]>
Commit | Line | Data |
---|---|---|
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 | * | |
1dc563a6 | 30 | * Copyright (c) 2011, 2015, Intel Corporation. |
d7e09d03 PT |
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" | |
d7e09d03 | 57 | |
d7e09d03 PT |
58 | #define MAX_OBD_DEVICES 8192 |
59 | ||
60 | struct osc_async_rc { | |
61 | int ar_rc; | |
62 | int ar_force_sync; | |
63 | __u64 ar_min_xid; | |
64 | }; | |
65 | ||
66 | struct lov_oinfo { /* per-stripe data structure */ | |
67 | struct ost_id loi_oi; /* object ID/Sequence on the target OST */ | |
68 | int loi_ost_idx; /* OST stripe index in lov_tgt_desc->tgts */ | |
69 | int loi_ost_gen; /* generation of this loi_ost_idx */ | |
70 | ||
71 | unsigned long loi_kms_valid:1; | |
72 | __u64 loi_kms; /* known minimum size */ | |
73 | struct ost_lvb loi_lvb; | |
74 | struct osc_async_rc loi_ar; | |
75 | }; | |
76 | ||
77 | static inline void loi_kms_set(struct lov_oinfo *oinfo, __u64 kms) | |
78 | { | |
79 | oinfo->loi_kms = kms; | |
80 | oinfo->loi_kms_valid = 1; | |
81 | } | |
82 | ||
83 | static inline void loi_init(struct lov_oinfo *loi) | |
84 | { | |
85 | } | |
86 | ||
87 | struct lov_stripe_md { | |
88 | atomic_t lsm_refc; | |
89 | spinlock_t lsm_lock; | |
90 | pid_t lsm_lock_owner; /* debugging */ | |
91 | ||
92 | /* maximum possible file size, might change as OSTs status changes, | |
c56e256d OD |
93 | * e.g. disconnected, deactivated |
94 | */ | |
d7e09d03 PT |
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 |
120 | static inline bool lsm_is_released(struct lov_stripe_md *lsm) |
121 | { | |
122 | return !!(lsm->lsm_pattern & LOV_PATTERN_F_RELEASED); | |
123 | } | |
124 | ||
125 | static inline bool lsm_has_objects(struct lov_stripe_md *lsm) | |
126 | { | |
d2a13989 | 127 | if (!lsm) |
5dd16419 JX |
128 | return false; |
129 | if (lsm_is_released(lsm)) | |
130 | return false; | |
131 | return true; | |
132 | } | |
133 | ||
081b7265 JH |
134 | static 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 |
141 | struct obd_info; |
142 | ||
143 | typedef int (*obd_enqueue_update_f)(void *cookie, int rc); | |
144 | ||
145 | /* obd info for a particular level (lov, osc). */ | |
146 | struct obd_info { | |
d7e09d03 PT |
147 | /* Flags used for set request specific flags: |
148 | - while lock handling, the flags obtained on the enqueue | |
149 | request are set here. | |
150 | - while stats, the flags used for control delay/resend. | |
151 | - while setattr, the flags used for distinguish punch operation | |
152 | */ | |
153 | __u64 oi_flags; | |
d7e09d03 PT |
154 | /* lsm data specific for every OSC. */ |
155 | struct lov_stripe_md *oi_md; | |
156 | /* obdo data specific for every OSC, if needed at all. */ | |
157 | struct obdo *oi_oa; | |
158 | /* statfs data specific for every OSC, if needed at all. */ | |
159 | struct obd_statfs *oi_osfs; | |
160 | /* An update callback which is called to update some data on upper | |
17891183 | 161 | * level. E.g. it is used for update lsm->lsm_oinfo at every received |
d7e09d03 | 162 | * request in osc level for enqueue requests. It is also possible to |
c56e256d OD |
163 | * update some caller data from LOV layer if needed. |
164 | */ | |
d7e09d03 | 165 | obd_enqueue_update_f oi_cb_up; |
d7e09d03 PT |
166 | }; |
167 | ||
d7e09d03 PT |
168 | void lov_stripe_lock(struct lov_stripe_md *md); |
169 | void lov_stripe_unlock(struct lov_stripe_md *md); | |
170 | ||
171 | struct obd_type { | |
172 | struct list_head typ_chain; | |
173 | struct obd_ops *typ_dt_ops; | |
174 | struct md_ops *typ_md_ops; | |
61e87ab0 | 175 | struct dentry *typ_debugfs_entry; |
d7e09d03 PT |
176 | char *typ_name; |
177 | int typ_refcnt; | |
178 | struct lu_device_type *typ_lu; | |
179 | spinlock_t obd_type_lock; | |
9b801302 | 180 | struct kobject *typ_kobj; |
d7e09d03 PT |
181 | }; |
182 | ||
183 | struct brw_page { | |
21aef7d9 | 184 | u64 off; |
d7e09d03 PT |
185 | struct page *pg; |
186 | int count; | |
21aef7d9 | 187 | u32 flag; |
d7e09d03 PT |
188 | }; |
189 | ||
d7e09d03 PT |
190 | /* llog contexts */ |
191 | enum llog_ctxt_id { | |
192 | LLOG_CONFIG_ORIG_CTXT = 0, | |
193 | LLOG_CONFIG_REPL_CTXT, | |
194 | LLOG_MDS_OST_ORIG_CTXT, | |
195 | LLOG_MDS_OST_REPL_CTXT, | |
196 | LLOG_SIZE_ORIG_CTXT, | |
197 | LLOG_SIZE_REPL_CTXT, | |
198 | LLOG_RD1_ORIG_CTXT, | |
199 | LLOG_RD1_REPL_CTXT, | |
200 | LLOG_TEST_ORIG_CTXT, | |
201 | LLOG_TEST_REPL_CTXT, | |
202 | LLOG_LOVEA_ORIG_CTXT, | |
203 | LLOG_LOVEA_REPL_CTXT, | |
99a92265 | 204 | LLOG_CHANGELOG_ORIG_CTXT, /**< changelog generation on mdd */ |
205 | LLOG_CHANGELOG_REPL_CTXT, /**< changelog access on clients */ | |
206 | LLOG_CHANGELOG_USER_ORIG_CTXT, /**< for multiple changelog consumers */ | |
207 | LLOG_AGENT_ORIG_CTXT, /**< agent requests generation on cdt */ | |
d7e09d03 PT |
208 | LLOG_MAX_CTXTS |
209 | }; | |
210 | ||
d7e09d03 PT |
211 | struct timeout_item { |
212 | enum timeout_event ti_event; | |
a649ad1d | 213 | unsigned long ti_timeout; |
d7e09d03 PT |
214 | timeout_cb_t ti_cb; |
215 | void *ti_cb_data; | |
216 | struct list_head ti_obd_list; | |
217 | struct list_head ti_chain; | |
218 | }; | |
219 | ||
220 | #define OSC_MAX_RIF_DEFAULT 8 | |
d7e09d03 PT |
221 | #define OSC_MAX_RIF_MAX 256 |
222 | #define OSC_MAX_DIRTY_DEFAULT (OSC_MAX_RIF_DEFAULT * 4) | |
223 | #define OSC_MAX_DIRTY_MB_MAX 2048 /* arbitrary, but < MAX_LONG bytes */ | |
224 | #define OSC_DEFAULT_RESENDS 10 | |
225 | ||
226 | /* possible values for fo_sync_lock_cancel */ | |
227 | enum { | |
228 | NEVER_SYNC_ON_CANCEL = 0, | |
229 | BLOCKING_SYNC_ON_CANCEL = 1, | |
230 | ALWAYS_SYNC_ON_CANCEL = 2, | |
231 | NUM_SYNC_ON_CANCEL_STATES | |
232 | }; | |
233 | ||
234 | #define MDC_MAX_RIF_DEFAULT 8 | |
235 | #define MDC_MAX_RIF_MAX 512 | |
236 | ||
237 | struct mdc_rpc_lock; | |
238 | struct obd_import; | |
239 | struct client_obd { | |
240 | struct rw_semaphore cl_sem; | |
241 | struct obd_uuid cl_target_uuid; | |
242 | struct obd_import *cl_import; /* ptlrpc connection state */ | |
243 | int cl_conn_count; | |
244 | /* max_mds_easize is purely a performance thing so we don't have to | |
c56e256d OD |
245 | * call obd_size_diskmd() all the time. |
246 | */ | |
44779340 BB |
247 | int cl_default_mds_easize; |
248 | int cl_max_mds_easize; | |
249 | int cl_default_mds_cookiesize; | |
250 | int cl_max_mds_cookiesize; | |
d7e09d03 PT |
251 | |
252 | enum lustre_sec_part cl_sp_me; | |
253 | enum lustre_sec_part cl_sp_to; | |
