projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch 'pm-cpufreq'
[linux-2.6-block.git] / drivers / staging / lustre / lustre / include / obd.h
CommitLineData
d7e09d03
PT		 1/*
2 * GPL HEADER START
3 *
4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
9 *
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
15 *
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
19 *
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
22 * have any questions.
23 *
24 * GPL HEADER END
25 */
26/*
27 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
29 *
30 * Copyright (c) 2011, 2012, Intel Corporation.
31 */
32/*
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
35 */
36
37#ifndef __OBD_H
38#define __OBD_H
39
1accaadf 40#include "linux/obd.h"
d7e09d03
PT		 41
42#define IOC_OSC_TYPE 'h'
43#define IOC_OSC_MIN_NR 20
44#define IOC_OSC_SET_ACTIVE _IOWR(IOC_OSC_TYPE, 21, struct obd_device *)
45#define IOC_OSC_MAX_NR 50
46
47#define IOC_MDC_TYPE 'i'
48#define IOC_MDC_MIN_NR 20
49#define IOC_MDC_MAX_NR 50
50
1accaadf
GKH		 51#include "lustre/lustre_idl.h"
52#include "lustre_lib.h"
1accaadf
GKH		 53#include "lu_ref.h"
54#include "lustre_export.h"
55#include "lustre_fid.h"
56#include "lustre_fld.h"
57#include "lustre_capa.h"
d7e09d03 58
d7e09d03
PT		 59#define MAX_OBD_DEVICES 8192
60
61struct osc_async_rc {
62 int ar_rc;
63 int ar_force_sync;
64 __u64 ar_min_xid;
65};
66
67struct lov_oinfo { /* per-stripe data structure */
68 struct ost_id loi_oi; /* object ID/Sequence on the target OST */
69 int loi_ost_idx; /* OST stripe index in lov_tgt_desc->tgts */
70 int loi_ost_gen; /* generation of this loi_ost_idx */
71
72 unsigned long loi_kms_valid:1;
73 __u64 loi_kms; /* known minimum size */
74 struct ost_lvb loi_lvb;
75 struct osc_async_rc loi_ar;
76};
77
78static inline void loi_kms_set(struct lov_oinfo *oinfo, __u64 kms)
79{
80 oinfo->loi_kms = kms;
81 oinfo->loi_kms_valid = 1;
82}
83
84static inline void loi_init(struct lov_oinfo *loi)
85{
86}
87
88struct lov_stripe_md {
89 atomic_t lsm_refc;
90 spinlock_t lsm_lock;
91 pid_t lsm_lock_owner; /* debugging */
92
93 /* maximum possible file size, might change as OSTs status changes,
94 * e.g. disconnected, deactivated */
95 __u64 lsm_maxbytes;
96 struct {
97 /* Public members. */
98 struct ost_id lw_object_oi; /* lov object id/seq */
99
100 /* LOV-private members start here -- only for use in lov/. */
101 __u32 lw_magic;
102 __u32 lw_stripe_size; /* size of the stripe */
103 __u32 lw_pattern; /* striping pattern (RAID0, RAID1) */
104 __u16 lw_stripe_count; /* number of objects being striped over */
105 __u16 lw_layout_gen; /* generation of the layout */
106 char lw_pool_name[LOV_MAXPOOLNAME]; /* pool name */
107 } lsm_wire;
108
109 struct lov_oinfo *lsm_oinfo[0];
110};
111
112#define lsm_oi lsm_wire.lw_object_oi
113#define lsm_magic lsm_wire.lw_magic
114#define lsm_layout_gen lsm_wire.lw_layout_gen
115#define lsm_stripe_size lsm_wire.lw_stripe_size
116#define lsm_pattern lsm_wire.lw_pattern
117#define lsm_stripe_count lsm_wire.lw_stripe_count
118#define lsm_pool_name lsm_wire.lw_pool_name
119
5dd16419
JX		 120static inline bool lsm_is_released(struct lov_stripe_md *lsm)
121{
122 return !!(lsm->lsm_pattern & LOV_PATTERN_F_RELEASED);
123}
124
125static inline bool lsm_has_objects(struct lov_stripe_md *lsm)
126{
127 if (lsm == NULL)
128 return false;
129 if (lsm_is_released(lsm))
130 return false;
131 return true;
132}
133
081b7265
JH		 134static inline int lov_stripe_md_size(unsigned int stripe_count)
135{
136 struct lov_stripe_md lsm;
137
138 return sizeof(lsm) + stripe_count * sizeof(lsm.lsm_oinfo[0]);
139}
140
d7e09d03
PT		 141struct obd_info;
142
143typedef int (*obd_enqueue_update_f)(void *cookie, int rc);
144
145/* obd info for a particular level (lov, osc). */
146struct obd_info {
147 /* Lock policy. It keeps an extent which is specific for a particular
148 * OSC. (e.g. lov_prep_enqueue_set initialises extent of the policy,
149 * and osc_enqueue passes it into ldlm_lock_match & ldlm_cli_enqueue. */
150 ldlm_policy_data_t oi_policy;
151 /* Flags used for set request specific flags:
152 - while lock handling, the flags obtained on the enqueue
153 request are set here.
154 - while stats, the flags used for control delay/resend.
155 - while setattr, the flags used for distinguish punch operation
156 */
157 __u64 oi_flags;
158 /* Lock handle specific for every OSC lock. */
159 struct lustre_handle *oi_lockh;
160 /* lsm data specific for every OSC. */
161 struct lov_stripe_md *oi_md;
162 /* obdo data specific for every OSC, if needed at all. */
163 struct obdo *oi_oa;
164 /* statfs data specific for every OSC, if needed at all. */
165 struct obd_statfs *oi_osfs;
166 /* An update callback which is called to update some data on upper
17891183 167 * level. E.g. it is used for update lsm->lsm_oinfo at every received
d7e09d03
PT		 168 * request in osc level for enqueue requests. It is also possible to
169 * update some caller data from LOV layer if needed. */
170 obd_enqueue_update_f oi_cb_up;
171 /* oss capability, its type is obd_capa in client to avoid copy.
172 * in contrary its type is lustre_capa in OSS. */
173 void *oi_capa;
174 /* transfer jobid from ost_sync() to filter_sync()... */
175 char *oi_jobid;
176};
177
178/* compare all relevant fields. */
179static inline int lov_stripe_md_cmp(struct lov_stripe_md *m1,
180 struct lov_stripe_md *m2)
181{
182 /*
183 * ->lsm_wire contains padding, but it should be zeroed out during
184 * allocation.
