]>
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) 2003, 2010, Oracle and/or its affiliates. All rights reserved. | |
28 | * Use is subject to license terms. | |
29 | * | |
30 | * Copyright (c) 2011, 2012, Intel Corporation. | |
31 | */ | |
32 | /* | |
33 | * This file is part of Lustre, http://www.lustre.org/ | |
34 | * Lustre is a trademark of Sun Microsystems, Inc. | |
35 | */ | |
36 | ||
37 | #ifndef LLITE_INTERNAL_H | |
38 | #define LLITE_INTERNAL_H | |
39 | #include <lustre_debug.h> | |
40 | #include <lustre_ver.h> | |
41 | #include <lustre_disk.h> /* for s2sbi */ | |
42 | #include <lustre_eacl.h> | |
43 | ||
44 | /* for struct cl_lock_descr and struct cl_io */ | |
45 | #include <cl_object.h> | |
46 | #include <lclient.h> | |
47 | #include <lustre_mdc.h> | |
48 | #include <linux/lustre_intent.h> | |
c7c99012 | 49 | #include <linux/compat.h> |
04eb2b7f | 50 | #include <linux/posix_acl_xattr.h> |
d7e09d03 PT |
51 | |
52 | #ifndef FMODE_EXEC | |
53 | #define FMODE_EXEC 0 | |
54 | #endif | |
55 | ||
56 | #ifndef VM_FAULT_RETRY | |
57 | #define VM_FAULT_RETRY 0 | |
58 | #endif | |
59 | ||
60 | /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it. | |
61 | * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */ | |
62 | #ifndef LOOKUP_CONTINUE | |
63 | #define LOOKUP_CONTINUE LOOKUP_PARENT | |
64 | #endif | |
65 | ||
66 | /** Only used on client-side for indicating the tail of dir hash/offset. */ | |
67 | #define LL_DIR_END_OFF 0x7fffffffffffffffULL | |
68 | #define LL_DIR_END_OFF_32BIT 0x7fffffffUL | |
69 | ||
70 | #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0") | |
71 | #define LUSTRE_FPRIVATE(file) ((file)->private_data) | |
72 | ||
73 | struct ll_dentry_data { | |
74 | int lld_cwd_count; | |
75 | int lld_mnt_count; | |
76 | struct obd_client_handle lld_cwd_och; | |
77 | struct obd_client_handle lld_mnt_och; | |
78 | struct lookup_intent *lld_it; | |
79 | unsigned int lld_sa_generation; | |
80 | unsigned int lld_invalid:1; | |
81 | struct rcu_head lld_rcu_head; | |
82 | }; | |
83 | ||
84 | #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata)) | |
85 | ||
86 | extern struct file_operations ll_pgcache_seq_fops; | |
87 | ||
88 | #define LLI_INODE_MAGIC 0x111d0de5 | |
89 | #define LLI_INODE_DEAD 0xdeadd00d | |
90 | ||
91 | /* remote client permission cache */ | |
92 | #define REMOTE_PERM_HASHSIZE 16 | |
93 | ||
94 | struct ll_getname_data { | |
0b09d381 | 95 | struct dir_context ctx; |
d7e09d03 PT |
96 | char *lgd_name; /* points to a buffer with NAME_MAX+1 size */ |
97 | struct lu_fid lgd_fid; /* target fid we are looking for */ | |
98 | int lgd_found; /* inode matched? */ | |
99 | }; | |
100 | ||
101 | /* llite setxid/access permission for user on remote client */ | |
102 | struct ll_remote_perm { | |
103 | struct hlist_node lrp_list; | |
104 | uid_t lrp_uid; | |
105 | gid_t lrp_gid; | |
106 | uid_t lrp_fsuid; | |
107 | gid_t lrp_fsgid; | |
108 | int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this | |
109 | is access permission with | |
110 | lrp_fsuid/lrp_fsgid. */ | |
111 | }; | |
112 | ||
113 | enum lli_flags { | |
114 | /* MDS has an authority for the Size-on-MDS attributes. */ | |
115 | LLIF_MDS_SIZE_LOCK = (1 << 0), | |
116 | /* Epoch close is postponed. */ | |