254 | struct sptlrpc_flavor cl_flvr_mgc; /* fixed flavor of mgc->mgs */ | |
255 | ||
256 | /* the grant values are protected by loi_list_lock below */ | |
257 | long cl_dirty; /* all _dirty_ in bytes */ | |
258 | long cl_dirty_max; /* allowed w/o rpc */ | |
259 | long cl_dirty_transit; /* dirty synchronous */ | |
260 | long cl_avail_grant; /* bytes of credit for ost */ | |
261 | long cl_lost_grant; /* lost credits (trunc) */ | |
262 | ||
263 | /* since we allocate grant by blocks, we don't know how many grant will | |
264 | * be used to add a page into cache. As a solution, we reserve maximum | |
265 | * grant before trying to dirty a page and unreserve the rest. | |
c56e256d OD |
266 | * See osc_{reserve|unreserve}_grant for details. |
267 | */ | |
d7e09d03 | 268 | long cl_reserved_grant; |
498374f7 HM |
269 | struct list_head cl_cache_waiters; /* waiting for cache/grant */ |
270 | unsigned long cl_next_shrink_grant; /* jiffies */ | |
271 | struct list_head cl_grant_shrink_list; /* Timeout event list */ | |
272 | int cl_grant_shrink_interval; /* seconds */ | |
d7e09d03 PT |
273 | |
274 | /* A chunk is an optimal size used by osc_extent to determine | |
c56e256d OD |
275 | * the extent size. A chunk is max(PAGE_CACHE_SIZE, OST block size) |
276 | */ | |
d7e09d03 PT |
277 | int cl_chunkbits; |
278 | int cl_chunk; | |
279 | int cl_extent_tax; /* extent overhead, by bytes */ | |
280 | ||
281 | /* keep track of objects that have lois that contain pages which | |
282 | * have been queued for async brw. this lock also protects the | |
c56e256d OD |
283 | * lists of osc_client_pages that hang off of the loi |
284 | */ | |
d7e09d03 PT |
285 | /* |
286 | * ->cl_loi_list_lock protects consistency of | |
287 | * ->cl_loi_{ready,read,write}_list. ->ap_make_ready() and | |
288 | * ->ap_completion() call-backs are executed under this lock. As we | |
289 | * cannot guarantee that these call-backs never block on all platforms | |
290 | * (as a matter of fact they do block on Mac OS X), type of | |
291 | * ->cl_loi_list_lock is platform dependent: it's a spin-lock on Linux | |
292 | * and blocking mutex on Mac OS X. (Alternative is to make this lock | |
293 | * blocking everywhere, but we don't want to slow down fast-path of | |
294 | * our main platform.) | |
295 | * | |
296 | * Exact type of ->cl_loi_list_lock is defined in arch/obd.h together | |
297 | * with client_obd_list_{un,}lock() and | |
298 | * client_obd_list_lock_{init,done}() functions. | |
299 | * | |
300 | * NB by Jinshan: though field names are still _loi_, but actually | |
301 | * osc_object{}s are in the list. | |
302 | */ | |
fef74397 | 303 | struct client_obd_lock cl_loi_list_lock; |
d7e09d03 PT |
304 | struct list_head cl_loi_ready_list; |
305 | struct list_head cl_loi_hp_ready_list; | |
306 | struct list_head cl_loi_write_list; | |
307 | struct list_head cl_loi_read_list; | |
308 | int cl_r_in_flight; | |
309 | int cl_w_in_flight; | |
406c1c7c | 310 | /* just a sum of the loi/lop pending numbers to be exported by sysfs */ |
d7e09d03 PT |
311 | atomic_t cl_pending_w_pages; |
312 | atomic_t cl_pending_r_pages; | |
313 | __u32 cl_max_pages_per_rpc; | |
314 | int cl_max_rpcs_in_flight; | |
315 | struct obd_histogram cl_read_rpc_hist; | |
316 | struct obd_histogram cl_write_rpc_hist; | |
317 | struct obd_histogram cl_read_page_hist; | |
318 | struct obd_histogram cl_write_page_hist; | |
319 | struct obd_histogram cl_read_offset_hist; | |
320 | struct obd_histogram cl_write_offset_hist; | |
321 | ||
322 | /* lru for osc caching pages */ | |
323 | struct cl_client_cache *cl_cache; | |
498374f7 | 324 | struct list_head cl_lru_osc; /* member of cl_cache->ccc_lru */ |
d7e09d03 PT |
325 | atomic_t *cl_lru_left; |
326 | atomic_t cl_lru_busy; | |
327 | atomic_t cl_lru_shrinkers; | |
328 | atomic_t cl_lru_in_list; | |
498374f7 | 329 | struct list_head cl_lru_list; /* lru page list */ |
fef74397 | 330 | struct client_obd_lock cl_lru_list_lock; /* page list protector */ |
d7e09d03 PT |
331 | |
332 | /* number of in flight destroy rpcs is limited to max_rpcs_in_flight */ | |
333 | atomic_t cl_destroy_in_flight; | |
334 | wait_queue_head_t cl_destroy_waitq; | |
335 | ||
336 | struct mdc_rpc_lock *cl_rpc_lock; | |
337 | struct mdc_rpc_lock *cl_close_lock; | |
338 | ||
339 | /* mgc datastruct */ | |
d7e09d03 PT |
340 | atomic_t cl_mgc_refcount; |
341 | struct obd_export *cl_mgc_mgsexp; | |
342 | ||
343 | /* checksumming for data sent over the network */ | |
344 | unsigned int cl_checksum:1; /* 0 = disabled, 1 = enabled */ | |
345 | /* supported checksum types that are worked out at connect time */ | |
346 | __u32 cl_supp_cksum_types; | |
347 | /* checksum algorithm to be used */ | |
d133210f | 348 | enum cksum_type cl_cksum_type; |
d7e09d03 PT |
349 | |
350 | /* also protected by the poorly named _loi_list_lock lock above */ | |
351 | struct osc_async_rc cl_ar; | |
352 | ||
353 | /* used by quotacheck when the servers are older than 2.4 */ | |
354 | int cl_qchk_stat; /* quotacheck stat of the peer */ | |
355 | #define CL_NOT_QUOTACHECKED 1 /* client->cl_qchk_stat init value */ | |
356 | #if LUSTRE_VERSION_CODE >= OBD_OCD_VERSION(2, 7, 50, 0) | |
357 | #warning "please consider removing quotacheck compatibility code" | |
358 | #endif | |
359 | ||
360 | /* sequence manager */ | |
361 | struct lu_client_seq *cl_seq; | |
362 | ||
363 | atomic_t cl_resends; /* resend count */ | |
364 | ||
365 | /* ptlrpc work for writeback in ptlrpcd context */ | |
366 | void *cl_writeback_work; | |
367 | /* hash tables for osc_quota_info */ | |
6da6eabe | 368 | struct cfs_hash *cl_quota_hash[MAXQUOTAS]; |
d7e09d03 | 369 | }; |
c9f6bb96 | 370 | |
d7e09d03 PT |
371 | #define obd2cli_tgt(obd) ((char *)(obd)->u.cli.cl_target_uuid.uuid) |
372 | ||
373 | struct obd_id_info { | |
374 | __u32 idx; | |
21aef7d9 | 375 | u64 *data; |
d7e09d03 PT |
376 | }; |
377 | ||
d7e09d03 PT |
378 | struct echo_client_obd { |
379 | struct obd_export *ec_exp; /* the local connection to osc/lov */ | |
380 | spinlock_t ec_lock; | |
381 | struct list_head ec_objects; | |
382 | struct list_head ec_locks; | |
d7e09d03 PT |
383 | __u64 ec_unique; |
384 | }; | |
385 | ||
d7e09d03 PT |
386 | /* Generic subset of OSTs */ |
387 | struct ost_pool { | |
c56e256d | 388 | __u32 *op_array; /* array of index of lov_obd->lov_tgts */ |
d7e09d03 PT |
389 | unsigned int op_count; /* number of OSTs in the array */ |
390 | unsigned int op_size; /* allocated size of lp_array */ | |