185 */
ec83e611 186 return memcmp(&m1->lsm_wire, &m2->lsm_wire, sizeof(m1->lsm_wire));
d7e09d03
PT		 187}
188
189static inline int lov_lum_lsm_cmp(struct lov_user_md *lum,
190 struct lov_stripe_md *lsm)
191{
192 if (lsm->lsm_magic != lum->lmm_magic)
193 return 1;
194 if ((lsm->lsm_stripe_count != 0) && (lum->lmm_stripe_count != 0) &&
195 (lsm->lsm_stripe_count != lum->lmm_stripe_count))
196 return 2;
197 if ((lsm->lsm_stripe_size != 0) && (lum->lmm_stripe_size != 0) &&
198 (lsm->lsm_stripe_size != lum->lmm_stripe_size))
199 return 3;
200 if ((lsm->lsm_pattern != 0) && (lum->lmm_pattern != 0) &&
201 (lsm->lsm_pattern != lum->lmm_pattern))
202 return 4;
203 if ((lsm->lsm_magic == LOV_MAGIC_V3) &&
204 (strncmp(lsm->lsm_pool_name,
205 ((struct lov_user_md_v3 *)lum)->lmm_pool_name,
206 LOV_MAXPOOLNAME) != 0))
207 return 5;
208 return 0;
209}
210
211static inline int lov_lum_swab_if_needed(struct lov_user_md_v3 *lumv3,
212 int *lmm_magic,
213 struct lov_user_md *lum)
214{
1d8cb70c 215 if (lum && copy_from_user(lumv3, lum, sizeof(struct lov_user_md_v1)))
d7e09d03
PT		 216 return -EFAULT;
217
218 *lmm_magic = lumv3->lmm_magic;
219
220 if (*lmm_magic == __swab32(LOV_USER_MAGIC_V1)) {
221 lustre_swab_lov_user_md_v1((struct lov_user_md_v1 *)lumv3);
222 *lmm_magic = LOV_USER_MAGIC_V1;
223 } else if (*lmm_magic == LOV_USER_MAGIC_V3) {
224 if (lum && copy_from_user(lumv3, lum, sizeof(*lumv3)))
225 return -EFAULT;
226 } else if (*lmm_magic == __swab32(LOV_USER_MAGIC_V3)) {
227 if (lum && copy_from_user(lumv3, lum, sizeof(*lumv3)))
228 return -EFAULT;
229 lustre_swab_lov_user_md_v3(lumv3);
230 *lmm_magic = LOV_USER_MAGIC_V3;
231 } else if (*lmm_magic != LOV_USER_MAGIC_V1) {
232 CDEBUG(D_IOCTL,
233 "bad userland LOV MAGIC: %#08x != %#08x nor %#08x\n",
234 *lmm_magic, LOV_USER_MAGIC_V1, LOV_USER_MAGIC_V3);
235 return -EINVAL;
236 }
237 return 0;
238}
239
240void lov_stripe_lock(struct lov_stripe_md *md);
241void lov_stripe_unlock(struct lov_stripe_md *md);
242
243struct obd_type {
244 struct list_head typ_chain;
245 struct obd_ops *typ_dt_ops;
246 struct md_ops *typ_md_ops;
61e87ab0 247 struct dentry *typ_debugfs_entry;
d7e09d03
PT		 248 char *typ_name;
249 int typ_refcnt;
250 struct lu_device_type *typ_lu;
251 spinlock_t obd_type_lock;
9b801302 252 struct kobject *typ_kobj;
d7e09d03
PT		 253};
254
255struct brw_page {
21aef7d9 256 u64 off;
d7e09d03
PT		 257 struct page *pg;
258 int count;
21aef7d9 259 u32 flag;
d7e09d03
PT		 260};
261
d7e09d03
PT		 262/* llog contexts */
263enum llog_ctxt_id {
264 LLOG_CONFIG_ORIG_CTXT = 0,
265 LLOG_CONFIG_REPL_CTXT,
266 LLOG_MDS_OST_ORIG_CTXT,
267 LLOG_MDS_OST_REPL_CTXT,
268 LLOG_SIZE_ORIG_CTXT,
269 LLOG_SIZE_REPL_CTXT,
270 LLOG_RD1_ORIG_CTXT,
271 LLOG_RD1_REPL_CTXT,
272 LLOG_TEST_ORIG_CTXT,
273 LLOG_TEST_REPL_CTXT,
274 LLOG_LOVEA_ORIG_CTXT,
275 LLOG_LOVEA_REPL_CTXT,
99a92265 276 LLOG_CHANGELOG_ORIG_CTXT, /**< changelog generation on mdd */
277 LLOG_CHANGELOG_REPL_CTXT, /**< changelog access on clients */
278 LLOG_CHANGELOG_USER_ORIG_CTXT, /**< for multiple changelog consumers */
279 LLOG_AGENT_ORIG_CTXT, /**< agent requests generation on cdt */
d7e09d03
PT		 280 LLOG_MAX_CTXTS
281};
282
d7e09d03
PT		 283struct timeout_item {
284 enum timeout_event ti_event;
a649ad1d 285 unsigned long ti_timeout;
d7e09d03
PT		 286 timeout_cb_t ti_cb;
287 void *ti_cb_data;
288 struct list_head ti_obd_list;
289 struct list_head ti_chain;
290};
291
292#define OSC_MAX_RIF_DEFAULT 8
293#define MDS_OSC_MAX_RIF_DEFAULT 50
294#define OSC_MAX_RIF_MAX 256
295#define OSC_MAX_DIRTY_DEFAULT (OSC_MAX_RIF_DEFAULT * 4)
296#define OSC_MAX_DIRTY_MB_MAX 2048 /* arbitrary, but < MAX_LONG bytes */
297#define OSC_DEFAULT_RESENDS 10
298
299/* possible values for fo_sync_lock_cancel */
300enum {
301 NEVER_SYNC_ON_CANCEL = 0,
302 BLOCKING_SYNC_ON_CANCEL = 1,
303 ALWAYS_SYNC_ON_CANCEL = 2,
304 NUM_SYNC_ON_CANCEL_STATES
305};
306
307#define MDC_MAX_RIF_DEFAULT 8
308#define MDC_MAX_RIF_MAX 512
309
310struct mdc_rpc_lock;
311struct obd_import;
312struct client_obd {
313 struct rw_semaphore cl_sem;
314 struct obd_uuid cl_target_uuid;
315 struct obd_import *cl_import; /* ptlrpc connection state */
316 int cl_conn_count;
317 /* max_mds_easize is purely a performance thing so we don't have to
318 * call obd_size_diskmd() all the time. */
44779340
BB		 319 int cl_default_mds_easize;
320 int cl_max_mds_easize;
321 int cl_default_mds_cookiesize;
322 int cl_max_mds_cookiesize;
d7e09d03
PT		 323
324 enum lustre_sec_part cl_sp_me;
325 enum lustre_sec_part cl_sp_to;
326 struct sptlrpc_flavor cl_flvr_mgc; /* fixed flavor of mgc->mgs */
327
328 /* the grant values are protected by loi_list_lock below */
329 long cl_dirty; /* all _dirty_ in bytes */
330 long cl_dirty_max; /* allowed w/o rpc */
331 long cl_dirty_transit; /* dirty synchronous */
332 long cl_avail_grant; /* bytes of credit for ost */
333 long cl_lost_grant; /* lost credits (trunc) */
334
335 /* since we allocate grant by blocks, we don't know how many grant will
336 * be used to add a page into cache. As a solution, we reserve maximum
337 * grant before trying to dirty a page and unreserve the rest.
338 * See osc_{reserve|unreserve}_grant for details. */
339 long cl_reserved_grant;
498374f7
HM		 340 struct list_head cl_cache_waiters; /* waiting for cache/grant */
341 unsigned long cl_next_shrink_grant; /* jiffies */
342 struct list_head cl_grant_shrink_list; /* Timeout event list */
343 int cl_grant_shrink_interval; /* seconds */
d7e09d03
PT		 344
345 /* A chunk is an optimal size used by osc_extent to determine
346 * the extent size. A chunk is max(PAGE_CACHE_SIZE, OST block size) */
347 int cl_chunkbits;
348 int cl_chunk;
349 int cl_extent_tax; /* extent overhead, by bytes */
350
351 /* keep track of objects that have lois that contain pages which
352 * have been queued for async brw. this lock also protects the
353 * lists of osc_client_pages that hang off of the loi */
354 /*
355 * ->cl_loi_list_lock protects consistency of
356 * ->cl_loi_{ready,read,write}_list. ->ap_make_ready() and
357 * ->ap_completion() call-backs are executed under this lock. As we
358 * cannot guarantee that these call-backs never block on all platforms
359 * (as a matter of fact they do block on Mac OS X), type of
360 * ->cl_loi_list_lock is platform dependent: it's a spin-lock on Linux
361 * and blocking mutex on Mac OS X. (Alternative is to make this lock
362 * blocking everywhere, but we don't want to slow down fast-path of
363 * our main platform.)
364 *
365 * Exact type of ->cl_loi_list_lock is defined in arch/obd.h together
366 * with client_obd_list_{un,}lock() and
367 * client_obd_list_lock_{init,done}() functions.
368 *
369 * NB by Jinshan: though field names are still _loi_, but actually
370 * osc_object{}s are in the list.
