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 | |
6a5b99a4 | 18 | * http://www.gnu.org/licenses/gpl-2.0.html |
d7e09d03 | 19 | * |
d7e09d03 PT |
20 | * GPL HEADER END |
21 | */ | |
22 | /* | |
23 | * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved. | |
24 | * Use is subject to license terms. | |
25 | * | |
1dc563a6 | 26 | * Copyright (c) 2011, 2015, Intel Corporation. |
d7e09d03 PT |
27 | */ |
28 | /* | |
29 | * This file is part of Lustre, http://www.lustre.org/ | |
30 | * Lustre is a trademark of Sun Microsystems, Inc. | |
31 | */ | |
32 | ||
33 | #ifndef __OBD_H | |
34 | #define __OBD_H | |
35 | ||
7d53d8f4 | 36 | #include <linux/spinlock.h> |
d7e09d03 | 37 | |
1accaadf GKH |
38 | #include "lustre/lustre_idl.h" |
39 | #include "lustre_lib.h" | |
1accaadf GKH |
40 | #include "lu_ref.h" |
41 | #include "lustre_export.h" | |
42 | #include "lustre_fid.h" | |
43 | #include "lustre_fld.h" | |
bb41292b | 44 | #include "lustre_intent.h" |
d7e09d03 | 45 | |
d7e09d03 PT |
46 | #define MAX_OBD_DEVICES 8192 |
47 | ||
48 | struct osc_async_rc { | |
49 | int ar_rc; | |
50 | int ar_force_sync; | |
51 | __u64 ar_min_xid; | |
52 | }; | |
53 | ||
54 | struct lov_oinfo { /* per-stripe data structure */ | |
55 | struct ost_id loi_oi; /* object ID/Sequence on the target OST */ | |
56 | int loi_ost_idx; /* OST stripe index in lov_tgt_desc->tgts */ | |
57 | int loi_ost_gen; /* generation of this loi_ost_idx */ | |
58 | ||
59 | unsigned long loi_kms_valid:1; | |
60 | __u64 loi_kms; /* known minimum size */ | |
61 | struct ost_lvb loi_lvb; | |
62 | struct osc_async_rc loi_ar; | |
63 | }; | |
64 | ||
65 | static inline void loi_kms_set(struct lov_oinfo *oinfo, __u64 kms) | |
66 | { | |
67 | oinfo->loi_kms = kms; | |
68 | oinfo->loi_kms_valid = 1; | |
69 | } | |
70 | ||
71 | static inline void loi_init(struct lov_oinfo *loi) | |
72 | { | |
73 | } | |
74 | ||
75 | struct lov_stripe_md { | |
76 | atomic_t lsm_refc; | |
77 | spinlock_t lsm_lock; | |
78 | pid_t lsm_lock_owner; /* debugging */ | |
79 | ||
80 | /* maximum possible file size, might change as OSTs status changes, | |
c56e256d OD |
81 | * e.g. disconnected, deactivated |
82 | */ | |
d7e09d03 PT |
83 | __u64 lsm_maxbytes; |
84 | struct { | |
85 | /* Public members. */ | |
86 | struct ost_id lw_object_oi; /* lov object id/seq */ | |
87 | ||
88 | /* LOV-private members start here -- only for use in lov/. */ | |
89 | __u32 lw_magic; | |
90 | __u32 lw_stripe_size; /* size of the stripe */ | |
91 | __u32 lw_pattern; /* striping pattern (RAID0, RAID1) */ | |
92 | __u16 lw_stripe_count; /* number of objects being striped over */ | |
93 | __u16 lw_layout_gen; /* generation of the layout */ | |
94 | char lw_pool_name[LOV_MAXPOOLNAME]; /* pool name */ | |
95 | } lsm_wire; | |
96 | ||
97 | struct lov_oinfo *lsm_oinfo[0]; | |
98 | }; | |
99 | ||
100 | #define lsm_oi lsm_wire.lw_object_oi | |
101 | #define lsm_magic lsm_wire.lw_magic | |
102 | #define lsm_layout_gen lsm_wire.lw_layout_gen | |
103 | #define lsm_stripe_size lsm_wire.lw_stripe_size | |
104 | #define lsm_pattern lsm_wire.lw_pattern | |
105 | #define lsm_stripe_count lsm_wire.lw_stripe_count | |
106 | #define lsm_pool_name lsm_wire.lw_pool_name | |
107 | ||
5dd16419 JX |
108 | static inline bool lsm_is_released(struct lov_stripe_md *lsm) |
109 | { | |
110 | return !!(lsm->lsm_pattern & LOV_PATTERN_F_RELEASED); | |
111 | } | |
112 | ||
113 | static inline bool lsm_has_objects(struct lov_stripe_md *lsm) | |
114 | { | |
d2a13989 | 115 | if (!lsm) |
5dd16419 JX |
116 | return false; |
117 | if (lsm_is_released(lsm)) | |
118 | return false; | |
119 | return true; | |
120 | } | |
121 | ||
081b7265 JH |
122 | static inline int lov_stripe_md_size(unsigned int stripe_count) |
123 | { | |
124 | struct lov_stripe_md lsm; | |
125 | ||
126 | return sizeof(lsm) + stripe_count * sizeof(lsm.lsm_oinfo[0]); | |
127 | } | |
128 | ||
d7e09d03 PT |
129 | struct obd_info; |
130 | ||
131 | typedef int (*obd_enqueue_update_f)(void *cookie, int rc); | |
132 | ||
133 | /* obd info for a particular level (lov, osc). */ | |
134 | struct obd_info { | |
d7e09d03 PT |
135 | /* Flags used for set request specific flags: |
136 | - while lock handling, the flags obtained on the enqueue | |
137 | request are set here. | |
138 | - while stats, the flags used for control delay/resend. | |
139 | - while setattr, the flags used for distinguish punch operation | |
140 | */ | |
141 | __u64 oi_flags; | |
d7e09d03 PT |
142 | /* lsm data specific for every OSC. */ |
143 | struct lov_stripe_md *oi_md; | |
144 | /* obdo data specific for every OSC, if needed at all. */ | |
145 | struct obdo *oi_oa; | |
146 | /* statfs data specific for every OSC, if needed at all. */ | |
147 | struct obd_statfs *oi_osfs; | |
148 | /* An update callback which is called to update some data on upper | |
17891183 | 149 | * level. E.g. it is used for update lsm->lsm_oinfo at every received |
d7e09d03 | 150 | * request in osc level for enqueue requests. It is also possible to |
c56e256d OD |
151 | * update some caller data from LOV layer if needed. |
152 | */ | |
d7e09d03 | 153 | obd_enqueue_update_f oi_cb_up; |
d7e09d03 PT |
154 | }; |
155 | ||
d7e09d03 PT |
156 | struct obd_type { |
157 | struct list_head typ_chain; | |
158 | struct obd_ops *typ_dt_ops; | |
159 | struct md_ops *typ_md_ops; | |
61e87ab0 | 160 | struct dentry *typ_debugfs_entry; |
d7e09d03 PT |
161 | char *typ_name; |
162 | int typ_refcnt; | |
163 | struct lu_device_type *typ_lu; | |
164 | spinlock_t obd_type_lock; | |
9b801302 | 165 | struct kobject *typ_kobj; |
d7e09d03 PT |
166 | }; |
167 | ||
168 | struct brw_page { | |
21aef7d9 | 169 | u64 off; |
d7e09d03 PT |
170 | struct page *pg; |
171 | int count; | |
21aef7d9 | 172 | u32 flag; |
d7e09d03 PT |
173 | }; |
174 | ||
d7e09d03 PT |
175 | struct timeout_item { |
176 | enum timeout_event ti_event; | |
a649ad1d | 177 | unsigned long ti_timeout; |
d7e09d03 PT |
178 | timeout_cb_t ti_cb; |
179 | void *ti_cb_data; | |
180 | struct list_head ti_obd_list; | |
181 | struct list_head ti_chain; | |
182 | }; | |
183 | ||
1d5d5ec1 FY |
184 | #define OBD_MAX_RIF_DEFAULT 8 |