391 | struct rw_semaphore op_rw_sem; /* to protect ost_pool use */ | |
392 | }; | |
393 | ||
394 | /* Round-robin allocator data */ | |
395 | struct lov_qos_rr { | |
396 | __u32 lqr_start_idx; /* start index of new inode */ | |
397 | __u32 lqr_offset_idx; /* aliasing for start_idx */ | |
398 | int lqr_start_count; /* reseed counter */ | |
399 | struct ost_pool lqr_pool; /* round-robin optimized list */ | |
400 | unsigned long lqr_dirty:1; /* recalc round-robin list */ | |
401 | }; | |
402 | ||
403 | /* allow statfs data caching for 1 second */ | |
404 | #define OBD_STATFS_CACHE_SECONDS 1 | |
405 | ||
406 | struct lov_statfs_data { | |
407 | struct obd_info lsd_oi; | |
408 | struct obd_statfs lsd_statfs; | |
409 | }; | |
c9f6bb96 | 410 | |
d7e09d03 PT |
411 | /* Stripe placement optimization */ |
412 | struct lov_qos { | |
498374f7 | 413 | struct list_head lq_oss_list; /* list of OSSs that targets use */ |
d7e09d03 | 414 | struct rw_semaphore lq_rw_sem; |
498374f7 | 415 | __u32 lq_active_oss_count; |
d7e09d03 PT |
416 | unsigned int lq_prio_free; /* priority for free space */ |
417 | unsigned int lq_threshold_rr;/* priority for rr */ | |
418 | struct lov_qos_rr lq_rr; /* round robin qos data */ | |
419 | unsigned long lq_dirty:1, /* recalc qos data */ | |
420 | lq_same_space:1,/* the ost's all have approx. | |
c56e256d OD |
421 | * the same space avail |
422 | */ | |
d7e09d03 PT |
423 | lq_reset:1, /* zero current penalties */ |
424 | lq_statfs_in_progress:1; /* statfs op in | |
425 | progress */ | |
426 | /* qos statfs data */ | |
427 | struct lov_statfs_data *lq_statfs_data; | |
c56e256d OD |
428 | wait_queue_head_t lq_statfs_waitq; /* waitqueue to notify statfs |
429 | * requests completion | |
430 | */ | |
d7e09d03 PT |
431 | }; |
432 | ||
433 | struct lov_tgt_desc { | |
434 | struct list_head ltd_kill; | |
435 | struct obd_uuid ltd_uuid; | |
436 | struct obd_device *ltd_obd; | |
437 | struct obd_export *ltd_exp; | |
d7e09d03 PT |
438 | __u32 ltd_gen; |
439 | __u32 ltd_index; /* index in lov_obd->tgts */ | |
440 | unsigned long ltd_active:1,/* is this target up for requests */ | |
441 | ltd_activate:1,/* should target be activated */ | |
442 | ltd_reap:1; /* should this target be deleted */ | |
443 | }; | |
444 | ||
445 | /* Pool metadata */ | |
446 | #define pool_tgt_size(_p) _p->pool_obds.op_size | |
447 | #define pool_tgt_count(_p) _p->pool_obds.op_count | |
448 | #define pool_tgt_array(_p) _p->pool_obds.op_array | |
449 | #define pool_tgt_rw_sem(_p) _p->pool_obds.op_rw_sem | |
450 | ||
451 | struct pool_desc { | |
452 | char pool_name[LOV_MAXPOOLNAME + 1]; /* name of pool */ | |
453 | struct ost_pool pool_obds; /* pool members */ | |
454 | atomic_t pool_refcount; /* pool ref. counter */ | |
455 | struct lov_qos_rr pool_rr; /* round robin qos */ | |
456 | struct hlist_node pool_hash; /* access by poolname */ | |
457 | struct list_head pool_list; /* serial access */ | |
406c1c7c | 458 | struct dentry *pool_debugfs_entry; /* file in debugfs */ |
498374f7 | 459 | struct obd_device *pool_lobd; /* obd of the lov/lod to which |
c56e256d OD |
460 | * this pool belongs |
461 | */ | |
d7e09d03 PT |
462 | }; |
463 | ||
464 | struct lov_obd { | |
465 | struct lov_desc desc; | |
466 | struct lov_tgt_desc **lov_tgts; /* sparse array */ | |
c56e256d | 467 | struct ost_pool lov_packed; /* all OSTs in a packed array */ |
d7e09d03 PT |
468 | struct mutex lov_lock; |
469 | struct obd_connect_data lov_ocd; | |
470 | atomic_t lov_refcount; | |
471 | __u32 lov_tgt_count; /* how many OBD's */ | |
472 | __u32 lov_active_tgt_count; /* how many active */ | |
473 | __u32 lov_death_row;/* tgts scheduled to be deleted */ | |
474 | __u32 lov_tgt_size; /* size of tgts array */ | |
475 | int lov_connects; | |
476 | int lov_pool_count; | |
6da6eabe | 477 | struct cfs_hash *lov_pools_hash_body; /* used for key access */ |
498374f7 | 478 | struct list_head lov_pool_list; /* used for sequential access */ |
61e87ab0 | 479 | struct dentry *lov_pool_debugfs_entry; |
d7e09d03 PT |
480 | enum lustre_sec_part lov_sp_me; |
481 | ||
c52f69c5 | 482 | /* Cached LRU pages from upper layer */ |
d7e09d03 PT |
483 | void *lov_cache; |
484 | ||
485 | struct rw_semaphore lov_notify_lock; | |
cb1debff OD |
486 | |
487 | struct kobject *lov_tgts_kobj; | |
d7e09d03 PT |
488 | }; |
489 | ||
490 | struct lmv_tgt_desc { | |
491 | struct obd_uuid ltd_uuid; | |
492 | struct obd_export *ltd_exp; | |
493 | int ltd_idx; | |
494 | struct mutex ltd_fid_mutex; | |
495 | unsigned long ltd_active:1; /* target up for requests */ | |
496 | }; | |
497 | ||
498 | enum placement_policy { | |
499 | PLACEMENT_CHAR_POLICY = 0, | |
500 | PLACEMENT_NID_POLICY = 1, | |
501 | PLACEMENT_INVAL_POLICY = 2, | |
502 | PLACEMENT_MAX_POLICY | |
503 | }; | |
504 | ||
d7e09d03 PT |
505 | struct lmv_obd { |
506 | int refcount; | |
507 | struct lu_client_fld lmv_fld; | |
508 | spinlock_t lmv_lock; | |
fe672997 | 509 | enum placement_policy lmv_placement; |
d7e09d03 PT |
510 | struct lmv_desc desc; |
511 | struct obd_uuid cluuid; | |
512 | struct obd_export *exp; | |
513 | ||
514 | struct mutex init_mutex; | |
515 | int connected; | |
516 | int max_easize; | |
517 | int max_def_easize; | |
518 | int max_cookiesize; | |
44779340 | 519 | int max_def_cookiesize; |
d7e09d03 PT |
520 | int server_timeout; |
521 | ||
522 | int tgts_size; /* size of tgts array */ | |
523 | struct lmv_tgt_desc **tgts; | |
524 | ||
525 | struct obd_connect_data conn_data; | |
b5fa70d7 | 526 | struct kobject *lmv_tgts_kobj; |
d7e09d03 PT |
527 | }; |
528 | ||
529 | struct niobuf_local { | |
530 | __u64 lnb_file_offset; | |
531 | __u32 lnb_page_offset; | |
532 | __u32 len; | |
533 | __u32 flags; | |
534 | struct page *page; | |
535 | struct dentry *dentry; | |
536 | int lnb_grant_used; | |
537 | int rc; | |
538 | }; | |
539 | ||
540 | #define LUSTRE_FLD_NAME "fld" | |
541 | #define LUSTRE_SEQ_NAME "seq" | |
542 | ||
543 | #define LUSTRE_MDD_NAME "mdd" | |
544 | #define LUSTRE_OSD_LDISKFS_NAME "osd-ldiskfs" | |
545 | #define LUSTRE_OSD_ZFS_NAME "osd-zfs" | |
546 | #define LUSTRE_VVP_NAME "vvp" | |
547 | #define LUSTRE_LMV_NAME "lmv" | |
548 | #define LUSTRE_SLP_NAME "slp" | |
549 | #define LUSTRE_LOD_NAME "lod" | |
550 | #define LUSTRE_OSP_NAME "osp" | |
551 | #define LUSTRE_LWP_NAME "lwp" | |
552 | ||
553 | /* obd device type names */ | |
554 | /* FIXME all the references to LUSTRE_MDS_NAME should be swapped with LUSTRE_MDT_NAME */ | |