371 */
372 client_obd_lock_t cl_loi_list_lock;
373 struct list_head cl_loi_ready_list;
374 struct list_head cl_loi_hp_ready_list;
375 struct list_head cl_loi_write_list;
376 struct list_head cl_loi_read_list;
377 int cl_r_in_flight;
378 int cl_w_in_flight;
379 /* just a sum of the loi/lop pending numbers to be exported by /proc */
380 atomic_t cl_pending_w_pages;
381 atomic_t cl_pending_r_pages;
382 __u32 cl_max_pages_per_rpc;
383 int cl_max_rpcs_in_flight;
384 struct obd_histogram cl_read_rpc_hist;
385 struct obd_histogram cl_write_rpc_hist;
386 struct obd_histogram cl_read_page_hist;
387 struct obd_histogram cl_write_page_hist;
388 struct obd_histogram cl_read_offset_hist;
389 struct obd_histogram cl_write_offset_hist;
390
391 /* lru for osc caching pages */
392 struct cl_client_cache *cl_cache;
498374f7 393 struct list_head cl_lru_osc; /* member of cl_cache->ccc_lru */
d7e09d03
PT		 394 atomic_t *cl_lru_left;
395 atomic_t cl_lru_busy;
396 atomic_t cl_lru_shrinkers;
397 atomic_t cl_lru_in_list;
498374f7 398 struct list_head cl_lru_list; /* lru page list */
d7e09d03
PT		 399 client_obd_lock_t cl_lru_list_lock; /* page list protector */
400
401 /* number of in flight destroy rpcs is limited to max_rpcs_in_flight */
402 atomic_t cl_destroy_in_flight;
403 wait_queue_head_t cl_destroy_waitq;
404
405 struct mdc_rpc_lock *cl_rpc_lock;
406 struct mdc_rpc_lock *cl_close_lock;
407
408 /* mgc datastruct */
6246dab1 409 struct mutex cl_mgc_mutex;
aa4e3c8a
MP		 410 struct local_oid_storage *cl_mgc_los;
411 struct dt_object *cl_mgc_configs_dir;
d7e09d03
PT		 412 atomic_t cl_mgc_refcount;
413 struct obd_export *cl_mgc_mgsexp;
414
415 /* checksumming for data sent over the network */
416 unsigned int cl_checksum:1; /* 0 = disabled, 1 = enabled */
417 /* supported checksum types that are worked out at connect time */
418 __u32 cl_supp_cksum_types;
419 /* checksum algorithm to be used */
420 cksum_type_t cl_cksum_type;
421
422 /* also protected by the poorly named _loi_list_lock lock above */
423 struct osc_async_rc cl_ar;
424
425 /* used by quotacheck when the servers are older than 2.4 */
426 int cl_qchk_stat; /* quotacheck stat of the peer */
427#define CL_NOT_QUOTACHECKED 1 /* client->cl_qchk_stat init value */
428#if LUSTRE_VERSION_CODE >= OBD_OCD_VERSION(2, 7, 50, 0)
429#warning "please consider removing quotacheck compatibility code"
430#endif
431
432 /* sequence manager */
433 struct lu_client_seq *cl_seq;
434
435 atomic_t cl_resends; /* resend count */
436
437 /* ptlrpc work for writeback in ptlrpcd context */
438 void *cl_writeback_work;
439 /* hash tables for osc_quota_info */
6da6eabe 440 struct cfs_hash *cl_quota_hash[MAXQUOTAS];
d7e09d03
PT		 441};
442#define obd2cli_tgt(obd) ((char *)(obd)->u.cli.cl_target_uuid.uuid)
443
444struct obd_id_info {
445 __u32 idx;
21aef7d9 446 u64 *data;
d7e09d03
PT		 447};
448
d7e09d03
PT		 449struct echo_client_obd {
450 struct obd_export *ec_exp; /* the local connection to osc/lov */
451 spinlock_t ec_lock;
452 struct list_head ec_objects;
453 struct list_head ec_locks;
454 int ec_nstripes;
455 __u64 ec_unique;
456};
457
458struct lov_qos_oss {
459 struct obd_uuid lqo_uuid; /* ptlrpc's c_remote_uuid */
460 struct list_head lqo_oss_list; /* link to lov_qos */
461 __u64 lqo_bavail; /* total bytes avail on OSS */
462 __u64 lqo_penalty; /* current penalty */
463 __u64 lqo_penalty_per_obj;/* penalty decrease every obj*/
464 time_t lqo_used; /* last used time, seconds */
465 __u32 lqo_ost_count; /* number of osts on this oss */
466};
467
468struct ltd_qos {
469 struct lov_qos_oss *ltq_oss; /* oss info */
470 __u64 ltq_penalty; /* current penalty */
471 __u64 ltq_penalty_per_obj; /* penalty decrease every obj*/
472 __u64 ltq_weight; /* net weighting */
473 time_t ltq_used; /* last used time, seconds */
474 unsigned int ltq_usable:1; /* usable for striping */
475};
476
477/* Generic subset of OSTs */
478struct ost_pool {
479 __u32 *op_array; /* array of index of
480 lov_obd->lov_tgts */
481 unsigned int op_count; /* number of OSTs in the array */
482 unsigned int op_size; /* allocated size of lp_array */
483 struct rw_semaphore op_rw_sem; /* to protect ost_pool use */
484};
485
486/* Round-robin allocator data */
487struct lov_qos_rr {
488 __u32 lqr_start_idx; /* start index of new inode */
489 __u32 lqr_offset_idx; /* aliasing for start_idx */
490 int lqr_start_count; /* reseed counter */
491 struct ost_pool lqr_pool; /* round-robin optimized list */
492 unsigned long lqr_dirty:1; /* recalc round-robin list */
493};
494
495/* allow statfs data caching for 1 second */
496#define OBD_STATFS_CACHE_SECONDS 1
497
498struct lov_statfs_data {
499 struct obd_info lsd_oi;
500 struct obd_statfs lsd_statfs;
501};
502/* Stripe placement optimization */
503struct lov_qos {
498374f7 504 struct list_head lq_oss_list; /* list of OSSs that targets use */
d7e09d03 505 struct rw_semaphore lq_rw_sem;
498374f7 506 __u32 lq_active_oss_count;
d7e09d03
PT		 507 unsigned int lq_prio_free; /* priority for free space */
508 unsigned int lq_threshold_rr;/* priority for rr */
509 struct lov_qos_rr lq_rr; /* round robin qos data */
510 unsigned long lq_dirty:1, /* recalc qos data */
511 lq_same_space:1,/* the ost's all have approx.
512 the same space avail */
513 lq_reset:1, /* zero current penalties */
514 lq_statfs_in_progress:1; /* statfs op in
515 progress */
516 /* qos statfs data */
517 struct lov_statfs_data *lq_statfs_data;
518 wait_queue_head_t lq_statfs_waitq; /* waitqueue to notify statfs
519 * requests completion */
520};
521
522struct lov_tgt_desc {
523 struct list_head ltd_kill;
524 struct obd_uuid ltd_uuid;
525 struct obd_device *ltd_obd;
526 struct obd_export *ltd_exp;
527 struct ltd_qos ltd_qos; /* qos info per target */
528 __u32 ltd_gen;
529 __u32 ltd_index; /* index in lov_obd->tgts */
530 unsigned long ltd_active:1,/* is this target up for requests */
531 ltd_activate:1,/* should target be activated */
532 ltd_reap:1; /* should this target be deleted */
533};
534
535/* Pool metadata */
536#define pool_tgt_size(_p) _p->pool_obds.op_size
537#define pool_tgt_count(_p) _p->pool_obds.op_count
538#define pool_tgt_array(_p) _p->pool_obds.op_array
539#define pool_tgt_rw_sem(_p) _p->pool_obds.op_rw_sem
540
541struct pool_desc {
542 char pool_name[LOV_MAXPOOLNAME + 1]; /* name of pool */
543 struct ost_pool pool_obds; /* pool members */
544 atomic_t pool_refcount; /* pool ref. counter */
545 struct lov_qos_rr pool_rr; /* round robin qos */
546 struct hlist_node pool_hash; /* access by poolname */
547 struct list_head pool_list; /* serial access */
61e87ab0 548 struct dentry *pool_debugfs_entry; /* file in /proc */
498374f7
HM		 549 struct obd_device *pool_lobd; /* obd of the lov/lod to which