185 | #define OBD_MAX_RIF_MAX 512 | |
186 | #define OSC_MAX_RIF_MAX 256 | |
187 | #define OSC_MAX_DIRTY_DEFAULT (OBD_MAX_RIF_DEFAULT * 4) | |
188 | #define OSC_MAX_DIRTY_MB_MAX 2048 /* arbitrary, but < MAX_LONG bytes */ | |
189 | #define OSC_DEFAULT_RESENDS 10 | |
d7e09d03 PT |
190 | |
191 | /* possible values for fo_sync_lock_cancel */ | |
192 | enum { | |
193 | NEVER_SYNC_ON_CANCEL = 0, | |
194 | BLOCKING_SYNC_ON_CANCEL = 1, | |
195 | ALWAYS_SYNC_ON_CANCEL = 2, | |
196 | NUM_SYNC_ON_CANCEL_STATES | |
197 | }; | |
198 | ||
9936913e AD |
199 | enum obd_cl_sem_lock_class { |
200 | OBD_CLI_SEM_NORMAL, | |
201 | OBD_CLI_SEM_MGC, | |
202 | OBD_CLI_SEM_MDCOSC, | |
203 | }; | |
204 | ||
d7e09d03 PT |
205 | struct mdc_rpc_lock; |
206 | struct obd_import; | |
207 | struct client_obd { | |
208 | struct rw_semaphore cl_sem; | |
209 | struct obd_uuid cl_target_uuid; | |
210 | struct obd_import *cl_import; /* ptlrpc connection state */ | |
211 | int cl_conn_count; | |
212 | /* max_mds_easize is purely a performance thing so we don't have to | |
c56e256d OD |
213 | * call obd_size_diskmd() all the time. |
214 | */ | |
44779340 BB |
215 | int cl_default_mds_easize; |
216 | int cl_max_mds_easize; | |
217 | int cl_default_mds_cookiesize; | |
218 | int cl_max_mds_cookiesize; | |
d7e09d03 PT |
219 | |
220 | enum lustre_sec_part cl_sp_me; | |
221 | enum lustre_sec_part cl_sp_to; | |
222 | struct sptlrpc_flavor cl_flvr_mgc; /* fixed flavor of mgc->mgs */ | |
223 | ||
224 | /* the grant values are protected by loi_list_lock below */ | |
3147b268 HZ |
225 | long cl_dirty_pages; /* all _dirty_ in pahges */ |
226 | long cl_dirty_max_pages;/* allowed w/o rpc */ | |
d7e09d03 PT |
227 | long cl_dirty_transit; /* dirty synchronous */ |
228 | long cl_avail_grant; /* bytes of credit for ost */ | |
229 | long cl_lost_grant; /* lost credits (trunc) */ | |
230 | ||
231 | /* since we allocate grant by blocks, we don't know how many grant will | |
232 | * be used to add a page into cache. As a solution, we reserve maximum | |
233 | * grant before trying to dirty a page and unreserve the rest. | |
c56e256d OD |
234 | * See osc_{reserve|unreserve}_grant for details. |
235 | */ | |
d7e09d03 | 236 | long cl_reserved_grant; |
498374f7 HM |
237 | struct list_head cl_cache_waiters; /* waiting for cache/grant */ |
238 | unsigned long cl_next_shrink_grant; /* jiffies */ | |
239 | struct list_head cl_grant_shrink_list; /* Timeout event list */ | |
240 | int cl_grant_shrink_interval; /* seconds */ | |
d7e09d03 PT |
241 | |
242 | /* A chunk is an optimal size used by osc_extent to determine | |
ea1754a0 | 243 | * the extent size. A chunk is max(PAGE_SIZE, OST block size) |
c56e256d | 244 | */ |
d7e09d03 | 245 | int cl_chunkbits; |
d7e09d03 PT |
246 | int cl_extent_tax; /* extent overhead, by bytes */ |
247 | ||
248 | /* keep track of objects that have lois that contain pages which | |
249 | * have been queued for async brw. this lock also protects the | |
c56e256d OD |
250 | * lists of osc_client_pages that hang off of the loi |
251 | */ | |
d7e09d03 PT |
252 | /* |
253 | * ->cl_loi_list_lock protects consistency of | |
254 | * ->cl_loi_{ready,read,write}_list. ->ap_make_ready() and | |
255 | * ->ap_completion() call-backs are executed under this lock. As we | |
256 | * cannot guarantee that these call-backs never block on all platforms | |
257 | * (as a matter of fact they do block on Mac OS X), type of | |
258 | * ->cl_loi_list_lock is platform dependent: it's a spin-lock on Linux | |
259 | * and blocking mutex on Mac OS X. (Alternative is to make this lock | |
260 | * blocking everywhere, but we don't want to slow down fast-path of | |
261 | * our main platform.) | |
262 | * | |
d7e09d03 PT |
263 | * NB by Jinshan: though field names are still _loi_, but actually |
264 | * osc_object{}s are in the list. | |
265 | */ | |
7d53d8f4 | 266 | spinlock_t cl_loi_list_lock; |
d7e09d03 PT |
267 | struct list_head cl_loi_ready_list; |
268 | struct list_head cl_loi_hp_ready_list; | |
269 | struct list_head cl_loi_write_list; | |
270 | struct list_head cl_loi_read_list; | |
271 | int cl_r_in_flight; | |
272 | int cl_w_in_flight; | |
406c1c7c | 273 | /* just a sum of the loi/lop pending numbers to be exported by sysfs */ |
d7e09d03 PT |
274 | atomic_t cl_pending_w_pages; |
275 | atomic_t cl_pending_r_pages; | |
276 | __u32 cl_max_pages_per_rpc; | |
277 | int cl_max_rpcs_in_flight; | |
278 | struct obd_histogram cl_read_rpc_hist; | |
279 | struct obd_histogram cl_write_rpc_hist; | |
280 | struct obd_histogram cl_read_page_hist; | |
281 | struct obd_histogram cl_write_page_hist; | |
282 | struct obd_histogram cl_read_offset_hist; | |
283 | struct obd_histogram cl_write_offset_hist; | |
284 | ||
285 | /* lru for osc caching pages */ | |
286 | struct cl_client_cache *cl_cache; | |
498374f7 | 287 | struct list_head cl_lru_osc; /* member of cl_cache->ccc_lru */ |
d7e09d03 PT |
288 | atomic_t *cl_lru_left; |
289 | atomic_t cl_lru_busy; | |
290 | atomic_t cl_lru_shrinkers; | |
291 | atomic_t cl_lru_in_list; | |
498374f7 | 292 | struct list_head cl_lru_list; /* lru page list */ |
7d53d8f4 | 293 | spinlock_t cl_lru_list_lock; /* page list protector */ |
dece0b95 | 294 | atomic_t cl_unstable_count; |
d7e09d03 PT |
295 | |
296 | /* number of in flight destroy rpcs is limited to max_rpcs_in_flight */ | |
297 | atomic_t cl_destroy_in_flight; | |
298 | wait_queue_head_t cl_destroy_waitq; | |
299 | ||
300 | struct mdc_rpc_lock *cl_rpc_lock; | |
301 | struct mdc_rpc_lock *cl_close_lock; | |
302 | ||
303 | /* mgc datastruct */ | |
d7e09d03 PT |
304 | atomic_t cl_mgc_refcount; |
305 | struct obd_export *cl_mgc_mgsexp; | |
306 | ||
307 | /* checksumming for data sent over the network */ | |
308 | unsigned int cl_checksum:1; /* 0 = disabled, 1 = enabled */ | |
309 | /* supported checksum types that are worked out at connect time */ | |
310 | __u32 cl_supp_cksum_types; | |
311 | /* checksum algorithm to be used */ | |
d133210f | 312 | enum cksum_type cl_cksum_type; |
d7e09d03 PT |
313 | |
314 | /* also protected by the poorly named _loi_list_lock lock above */ | |
315 | struct osc_async_rc cl_ar; | |
316 | ||