555 | #define LUSTRE_MDS_NAME "mds" | |
556 | #define LUSTRE_MDT_NAME "mdt" | |
557 | #define LUSTRE_MDC_NAME "mdc" | |
558 | #define LUSTRE_OSS_NAME "ost" /* FIXME change name to oss */ | |
559 | #define LUSTRE_OST_NAME "obdfilter" /* FIXME change name to ost */ | |
560 | #define LUSTRE_OSC_NAME "osc" | |
561 | #define LUSTRE_LOV_NAME "lov" | |
562 | #define LUSTRE_MGS_NAME "mgs" | |
563 | #define LUSTRE_MGC_NAME "mgc" | |
564 | ||
565 | #define LUSTRE_ECHO_NAME "obdecho" | |
566 | #define LUSTRE_ECHO_CLIENT_NAME "echo_client" | |
567 | #define LUSTRE_QMT_NAME "qmt" | |
568 | ||
569 | /* Constant obd names (post-rename) */ | |
570 | #define LUSTRE_MDS_OBDNAME "MDS" | |
571 | #define LUSTRE_OSS_OBDNAME "OSS" | |
572 | #define LUSTRE_MGS_OBDNAME "MGS" | |
573 | #define LUSTRE_MGC_OBDNAME "MGC" | |
574 | ||
d7e09d03 PT |
575 | /* Don't conflict with on-wire flags OBD_BRW_WRITE, etc */ |
576 | #define N_LOCAL_TEMP_PAGE 0x10000000 | |
577 | ||
578 | struct obd_trans_info { | |
579 | __u64 oti_transno; | |
580 | __u64 oti_xid; | |
581 | /* Only used on the server side for tracking acks. */ | |
582 | struct oti_req_ack_lock { | |
583 | struct lustre_handle lock; | |
584 | __u32 mode; | |
585 | } oti_ack_locks[4]; | |
586 | void *oti_handle; | |
587 | struct llog_cookie oti_onecookie; | |
588 | struct llog_cookie *oti_logcookies; | |
589 | int oti_numcookies; | |
590 | /** synchronous write is needed */ | |
591 | unsigned long oti_sync_write:1; | |
592 | ||
593 | /* initial thread handling transaction */ | |
498374f7 | 594 | struct ptlrpc_thread *oti_thread; |
d7e09d03 PT |
595 | __u32 oti_conn_cnt; |
596 | /** VBR: versions */ | |
597 | __u64 oti_pre_version; | |
598 | /** JobID */ | |
599 | char *oti_jobid; | |
600 | ||
601 | struct obd_uuid *oti_ost_uuid; | |
602 | }; | |
603 | ||
1d8cb70c GD |
604 | static inline void oti_alloc_cookies(struct obd_trans_info *oti, |
605 | int num_cookies) | |
d7e09d03 PT |
606 | { |
607 | if (!oti) | |
608 | return; | |
609 | ||
610 | if (num_cookies == 1) | |
611 | oti->oti_logcookies = &oti->oti_onecookie; | |
612 | else | |
df29ca21 OD |
613 | oti->oti_logcookies = libcfs_kvzalloc(num_cookies * sizeof(oti->oti_onecookie), |
614 | GFP_NOFS); | |
d7e09d03 PT |
615 | |
616 | oti->oti_numcookies = num_cookies; | |
617 | } | |
618 | ||
619 | static inline void oti_free_cookies(struct obd_trans_info *oti) | |
620 | { | |
621 | if (!oti || !oti->oti_logcookies) | |
622 | return; | |
623 | ||
624 | if (oti->oti_logcookies == &oti->oti_onecookie) | |
625 | LASSERT(oti->oti_numcookies == 1); | |
626 | else | |
df29ca21 OD |
627 | kvfree(oti->oti_logcookies); |
628 | ||
d7e09d03 PT |
629 | oti->oti_logcookies = NULL; |
630 | oti->oti_numcookies = 0; | |
631 | } | |
632 | ||
633 | /* | |
634 | * Events signalled through obd_notify() upcall-chain. | |
635 | */ | |
636 | enum obd_notify_event { | |
637 | /* target added */ | |
638 | OBD_NOTIFY_CREATE, | |
639 | /* Device connect start */ | |
640 | OBD_NOTIFY_CONNECT, | |
641 | /* Device activated */ | |
642 | OBD_NOTIFY_ACTIVE, | |
643 | /* Device deactivated */ | |
644 | OBD_NOTIFY_INACTIVE, | |
645 | /* Device disconnected */ | |
646 | OBD_NOTIFY_DISCON, | |
647 | /* Connect data for import were changed */ | |
648 | OBD_NOTIFY_OCD, | |
649 | /* Sync request */ | |
650 | OBD_NOTIFY_SYNC_NONBLOCK, | |
651 | OBD_NOTIFY_SYNC, | |
652 | /* Configuration event */ | |
653 | OBD_NOTIFY_CONFIG, | |
654 | /* Administratively deactivate/activate event */ | |
655 | OBD_NOTIFY_DEACTIVATE, | |
656 | OBD_NOTIFY_ACTIVATE | |
657 | }; | |
658 | ||
d7e09d03 PT |
659 | /* |
660 | * Data structure used to pass obd_notify()-event to non-obd listeners (llite | |
85f552df | 661 | * being main example). |
d7e09d03 PT |
662 | */ |
663 | struct obd_notify_upcall { | |
664 | int (*onu_upcall)(struct obd_device *host, struct obd_device *watched, | |
665 | enum obd_notify_event ev, void *owner, void *data); | |
666 | /* Opaque datum supplied by upper layer listener */ | |
667 | void *onu_owner; | |
668 | }; | |
669 | ||
670 | struct target_recovery_data { | |
671 | svc_handler_t trd_recovery_handler; | |
672 | pid_t trd_processing_task; | |
673 | struct completion trd_starting; | |
674 | struct completion trd_finishing; | |
675 | }; | |
676 | ||
677 | struct obd_llog_group { | |
678 | int olg_seq; | |
679 | struct llog_ctxt *olg_ctxts[LLOG_MAX_CTXTS]; | |
680 | wait_queue_head_t olg_waitq; | |
681 | spinlock_t olg_lock; | |
682 | struct mutex olg_cat_processing; | |
683 | }; | |
684 | ||
685 | /* corresponds to one of the obd's */ | |
686 | #define OBD_DEVICE_MAGIC 0XAB5CD6EF | |
687 | #define OBD_DEV_BY_DEVNAME 0xffffd0de | |
688 | ||
91b3aaf9 JH |
689 | struct lvfs_run_ctxt { |
690 | struct dt_device *dt; | |
691 | }; | |
692 | ||
d7e09d03 PT |
693 | struct obd_device { |
694 | struct obd_type *obd_type; | |
695 | __u32 obd_magic; | |
696 | ||
697 | /* common and UUID name of this device */ | |
698 | char obd_name[MAX_OBD_NAME]; | |
699 | struct obd_uuid obd_uuid; | |
700 | ||
701 | struct lu_device *obd_lu_dev; | |
702 | ||
703 | int obd_minor; | |
704 | /* bitfield modification is protected by obd_dev_lock */ | |
705 | unsigned long obd_attached:1, /* finished attach */ | |
706 | obd_set_up:1, /* finished setup */ | |
d7e09d03 | 707 | obd_version_recov:1, /* obd uses version checking */ |
c56e256d OD |
708 | obd_replayable:1,/* recovery is enabled; inform clients */ |
709 | obd_no_transno:1, /* no committed-transno notification */ | |
d7e09d03 PT |
710 | obd_no_recov:1, /* fail instead of retry messages */ |
711 | obd_stopping:1, /* started cleanup */ | |
712 | obd_starting:1, /* started setup */ | |
713 | obd_force:1, /* cleanup with > 0 obd refcount */ | |
c56e256d OD |
714 | obd_fail:1, /* cleanup with failover */ |
715 | obd_async_recov:1, /* allow asynchronous orphan cleanup */ | |
d7e09d03 PT |
716 | obd_no_conn:1, /* deny new connections */ |
717 | obd_inactive:1, /* device active/inactive | |
406c1c7c OD |
718 | * (for sysfs status only!!) |
719 | */ | |
d7e09d03 PT |
720 | obd_no_ir:1, /* no imperative recovery. */ |
721 | obd_process_conf:1; /* device is processing mgs config */ | |
722 | /* use separate field as it is set in interrupt to don't mess with | |
c56e256d OD |
723 | * protection of other bits using _bh lock |
724 | */ | |
d7e09d03 PT |
725 | unsigned long obd_recovery_expired:1; |
726 | /* uuid-export hash body */ | |
6da6eabe | 727 | struct cfs_hash *obd_uuid_hash; |
d7e09d03 PT |