550 * this pool belongs */
d7e09d03
PT		 551};
552
553struct lov_obd {
554 struct lov_desc desc;
555 struct lov_tgt_desc **lov_tgts; /* sparse array */
556 struct ost_pool lov_packed; /* all OSTs in a packed
557 array */
558 struct mutex lov_lock;
559 struct obd_connect_data lov_ocd;
560 atomic_t lov_refcount;
561 __u32 lov_tgt_count; /* how many OBD's */
562 __u32 lov_active_tgt_count; /* how many active */
563 __u32 lov_death_row;/* tgts scheduled to be deleted */
564 __u32 lov_tgt_size; /* size of tgts array */
565 int lov_connects;
566 int lov_pool_count;
6da6eabe 567 struct cfs_hash *lov_pools_hash_body; /* used for key access */
498374f7 568 struct list_head lov_pool_list; /* used for sequential access */
61e87ab0 569 struct dentry *lov_pool_debugfs_entry;
d7e09d03
PT		 570 enum lustre_sec_part lov_sp_me;
571
c52f69c5 572 /* Cached LRU pages from upper layer */
d7e09d03
PT		 573 void *lov_cache;
574
575 struct rw_semaphore lov_notify_lock;
cb1debff
OD		 576
577 struct kobject *lov_tgts_kobj;
d7e09d03
PT		 578};
579
580struct lmv_tgt_desc {
581 struct obd_uuid ltd_uuid;
582 struct obd_export *ltd_exp;
583 int ltd_idx;
584 struct mutex ltd_fid_mutex;
585 unsigned long ltd_active:1; /* target up for requests */
586};
587
588enum placement_policy {
589 PLACEMENT_CHAR_POLICY = 0,
590 PLACEMENT_NID_POLICY = 1,
591 PLACEMENT_INVAL_POLICY = 2,
592 PLACEMENT_MAX_POLICY
593};
594
d7e09d03
PT		 595struct lmv_obd {
596 int refcount;
597 struct lu_client_fld lmv_fld;
598 spinlock_t lmv_lock;
fe672997 599 enum placement_policy lmv_placement;
d7e09d03
PT		 600 struct lmv_desc desc;
601 struct obd_uuid cluuid;
602 struct obd_export *exp;
603
604 struct mutex init_mutex;
605 int connected;
606 int max_easize;
607 int max_def_easize;
608 int max_cookiesize;
44779340 609 int max_def_cookiesize;
d7e09d03
PT		 610 int server_timeout;
611
612 int tgts_size; /* size of tgts array */
613 struct lmv_tgt_desc **tgts;
614
615 struct obd_connect_data conn_data;
b5fa70d7 616 struct kobject *lmv_tgts_kobj;
d7e09d03
PT		 617};
618
619struct niobuf_local {
620 __u64 lnb_file_offset;
621 __u32 lnb_page_offset;
622 __u32 len;
623 __u32 flags;
624 struct page *page;
625 struct dentry *dentry;
626 int lnb_grant_used;
627 int rc;
628};
629
630#define LUSTRE_FLD_NAME "fld"
631#define LUSTRE_SEQ_NAME "seq"
632
633#define LUSTRE_MDD_NAME "mdd"
634#define LUSTRE_OSD_LDISKFS_NAME "osd-ldiskfs"
635#define LUSTRE_OSD_ZFS_NAME "osd-zfs"
636#define LUSTRE_VVP_NAME "vvp"
637#define LUSTRE_LMV_NAME "lmv"
638#define LUSTRE_SLP_NAME "slp"
639#define LUSTRE_LOD_NAME "lod"
640#define LUSTRE_OSP_NAME "osp"
641#define LUSTRE_LWP_NAME "lwp"
642
643/* obd device type names */
644 /* FIXME all the references to LUSTRE_MDS_NAME should be swapped with LUSTRE_MDT_NAME */
645#define LUSTRE_MDS_NAME "mds"
646#define LUSTRE_MDT_NAME "mdt"
647#define LUSTRE_MDC_NAME "mdc"
648#define LUSTRE_OSS_NAME "ost" /* FIXME change name to oss */
649#define LUSTRE_OST_NAME "obdfilter" /* FIXME change name to ost */
650#define LUSTRE_OSC_NAME "osc"
651#define LUSTRE_LOV_NAME "lov"
652#define LUSTRE_MGS_NAME "mgs"
653#define LUSTRE_MGC_NAME "mgc"
654
655#define LUSTRE_ECHO_NAME "obdecho"
656#define LUSTRE_ECHO_CLIENT_NAME "echo_client"
657#define LUSTRE_QMT_NAME "qmt"
658
659/* Constant obd names (post-rename) */
660#define LUSTRE_MDS_OBDNAME "MDS"
661#define LUSTRE_OSS_OBDNAME "OSS"
662#define LUSTRE_MGS_OBDNAME "MGS"
663#define LUSTRE_MGC_OBDNAME "MGC"
664
d7e09d03
PT		 665/* Don't conflict with on-wire flags OBD_BRW_WRITE, etc */
666#define N_LOCAL_TEMP_PAGE 0x10000000
667
668struct obd_trans_info {
669 __u64 oti_transno;
670 __u64 oti_xid;
671 /* Only used on the server side for tracking acks. */
672 struct oti_req_ack_lock {
673 struct lustre_handle lock;
674 __u32 mode;
675 } oti_ack_locks[4];
676 void *oti_handle;
677 struct llog_cookie oti_onecookie;
678 struct llog_cookie *oti_logcookies;
679 int oti_numcookies;
680 /** synchronous write is needed */
681 unsigned long oti_sync_write:1;
682
683 /* initial thread handling transaction */
498374f7 684 struct ptlrpc_thread *oti_thread;
d7e09d03
PT		 685 __u32 oti_conn_cnt;
686 /** VBR: versions */
687 __u64 oti_pre_version;
688 /** JobID */
689 char *oti_jobid;
690
691 struct obd_uuid *oti_ost_uuid;
692};
693
694static inline void oti_init(struct obd_trans_info *oti,
695 struct ptlrpc_request *req)
696{
697 if (oti == NULL)
698 return;
699 memset(oti, 0, sizeof(*oti));
700
701 if (req == NULL)
702 return;
703
704 oti->oti_xid = req->rq_xid;
705 /** VBR: take versions from request */
706 if (req->rq_reqmsg != NULL &&
707 lustre_msg_get_flags(req->rq_reqmsg) & MSG_REPLAY) {
708 __u64 *pre_version = lustre_msg_get_versions(req->rq_reqmsg);
498374f7 709
d7e09d03
PT		 710 oti->oti_pre_version = pre_version ? pre_version[0] : 0;
711 oti->oti_transno = lustre_msg_get_transno(req->rq_reqmsg);
712 }
713
714 /** called from mds_create_objects */
715 if (req->rq_repmsg != NULL)
716 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
717 oti->oti_thread = req->rq_svc_thread;
718 if (req->rq_reqmsg != NULL)
719 oti->oti_conn_cnt = lustre_msg_get_conn_cnt(req->rq_reqmsg);
720}
721
1d8cb70c
GD		 722static inline void oti_alloc_cookies(struct obd_trans_info *oti,
723 int num_cookies)
d7e09d03
PT		 724{
725 if (!oti)
726 return;
727
728 if (num_cookies == 1)
729 oti->oti_logcookies = &oti->oti_onecookie;
730 else
731 OBD_ALLOC_LARGE(oti->oti_logcookies,
732 num_cookies * sizeof(oti->oti_onecookie));
733
734 oti->oti_numcookies = num_cookies;
735}
736
737static inline void oti_free_cookies(struct obd_trans_info *oti)
738{
739 if (!oti || !oti->oti_logcookies)
740 return;
741
742 if (oti->oti_logcookies == &oti->oti_onecookie)
743 LASSERT(oti->oti_numcookies == 1);
744 else
745 OBD_FREE_LARGE(oti->oti_logcookies,
746 oti->oti_numcookies*sizeof(oti->oti_onecookie));
747 oti->oti_logcookies = NULL;
748 oti->oti_numcookies = 0;
749}
750
751/*
752 * Events signalled through obd_notify() upcall-chain.
753 */
754enum obd_notify_event {
755 /* target added */
756 OBD_NOTIFY_CREATE,
757 /* Device connect start */
758 OBD_NOTIFY_CONNECT,
759 /* Device activated */
760 OBD_NOTIFY_ACTIVE,
761 /* Device deactivated */
762 OBD_NOTIFY_INACTIVE,
763 /* Device disconnected */
764 OBD_NOTIFY_DISCON,
765 /* Connect data for import were changed */
766 OBD_NOTIFY_OCD,
767 /* Sync request */
768 OBD_NOTIFY_SYNC_NONBLOCK,
769 OBD_NOTIFY_SYNC,
770 /* Configuration event */
771 OBD_NOTIFY_CONFIG,
772 /* Administratively deactivate/activate event */
773 OBD_NOTIFY_DEACTIVATE,
774 OBD_NOTIFY_ACTIVATE
775};
776
d7e09d03
PT		 777/*
778 * Data structure used to pass obd_notify()-event to non-obd listeners (llite
779 * and liblustre being main examples).