317 | /* used by quotacheck when the servers are older than 2.4 */ | |
318 | int cl_qchk_stat; /* quotacheck stat of the peer */ | |
319 | #define CL_NOT_QUOTACHECKED 1 /* client->cl_qchk_stat init value */ | |
320 | #if LUSTRE_VERSION_CODE >= OBD_OCD_VERSION(2, 7, 50, 0) | |
321 | #warning "please consider removing quotacheck compatibility code" | |
322 | #endif | |
323 | ||
324 | /* sequence manager */ | |
325 | struct lu_client_seq *cl_seq; | |
326 | ||
327 | atomic_t cl_resends; /* resend count */ | |
328 | ||
329 | /* ptlrpc work for writeback in ptlrpcd context */ | |
330 | void *cl_writeback_work; | |
2579d8d0 | 331 | void *cl_lru_work; |
d7e09d03 | 332 | /* hash tables for osc_quota_info */ |
6da6eabe | 333 | struct cfs_hash *cl_quota_hash[MAXQUOTAS]; |
d7e09d03 | 334 | }; |
c9f6bb96 | 335 | |
d7e09d03 PT |
336 | #define obd2cli_tgt(obd) ((char *)(obd)->u.cli.cl_target_uuid.uuid) |
337 | ||
338 | struct obd_id_info { | |
339 | __u32 idx; | |
21aef7d9 | 340 | u64 *data; |
d7e09d03 PT |
341 | }; |
342 | ||
d7e09d03 PT |
343 | struct echo_client_obd { |
344 | struct obd_export *ec_exp; /* the local connection to osc/lov */ | |
345 | spinlock_t ec_lock; | |
346 | struct list_head ec_objects; | |
347 | struct list_head ec_locks; | |
d7e09d03 PT |
348 | __u64 ec_unique; |
349 | }; | |
350 | ||
d7e09d03 PT |
351 | /* Generic subset of OSTs */ |
352 | struct ost_pool { | |
c56e256d | 353 | __u32 *op_array; /* array of index of lov_obd->lov_tgts */ |
d7e09d03 PT |
354 | unsigned int op_count; /* number of OSTs in the array */ |
355 | unsigned int op_size; /* allocated size of lp_array */ | |
356 | struct rw_semaphore op_rw_sem; /* to protect ost_pool use */ | |
357 | }; | |
358 | ||
d7e09d03 PT |
359 | /* allow statfs data caching for 1 second */ |
360 | #define OBD_STATFS_CACHE_SECONDS 1 | |
361 | ||
d7e09d03 PT |
362 | struct lov_tgt_desc { |
363 | struct list_head ltd_kill; | |
364 | struct obd_uuid ltd_uuid; | |
365 | struct obd_device *ltd_obd; | |
366 | struct obd_export *ltd_exp; | |
d7e09d03 PT |
367 | __u32 ltd_gen; |
368 | __u32 ltd_index; /* index in lov_obd->tgts */ | |
369 | unsigned long ltd_active:1,/* is this target up for requests */ | |
370 | ltd_activate:1,/* should target be activated */ | |
371 | ltd_reap:1; /* should this target be deleted */ | |
372 | }; | |
373 | ||
d7e09d03 PT |
374 | struct lov_obd { |
375 | struct lov_desc desc; | |
376 | struct lov_tgt_desc **lov_tgts; /* sparse array */ | |
c56e256d | 377 | struct ost_pool lov_packed; /* all OSTs in a packed array */ |
d7e09d03 PT |
378 | struct mutex lov_lock; |
379 | struct obd_connect_data lov_ocd; | |
380 | atomic_t lov_refcount; | |
d7e09d03 PT |
381 | __u32 lov_death_row;/* tgts scheduled to be deleted */ |
382 | __u32 lov_tgt_size; /* size of tgts array */ | |
383 | int lov_connects; | |
384 | int lov_pool_count; | |
6da6eabe | 385 | struct cfs_hash *lov_pools_hash_body; /* used for key access */ |
498374f7 | 386 | struct list_head lov_pool_list; /* used for sequential access */ |
61e87ab0 | 387 | struct dentry *lov_pool_debugfs_entry; |
d7e09d03 PT |
388 | enum lustre_sec_part lov_sp_me; |
389 | ||
ac5b1481 | 390 | /* Cached LRU and unstable data from upper layer */ |
1b02bde3 | 391 | struct cl_client_cache *lov_cache; |
d7e09d03 PT |
392 | |
393 | struct rw_semaphore lov_notify_lock; | |
cb1debff OD |
394 | |
395 | struct kobject *lov_tgts_kobj; | |
d7e09d03 PT |
396 | }; |
397 | ||
398 | struct lmv_tgt_desc { | |
399 | struct obd_uuid ltd_uuid; | |
400 | struct obd_export *ltd_exp; | |
401 | int ltd_idx; | |
402 | struct mutex ltd_fid_mutex; | |
403 | unsigned long ltd_active:1; /* target up for requests */ | |
404 | }; | |
405 | ||
406 | enum placement_policy { | |
407 | PLACEMENT_CHAR_POLICY = 0, | |
408 | PLACEMENT_NID_POLICY = 1, | |
409 | PLACEMENT_INVAL_POLICY = 2, | |
410 | PLACEMENT_MAX_POLICY | |
411 | }; | |
412 | ||
d7e09d03 PT |
413 | struct lmv_obd { |
414 | int refcount; | |
415 | struct lu_client_fld lmv_fld; | |
416 | spinlock_t lmv_lock; | |
fe672997 | 417 | enum placement_policy lmv_placement; |
d7e09d03 PT |
418 | struct lmv_desc desc; |
419 | struct obd_uuid cluuid; | |
420 | struct obd_export *exp; | |
421 | ||
019e9351 | 422 | struct mutex lmv_init_mutex; |
d7e09d03 PT |
423 | int connected; |
424 | int max_easize; | |
425 | int max_def_easize; | |
426 | int max_cookiesize; | |
44779340 | 427 | int max_def_cookiesize; |
d7e09d03 PT |
428 | int server_timeout; |
429 | ||
430 | int tgts_size; /* size of tgts array */ | |
431 | struct lmv_tgt_desc **tgts; | |
432 | ||
433 | struct obd_connect_data conn_data; | |
b5fa70d7 | 434 | struct kobject *lmv_tgts_kobj; |
d7e09d03 PT |
435 | }; |
436 | ||
437 | struct niobuf_local { | |
438 | __u64 lnb_file_offset; | |
439 | __u32 lnb_page_offset; | |
440 | __u32 len; | |
441 | __u32 flags; | |
442 | struct page *page; | |
443 | struct dentry *dentry; | |
d7e09d03 PT |
444 | int rc; |
445 | }; | |
446 | ||
447 | #define LUSTRE_FLD_NAME "fld" | |
448 | #define LUSTRE_SEQ_NAME "seq" | |
449 | ||
450 | #define LUSTRE_MDD_NAME "mdd" | |
451 | #define LUSTRE_OSD_LDISKFS_NAME "osd-ldiskfs" | |
452 | #define LUSTRE_OSD_ZFS_NAME "osd-zfs" | |
453 | #define LUSTRE_VVP_NAME "vvp" | |
454 | #define LUSTRE_LMV_NAME "lmv" | |
455 | #define LUSTRE_SLP_NAME "slp" | |
456 | #define LUSTRE_LOD_NAME "lod" | |
457 | #define LUSTRE_OSP_NAME "osp" | |
458 | #define LUSTRE_LWP_NAME "lwp" | |
459 | ||
460 | /* obd device type names */ | |
461 | /* FIXME all the references to LUSTRE_MDS_NAME should be swapped with LUSTRE_MDT_NAME */ | |
462 | #define LUSTRE_MDS_NAME "mds" | |
463 | #define LUSTRE_MDT_NAME "mdt" | |
464 | #define LUSTRE_MDC_NAME "mdc" | |
465 | #define LUSTRE_OSS_NAME "ost" /* FIXME change name to oss */ | |
466 | #define LUSTRE_OST_NAME "obdfilter" /* FIXME change name to ost */ | |
467 | #define LUSTRE_OSC_NAME "osc" | |
468 | #define LUSTRE_LOV_NAME "lov" | |
469 | #define LUSTRE_MGS_NAME "mgs" | |
470 | #define LUSTRE_MGC_NAME "mgc" | |
471 | ||
472 | #define LUSTRE_ECHO_NAME "obdecho" | |
473 | #define LUSTRE_ECHO_CLIENT_NAME "echo_client" | |
474 | #define LUSTRE_QMT_NAME "qmt" | |
475 | ||