728 | atomic_t obd_refcount; |
729 | wait_queue_head_t obd_refcount_waitq; | |
730 | struct list_head obd_exports; | |
731 | struct list_head obd_unlinked_exports; | |
732 | struct list_head obd_delayed_exports; | |
733 | int obd_num_exports; | |
734 | spinlock_t obd_nid_lock; | |
735 | struct ldlm_namespace *obd_namespace; | |
736 | struct ptlrpc_client obd_ldlm_client; /* XXX OST/MDS only */ | |
737 | /* a spinlock is OK for what we do now, may need a semaphore later */ | |
738 | spinlock_t obd_dev_lock; /* protect OBD bitfield above */ | |
739 | struct mutex obd_dev_mutex; | |
740 | __u64 obd_last_committed; | |
d7e09d03 PT |
741 | spinlock_t obd_osfs_lock; |
742 | struct obd_statfs obd_osfs; /* locked by obd_osfs_lock */ | |
743 | __u64 obd_osfs_age; | |
744 | struct lvfs_run_ctxt obd_lvfs_ctxt; | |
745 | struct obd_llog_group obd_olg; /* default llog group */ | |
746 | struct obd_device *obd_observer; | |
747 | struct rw_semaphore obd_observer_link_sem; | |
748 | struct obd_notify_upcall obd_upcall; | |
749 | struct obd_export *obd_self_export; | |
d7e09d03 | 750 | |
d7e09d03 | 751 | union { |
d7e09d03 | 752 | struct client_obd cli; |
d7e09d03 | 753 | struct echo_client_obd echo_client; |
d7e09d03 PT |
754 | struct lov_obd lov; |
755 | struct lmv_obd lmv; | |
756 | } u; | |
757 | /* Fields used by LProcFS */ | |
758 | unsigned int obd_cntr_base; | |
759 | struct lprocfs_stats *obd_stats; | |
760 | ||
761 | unsigned int md_cntr_base; | |
762 | struct lprocfs_stats *md_stats; | |
763 | ||
61e87ab0 | 764 | struct dentry *obd_debugfs_entry; |
61e87ab0 | 765 | struct dentry *obd_svc_debugfs_entry; |
d7e09d03 PT |
766 | struct lprocfs_stats *obd_svc_stats; |
767 | atomic_t obd_evict_inprogress; | |
768 | wait_queue_head_t obd_evict_inprogress_waitq; | |
498374f7 | 769 | struct list_head obd_evict_list; /* protected with pet_lock */ |
d7e09d03 PT |
770 | |
771 | /** | |
772 | * Ldlm pool part. Save last calculated SLV and Limit. | |
773 | */ | |
774 | rwlock_t obd_pool_lock; | |
775 | int obd_pool_limit; | |
776 | __u64 obd_pool_slv; | |
777 | ||
778 | /** | |
779 | * A list of outstanding class_incref()'s against this obd. For | |
780 | * debugging. | |
781 | */ | |
782 | struct lu_ref obd_reference; | |
783 | ||
784 | int obd_conn_inprogress; | |
9b801302 OD |
785 | |
786 | struct kobject obd_kobj; /* sysfs object */ | |
787 | struct completion obd_kobj_unregister; | |
d7e09d03 PT |
788 | }; |
789 | ||
790 | #define OBD_LLOG_FL_SENDNOW 0x0001 | |
791 | #define OBD_LLOG_FL_EXIT 0x0002 | |
792 | ||
793 | enum obd_cleanup_stage { | |
794 | /* Special case hack for MDS LOVs */ | |
795 | OBD_CLEANUP_EARLY, | |
796 | /* can be directly mapped to .ldto_device_fini() */ | |
797 | OBD_CLEANUP_EXPORTS, | |
798 | }; | |
799 | ||
800 | /* get/set_info keys */ | |
801 | #define KEY_ASYNC "async" | |
802 | #define KEY_BLOCKSIZE_BITS "blocksize_bits" | |
803 | #define KEY_BLOCKSIZE "blocksize" | |
d7e09d03 PT |
804 | #define KEY_CHANGELOG_CLEAR "changelog_clear" |
805 | #define KEY_FID2PATH "fid2path" | |
806 | #define KEY_CHECKSUM "checksum" | |
807 | #define KEY_CLEAR_FS "clear_fs" | |
808 | #define KEY_CONN_DATA "conn_data" | |
809 | #define KEY_EVICT_BY_NID "evict_by_nid" | |
810 | #define KEY_FIEMAP "fiemap" | |
811 | #define KEY_FLUSH_CTX "flush_ctx" | |
812 | #define KEY_GRANT_SHRINK "grant_shrink" | |
813 | #define KEY_HSM_COPYTOOL_SEND "hsm_send" | |
814 | #define KEY_INIT_RECOV_BACKUP "init_recov_bk" | |
815 | #define KEY_INIT_RECOV "initial_recov" | |
816 | #define KEY_INTERMDS "inter_mds" | |
817 | #define KEY_LAST_ID "last_id" | |
818 | #define KEY_LAST_FID "last_fid" | |
819 | #define KEY_LOCK_TO_STRIPE "lock_to_stripe" | |
820 | #define KEY_LOVDESC "lovdesc" | |
821 | #define KEY_LOV_IDX "lov_idx" | |
44779340 BB |
822 | #define KEY_MAX_EASIZE "max_easize" |
823 | #define KEY_DEFAULT_EASIZE "default_easize" | |
d7e09d03 PT |
824 | #define KEY_MDS_CONN "mds_conn" |
825 | #define KEY_MGSSEC "mgssec" | |
826 | #define KEY_NEXT_ID "next_id" | |
827 | #define KEY_READ_ONLY "read-only" | |
828 | #define KEY_REGISTER_TARGET "register_target" | |
829 | #define KEY_SET_FS "set_fs" | |
830 | #define KEY_TGT_COUNT "tgt_count" | |
831 | /* KEY_SET_INFO in lustre_idl.h */ | |
832 | #define KEY_SPTLRPC_CONF "sptlrpc_conf" | |
833 | #define KEY_CONNECT_FLAG "connect_flags" | |
834 | #define KEY_SYNC_LOCK_CANCEL "sync_lock_cancel" | |
835 | ||
836 | #define KEY_CACHE_SET "cache_set" | |
837 | #define KEY_CACHE_LRU_SHRINK "cache_lru_shrink" | |
838 | #define KEY_CHANGELOG_INDEX "changelog_index" | |
839 | ||
840 | struct lu_context; | |
841 | ||
842 | /* /!\ must be coherent with include/linux/namei.h on patched kernel */ | |
843 | #define IT_OPEN (1 << 0) | |
844 | #define IT_CREAT (1 << 1) | |
845 | #define IT_READDIR (1 << 2) | |
846 | #define IT_GETATTR (1 << 3) | |
847 | #define IT_LOOKUP (1 << 4) | |
848 | #define IT_UNLINK (1 << 5) | |
849 | #define IT_TRUNC (1 << 6) | |
850 | #define IT_GETXATTR (1 << 7) | |
851 | #define IT_EXEC (1 << 8) | |
852 | #define IT_PIN (1 << 9) | |
853 | #define IT_LAYOUT (1 << 10) | |
854 | #define IT_QUOTA_DQACQ (1 << 11) | |
855 | #define IT_QUOTA_CONN (1 << 12) | |
7fc1f831 | 856 | #define IT_SETXATTR (1 << 13) |
d7e09d03 PT |
857 | |
858 | static inline int it_to_lock_mode(struct lookup_intent *it) | |
859 | { | |
860 | /* CREAT needs to be tested before open (both could be set) */ | |
861 | if (it->it_op & IT_CREAT) | |
862 | return LCK_CW; | |
863 | else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP | | |
864 | IT_LAYOUT)) | |
865 | return LCK_CR; | |
7fc1f831 AP |
866 | else if (it->it_op & IT_GETXATTR) |
867 | return LCK_PR; | |
868 | else if (it->it_op & IT_SETXATTR) | |
869 | return LCK_PW; | |
d7e09d03 PT |
870 | |
871 | LASSERTF(0, "Invalid it_op: %d\n", it->it_op); | |
872 | return -EINVAL; | |
873 | } | |
874 | ||
875 | struct md_op_data { | |
17891183 MI |
876 | struct lu_fid op_fid1; /* operation fid1 (usually parent) */ |
877 | struct lu_fid op_fid2; /* operation fid2 (usually child) */ | |
d7e09d03 PT |
878 | struct lu_fid op_fid3; /* 2 extra fids to find conflicting */ |
879 | struct lu_fid op_fid4; /* to the operation locks. */ | |
114acca8 | 880 | u32 op_mds; /* what mds server open will go to */ |
d7e09d03 | 881 | struct lustre_handle op_handle; |
21aef7d9 | 882 | s64 op_mod_time; |
d7e09d03 PT |
883 | const char *op_name; |
884 | int op_namelen; | |
885 | __u32 op_mode; | |