780 */
781struct obd_notify_upcall {
782 int (*onu_upcall)(struct obd_device *host, struct obd_device *watched,
783 enum obd_notify_event ev, void *owner, void *data);
784 /* Opaque datum supplied by upper layer listener */
785 void *onu_owner;
786};
787
788struct target_recovery_data {
789 svc_handler_t trd_recovery_handler;
790 pid_t trd_processing_task;
791 struct completion trd_starting;
792 struct completion trd_finishing;
793};
794
795struct obd_llog_group {
796 int olg_seq;
797 struct llog_ctxt *olg_ctxts[LLOG_MAX_CTXTS];
798 wait_queue_head_t olg_waitq;
799 spinlock_t olg_lock;
800 struct mutex olg_cat_processing;
801};
802
803/* corresponds to one of the obd's */
804#define OBD_DEVICE_MAGIC 0XAB5CD6EF
805#define OBD_DEV_BY_DEVNAME 0xffffd0de
806
91b3aaf9
JH		 807struct lvfs_run_ctxt {
808 struct dt_device *dt;
809};
810
d7e09d03
PT		 811struct obd_device {
812 struct obd_type *obd_type;
813 __u32 obd_magic;
814
815 /* common and UUID name of this device */
816 char obd_name[MAX_OBD_NAME];
817 struct obd_uuid obd_uuid;
818
819 struct lu_device *obd_lu_dev;
820
821 int obd_minor;
822 /* bitfield modification is protected by obd_dev_lock */
823 unsigned long obd_attached:1, /* finished attach */
824 obd_set_up:1, /* finished setup */
825 obd_recovering:1, /* there are recoverable clients */
826 obd_abort_recovery:1,/* recovery expired */
827 obd_version_recov:1, /* obd uses version checking */
828 obd_replayable:1, /* recovery is enabled; inform clients */
829 obd_no_transno:1, /* no committed-transno notification */
830 obd_no_recov:1, /* fail instead of retry messages */
831 obd_stopping:1, /* started cleanup */
832 obd_starting:1, /* started setup */
833 obd_force:1, /* cleanup with > 0 obd refcount */
834 obd_fail:1, /* cleanup with failover */
835 obd_async_recov:1, /* allow asynchronous orphan cleanup */
836 obd_no_conn:1, /* deny new connections */
837 obd_inactive:1, /* device active/inactive
838 * (for /proc/status only!!) */
839 obd_no_ir:1, /* no imperative recovery. */
840 obd_process_conf:1; /* device is processing mgs config */
841 /* use separate field as it is set in interrupt to don't mess with
842 * protection of other bits using _bh lock */
843 unsigned long obd_recovery_expired:1;
844 /* uuid-export hash body */
6da6eabe 845 struct cfs_hash *obd_uuid_hash;
d7e09d03 846 /* nid-export hash body */
6da6eabe 847 struct cfs_hash *obd_nid_hash;
d7e09d03
PT		 848 atomic_t obd_refcount;
849 wait_queue_head_t obd_refcount_waitq;
850 struct list_head obd_exports;
851 struct list_head obd_unlinked_exports;
852 struct list_head obd_delayed_exports;
853 int obd_num_exports;
854 spinlock_t obd_nid_lock;
855 struct ldlm_namespace *obd_namespace;
856 struct ptlrpc_client obd_ldlm_client; /* XXX OST/MDS only */
857 /* a spinlock is OK for what we do now, may need a semaphore later */
858 spinlock_t obd_dev_lock; /* protect OBD bitfield above */
859 struct mutex obd_dev_mutex;
860 __u64 obd_last_committed;
d7e09d03
PT		 861 spinlock_t obd_osfs_lock;
862 struct obd_statfs obd_osfs; /* locked by obd_osfs_lock */
863 __u64 obd_osfs_age;
864 struct lvfs_run_ctxt obd_lvfs_ctxt;
865 struct obd_llog_group obd_olg; /* default llog group */
866 struct obd_device *obd_observer;
867 struct rw_semaphore obd_observer_link_sem;
868 struct obd_notify_upcall obd_upcall;
869 struct obd_export *obd_self_export;
870 /* list of exports in LRU order, for ping evictor, with obd_dev_lock */
871 struct list_head obd_exports_timed;
872 time_t obd_eviction_timer; /* for ping evictor */
873
874 int obd_max_recoverable_clients;
875 atomic_t obd_connected_clients;
876 int obd_stale_clients;
877 int obd_delayed_clients;
878 /* this lock protects all recovery list_heads, timer and
879 * obd_next_recovery_transno value */
880 spinlock_t obd_recovery_task_lock;
881 __u64 obd_next_recovery_transno;
882 int obd_replayed_requests;
883 int obd_requests_queued_for_recovery;
884 wait_queue_head_t obd_next_transno_waitq;
885 /* protected by obd_recovery_task_lock */
498374f7
HM		 886 struct timer_list obd_recovery_timer;
887 time_t obd_recovery_start; /* seconds */
888 time_t obd_recovery_end; /* seconds, for lprocfs_status */
d7e09d03
PT		 889 int obd_recovery_time_hard;
890 int obd_recovery_timeout;
891 int obd_recovery_ir_factor;
892
893 /* new recovery stuff from CMD2 */
894 struct target_recovery_data obd_recovery_data;
895 int obd_replayed_locks;
896 atomic_t obd_req_replay_clients;
897 atomic_t obd_lock_replay_clients;
898 /* all lists are protected by obd_recovery_task_lock */
899 struct list_head obd_req_replay_queue;
900 struct list_head obd_lock_replay_queue;
901 struct list_head obd_final_req_queue;
902 int obd_recovery_stage;
903
904 union {
d7e09d03 905 struct client_obd cli;
d7e09d03 906 struct echo_client_obd echo_client;
d7e09d03
PT		 907 struct lov_obd lov;
908 struct lmv_obd lmv;
909 } u;
910 /* Fields used by LProcFS */
911 unsigned int obd_cntr_base;
912 struct lprocfs_stats *obd_stats;
913
914 unsigned int md_cntr_base;
915 struct lprocfs_stats *md_stats;
916
61e87ab0 917 struct dentry *obd_debugfs_entry;
61e87ab0 918 struct dentry *obd_svc_debugfs_entry;
d7e09d03
PT		 919 struct lprocfs_stats *obd_svc_stats;
920 atomic_t obd_evict_inprogress;
921 wait_queue_head_t obd_evict_inprogress_waitq;
498374f7 922 struct list_head obd_evict_list; /* protected with pet_lock */
d7e09d03
PT		 923
924 /**
925 * Ldlm pool part. Save last calculated SLV and Limit.
926 */
927 rwlock_t obd_pool_lock;
928 int obd_pool_limit;
929 __u64 obd_pool_slv;
930
931 /**
932 * A list of outstanding class_incref()'s against this obd. For
933 * debugging.