476 | /* Constant obd names (post-rename) */ | |
477 | #define LUSTRE_MDS_OBDNAME "MDS" | |
478 | #define LUSTRE_OSS_OBDNAME "OSS" | |
479 | #define LUSTRE_MGS_OBDNAME "MGS" | |
480 | #define LUSTRE_MGC_OBDNAME "MGC" | |
481 | ||
d7e09d03 PT |
482 | /* Don't conflict with on-wire flags OBD_BRW_WRITE, etc */ |
483 | #define N_LOCAL_TEMP_PAGE 0x10000000 | |
484 | ||
485 | struct obd_trans_info { | |
d7e09d03 PT |
486 | __u64 oti_xid; |
487 | /* Only used on the server side for tracking acks. */ | |
488 | struct oti_req_ack_lock { | |
489 | struct lustre_handle lock; | |
490 | __u32 mode; | |
491 | } oti_ack_locks[4]; | |
492 | void *oti_handle; | |
493 | struct llog_cookie oti_onecookie; | |
494 | struct llog_cookie *oti_logcookies; | |
d7e09d03 | 495 | |
d7e09d03 PT |
496 | /** VBR: versions */ |
497 | __u64 oti_pre_version; | |
d7e09d03 PT |
498 | }; |
499 | ||
d7e09d03 PT |
500 | /* |
501 | * Events signalled through obd_notify() upcall-chain. | |
502 | */ | |
503 | enum obd_notify_event { | |
504 | /* target added */ | |
505 | OBD_NOTIFY_CREATE, | |
506 | /* Device connect start */ | |
507 | OBD_NOTIFY_CONNECT, | |
508 | /* Device activated */ | |
509 | OBD_NOTIFY_ACTIVE, | |
510 | /* Device deactivated */ | |
511 | OBD_NOTIFY_INACTIVE, | |
512 | /* Device disconnected */ | |
513 | OBD_NOTIFY_DISCON, | |
514 | /* Connect data for import were changed */ | |
515 | OBD_NOTIFY_OCD, | |
516 | /* Sync request */ | |
517 | OBD_NOTIFY_SYNC_NONBLOCK, | |
518 | OBD_NOTIFY_SYNC, | |
519 | /* Configuration event */ | |
520 | OBD_NOTIFY_CONFIG, | |
521 | /* Administratively deactivate/activate event */ | |
522 | OBD_NOTIFY_DEACTIVATE, | |
523 | OBD_NOTIFY_ACTIVATE | |
524 | }; | |
525 | ||
d7e09d03 PT |
526 | /* |
527 | * Data structure used to pass obd_notify()-event to non-obd listeners (llite | |
85f552df | 528 | * being main example). |
d7e09d03 PT |
529 | */ |
530 | struct obd_notify_upcall { | |
531 | int (*onu_upcall)(struct obd_device *host, struct obd_device *watched, | |
532 | enum obd_notify_event ev, void *owner, void *data); | |
533 | /* Opaque datum supplied by upper layer listener */ | |
534 | void *onu_owner; | |
535 | }; | |
536 | ||
537 | struct target_recovery_data { | |
538 | svc_handler_t trd_recovery_handler; | |
539 | pid_t trd_processing_task; | |
540 | struct completion trd_starting; | |
541 | struct completion trd_finishing; | |
542 | }; | |
543 | ||
544 | struct obd_llog_group { | |
d7e09d03 PT |
545 | struct llog_ctxt *olg_ctxts[LLOG_MAX_CTXTS]; |
546 | wait_queue_head_t olg_waitq; | |
547 | spinlock_t olg_lock; | |
548 | struct mutex olg_cat_processing; | |
549 | }; | |
550 | ||
551 | /* corresponds to one of the obd's */ | |
552 | #define OBD_DEVICE_MAGIC 0XAB5CD6EF | |
d7e09d03 | 553 | |
91b3aaf9 JH |
554 | struct lvfs_run_ctxt { |
555 | struct dt_device *dt; | |
556 | }; | |
557 | ||
d7e09d03 PT |
558 | struct obd_device { |
559 | struct obd_type *obd_type; | |
560 | __u32 obd_magic; | |
561 | ||
562 | /* common and UUID name of this device */ | |
563 | char obd_name[MAX_OBD_NAME]; | |
564 | struct obd_uuid obd_uuid; | |
565 | ||
566 | struct lu_device *obd_lu_dev; | |
567 | ||
568 | int obd_minor; | |
569 | /* bitfield modification is protected by obd_dev_lock */ | |
570 | unsigned long obd_attached:1, /* finished attach */ | |
571 | obd_set_up:1, /* finished setup */ | |
d7e09d03 | 572 | obd_version_recov:1, /* obd uses version checking */ |
c56e256d OD |
573 | obd_replayable:1,/* recovery is enabled; inform clients */ |
574 | obd_no_transno:1, /* no committed-transno notification */ | |
d7e09d03 PT |
575 | obd_no_recov:1, /* fail instead of retry messages */ |
576 | obd_stopping:1, /* started cleanup */ | |
577 | obd_starting:1, /* started setup */ | |
578 | obd_force:1, /* cleanup with > 0 obd refcount */ | |
c56e256d | 579 | obd_fail:1, /* cleanup with failover */ |
d7e09d03 PT |
580 | obd_no_conn:1, /* deny new connections */ |
581 | obd_inactive:1, /* device active/inactive | |
406c1c7c OD |
582 | * (for sysfs status only!!) |
583 | */ | |
d7e09d03 PT |
584 | obd_no_ir:1, /* no imperative recovery. */ |
585 | obd_process_conf:1; /* device is processing mgs config */ | |
586 | /* use separate field as it is set in interrupt to don't mess with | |
c56e256d OD |
587 | * protection of other bits using _bh lock |
588 | */ | |
d7e09d03 PT |
589 | unsigned long obd_recovery_expired:1; |
590 | /* uuid-export hash body */ | |
6da6eabe | 591 | struct cfs_hash *obd_uuid_hash; |
d7e09d03 PT |
592 | atomic_t obd_refcount; |
593 | wait_queue_head_t obd_refcount_waitq; | |
594 | struct list_head obd_exports; | |
595 | struct list_head obd_unlinked_exports; | |
596 | struct list_head obd_delayed_exports; | |
597 | int obd_num_exports; | |
598 | spinlock_t obd_nid_lock; | |
599 | struct ldlm_namespace *obd_namespace; | |
600 | struct ptlrpc_client obd_ldlm_client; /* XXX OST/MDS only */ | |
601 | /* a spinlock is OK for what we do now, may need a semaphore later */ | |
602 | spinlock_t obd_dev_lock; /* protect OBD bitfield above */ | |
603 | struct mutex obd_dev_mutex; | |
604 | __u64 obd_last_committed; | |
d7e09d03 PT |
605 | spinlock_t obd_osfs_lock; |
606 | struct obd_statfs obd_osfs; /* locked by obd_osfs_lock */ | |
607 | __u64 obd_osfs_age; | |
608 | struct lvfs_run_ctxt obd_lvfs_ctxt; | |
609 | struct obd_llog_group obd_olg; /* default llog group */ | |
610 | struct obd_device *obd_observer; | |
611 | struct rw_semaphore obd_observer_link_sem; | |
612 | struct obd_notify_upcall obd_upcall; | |
613 | struct obd_export *obd_self_export; | |
d7e09d03 | 614 | |
d7e09d03 | 615 | union { |
d7e09d03 | 616 | struct client_obd cli; |
d7e09d03 | 617 | struct echo_client_obd echo_client; |
d7e09d03 PT |
618 | struct lov_obd lov; |
619 | struct lmv_obd lmv; | |
620 | } u; | |
621 | /* Fields used by LProcFS */ | |
622 | unsigned int obd_cntr_base; | |
623 | struct lprocfs_stats *obd_stats; | |
624 | ||
625 | unsigned int md_cntr_base; | |
626 | struct lprocfs_stats *md_stats; | |
627 | ||
61e87ab0 | 628 | struct dentry *obd_debugfs_entry; |
61e87ab0 | 629 | struct dentry *obd_svc_debugfs_entry; |
d7e09d03 PT |