886 | struct lmv_stripe_md *op_mea1; | |
887 | struct lmv_stripe_md *op_mea2; | |
888 | __u32 op_suppgids[2]; | |
889 | __u32 op_fsuid; | |
890 | __u32 op_fsgid; | |
891 | cfs_cap_t op_cap; | |
892 | void *op_data; | |
893 | ||
894 | /* iattr fields and blocks. */ | |
895 | struct iattr op_attr; | |
896 | unsigned int op_attr_flags; | |
897 | __u64 op_valid; | |
898 | loff_t op_attr_blocks; | |
899 | ||
900 | /* Size-on-MDS epoch and flags. */ | |
901 | __u64 op_ioepoch; | |
902 | __u32 op_flags; | |
903 | ||
d7e09d03 | 904 | /* Various operation flags. */ |
48d23e61 | 905 | enum mds_op_bias op_bias; |
d7e09d03 PT |
906 | |
907 | /* Operation type */ | |
908 | __u32 op_opc; | |
909 | ||
910 | /* Used by readdir */ | |
911 | __u64 op_offset; | |
912 | ||
913 | /* Used by readdir */ | |
914 | __u32 op_npages; | |
915 | ||
916 | /* used to transfer info between the stacks of MD client | |
c56e256d OD |
917 | * see enum op_cli_flags |
918 | */ | |
d7e09d03 | 919 | __u32 op_cli_flags; |
48d23e61 JX |
920 | |
921 | /* File object data version for HSM release, on client */ | |
922 | __u64 op_data_version; | |
923 | struct lustre_handle op_lease_handle; | |
d7e09d03 PT |
924 | }; |
925 | ||
926 | enum op_cli_flags { | |
927 | CLI_SET_MEA = 1 << 0, | |
928 | CLI_RM_ENTRY = 1 << 1, | |
929 | }; | |
930 | ||
931 | struct md_enqueue_info; | |
932 | /* metadata stat-ahead */ | |
d7e09d03 | 933 | |
d7e09d03 PT |
934 | struct md_enqueue_info { |
935 | struct md_op_data mi_data; | |
936 | struct lookup_intent mi_it; | |
937 | struct lustre_handle mi_lockh; | |
938 | struct inode *mi_dir; | |
fe672997 | 939 | int (*mi_cb)(struct ptlrpc_request *req, |
10457d4b | 940 | struct md_enqueue_info *minfo, int rc); |
d7e09d03 PT |
941 | __u64 mi_cbdata; |
942 | unsigned int mi_generation; | |
943 | }; | |
944 | ||
945 | struct obd_ops { | |
a13b1f32 DC |
946 | struct module *owner; |
947 | int (*iocontrol)(unsigned int cmd, struct obd_export *exp, int len, | |
e09bee34 | 948 | void *karg, void __user *uarg); |
a13b1f32 DC |
949 | int (*get_info)(const struct lu_env *env, struct obd_export *, |
950 | __u32 keylen, void *key, __u32 *vallen, void *val, | |
951 | struct lov_stripe_md *lsm); | |
952 | int (*set_info_async)(const struct lu_env *, struct obd_export *, | |
953 | __u32 keylen, void *key, | |
954 | __u32 vallen, void *val, | |
955 | struct ptlrpc_request_set *set); | |
956 | int (*attach)(struct obd_device *dev, u32 len, void *data); | |
957 | int (*detach)(struct obd_device *dev); | |
958 | int (*setup)(struct obd_device *dev, struct lustre_cfg *cfg); | |
959 | int (*precleanup)(struct obd_device *dev, | |
960 | enum obd_cleanup_stage cleanup_stage); | |
961 | int (*cleanup)(struct obd_device *dev); | |
962 | int (*process_config)(struct obd_device *dev, u32 len, void *data); | |
963 | int (*postrecov)(struct obd_device *dev); | |
964 | int (*add_conn)(struct obd_import *imp, struct obd_uuid *uuid, | |
965 | int priority); | |
966 | int (*del_conn)(struct obd_import *imp, struct obd_uuid *uuid); | |
d7e09d03 PT |
967 | /* connect to the target device with given connection |
968 | * data. @ocd->ocd_connect_flags is modified to reflect flags actually | |
969 | * granted by the target, which are guaranteed to be a subset of flags | |
c56e256d OD |
970 | * asked for. If @ocd == NULL, use default parameters. |
971 | */ | |
a13b1f32 DC |
972 | int (*connect)(const struct lu_env *env, |
973 | struct obd_export **exp, struct obd_device *src, | |
974 | struct obd_uuid *cluuid, struct obd_connect_data *ocd, | |
975 | void *localdata); | |
976 | int (*reconnect)(const struct lu_env *env, | |
977 | struct obd_export *exp, struct obd_device *src, | |
978 | struct obd_uuid *cluuid, | |
979 | struct obd_connect_data *ocd, | |
d7e09d03 | 980 | void *localdata); |
a13b1f32 | 981 | int (*disconnect)(struct obd_export *exp); |
d7e09d03 PT |
982 | |
983 | /* Initialize/finalize fids infrastructure. */ | |
a13b1f32 DC |
984 | int (*fid_init)(struct obd_device *obd, |
985 | struct obd_export *exp, enum lu_cli_type type); | |
986 | int (*fid_fini)(struct obd_device *obd); | |
d7e09d03 PT |
987 | |
988 | /* Allocate new fid according to passed @hint. */ | |
a13b1f32 DC |
989 | int (*fid_alloc)(struct obd_export *exp, struct lu_fid *fid, |
990 | struct md_op_data *op_data); | |
d7e09d03 PT |
991 | |
992 | /* | |
993 | * Object with @fid is getting deleted, we may want to do something | |
994 | * about this. | |
995 | */ | |
a13b1f32 DC |
996 | int (*statfs)(const struct lu_env *, struct obd_export *exp, |
997 | struct obd_statfs *osfs, __u64 max_age, __u32 flags); | |
998 | int (*statfs_async)(struct obd_export *exp, struct obd_info *oinfo, | |
999 | __u64 max_age, struct ptlrpc_request_set *set); | |
1000 | int (*packmd)(struct obd_export *exp, struct lov_mds_md **disk_tgt, | |
1001 | struct lov_stripe_md *mem_src); | |
1002 | int (*unpackmd)(struct obd_export *exp, | |
1003 | struct lov_stripe_md **mem_tgt, | |
1004 | struct lov_mds_md *disk_src, int disk_len); | |
1005 | int (*preallocate)(struct lustre_handle *, u32 *req, u64 *ids); | |
1006 | int (*create)(const struct lu_env *env, struct obd_export *exp, | |
1007 | struct obdo *oa, struct lov_stripe_md **ea, | |
1008 | struct obd_trans_info *oti); | |
1009 | int (*destroy)(const struct lu_env *env, struct obd_export *exp, | |
1010 | struct obdo *oa, struct lov_stripe_md *ea, | |
1011 | struct obd_trans_info *oti, struct obd_export *md_exp); | |
1012 | int (*setattr)(const struct lu_env *, struct obd_export *exp, | |
1013 | struct obd_info *oinfo, struct obd_trans_info *oti); | |
1014 | int (*setattr_async)(struct obd_export *exp, struct obd_info *oinfo, | |
1015 | struct obd_trans_info *oti, | |
1016 | struct ptlrpc_request_set *rqset); | |
1017 | int (*getattr)(const struct lu_env *env, struct obd_export *exp, | |
1018 | struct obd_info *oinfo); | |
1019 | int (*getattr_async)(struct obd_export *exp, struct obd_info *oinfo, | |
1020 | struct ptlrpc_request_set *set); | |
1021 | int (*adjust_kms)(struct obd_export *exp, struct lov_stripe_md *lsm, | |
1022 | u64 size, int shrink); | |
1023 | int (*preprw)(const struct lu_env *env, int cmd, | |
1024 | struct obd_export *exp, struct obdo *oa, int objcount, | |
1025 | struct obd_ioobj *obj, struct niobuf_remote *remote, | |
1026 | int *nr_pages, struct niobuf_local *local, | |
1027 | struct obd_trans_info *oti); | |
1028 | int (*commitrw)(const struct lu_env *env, int cmd, | |