117 | LLIF_EPOCH_PENDING = (1 << 1), | |
118 | /* DONE WRITING is allowed. */ | |
119 | LLIF_DONE_WRITING = (1 << 2), | |
120 | /* Sizeon-on-MDS attributes are changed. An attribute update needs to | |
121 | * be sent to MDS. */ | |
122 | LLIF_SOM_DIRTY = (1 << 3), | |
123 | /* File is contented */ | |
124 | LLIF_CONTENDED = (1 << 4), | |
125 | /* Truncate uses server lock for this file */ | |
126 | LLIF_SRVLOCK = (1 << 5), | |
127 | /* File data is modified. */ | |
128 | LLIF_DATA_MODIFIED = (1 << 6), | |
5ea17d6c JL |
129 | /* File is being restored */ |
130 | LLIF_FILE_RESTORING = (1 << 7), | |
7fc1f831 AP |
131 | /* Xattr cache is attached to the file */ |
132 | LLIF_XATTR_CACHE = (1 << 8), | |
d7e09d03 PT |
133 | }; |
134 | ||
135 | struct ll_inode_info { | |
136 | __u32 lli_inode_magic; | |
137 | __u32 lli_flags; | |
138 | __u64 lli_ioepoch; | |
139 | ||
140 | spinlock_t lli_lock; | |
141 | struct posix_acl *lli_posix_acl; | |
142 | ||
143 | struct hlist_head *lli_remote_perms; | |
144 | struct mutex lli_rmtperm_mutex; | |
145 | ||
146 | /* identifying fields for both metadata and data stacks. */ | |
147 | struct lu_fid lli_fid; | |
148 | /* Parent fid for accessing default stripe data on parent directory | |
149 | * for allocating OST objects after a mknod() and later open-by-FID. */ | |
150 | struct lu_fid lli_pfid; | |
151 | ||
152 | struct list_head lli_close_list; | |
153 | struct list_head lli_oss_capas; | |
154 | /* open count currently used by capability only, indicate whether | |
155 | * capability needs renewal */ | |
156 | atomic_t lli_open_count; | |
157 | struct obd_capa *lli_mds_capa; | |
158 | cfs_time_t lli_rmtperm_time; | |
159 | ||
160 | /* handle is to be sent to MDS later on done_writing and setattr. | |
161 | * Open handle data are needed for the recovery to reconstruct | |
162 | * the inode state on the MDS. XXX: recovery is not ready yet. */ | |
163 | struct obd_client_handle *lli_pending_och; | |
164 | ||
165 | /* We need all three because every inode may be opened in different | |
166 | * modes */ | |
167 | struct obd_client_handle *lli_mds_read_och; | |
168 | struct obd_client_handle *lli_mds_write_och; | |
169 | struct obd_client_handle *lli_mds_exec_och; | |
170 | __u64 lli_open_fd_read_count; | |
171 | __u64 lli_open_fd_write_count; | |
172 | __u64 lli_open_fd_exec_count; | |
173 | /* Protects access to och pointers and their usage counters */ | |
174 | struct mutex lli_och_mutex; | |
175 | ||
176 | struct inode lli_vfs_inode; | |
177 | ||
178 | /* the most recent timestamps obtained from mds */ | |
179 | struct ost_lvb lli_lvb; | |
180 | spinlock_t lli_agl_lock; | |
181 | ||
182 | /* Try to make the d::member and f::member are aligned. Before using | |
183 | * these members, make clear whether it is directory or not. */ | |
184 | union { | |
185 | /* for directory */ | |
186 | struct { | |
187 | /* serialize normal readdir and statahead-readdir. */ | |
188 | struct mutex d_readdir_mutex; | |
189 | ||
190 | /* metadata statahead */ | |
191 | /* since parent-child threads can share the same @file | |
192 | * struct, "opendir_key" is the token when dir close for | |
193 | * case of parent exit before child -- it is me should | |
194 | * cleanup the dir readahead. */ | |
195 | void *d_opendir_key; | |
