lc, enr);
285 return e && e->refcnt;
286}
287
288#define lc_entry(ptr, type, member) \
289 container_of(ptr, type, member)
290
291extern struct lc_element *lc_element_by_index(struct lru_cache *lc, unsigned i);
292extern unsigned int lc_index_of(struct lru_cache *lc, struct lc_element *e);
293
294#endif
diff --git a/include/linux/writeback.h b/include/linux/writeback.h
index 66ebddcff664..705f01fe413a 100644
--- a/include/linux/writeback.h
+++ b/include/linux/writeback.h
@@ -49,6 +49,7 @@ struct writeback_control {
49 unsigned nonblocking:1; /* Don't get stuck on request queues */ 49 unsigned nonblocking:1; /* Don't get stuck on request queues */
50 unsigned encountered_congestion:1; /* An output: a queue is full */ 50 unsigned encountered_congestion:1; /* An output: a queue is full */
51 unsigned for_kupdate:1; /* A kupdate writeback */ 51 unsigned for_kupdate:1; /* A kupdate writeback */
52 unsigned for_background:1; /* A background writeback */
52 unsigned for_reclaim:1; /* Invoked from the page allocator */ 53 unsigned for_reclaim:1; /* Invoked from the page allocator */
53 unsigned range_cyclic:1; /* range_start is cyclic */ 54 unsigned range_cyclic:1; /* range_start is cyclic */
54 unsigned more_io:1; /* more io to be dispatched */ 55 unsigned more_io:1; /* more io to be dispatched */
generated by cgit v1.2.2 (git 2.25.0) at 2024-12-18 07:01:49 -0500