1029 | struct obd_export *exp, struct obdo *oa, | |
1030 | int objcount, struct obd_ioobj *obj, | |
1031 | struct niobuf_remote *remote, int pages, | |
1032 | struct niobuf_local *local, | |
1033 | struct obd_trans_info *oti, int rc); | |
1034 | int (*find_cbdata)(struct obd_export *, struct lov_stripe_md *, | |
1035 | ldlm_iterator_t it, void *data); | |
1036 | int (*init_export)(struct obd_export *exp); | |
1037 | int (*destroy_export)(struct obd_export *exp); | |
d7e09d03 PT |
1038 | |
1039 | /* metadata-only methods */ | |
a13b1f32 DC |
1040 | int (*import_event)(struct obd_device *, struct obd_import *, |
1041 | enum obd_import_event); | |
d7e09d03 | 1042 | |
a13b1f32 DC |
1043 | int (*notify)(struct obd_device *obd, struct obd_device *watched, |
1044 | enum obd_notify_event ev, void *data); | |
d7e09d03 | 1045 | |
a13b1f32 DC |
1046 | int (*health_check)(const struct lu_env *env, struct obd_device *); |
1047 | struct obd_uuid *(*get_uuid)(struct obd_export *exp); | |
d7e09d03 PT |
1048 | |
1049 | /* quota methods */ | |
a13b1f32 | 1050 | int (*quotacheck)(struct obd_device *, struct obd_export *, |
d7e09d03 | 1051 | struct obd_quotactl *); |
a13b1f32 DC |
1052 | int (*quotactl)(struct obd_device *, struct obd_export *, |
1053 | struct obd_quotactl *); | |
d7e09d03 | 1054 | |
d7e09d03 | 1055 | /* pools methods */ |
a13b1f32 DC |
1056 | int (*pool_new)(struct obd_device *obd, char *poolname); |
1057 | int (*pool_del)(struct obd_device *obd, char *poolname); | |
1058 | int (*pool_add)(struct obd_device *obd, char *poolname, | |
1059 | char *ostname); | |
1060 | int (*pool_rem)(struct obd_device *obd, char *poolname, | |
1061 | char *ostname); | |
1062 | void (*getref)(struct obd_device *obd); | |
1063 | void (*putref)(struct obd_device *obd); | |
d7e09d03 PT |
1064 | /* |
1065 | * NOTE: If adding ops, add another LPROCFS_OBD_OP_INIT() line | |
1066 | * to lprocfs_alloc_obd_stats() in obdclass/lprocfs_status.c. | |
c56e256d OD |
1067 | * Also, add a wrapper function in include/linux/obd_class.h. |
1068 | */ | |
d7e09d03 PT |
1069 | }; |
1070 | ||
1071 | enum { | |
1072 | LUSTRE_OPC_MKDIR = (1 << 0), | |
1073 | LUSTRE_OPC_SYMLINK = (1 << 1), | |
1074 | LUSTRE_OPC_MKNOD = (1 << 2), | |
1075 | LUSTRE_OPC_CREATE = (1 << 3), | |
1076 | LUSTRE_OPC_ANY = (1 << 4) | |
1077 | }; | |
1078 | ||
1079 | /* lmv structures */ | |
1080 | #define MEA_MAGIC_LAST_CHAR 0xb2221ca1 | |
1081 | #define MEA_MAGIC_ALL_CHARS 0xb222a11c | |
1082 | #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b | |
1083 | ||
1084 | #define MAX_HASH_SIZE_32 0x7fffffffUL | |
1085 | #define MAX_HASH_SIZE 0x7fffffffffffffffULL | |
1086 | #define MAX_HASH_HIGHEST_BIT 0x1000000000000000ULL | |
1087 | ||
1088 | struct lustre_md { | |
1089 | struct mdt_body *body; | |
1090 | struct lov_stripe_md *lsm; | |
1091 | struct lmv_stripe_md *mea; | |
1092 | #ifdef CONFIG_FS_POSIX_ACL | |
1093 | struct posix_acl *posix_acl; | |
1094 | #endif | |
1095 | struct mdt_remote_perm *remote_perm; | |
d7e09d03 PT |
1096 | }; |
1097 | ||
1098 | struct md_open_data { | |
1099 | struct obd_client_handle *mod_och; | |
1100 | struct ptlrpc_request *mod_open_req; | |
1101 | struct ptlrpc_request *mod_close_req; | |
63d42578 HZ |
1102 | atomic_t mod_refcount; |
1103 | bool mod_is_create; | |
d7e09d03 PT |
1104 | }; |
1105 | ||
1106 | struct lookup_intent; | |
1107 | ||
1108 | struct md_ops { | |
df18a80a DC |
1109 | int (*getstatus)(struct obd_export *, struct lu_fid *); |
1110 | int (*null_inode)(struct obd_export *, const struct lu_fid *); | |
1111 | int (*find_cbdata)(struct obd_export *, const struct lu_fid *, | |
1112 | ldlm_iterator_t, void *); | |
1113 | int (*close)(struct obd_export *, struct md_op_data *, | |
1114 | struct md_open_data *, struct ptlrpc_request **); | |
1115 | int (*create)(struct obd_export *, struct md_op_data *, | |
1116 | const void *, int, int, __u32, __u32, cfs_cap_t, | |
1117 | __u64, struct ptlrpc_request **); | |
1118 | int (*done_writing)(struct obd_export *, struct md_op_data *, | |
1119 | struct md_open_data *); | |
1120 | int (*enqueue)(struct obd_export *, struct ldlm_enqueue_info *, | |
1121 | struct lookup_intent *, struct md_op_data *, | |
1122 | struct lustre_handle *, void *, int, | |
1123 | struct ptlrpc_request **, __u64); | |
1124 | int (*getattr)(struct obd_export *, struct md_op_data *, | |
1125 | struct ptlrpc_request **); | |
1126 | int (*getattr_name)(struct obd_export *, struct md_op_data *, | |
1127 | struct ptlrpc_request **); | |
1128 | int (*intent_lock)(struct obd_export *, struct md_op_data *, | |
1129 | void *, int, struct lookup_intent *, int, | |
1130 | struct ptlrpc_request **, | |
1131 | ldlm_blocking_callback, __u64); | |
1132 | int (*link)(struct obd_export *, struct md_op_data *, | |
1133 | struct ptlrpc_request **); | |
1134 | int (*rename)(struct obd_export *, struct md_op_data *, | |
1135 | const char *, int, const char *, int, | |
d7e09d03 | 1136 | struct ptlrpc_request **); |
df18a80a DC |
1137 | int (*is_subdir)(struct obd_export *, const struct lu_fid *, |
1138 | const struct lu_fid *, | |
d7e09d03 | 1139 | struct ptlrpc_request **); |
df18a80a DC |
1140 | int (*setattr)(struct obd_export *, struct md_op_data *, void *, |
1141 | int, void *, int, struct ptlrpc_request **, | |
d7e09d03 | 1142 | struct md_open_data **mod); |
df18a80a DC |
1143 | int (*sync)(struct obd_export *, const struct lu_fid *, |
1144 | struct ptlrpc_request **); | |
1145 | int (*readpage)(struct obd_export *, struct md_op_data *, | |
1146 | struct page **, struct ptlrpc_request **); | |
1147 | ||
1148 | int (*unlink)(struct obd_export *, struct md_op_data *, | |
ef2e0f55 | 1149 | struct ptlrpc_request **); |
d7e09d03 | 1150 | |
df18a80a DC |
1151 | int (*setxattr)(struct obd_export *, const struct lu_fid *, |
1152 | u64, const char *, const char *, int, int, int, __u32, | |
d7e09d03 PT |
1153 | struct ptlrpc_request **); |
1154 | ||
df18a80a DC |
1155 | int (*getxattr)(struct obd_export *, const struct lu_fid *, |
1156 | u64, const char *, const char *, int, int, int, | |
1157 | struct ptlrpc_request **); | |
d7e09d03 | 1158 | |
df18a80a | 1159 | int (*init_ea_size)(struct obd_export *, int, int, int, int); |
d7e09d03 | 1160 | |
df18a80a DC |
1161 | int (*get_lustre_md)(struct obd_export *, struct ptlrpc_request *, |
1162 | struct obd_export *, struct obd_export *, | |
1163 | struct lustre_md *); | |
d7e09d03 | 1164 | |
df18a80a | 1165 | int (*free_lustre_md)(struct obd_export *, struct lustre_md *); |