630 | struct lprocfs_stats *obd_svc_stats; |
631 | atomic_t obd_evict_inprogress; | |
632 | wait_queue_head_t obd_evict_inprogress_waitq; | |
498374f7 | 633 | struct list_head obd_evict_list; /* protected with pet_lock */ |
d7e09d03 PT |
634 | |
635 | /** | |
636 | * Ldlm pool part. Save last calculated SLV and Limit. | |
637 | */ | |
638 | rwlock_t obd_pool_lock; | |
639 | int obd_pool_limit; | |
640 | __u64 obd_pool_slv; | |
641 | ||
642 | /** | |
643 | * A list of outstanding class_incref()'s against this obd. For | |
644 | * debugging. | |
645 | */ | |
646 | struct lu_ref obd_reference; | |
647 | ||
648 | int obd_conn_inprogress; | |
9b801302 OD |
649 | |
650 | struct kobject obd_kobj; /* sysfs object */ | |
651 | struct completion obd_kobj_unregister; | |
d7e09d03 PT |
652 | }; |
653 | ||
d7e09d03 PT |
654 | enum obd_cleanup_stage { |
655 | /* Special case hack for MDS LOVs */ | |
656 | OBD_CLEANUP_EARLY, | |
657 | /* can be directly mapped to .ldto_device_fini() */ | |
658 | OBD_CLEANUP_EXPORTS, | |
659 | }; | |
660 | ||
661 | /* get/set_info keys */ | |
662 | #define KEY_ASYNC "async" | |
d7e09d03 PT |
663 | #define KEY_CHANGELOG_CLEAR "changelog_clear" |
664 | #define KEY_FID2PATH "fid2path" | |
665 | #define KEY_CHECKSUM "checksum" | |
666 | #define KEY_CLEAR_FS "clear_fs" | |
667 | #define KEY_CONN_DATA "conn_data" | |
668 | #define KEY_EVICT_BY_NID "evict_by_nid" | |
669 | #define KEY_FIEMAP "fiemap" | |
670 | #define KEY_FLUSH_CTX "flush_ctx" | |
671 | #define KEY_GRANT_SHRINK "grant_shrink" | |
672 | #define KEY_HSM_COPYTOOL_SEND "hsm_send" | |
673 | #define KEY_INIT_RECOV_BACKUP "init_recov_bk" | |
d7e09d03 PT |
674 | #define KEY_INTERMDS "inter_mds" |
675 | #define KEY_LAST_ID "last_id" | |
676 | #define KEY_LAST_FID "last_fid" | |
677 | #define KEY_LOCK_TO_STRIPE "lock_to_stripe" | |
678 | #define KEY_LOVDESC "lovdesc" | |
44779340 BB |
679 | #define KEY_MAX_EASIZE "max_easize" |
680 | #define KEY_DEFAULT_EASIZE "default_easize" | |
d7e09d03 PT |
681 | #define KEY_MDS_CONN "mds_conn" |
682 | #define KEY_MGSSEC "mgssec" | |
683 | #define KEY_NEXT_ID "next_id" | |
684 | #define KEY_READ_ONLY "read-only" | |
685 | #define KEY_REGISTER_TARGET "register_target" | |
686 | #define KEY_SET_FS "set_fs" | |
687 | #define KEY_TGT_COUNT "tgt_count" | |
688 | /* KEY_SET_INFO in lustre_idl.h */ | |
689 | #define KEY_SPTLRPC_CONF "sptlrpc_conf" | |
690 | #define KEY_CONNECT_FLAG "connect_flags" | |
d7e09d03 PT |
691 | |
692 | #define KEY_CACHE_SET "cache_set" | |
693 | #define KEY_CACHE_LRU_SHRINK "cache_lru_shrink" | |
d7e09d03 PT |
694 | |
695 | struct lu_context; | |
696 | ||
697 | /* /!\ must be coherent with include/linux/namei.h on patched kernel */ | |
698 | #define IT_OPEN (1 << 0) | |
699 | #define IT_CREAT (1 << 1) | |
700 | #define IT_READDIR (1 << 2) | |
701 | #define IT_GETATTR (1 << 3) | |
702 | #define IT_LOOKUP (1 << 4) | |
703 | #define IT_UNLINK (1 << 5) | |
704 | #define IT_TRUNC (1 << 6) | |
705 | #define IT_GETXATTR (1 << 7) | |
706 | #define IT_EXEC (1 << 8) | |
707 | #define IT_PIN (1 << 9) | |
708 | #define IT_LAYOUT (1 << 10) | |
709 | #define IT_QUOTA_DQACQ (1 << 11) | |
710 | #define IT_QUOTA_CONN (1 << 12) | |
7fc1f831 | 711 | #define IT_SETXATTR (1 << 13) |
d7e09d03 PT |
712 | |
713 | static inline int it_to_lock_mode(struct lookup_intent *it) | |
714 | { | |
715 | /* CREAT needs to be tested before open (both could be set) */ | |
716 | if (it->it_op & IT_CREAT) | |
717 | return LCK_CW; | |
34a60457 | 718 | else if (it->it_op & (IT_GETATTR | IT_OPEN | IT_LOOKUP | |
d7e09d03 PT |
719 | IT_LAYOUT)) |
720 | return LCK_CR; | |
34a60457 | 721 | else if (it->it_op & IT_READDIR) |
722 | return LCK_PR; | |
7fc1f831 AP |
723 | else if (it->it_op & IT_GETXATTR) |
724 | return LCK_PR; | |
725 | else if (it->it_op & IT_SETXATTR) | |
726 | return LCK_PW; | |
d7e09d03 PT |
727 | |
728 | LASSERTF(0, "Invalid it_op: %d\n", it->it_op); | |
729 | return -EINVAL; | |
730 | } | |
731 | ||
732 | struct md_op_data { | |
17891183 MI |
733 | struct lu_fid op_fid1; /* operation fid1 (usually parent) */ |
734 | struct lu_fid op_fid2; /* operation fid2 (usually child) */ | |
d7e09d03 PT |
735 | struct lu_fid op_fid3; /* 2 extra fids to find conflicting */ |
736 | struct lu_fid op_fid4; /* to the operation locks. */ | |
114acca8 | 737 | u32 op_mds; /* what mds server open will go to */ |
d7e09d03 | 738 | struct lustre_handle op_handle; |
21aef7d9 | 739 | s64 op_mod_time; |
d7e09d03 PT |
740 | const char *op_name; |
741 | int op_namelen; | |
742 | __u32 op_mode; | |
743 | struct lmv_stripe_md *op_mea1; | |
744 | struct lmv_stripe_md *op_mea2; | |
745 | __u32 op_suppgids[2]; | |
746 | __u32 op_fsuid; | |
747 | __u32 op_fsgid; | |
748 | cfs_cap_t op_cap; | |
749 | void *op_data; | |
70a251f6 | 750 | size_t op_data_size; |
d7e09d03 PT |
751 | |
752 | /* iattr fields and blocks. */ | |
753 | struct iattr op_attr; | |
754 | unsigned int op_attr_flags; | |
755 | __u64 op_valid; | |
756 | loff_t op_attr_blocks; | |
757 | ||
758 | /* Size-on-MDS epoch and flags. */ | |
759 | __u64 op_ioepoch; | |
760 | __u32 op_flags; | |
761 | ||
d7e09d03 | 762 | /* Various operation flags. */ |
48d23e61 | 763 | enum mds_op_bias op_bias; |
d7e09d03 | 764 | |
d7e09d03 PT |
765 | /* Used by readdir */ |
766 | __u64 op_offset; | |
767 | ||
768 | /* Used by readdir */ | |
769 | __u32 op_npages; | |
770 | ||
771 | /* used to transfer info between the stacks of MD client | |
c56e256d OD |
772 | * see enum op_cli_flags |
773 | */ | |
d7e09d03 | 774 | __u32 op_cli_flags; |
48d23e61 JX |
775 | |
776 | /* File object data version for HSM release, on client */ | |
777 | __u64 op_data_version; | |
778 | struct lustre_handle op_lease_handle; | |
d7e09d03 PT |
779 | }; |
780 | ||
049e215e | 781 | #define op_stripe_offset op_ioepoch |
782 | #define op_max_pages op_valid | |
783 | ||
784 | struct md_callback { | |
785 | int (*md_blocking_ast)(struct ldlm_lock *lock, | |
786 | struct ldlm_lock_desc *desc, | |
787 | void *data, int flag); | |
788 | }; | |
789 | ||
d7e09d03 PT |
790 | enum op_cli_flags { |
791 | CLI_SET_MEA = 1 << 0, | |
792 | CLI_RM_ENTRY = 1 << 1, | |