934 */
935 struct lu_ref obd_reference;
936
937 int obd_conn_inprogress;
9b801302
OD		 938
939 struct kobject obd_kobj; /* sysfs object */
940 struct completion obd_kobj_unregister;
d7e09d03
PT		 941};
942
943#define OBD_LLOG_FL_SENDNOW 0x0001
944#define OBD_LLOG_FL_EXIT 0x0002
945
946enum obd_cleanup_stage {
947/* Special case hack for MDS LOVs */
948 OBD_CLEANUP_EARLY,
949/* can be directly mapped to .ldto_device_fini() */
950 OBD_CLEANUP_EXPORTS,
951};
952
953/* get/set_info keys */
954#define KEY_ASYNC "async"
955#define KEY_BLOCKSIZE_BITS "blocksize_bits"
956#define KEY_BLOCKSIZE "blocksize"
957#define KEY_CAPA_KEY "capa_key"
958#define KEY_CHANGELOG_CLEAR "changelog_clear"
959#define KEY_FID2PATH "fid2path"
960#define KEY_CHECKSUM "checksum"
961#define KEY_CLEAR_FS "clear_fs"
962#define KEY_CONN_DATA "conn_data"
963#define KEY_EVICT_BY_NID "evict_by_nid"
964#define KEY_FIEMAP "fiemap"
965#define KEY_FLUSH_CTX "flush_ctx"
966#define KEY_GRANT_SHRINK "grant_shrink"
967#define KEY_HSM_COPYTOOL_SEND "hsm_send"
968#define KEY_INIT_RECOV_BACKUP "init_recov_bk"
969#define KEY_INIT_RECOV "initial_recov"
970#define KEY_INTERMDS "inter_mds"
971#define KEY_LAST_ID "last_id"
972#define KEY_LAST_FID "last_fid"
973#define KEY_LOCK_TO_STRIPE "lock_to_stripe"
974#define KEY_LOVDESC "lovdesc"
975#define KEY_LOV_IDX "lov_idx"
44779340
BB		 976#define KEY_MAX_EASIZE "max_easize"
977#define KEY_DEFAULT_EASIZE "default_easize"
978#define KEY_MAX_COOKIESIZE "max_cookiesize"
979#define KEY_DEFAULT_COOKIESIZE "default_cookiesize"
d7e09d03
PT		 980#define KEY_MDS_CONN "mds_conn"
981#define KEY_MGSSEC "mgssec"
982#define KEY_NEXT_ID "next_id"
983#define KEY_READ_ONLY "read-only"
984#define KEY_REGISTER_TARGET "register_target"
985#define KEY_SET_FS "set_fs"
986#define KEY_TGT_COUNT "tgt_count"
987/* KEY_SET_INFO in lustre_idl.h */
988#define KEY_SPTLRPC_CONF "sptlrpc_conf"
989#define KEY_CONNECT_FLAG "connect_flags"
990#define KEY_SYNC_LOCK_CANCEL "sync_lock_cancel"
991
992#define KEY_CACHE_SET "cache_set"
993#define KEY_CACHE_LRU_SHRINK "cache_lru_shrink"
994#define KEY_CHANGELOG_INDEX "changelog_index"
995
996struct lu_context;
997
998/* /!\ must be coherent with include/linux/namei.h on patched kernel */
999#define IT_OPEN (1 << 0)
1000#define IT_CREAT (1 << 1)
1001#define IT_READDIR (1 << 2)
1002#define IT_GETATTR (1 << 3)
1003#define IT_LOOKUP (1 << 4)
1004#define IT_UNLINK (1 << 5)
1005#define IT_TRUNC (1 << 6)
1006#define IT_GETXATTR (1 << 7)
1007#define IT_EXEC (1 << 8)
1008#define IT_PIN (1 << 9)
1009#define IT_LAYOUT (1 << 10)
1010#define IT_QUOTA_DQACQ (1 << 11)
1011#define IT_QUOTA_CONN (1 << 12)
7fc1f831 1012#define IT_SETXATTR (1 << 13)
d7e09d03
PT		 1013
1014static inline int it_to_lock_mode(struct lookup_intent *it)
1015{
1016 /* CREAT needs to be tested before open (both could be set) */
1017 if (it->it_op & IT_CREAT)
1018 return LCK_CW;
1019 else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP |
1020 IT_LAYOUT))
1021 return LCK_CR;
7fc1f831
AP		 1022 else if (it->it_op & IT_GETXATTR)
1023 return LCK_PR;
1024 else if (it->it_op & IT_SETXATTR)
1025 return LCK_PW;
d7e09d03
PT		 1026
1027 LASSERTF(0, "Invalid it_op: %d\n", it->it_op);
1028 return -EINVAL;
1029}
1030
1031struct md_op_data {
17891183
MI		 1032 struct lu_fid op_fid1; /* operation fid1 (usually parent) */
1033 struct lu_fid op_fid2; /* operation fid2 (usually child) */
d7e09d03
PT		 1034 struct lu_fid op_fid3; /* 2 extra fids to find conflicting */
1035 struct lu_fid op_fid4; /* to the operation locks. */
114acca8 1036 u32 op_mds; /* what mds server open will go to */
d7e09d03 1037 struct lustre_handle op_handle;
21aef7d9 1038 s64 op_mod_time;
d7e09d03
PT		 1039 const char *op_name;
1040 int op_namelen;
1041 __u32 op_mode;
1042 struct lmv_stripe_md *op_mea1;
1043 struct lmv_stripe_md *op_mea2;
1044 __u32 op_suppgids[2];
1045 __u32 op_fsuid;
1046 __u32 op_fsgid;
1047 cfs_cap_t op_cap;
1048 void *op_data;
1049
1050 /* iattr fields and blocks. */
1051 struct iattr op_attr;
1052 unsigned int op_attr_flags;
1053 __u64 op_valid;
1054 loff_t op_attr_blocks;
1055
1056 /* Size-on-MDS epoch and flags. */
1057 __u64 op_ioepoch;
1058 __u32 op_flags;
1059
1060 /* Capa fields */
1061 struct obd_capa *op_capa1;
1062 struct obd_capa *op_capa2;
1063
1064 /* Various operation flags. */
48d23e61 1065 enum mds_op_bias op_bias;
d7e09d03
PT		 1066
1067 /* Operation type */
1068 __u32 op_opc;
1069
1070 /* Used by readdir */
1071 __u64 op_offset;
1072
1073 /* Used by readdir */
1074 __u32 op_npages;
1075
1076 /* used to transfer info between the stacks of MD client
1077 * see enum op_cli_flags */
1078 __u32 op_cli_flags;
48d23e61
JX		 1079
1080 /* File object data version for HSM release, on client */
1081 __u64 op_data_version;
1082 struct lustre_handle op_lease_handle;
d7e09d03
PT		 1083};
1084
1085enum op_cli_flags {
1086 CLI_SET_MEA = 1 << 0,
1087 CLI_RM_ENTRY = 1 << 1,
1088};
1089
1090struct md_enqueue_info;
1091/* metadata stat-ahead */
d7e09d03 1092
d7e09d03
PT		 1093struct md_enqueue_info {
1094 struct md_op_data mi_data;
1095 struct lookup_intent mi_it;
1096 struct lustre_handle mi_lockh;
1097 struct inode *mi_dir;
fe672997
HM		 1098 int (*mi_cb)(struct ptlrpc_request *req,
1099 struct md_enqueue_info *minfo, int rc);
d7e09d03
PT		 1100 __u64 mi_cbdata;
1101 unsigned int mi_generation;
1102};
1103
1104struct obd_ops {
c34d9cd8 1105 struct module *o_owner;
d7e09d03
PT		 1106 int (*o_iocontrol)(unsigned int cmd, struct obd_export *exp, int len,
1107 void *karg, void *uarg);
1108 int (*o_get_info)(const struct lu_env *env, struct obd_export *,
1109 __u32 keylen, void *key, __u32 *vallen, void *val,
1110 struct lov_stripe_md *lsm);
1111 int (*o_set_info_async)(const struct lu_env *, struct obd_export *,
1112 __u32 keylen, void *key,
1113 __u32 vallen, void *val,
1114 struct ptlrpc_request_set *set);
21aef7d9 1115 int (*o_attach)(struct obd_device *dev, u32 len, void *data);
d7e09d03 1116 int (*o_detach)(struct obd_device *dev);
b037976f 1117 int (*o_setup)(struct obd_device *dev, struct lustre_cfg *cfg);
d7e09d03
PT		 1118 int (*o_precleanup)(struct obd_device *dev,
1119 enum obd_cleanup_stage cleanup_stage);
1120 int (*o_cleanup)(struct obd_device *dev);
21aef7d9 1121 int (*o_process_config)(struct obd_device *dev, u32 len, void *data);
d7e09d03
PT		 1122 int (*o_postrecov)(struct obd_device *dev);
1123 int (*o_add_conn)(struct obd_import *imp, struct obd_uuid *uuid,
1124 int priority);
1125 int (*o_del_conn)(struct obd_import *imp, struct obd_uuid *uuid);
1126 /* connect to the target device with given connection
1127 * data. @ocd->ocd_connect_flags is modified to reflect flags actually
1128 * granted by the target, which are guaranteed to be a subset of flags
1129 * asked for. If @ocd == NULL, use default parameters. */
1130 int (*o_connect)(const struct lu_env *env,
1131 struct obd_export **exp, struct obd_device *src,
1132 struct obd_uuid *cluuid, struct obd_connect_data *ocd,
1133 void *localdata);
1134 int (*o_reconnect)(const struct lu_env *env,
1135 struct obd_export *exp, struct obd_device *src,
1136 struct obd_uuid *cluuid,
1137 struct obd_connect_data *ocd,
1138 void *localdata);
1139 int (*o_disconnect)(struct obd_export *exp);
1140
1141 /* Initialize/finalize fids infrastructure. */
1142 int (*o_fid_init)(struct obd_device *obd,
1143 struct obd_export *exp, enum lu_cli_type type);
1144 int (*o_fid_fini)(struct obd_device *obd);
1145
1146 /* Allocate new fid according to passed @hint. */
1147 int (*o_fid_alloc)(struct obd_export *exp, struct lu_fid *fid,
1148 struct md_op_data *op_data);
1149
1150 /*
1151 * Object with @fid is getting deleted, we may want to do something
1152 * about this.