d7e09d03 | 1166 | |
df18a80a DC |
1167 | int (*set_open_replay_data)(struct obd_export *, |
1168 | struct obd_client_handle *, | |
1169 | struct lookup_intent *); | |
1170 | int (*clear_open_replay_data)(struct obd_export *, | |
1171 | struct obd_client_handle *); | |
1172 | int (*set_lock_data)(struct obd_export *, __u64 *, void *, __u64 *); | |
d7e09d03 | 1173 | |
52ee0d20 OD |
1174 | enum ldlm_mode (*lock_match)(struct obd_export *, __u64, |
1175 | const struct lu_fid *, enum ldlm_type, | |
1176 | ldlm_policy_data_t *, enum ldlm_mode, | |
1177 | struct lustre_handle *); | |
d7e09d03 | 1178 | |
df18a80a | 1179 | int (*cancel_unused)(struct obd_export *, const struct lu_fid *, |
52ee0d20 | 1180 | ldlm_policy_data_t *, enum ldlm_mode, |
f833ee42 | 1181 | enum ldlm_cancel_flags flags, void *opaque); |
d7e09d03 | 1182 | |
df18a80a DC |
1183 | int (*get_remote_perm)(struct obd_export *, const struct lu_fid *, |
1184 | __u32, struct ptlrpc_request **); | |
d7e09d03 | 1185 | |
df18a80a DC |
1186 | int (*intent_getattr_async)(struct obd_export *, |
1187 | struct md_enqueue_info *, | |
1188 | struct ldlm_enqueue_info *); | |
d7e09d03 | 1189 | |
df18a80a DC |
1190 | int (*revalidate_lock)(struct obd_export *, struct lookup_intent *, |
1191 | struct lu_fid *, __u64 *bits); | |
d7e09d03 PT |
1192 | |
1193 | /* | |
1194 | * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to | |
1195 | * lprocfs_alloc_md_stats() in obdclass/lprocfs_status.c. Also, add a | |
1196 | * wrapper function in include/linux/obd_class.h. | |
1197 | */ | |
1198 | }; | |
1199 | ||
1200 | struct lsm_operations { | |
1201 | void (*lsm_free)(struct lov_stripe_md *); | |
1202 | int (*lsm_destroy)(struct lov_stripe_md *, struct obdo *oa, | |
1203 | struct obd_export *md_exp); | |
21aef7d9 OD |
1204 | void (*lsm_stripe_by_index)(struct lov_stripe_md *, int *, u64 *, |
1205 | u64 *); | |
1206 | void (*lsm_stripe_by_offset)(struct lov_stripe_md *, int *, u64 *, | |
1207 | u64 *); | |
b037976f | 1208 | int (*lsm_lmm_verify)(struct lov_mds_md *lmm, int lmm_bytes, |
10457d4b | 1209 | __u16 *stripe_count); |
b037976f | 1210 | int (*lsm_unpackmd)(struct lov_obd *lov, struct lov_stripe_md *lsm, |
10457d4b | 1211 | struct lov_mds_md *lmm); |
d7e09d03 PT |
1212 | }; |
1213 | ||
1214 | extern const struct lsm_operations lsm_v1_ops; | |
1215 | extern const struct lsm_operations lsm_v3_ops; | |
1216 | static inline const struct lsm_operations *lsm_op_find(int magic) | |
1217 | { | |
a58a38ac | 1218 | switch (magic) { |
d7e09d03 PT |
1219 | case LOV_MAGIC_V1: |
1220 | return &lsm_v1_ops; | |
1221 | case LOV_MAGIC_V3: | |
1222 | return &lsm_v3_ops; | |
1223 | default: | |
1224 | CERROR("Cannot recognize lsm_magic %08x\n", magic); | |
1225 | return NULL; | |
1226 | } | |
1227 | } | |
1228 | ||
1229 | /* Requests for obd_extent_calc() */ | |
1230 | #define OBD_CALC_STRIPE_START 1 | |
1231 | #define OBD_CALC_STRIPE_END 2 | |
1232 | ||
d7e09d03 PT |
1233 | static inline struct md_open_data *obd_mod_alloc(void) |
1234 | { | |
1235 | struct md_open_data *mod; | |
498374f7 | 1236 | |
af13af52 | 1237 | mod = kzalloc(sizeof(*mod), GFP_NOFS); |
d2a13989 | 1238 | if (!mod) |
d7e09d03 PT |
1239 | return NULL; |
1240 | atomic_set(&mod->mod_refcount, 1); | |
1241 | return mod; | |
1242 | } | |
1243 | ||
1244 | #define obd_mod_get(mod) atomic_inc(&(mod)->mod_refcount) | |
1245 | #define obd_mod_put(mod) \ | |
1246 | ({ \ | |
1247 | if (atomic_dec_and_test(&(mod)->mod_refcount)) { \ | |
1248 | if ((mod)->mod_open_req) \ | |
1249 | ptlrpc_req_finished((mod)->mod_open_req); \ | |
af13af52 | 1250 | kfree(mod); \ |
d7e09d03 PT |
1251 | } \ |
1252 | }) | |
1253 | ||
21aef7d9 | 1254 | void obdo_from_inode(struct obdo *dst, struct inode *src, u32 valid); |
d7e09d03 PT |
1255 | void obdo_set_parent_fid(struct obdo *dst, const struct lu_fid *parent); |
1256 | ||
1257 | /* return 1 if client should be resend request */ | |
1258 | static inline int client_should_resend(int resend, struct client_obd *cli) | |
1259 | { | |
1260 | return atomic_read(&cli->cl_resends) ? | |
1261 | atomic_read(&cli->cl_resends) > resend : 1; | |
1262 | } | |
1263 | ||
1264 | /** | |
1265 | * Return device name for this device | |
1266 | * | |
1267 | * XXX: lu_device is declared before obd_device, while a pointer pointing | |
1268 | * back to obd_device in lu_device, so this helper function defines here | |
1269 | * instead of in lu_object.h | |
1270 | */ | |
1271 | static inline const char *lu_dev_name(const struct lu_device *lu_dev) | |
1272 | { | |
1273 | return lu_dev->ld_obd->obd_name; | |
1274 | } | |
1275 | ||
1276 | static inline bool filename_is_volatile(const char *name, int namelen, int *idx) | |
1277 | { | |
1278 | const char *start; | |
1279 | char *end; | |
1280 | ||
1281 | if (strncmp(name, LUSTRE_VOLATILE_HDR, LUSTRE_VOLATILE_HDR_LEN) != 0) | |
1282 | return false; | |
1283 | ||
1284 | /* caller does not care of idx */ | |
d2a13989 | 1285 | if (!idx) |
d7e09d03 PT |
1286 | return true; |
1287 | ||
1288 | /* volatile file, the MDT can be set from name */ | |
1289 | /* name format is LUSTRE_VOLATILE_HDR:[idx]: */ | |
1290 | /* if no MDT is specified, use std way */ | |
1291 | if (namelen < LUSTRE_VOLATILE_HDR_LEN + 2) | |
1292 | goto bad_format; | |
1293 | /* test for no MDT idx case */ | |
1294 | if ((*(name + LUSTRE_VOLATILE_HDR_LEN) == ':') && | |
1295 | (*(name + LUSTRE_VOLATILE_HDR_LEN + 1) == ':')) { | |
1296 | *idx = -1; | |
1297 | return true; | |
1298 | } | |
1299 | /* we have an idx, read it */ | |
1300 | start = name + LUSTRE_VOLATILE_HDR_LEN + 1; | |
4ad2adc9 | 1301 | *idx = simple_strtoul(start, &end, 0); |
d7e09d03 PT |
1302 | /* error cases: |
1303 | * no digit, no trailing :, negative value | |
1304 | */ | |
1305 | if (((*idx == 0) && (end == start)) || | |
1306 | (*end != ':') || (*idx < 0)) | |
1307 | goto bad_format; | |
1308 | ||
1309 | return true; | |
1310 | bad_format: | |
1311 | /* bad format of mdt idx, we cannot return an error | |
c56e256d OD |
1312 | * to caller so we use hash algo |
1313 | */ | |
d7e09d03 PT |
1314 | CERROR("Bad volatile file name format: %s\n", |
1315 | name + LUSTRE_VOLATILE_HDR_LEN); | |
1316 | return false; | |
1317 | } | |
1318 | ||
1319 | static inline int cli_brw_size(struct obd_device *obd) | |
1320 | { | |
09cbfeaf | 1321 | return obd->u.cli.cl_max_pages_per_rpc << PAGE_SHIFT; |
d7e09d03 PT |
1322 | } |
1323 | ||
1324 | #endif /* __OBD_H */ |