c08db984 | 793 | CLI_HASH64 = BIT(2), |
794 | CLI_API32 = BIT(3), | |
79496845 | 795 | CLI_MIGRATE = BIT(4), |
d7e09d03 PT |
796 | }; |
797 | ||
798 | struct md_enqueue_info; | |
799 | /* metadata stat-ahead */ | |
d7e09d03 | 800 | |
d7e09d03 PT |
801 | struct md_enqueue_info { |
802 | struct md_op_data mi_data; | |
803 | struct lookup_intent mi_it; | |
804 | struct lustre_handle mi_lockh; | |
805 | struct inode *mi_dir; | |
fe672997 | 806 | int (*mi_cb)(struct ptlrpc_request *req, |
10457d4b | 807 | struct md_enqueue_info *minfo, int rc); |
d7e09d03 PT |
808 | __u64 mi_cbdata; |
809 | unsigned int mi_generation; | |
810 | }; | |
811 | ||
812 | struct obd_ops { | |
a13b1f32 DC |
813 | struct module *owner; |
814 | int (*iocontrol)(unsigned int cmd, struct obd_export *exp, int len, | |
e09bee34 | 815 | void *karg, void __user *uarg); |
a13b1f32 DC |
816 | int (*get_info)(const struct lu_env *env, struct obd_export *, |
817 | __u32 keylen, void *key, __u32 *vallen, void *val, | |
818 | struct lov_stripe_md *lsm); | |
819 | int (*set_info_async)(const struct lu_env *, struct obd_export *, | |
820 | __u32 keylen, void *key, | |
821 | __u32 vallen, void *val, | |
822 | struct ptlrpc_request_set *set); | |
a13b1f32 DC |
823 | int (*setup)(struct obd_device *dev, struct lustre_cfg *cfg); |
824 | int (*precleanup)(struct obd_device *dev, | |
825 | enum obd_cleanup_stage cleanup_stage); | |
826 | int (*cleanup)(struct obd_device *dev); | |
827 | int (*process_config)(struct obd_device *dev, u32 len, void *data); | |
828 | int (*postrecov)(struct obd_device *dev); | |
829 | int (*add_conn)(struct obd_import *imp, struct obd_uuid *uuid, | |
830 | int priority); | |
831 | int (*del_conn)(struct obd_import *imp, struct obd_uuid *uuid); | |
d7e09d03 PT |
832 | /* connect to the target device with given connection |
833 | * data. @ocd->ocd_connect_flags is modified to reflect flags actually | |
834 | * granted by the target, which are guaranteed to be a subset of flags | |
c56e256d OD |
835 | * asked for. If @ocd == NULL, use default parameters. |
836 | */ | |
a13b1f32 DC |
837 | int (*connect)(const struct lu_env *env, |
838 | struct obd_export **exp, struct obd_device *src, | |
839 | struct obd_uuid *cluuid, struct obd_connect_data *ocd, | |
840 | void *localdata); | |
841 | int (*reconnect)(const struct lu_env *env, | |
842 | struct obd_export *exp, struct obd_device *src, | |
843 | struct obd_uuid *cluuid, | |
844 | struct obd_connect_data *ocd, | |
d7e09d03 | 845 | void *localdata); |
a13b1f32 | 846 | int (*disconnect)(struct obd_export *exp); |
d7e09d03 PT |
847 | |
848 | /* Initialize/finalize fids infrastructure. */ | |
a13b1f32 DC |
849 | int (*fid_init)(struct obd_device *obd, |
850 | struct obd_export *exp, enum lu_cli_type type); | |
851 | int (*fid_fini)(struct obd_device *obd); | |
d7e09d03 PT |
852 | |
853 | /* Allocate new fid according to passed @hint. */ | |
8f18c8a4 | 854 | int (*fid_alloc)(const struct lu_env *env, struct obd_export *exp, |
855 | struct lu_fid *fid, struct md_op_data *op_data); | |
d7e09d03 PT |
856 | |
857 | /* | |
858 | * Object with @fid is getting deleted, we may want to do something | |
859 | * about this. | |
860 | */ | |
a13b1f32 DC |
861 | int (*statfs)(const struct lu_env *, struct obd_export *exp, |
862 | struct obd_statfs *osfs, __u64 max_age, __u32 flags); | |
863 | int (*statfs_async)(struct obd_export *exp, struct obd_info *oinfo, | |
864 | __u64 max_age, struct ptlrpc_request_set *set); | |
865 | int (*packmd)(struct obd_export *exp, struct lov_mds_md **disk_tgt, | |
866 | struct lov_stripe_md *mem_src); | |
867 | int (*unpackmd)(struct obd_export *exp, | |
868 | struct lov_stripe_md **mem_tgt, | |
869 | struct lov_mds_md *disk_src, int disk_len); | |
870 | int (*preallocate)(struct lustre_handle *, u32 *req, u64 *ids); | |
871 | int (*create)(const struct lu_env *env, struct obd_export *exp, | |
872 | struct obdo *oa, struct lov_stripe_md **ea, | |
873 | struct obd_trans_info *oti); | |
874 | int (*destroy)(const struct lu_env *env, struct obd_export *exp, | |
875 | struct obdo *oa, struct lov_stripe_md *ea, | |
876 | struct obd_trans_info *oti, struct obd_export *md_exp); | |
877 | int (*setattr)(const struct lu_env *, struct obd_export *exp, | |
878 | struct obd_info *oinfo, struct obd_trans_info *oti); | |
879 | int (*setattr_async)(struct obd_export *exp, struct obd_info *oinfo, | |
880 | struct obd_trans_info *oti, | |
881 | struct ptlrpc_request_set *rqset); | |
882 | int (*getattr)(const struct lu_env *env, struct obd_export *exp, | |
883 | struct obd_info *oinfo); | |
884 | int (*getattr_async)(struct obd_export *exp, struct obd_info *oinfo, | |
885 | struct ptlrpc_request_set *set); | |
886 | int (*adjust_kms)(struct obd_export *exp, struct lov_stripe_md *lsm, | |
887 | u64 size, int shrink); | |
888 | int (*preprw)(const struct lu_env *env, int cmd, | |
889 | struct obd_export *exp, struct obdo *oa, int objcount, | |
890 | struct obd_ioobj *obj, struct niobuf_remote *remote, | |
891 | int *nr_pages, struct niobuf_local *local, | |
892 | struct obd_trans_info *oti); | |
893 | int (*commitrw)(const struct lu_env *env, int cmd, | |
894 | struct obd_export *exp, struct obdo *oa, | |
895 | int objcount, struct obd_ioobj *obj, | |
896 | struct niobuf_remote *remote, int pages, | |
897 | struct niobuf_local *local, | |
898 | struct obd_trans_info *oti, int rc); | |
a13b1f32 DC |
899 | int (*init_export)(struct obd_export *exp); |
900 | int (*destroy_export)(struct obd_export *exp); | |
d7e09d03 PT |
901 | |
902 | /* metadata-only methods */ | |
a13b1f32 DC |
903 | int (*import_event)(struct obd_device *, struct obd_import *, |
904 | enum obd_import_event); | |
d7e09d03 | 905 | |
a13b1f32 DC |
906 | int (*notify)(struct obd_device *obd, struct obd_device *watched, |
907 | enum obd_notify_event ev, void *data); | |
d7e09d03 | 908 | |
a13b1f32 DC |
909 | int (*health_check)(const struct lu_env *env, struct obd_device *); |
910 | struct obd_uuid *(*get_uuid)(struct obd_export *exp); | |
d7e09d03 PT |
911 | |
912 | /* quota methods */ | |
a13b1f32 | 913 | int (*quotacheck)(struct obd_device *, struct obd_export *, |