1153 */
1154 int (*o_statfs)(const struct lu_env *, struct obd_export *exp,
1155 struct obd_statfs *osfs, __u64 max_age, __u32 flags);
1156 int (*o_statfs_async)(struct obd_export *exp, struct obd_info *oinfo,
1157 __u64 max_age, struct ptlrpc_request_set *set);
1158 int (*o_packmd)(struct obd_export *exp, struct lov_mds_md **disk_tgt,
1159 struct lov_stripe_md *mem_src);
1d8cb70c
GD		 1160 int (*o_unpackmd)(struct obd_export *exp,
1161 struct lov_stripe_md **mem_tgt,
d7e09d03 1162 struct lov_mds_md *disk_src, int disk_len);
21aef7d9 1163 int (*o_preallocate)(struct lustre_handle *, u32 *req, u64 *ids);
d7e09d03 1164 /* FIXME: add fid capability support for create & destroy! */
d7e09d03
PT		 1165 int (*o_create)(const struct lu_env *env, struct obd_export *exp,
1166 struct obdo *oa, struct lov_stripe_md **ea,
1167 struct obd_trans_info *oti);
d7e09d03
PT		 1168 int (*o_destroy)(const struct lu_env *env, struct obd_export *exp,
1169 struct obdo *oa, struct lov_stripe_md *ea,
1170 struct obd_trans_info *oti, struct obd_export *md_exp,
1171 void *capa);
1172 int (*o_setattr)(const struct lu_env *, struct obd_export *exp,
1173 struct obd_info *oinfo, struct obd_trans_info *oti);
1174 int (*o_setattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1175 struct obd_trans_info *oti,
1176 struct ptlrpc_request_set *rqset);
1177 int (*o_getattr)(const struct lu_env *env, struct obd_export *exp,
1178 struct obd_info *oinfo);
1179 int (*o_getattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1180 struct ptlrpc_request_set *set);
d7e09d03 1181 int (*o_adjust_kms)(struct obd_export *exp, struct lov_stripe_md *lsm,
21aef7d9 1182 u64 size, int shrink);
d7e09d03
PT		 1183 int (*o_preprw)(const struct lu_env *env, int cmd,
1184 struct obd_export *exp, struct obdo *oa, int objcount,
1185 struct obd_ioobj *obj, struct niobuf_remote *remote,
1186 int *nr_pages, struct niobuf_local *local,
1187 struct obd_trans_info *oti, struct lustre_capa *capa);
1188 int (*o_commitrw)(const struct lu_env *env, int cmd,
1189 struct obd_export *exp, struct obdo *oa,
1190 int objcount, struct obd_ioobj *obj,
1191 struct niobuf_remote *remote, int pages,
1192 struct niobuf_local *local,
1193 struct obd_trans_info *oti, int rc);
d7e09d03
PT		 1194 int (*o_find_cbdata)(struct obd_export *, struct lov_stripe_md *,
1195 ldlm_iterator_t it, void *data);
d7e09d03
PT		 1196 int (*o_init_export)(struct obd_export *exp);
1197 int (*o_destroy_export)(struct obd_export *exp);
d7e09d03
PT		 1198
1199 /* metadata-only methods */
d7e09d03
PT		 1200 int (*o_import_event)(struct obd_device *, struct obd_import *,
1201 enum obd_import_event);
1202
1203 int (*o_notify)(struct obd_device *obd, struct obd_device *watched,
1204 enum obd_notify_event ev, void *data);
1205
1206 int (*o_health_check)(const struct lu_env *env, struct obd_device *);
b037976f 1207 struct obd_uuid *(*o_get_uuid)(struct obd_export *exp);
d7e09d03
PT		 1208
1209 /* quota methods */
1210 int (*o_quotacheck)(struct obd_device *, struct obd_export *,
1211 struct obd_quotactl *);
1212 int (*o_quotactl)(struct obd_device *, struct obd_export *,
1213 struct obd_quotactl *);
1214
d7e09d03
PT		 1215 /* pools methods */
1216 int (*o_pool_new)(struct obd_device *obd, char *poolname);
1217 int (*o_pool_del)(struct obd_device *obd, char *poolname);
1218 int (*o_pool_add)(struct obd_device *obd, char *poolname,
1219 char *ostname);
1220 int (*o_pool_rem)(struct obd_device *obd, char *poolname,
1221 char *ostname);
1222 void (*o_getref)(struct obd_device *obd);
1223 void (*o_putref)(struct obd_device *obd);
1224 /*
1225 * NOTE: If adding ops, add another LPROCFS_OBD_OP_INIT() line
1226 * to lprocfs_alloc_obd_stats() in obdclass/lprocfs_status.c.
1227 * Also, add a wrapper function in include/linux/obd_class.h. */
1228};
1229
1230enum {
1231 LUSTRE_OPC_MKDIR = (1 << 0),
1232 LUSTRE_OPC_SYMLINK = (1 << 1),
1233 LUSTRE_OPC_MKNOD = (1 << 2),
1234 LUSTRE_OPC_CREATE = (1 << 3),
1235 LUSTRE_OPC_ANY = (1 << 4)
1236};
1237
1238/* lmv structures */
1239#define MEA_MAGIC_LAST_CHAR 0xb2221ca1
1240#define MEA_MAGIC_ALL_CHARS 0xb222a11c
1241#define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
1242
1243#define MAX_HASH_SIZE_32 0x7fffffffUL
1244#define MAX_HASH_SIZE 0x7fffffffffffffffULL
1245#define MAX_HASH_HIGHEST_BIT 0x1000000000000000ULL
1246
1247struct lustre_md {
1248 struct mdt_body *body;
1249 struct lov_stripe_md *lsm;
1250 struct lmv_stripe_md *mea;
1251#ifdef CONFIG_FS_POSIX_ACL
1252 struct posix_acl *posix_acl;
1253#endif
1254 struct mdt_remote_perm *remote_perm;
1255 struct obd_capa *mds_capa;
1256 struct obd_capa *oss_capa;
1257};
1258
1259struct md_open_data {
1260 struct obd_client_handle *mod_och;
1261 struct ptlrpc_request *mod_open_req;
1262 struct ptlrpc_request *mod_close_req;
63d42578
HZ		 1263 atomic_t mod_refcount;
1264 bool mod_is_create;
d7e09d03
PT		 1265};
1266
1267struct lookup_intent;
1268
1269struct md_ops {
1270 int (*m_getstatus)(struct obd_export *, struct lu_fid *,
1271 struct obd_capa **);
1272 int (*m_null_inode)(struct obd_export *, const struct lu_fid *);
1273 int (*m_find_cbdata)(struct obd_export *, const struct lu_fid *,
1274 ldlm_iterator_t, void *);
1275 int (*m_close)(struct obd_export *, struct md_op_data *,
1276 struct md_open_data *, struct ptlrpc_request **);
1277 int (*m_create)(struct obd_export *, struct md_op_data *,
1278 const void *, int, int, __u32, __u32, cfs_cap_t,
1279 __u64, struct ptlrpc_request **);
1280 int (*m_done_writing)(struct obd_export *, struct md_op_data *,
1281 struct md_open_data *);
1282 int (*m_enqueue)(struct obd_export *, struct ldlm_enqueue_info *,
1283 struct lookup_intent *, struct md_op_data *,
1284 struct lustre_handle *, void *, int,
1285 struct ptlrpc_request **, __u64);
1286 int (*m_getattr)(struct obd_export *, struct md_op_data *,
1287 struct ptlrpc_request **);
1288 int (*m_getattr_name)(struct obd_export *, struct md_op_data *,
1289 struct ptlrpc_request **);
1290 int (*m_intent_lock)(struct obd_export *, struct md_op_data *,
1291 void *, int, struct lookup_intent *, int,
1292 struct ptlrpc_request **,
1293 ldlm_blocking_callback, __u64);
1294 int (*m_link)(struct obd_export *, struct md_op_data *,
1295 struct ptlrpc_request **);
1296 int (*m_rename)(struct obd_export *, struct md_op_data *,
1297 const char *, int, const char *, int,
1298 struct ptlrpc_request **);
1299 int (*m_is_subdir)(struct obd_export *, const struct lu_fid *,
1300 const struct lu_fid *,
1301 struct ptlrpc_request **);
1302 int (*m_setattr)(struct obd_export *, struct md_op_data *, void *,
1303 int , void *, int, struct ptlrpc_request **,
1304 struct md_open_data **mod);