d7e09d03 | 914 | struct obd_quotactl *); |
a13b1f32 DC |
915 | int (*quotactl)(struct obd_device *, struct obd_export *, |
916 | struct obd_quotactl *); | |
d7e09d03 | 917 | |
d7e09d03 | 918 | /* pools methods */ |
a13b1f32 DC |
919 | int (*pool_new)(struct obd_device *obd, char *poolname); |
920 | int (*pool_del)(struct obd_device *obd, char *poolname); | |
921 | int (*pool_add)(struct obd_device *obd, char *poolname, | |
922 | char *ostname); | |
923 | int (*pool_rem)(struct obd_device *obd, char *poolname, | |
924 | char *ostname); | |
925 | void (*getref)(struct obd_device *obd); | |
926 | void (*putref)(struct obd_device *obd); | |
d7e09d03 PT |
927 | /* |
928 | * NOTE: If adding ops, add another LPROCFS_OBD_OP_INIT() line | |
929 | * to lprocfs_alloc_obd_stats() in obdclass/lprocfs_status.c. | |
c56e256d OD |
930 | * Also, add a wrapper function in include/linux/obd_class.h. |
931 | */ | |
d7e09d03 PT |
932 | }; |
933 | ||
d7e09d03 | 934 | /* lmv structures */ |
d7e09d03 PT |
935 | struct lustre_md { |
936 | struct mdt_body *body; | |
937 | struct lov_stripe_md *lsm; | |
f6718d1d | 938 | struct lmv_stripe_md *lmv; |
d7e09d03 PT |
939 | #ifdef CONFIG_FS_POSIX_ACL |
940 | struct posix_acl *posix_acl; | |
941 | #endif | |
942 | struct mdt_remote_perm *remote_perm; | |
d7e09d03 PT |
943 | }; |
944 | ||
945 | struct md_open_data { | |
946 | struct obd_client_handle *mod_och; | |
947 | struct ptlrpc_request *mod_open_req; | |
948 | struct ptlrpc_request *mod_close_req; | |
63d42578 HZ |
949 | atomic_t mod_refcount; |
950 | bool mod_is_create; | |
d7e09d03 PT |
951 | }; |
952 | ||
953 | struct lookup_intent; | |
2de35386 | 954 | struct cl_attr; |
d7e09d03 PT |
955 | |
956 | struct md_ops { | |
df18a80a DC |
957 | int (*getstatus)(struct obd_export *, struct lu_fid *); |
958 | int (*null_inode)(struct obd_export *, const struct lu_fid *); | |
df18a80a DC |
959 | int (*close)(struct obd_export *, struct md_op_data *, |
960 | struct md_open_data *, struct ptlrpc_request **); | |
961 | int (*create)(struct obd_export *, struct md_op_data *, | |
962 | const void *, int, int, __u32, __u32, cfs_cap_t, | |
963 | __u64, struct ptlrpc_request **); | |
964 | int (*done_writing)(struct obd_export *, struct md_op_data *, | |
965 | struct md_open_data *); | |
966 | int (*enqueue)(struct obd_export *, struct ldlm_enqueue_info *, | |
70a251f6 | 967 | const ldlm_policy_data_t *, |
df18a80a | 968 | struct lookup_intent *, struct md_op_data *, |
70a251f6 | 969 | struct lustre_handle *, __u64); |
df18a80a DC |
970 | int (*getattr)(struct obd_export *, struct md_op_data *, |
971 | struct ptlrpc_request **); | |
972 | int (*getattr_name)(struct obd_export *, struct md_op_data *, | |
973 | struct ptlrpc_request **); | |
974 | int (*intent_lock)(struct obd_export *, struct md_op_data *, | |
70a251f6 | 975 | struct lookup_intent *, |
df18a80a DC |
976 | struct ptlrpc_request **, |
977 | ldlm_blocking_callback, __u64); | |
978 | int (*link)(struct obd_export *, struct md_op_data *, | |
979 | struct ptlrpc_request **); | |
980 | int (*rename)(struct obd_export *, struct md_op_data *, | |
981 | const char *, int, const char *, int, | |
d7e09d03 | 982 | struct ptlrpc_request **); |
df18a80a DC |
983 | int (*setattr)(struct obd_export *, struct md_op_data *, void *, |
984 | int, void *, int, struct ptlrpc_request **, | |
d7e09d03 | 985 | struct md_open_data **mod); |
df18a80a DC |
986 | int (*sync)(struct obd_export *, const struct lu_fid *, |
987 | struct ptlrpc_request **); | |
049e215e | 988 | int (*read_page)(struct obd_export *, struct md_op_data *, |
989 | struct md_callback *cb_op, __u64 hash_offset, | |
990 | struct page **ppage); | |
df18a80a | 991 | int (*unlink)(struct obd_export *, struct md_op_data *, |
ef2e0f55 | 992 | struct ptlrpc_request **); |
d7e09d03 | 993 | |
df18a80a DC |
994 | int (*setxattr)(struct obd_export *, const struct lu_fid *, |
995 | u64, const char *, const char *, int, int, int, __u32, | |
d7e09d03 PT |
996 | struct ptlrpc_request **); |
997 | ||
df18a80a DC |
998 | int (*getxattr)(struct obd_export *, const struct lu_fid *, |
999 | u64, const char *, const char *, int, int, int, | |
1000 | struct ptlrpc_request **); | |
d7e09d03 | 1001 | |
df18a80a | 1002 | int (*init_ea_size)(struct obd_export *, int, int, int, int); |
d7e09d03 | 1003 | |
df18a80a DC |
1004 | int (*get_lustre_md)(struct obd_export *, struct ptlrpc_request *, |
1005 | struct obd_export *, struct obd_export *, | |
1006 | struct lustre_md *); | |
d7e09d03 | 1007 | |
df18a80a | 1008 | int (*free_lustre_md)(struct obd_export *, struct lustre_md *); |
d7e09d03 | 1009 | |
2de35386 | 1010 | int (*merge_attr)(struct obd_export *, |
1011 | const struct lmv_stripe_md *lsm, | |
1012 | struct cl_attr *attr); | |
1013 | ||
1014 | int (*update_lsm_md)(struct obd_export *, struct lmv_stripe_md *lsm, | |
1015 | struct mdt_body *, ldlm_blocking_callback); | |
1016 | ||
df18a80a DC |
1017 | int (*set_open_replay_data)(struct obd_export *, |
1018 | struct obd_client_handle *, | |
1019 | struct lookup_intent *); | |
1020 | int (*clear_open_replay_data)(struct obd_export *, | |
1021 | struct obd_client_handle *); | |
bc30c172 JH |
1022 | int (*set_lock_data)(struct obd_export *, const struct lustre_handle *, |
1023 | void *, __u64 *); | |
d7e09d03 | 1024 | |
52ee0d20 OD |
1025 | enum ldlm_mode (*lock_match)(struct obd_export *, __u64, |
1026 | const struct lu_fid *, enum ldlm_type, | |
1027 | ldlm_policy_data_t *, enum ldlm_mode, | |
1028 | struct lustre_handle *); | |
d7e09d03 | 1029 | |
df18a80a | 1030 | int (*cancel_unused)(struct obd_export *, const struct lu_fid *, |
52ee0d20 | 1031 | ldlm_policy_data_t *, enum ldlm_mode, |
f833ee42 | 1032 | enum ldlm_cancel_flags flags, void *opaque); |
d7e09d03 | 1033 | |
a609c393 | 1034 | int (*get_fid_from_lsm)(struct obd_export *, |
1035 | const struct lmv_stripe_md *, | |
1036 | const char *name, int namelen, | |
1037 | struct lu_fid *fid); | |
1038 | ||
df18a80a DC |
1039 | int (*intent_getattr_async)(struct obd_export *, |
1040 | struct md_enqueue_info *, | |
1041 | struct ldlm_enqueue_info *); | |
d7e09d03 | 1042 | |