1305 int (*m_sync)(struct obd_export *, const struct lu_fid *,
1306 struct obd_capa *, struct ptlrpc_request **);
1307 int (*m_readpage)(struct obd_export *, struct md_op_data *,
1308 struct page **, struct ptlrpc_request **);
1309
1310 int (*m_unlink)(struct obd_export *, struct md_op_data *,
1311 struct ptlrpc_request **);
1312
1313 int (*m_setxattr)(struct obd_export *, const struct lu_fid *,
21aef7d9 1314 struct obd_capa *, u64, const char *,
d7e09d03
PT		 1315 const char *, int, int, int, __u32,
1316 struct ptlrpc_request **);
1317
1318 int (*m_getxattr)(struct obd_export *, const struct lu_fid *,
21aef7d9 1319 struct obd_capa *, u64, const char *,
d7e09d03
PT		 1320 const char *, int, int, int,
1321 struct ptlrpc_request **);
1322
44779340 1323 int (*m_init_ea_size)(struct obd_export *, int, int, int, int);
d7e09d03
PT		 1324
1325 int (*m_get_lustre_md)(struct obd_export *, struct ptlrpc_request *,
1326 struct obd_export *, struct obd_export *,
1327 struct lustre_md *);
1328
1329 int (*m_free_lustre_md)(struct obd_export *, struct lustre_md *);
1330
1331 int (*m_set_open_replay_data)(struct obd_export *,
1332 struct obd_client_handle *,
63d42578 1333 struct lookup_intent *);
d7e09d03
PT		 1334 int (*m_clear_open_replay_data)(struct obd_export *,
1335 struct obd_client_handle *);
1336 int (*m_set_lock_data)(struct obd_export *, __u64 *, void *, __u64 *);
1337
1338 ldlm_mode_t (*m_lock_match)(struct obd_export *, __u64,
1339 const struct lu_fid *, ldlm_type_t,
1340 ldlm_policy_data_t *, ldlm_mode_t,
1341 struct lustre_handle *);
1342
1343 int (*m_cancel_unused)(struct obd_export *, const struct lu_fid *,
1344 ldlm_policy_data_t *, ldlm_mode_t,
1345 ldlm_cancel_flags_t flags, void *opaque);
1346 int (*m_renew_capa)(struct obd_export *, struct obd_capa *oc,
1347 renew_capa_cb_t cb);
1348 int (*m_unpack_capa)(struct obd_export *, struct ptlrpc_request *,
1349 const struct req_msg_field *, struct obd_capa **);
1350
1351 int (*m_get_remote_perm)(struct obd_export *, const struct lu_fid *,
1352 struct obd_capa *, __u32,
1353 struct ptlrpc_request **);
1354
1355 int (*m_intent_getattr_async)(struct obd_export *,
1356 struct md_enqueue_info *,
1357 struct ldlm_enqueue_info *);
1358
1359 int (*m_revalidate_lock)(struct obd_export *, struct lookup_intent *,
1360 struct lu_fid *, __u64 *bits);
1361
1362 /*
1363 * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to
1364 * lprocfs_alloc_md_stats() in obdclass/lprocfs_status.c. Also, add a
1365 * wrapper function in include/linux/obd_class.h.
1366 */
1367};
1368
1369struct lsm_operations {
1370 void (*lsm_free)(struct lov_stripe_md *);
1371 int (*lsm_destroy)(struct lov_stripe_md *, struct obdo *oa,
1372 struct obd_export *md_exp);
21aef7d9
OD		 1373 void (*lsm_stripe_by_index)(struct lov_stripe_md *, int *, u64 *,
1374 u64 *);
1375 void (*lsm_stripe_by_offset)(struct lov_stripe_md *, int *, u64 *,
1376 u64 *);
b037976f 1377 int (*lsm_lmm_verify)(struct lov_mds_md *lmm, int lmm_bytes,
d7e09d03 1378 __u16 *stripe_count);
b037976f 1379 int (*lsm_unpackmd)(struct lov_obd *lov, struct lov_stripe_md *lsm,
d7e09d03
PT		 1380 struct lov_mds_md *lmm);
1381};
1382
1383extern const struct lsm_operations lsm_v1_ops;
1384extern const struct lsm_operations lsm_v3_ops;
1385static inline const struct lsm_operations *lsm_op_find(int magic)
1386{
a58a38ac 1387 switch (magic) {
d7e09d03
PT		 1388 case LOV_MAGIC_V1:
1389 return &lsm_v1_ops;
1390 case LOV_MAGIC_V3:
1391 return &lsm_v3_ops;
1392 default:
1393 CERROR("Cannot recognize lsm_magic %08x\n", magic);
1394 return NULL;
1395 }
1396}
1397
1398/* Requests for obd_extent_calc() */
1399#define OBD_CALC_STRIPE_START 1
1400#define OBD_CALC_STRIPE_END 2
1401
1402static inline struct lustre_capa *oinfo_capa(struct obd_info *oinfo)
1403{
1404 return oinfo->oi_capa;
1405}
1406
1407static inline struct md_open_data *obd_mod_alloc(void)
1408{
1409 struct md_open_data *mod;
498374f7 1410
d7e09d03
PT		 1411 OBD_ALLOC_PTR(mod);
1412 if (mod == NULL)
1413 return NULL;
1414 atomic_set(&mod->mod_refcount, 1);
1415 return mod;
1416}
1417
1418#define obd_mod_get(mod) atomic_inc(&(mod)->mod_refcount)
1419#define obd_mod_put(mod) \
1420({ \
1421 if (atomic_dec_and_test(&(mod)->mod_refcount)) { \
1422 if ((mod)->mod_open_req) \
1423 ptlrpc_req_finished((mod)->mod_open_req); \
1424 OBD_FREE_PTR(mod); \
1425 } \
1426})
1427
21aef7d9 1428void obdo_from_inode(struct obdo *dst, struct inode *src, u32 valid);
d7e09d03
PT		 1429void obdo_set_parent_fid(struct obdo *dst, const struct lu_fid *parent);
1430
1431/* return 1 if client should be resend request */
1432static inline int client_should_resend(int resend, struct client_obd *cli)
1433{
1434 return atomic_read(&cli->cl_resends) ?
1435 atomic_read(&cli->cl_resends) > resend : 1;
1436}
1437
1438/**
1439 * Return device name for this device
1440 *
1441 * XXX: lu_device is declared before obd_device, while a pointer pointing
1442 * back to obd_device in lu_device, so this helper function defines here
1443 * instead of in lu_object.h
1444 */
1445static inline const char *lu_dev_name(const struct lu_device *lu_dev)
1446{
1447 return lu_dev->ld_obd->obd_name;
1448}
1449
1450static inline bool filename_is_volatile(const char *name, int namelen, int *idx)
1451{
1452 const char *start;
1453 char *end;
1454
1455 if (strncmp(name, LUSTRE_VOLATILE_HDR, LUSTRE_VOLATILE_HDR_LEN) != 0)
1456 return false;
1457
1458 /* caller does not care of idx */
1459 if (idx == NULL)
1460 return true;
1461
1462 /* volatile file, the MDT can be set from name */
1463 /* name format is LUSTRE_VOLATILE_HDR:[idx]: */
1464 /* if no MDT is specified, use std way */
1465 if (namelen < LUSTRE_VOLATILE_HDR_LEN + 2)
1466 goto bad_format;
1467 /* test for no MDT idx case */
1468 if ((*(name + LUSTRE_VOLATILE_HDR_LEN) == ':') &&
1469 (*(name + LUSTRE_VOLATILE_HDR_LEN + 1) == ':')) {
1470 *idx = -1;
1471 return true;
1472 }
1473 /* we have an idx, read it */
1474 start = name + LUSTRE_VOLATILE_HDR_LEN + 1;
4ad2adc9 1475 *idx = simple_strtoul(start, &end, 0);
d7e09d03
PT		 1476 /* error cases:
1477 * no digit, no trailing :, negative value
1478 */
1479 if (((*idx == 0) && (end == start)) ||
1480 (*end != ':') || (*idx < 0))
1481 goto bad_format;
1482
1483 return true;
1484bad_format:
1485 /* bad format of mdt idx, we cannot return an error
1486 * to caller so we use hash algo */
1487 CERROR("Bad volatile file name format: %s\n",
1488 name + LUSTRE_VOLATILE_HDR_LEN);
1489 return false;
1490}
1491
1492static inline int cli_brw_size(struct obd_device *obd)
1493{
1494 LASSERT(obd != NULL);
1495 return obd->u.cli.cl_max_pages_per_rpc << PAGE_CACHE_SHIFT;
1496}
1497
1498#endif /* __OBD_H */
Linux 6.x block layer and io_uring tree(s)