df18a80a DC |
1043 | int (*revalidate_lock)(struct obd_export *, struct lookup_intent *, |
1044 | struct lu_fid *, __u64 *bits); | |
d7e09d03 PT |
1045 | |
1046 | /* | |
1047 | * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to | |
1048 | * lprocfs_alloc_md_stats() in obdclass/lprocfs_status.c. Also, add a | |
1049 | * wrapper function in include/linux/obd_class.h. | |
1050 | */ | |
1051 | }; | |
1052 | ||
1053 | struct lsm_operations { | |
1054 | void (*lsm_free)(struct lov_stripe_md *); | |
1055 | int (*lsm_destroy)(struct lov_stripe_md *, struct obdo *oa, | |
1056 | struct obd_export *md_exp); | |
21aef7d9 OD |
1057 | void (*lsm_stripe_by_index)(struct lov_stripe_md *, int *, u64 *, |
1058 | u64 *); | |
1059 | void (*lsm_stripe_by_offset)(struct lov_stripe_md *, int *, u64 *, | |
1060 | u64 *); | |
b037976f | 1061 | int (*lsm_lmm_verify)(struct lov_mds_md *lmm, int lmm_bytes, |
10457d4b | 1062 | __u16 *stripe_count); |
b037976f | 1063 | int (*lsm_unpackmd)(struct lov_obd *lov, struct lov_stripe_md *lsm, |
10457d4b | 1064 | struct lov_mds_md *lmm); |
d7e09d03 PT |
1065 | }; |
1066 | ||
1067 | extern const struct lsm_operations lsm_v1_ops; | |
1068 | extern const struct lsm_operations lsm_v3_ops; | |
1069 | static inline const struct lsm_operations *lsm_op_find(int magic) | |
1070 | { | |
a58a38ac | 1071 | switch (magic) { |
d7e09d03 PT |
1072 | case LOV_MAGIC_V1: |
1073 | return &lsm_v1_ops; | |
1074 | case LOV_MAGIC_V3: | |
1075 | return &lsm_v3_ops; | |
1076 | default: | |
1077 | CERROR("Cannot recognize lsm_magic %08x\n", magic); | |
1078 | return NULL; | |
1079 | } | |
1080 | } | |
1081 | ||
d7e09d03 PT |
1082 | static inline struct md_open_data *obd_mod_alloc(void) |
1083 | { | |
1084 | struct md_open_data *mod; | |
498374f7 | 1085 | |
af13af52 | 1086 | mod = kzalloc(sizeof(*mod), GFP_NOFS); |
d2a13989 | 1087 | if (!mod) |
d7e09d03 PT |
1088 | return NULL; |
1089 | atomic_set(&mod->mod_refcount, 1); | |
1090 | return mod; | |
1091 | } | |
1092 | ||
1093 | #define obd_mod_get(mod) atomic_inc(&(mod)->mod_refcount) | |
1094 | #define obd_mod_put(mod) \ | |
1095 | ({ \ | |
1096 | if (atomic_dec_and_test(&(mod)->mod_refcount)) { \ | |
1097 | if ((mod)->mod_open_req) \ | |
1098 | ptlrpc_req_finished((mod)->mod_open_req); \ | |
af13af52 | 1099 | kfree(mod); \ |
d7e09d03 PT |
1100 | } \ |
1101 | }) | |
1102 | ||
21aef7d9 | 1103 | void obdo_from_inode(struct obdo *dst, struct inode *src, u32 valid); |
d7e09d03 PT |
1104 | void obdo_set_parent_fid(struct obdo *dst, const struct lu_fid *parent); |
1105 | ||
1106 | /* return 1 if client should be resend request */ | |
1107 | static inline int client_should_resend(int resend, struct client_obd *cli) | |
1108 | { | |
1109 | return atomic_read(&cli->cl_resends) ? | |
1110 | atomic_read(&cli->cl_resends) > resend : 1; | |
1111 | } | |
1112 | ||
1113 | /** | |
1114 | * Return device name for this device | |
1115 | * | |
1116 | * XXX: lu_device is declared before obd_device, while a pointer pointing | |
1117 | * back to obd_device in lu_device, so this helper function defines here | |
1118 | * instead of in lu_object.h | |
1119 | */ | |
1120 | static inline const char *lu_dev_name(const struct lu_device *lu_dev) | |
1121 | { | |
1122 | return lu_dev->ld_obd->obd_name; | |
1123 | } | |
1124 | ||
1125 | static inline bool filename_is_volatile(const char *name, int namelen, int *idx) | |
1126 | { | |
1127 | const char *start; | |
1128 | char *end; | |
1129 | ||
1130 | if (strncmp(name, LUSTRE_VOLATILE_HDR, LUSTRE_VOLATILE_HDR_LEN) != 0) | |
1131 | return false; | |
1132 | ||
1133 | /* caller does not care of idx */ | |
d2a13989 | 1134 | if (!idx) |
d7e09d03 PT |
1135 | return true; |
1136 | ||
1137 | /* volatile file, the MDT can be set from name */ | |
1138 | /* name format is LUSTRE_VOLATILE_HDR:[idx]: */ | |
1139 | /* if no MDT is specified, use std way */ | |
1140 | if (namelen < LUSTRE_VOLATILE_HDR_LEN + 2) | |
1141 | goto bad_format; | |
1142 | /* test for no MDT idx case */ | |
1143 | if ((*(name + LUSTRE_VOLATILE_HDR_LEN) == ':') && | |
1144 | (*(name + LUSTRE_VOLATILE_HDR_LEN + 1) == ':')) { | |
1145 | *idx = -1; | |
1146 | return true; | |
1147 | } | |
1148 | /* we have an idx, read it */ | |
1149 | start = name + LUSTRE_VOLATILE_HDR_LEN + 1; | |
4ad2adc9 | 1150 | *idx = simple_strtoul(start, &end, 0); |
d7e09d03 PT |
1151 | /* error cases: |
1152 | * no digit, no trailing :, negative value | |
1153 | */ | |
1154 | if (((*idx == 0) && (end == start)) || | |
1155 | (*end != ':') || (*idx < 0)) | |
1156 | goto bad_format; | |
1157 | ||
1158 | return true; | |
1159 | bad_format: | |
1160 | /* bad format of mdt idx, we cannot return an error | |
c56e256d OD |
1161 | * to caller so we use hash algo |
1162 | */ | |
d7e09d03 PT |
1163 | CERROR("Bad volatile file name format: %s\n", |
1164 | name + LUSTRE_VOLATILE_HDR_LEN); | |
1165 | return false; | |
1166 | } | |
1167 | ||
1168 | static inline int cli_brw_size(struct obd_device *obd) | |
1169 | { | |
09cbfeaf | 1170 | return obd->u.cli.cl_max_pages_per_rpc << PAGE_SHIFT; |
d7e09d03 PT |
1171 | } |
1172 | ||
3147b268 HZ |
1173 | /* |
1174 | * when RPC size or the max RPCs in flight is increased, the max dirty pages | |
1175 | * of the client should be increased accordingly to avoid sending fragmented | |
1176 | * RPCs over the network when the client runs out of the maximum dirty space | |
1177 | * when so many RPCs are being generated. | |
1178 | */ | |
1179 | static inline void client_adjust_max_dirty(struct client_obd *cli) | |
1180 | { | |
1181 | /* initializing */ | |
1182 | if (cli->cl_dirty_max_pages <= 0) | |
1183 | cli->cl_dirty_max_pages = | |
1184 | (OSC_MAX_DIRTY_DEFAULT * 1024 * 1024) >> PAGE_SHIFT; | |
1185 | else { | |
1186 | long dirty_max = cli->cl_max_rpcs_in_flight * | |
1187 | cli->cl_max_pages_per_rpc; | |
1188 | ||
1189 | if (dirty_max > cli->cl_dirty_max_pages) | |
1190 | cli->cl_dirty_max_pages = dirty_max; | |
1191 | } | |
1192 | ||
1193 | if (cli->cl_dirty_max_pages > totalram_pages / 8) | |
1194 | cli->cl_dirty_max_pages = totalram_pages / 8; | |
1195 | } | |
1196 | ||
d7e09d03 | 1197 | #endif /* __OBD_H */ |