196 | struct ll_statahead_info *d_sai; | |
197 | struct posix_acl *d_def_acl; | |
198 | /* protect statahead stuff. */ | |
199 | spinlock_t d_sa_lock; | |
200 | /* "opendir_pid" is the token when lookup/revalid | |
201 | * -- I am the owner of dir statahead. */ | |
202 | pid_t d_opendir_pid; | |
203 | } d; | |
204 | ||
205 | #define lli_readdir_mutex u.d.d_readdir_mutex | |
206 | #define lli_opendir_key u.d.d_opendir_key | |
207 | #define lli_sai u.d.d_sai | |
208 | #define lli_def_acl u.d.d_def_acl | |
209 | #define lli_sa_lock u.d.d_sa_lock | |
210 | #define lli_opendir_pid u.d.d_opendir_pid | |
211 | ||
212 | /* for non-directory */ | |
213 | struct { | |
47a57bde | 214 | struct mutex f_size_mutex; |
d7e09d03 PT |
215 | char *f_symlink_name; |
216 | __u64 f_maxbytes; | |
217 | /* | |
218 | * struct rw_semaphore { | |
219 | * signed long count; // align d.d_def_acl | |
220 | * spinlock_t wait_lock; // align d.d_sa_lock | |
221 | * struct list_head wait_list; | |
222 | * } | |
223 | */ | |
224 | struct rw_semaphore f_trunc_sem; | |
225 | struct mutex f_write_mutex; | |
226 | ||
227 | struct rw_semaphore f_glimpse_sem; | |
228 | cfs_time_t f_glimpse_time; | |
229 | struct list_head f_agl_list; | |
230 | __u64 f_agl_index; | |
231 | ||
232 | /* for writepage() only to communicate to fsync */ | |
233 | int f_async_rc; | |
234 | ||
235 | /* volatile file criteria is based on file name, this | |
236 | * flag is used to keep the test result, so the strcmp | |
237 | * is done only once | |
238 | */ | |
239 | bool f_volatile; | |
240 | /* | |
241 | * whenever a process try to read/write the file, the | |
242 | * jobid of the process will be saved here, and it'll | |
243 | * be packed into the write PRC when flush later. | |
244 | * | |
245 | * so the read/write statistics for jobid will not be | |
246 | * accurate if the file is shared by different jobs. | |
247 | */ | |
248 | char f_jobid[JOBSTATS_JOBID_SIZE]; | |
249 | } f; | |
250 | ||
47a57bde | 251 | #define lli_size_mutex u.f.f_size_mutex |
d7e09d03 PT |
252 | #define lli_symlink_name u.f.f_symlink_name |
253 | #define lli_maxbytes u.f.f_maxbytes | |
254 | #define lli_trunc_sem u.f.f_trunc_sem | |
255 | #define lli_write_mutex u.f.f_write_mutex | |
256 | #define lli_glimpse_sem u.f.f_glimpse_sem | |
257 | #define lli_glimpse_time u.f.f_glimpse_time | |
258 | #define lli_agl_list u.f.f_agl_list | |
259 | #define lli_agl_index u.f.f_agl_index | |
260 | #define lli_async_rc u.f.f_async_rc | |
261 | #define lli_jobid u.f.f_jobid | |
262 | #define lli_volatile u.f.f_volatile | |
263 | ||
264 | } u; | |
265 | ||
266 | /* XXX: For following frequent used members, although they maybe special | |
267 | * used for non-directory object, it is some time-wasting to check | |
268 | * whether the object is directory or not before using them. On the | |
269 | * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce | |
270 | * the "ll_inode_info" size even if moving those members into u.f. | |
271 | * So keep them out side. | |
272 | * | |
273 | * In the future, if more members are added only for directory, | |
274 | * some of the following members can be moved into u.f. | |
275 | */ | |
276 | bool lli_has_smd; | |
277 | struct cl_object *lli_clob; | |
278 | ||
279 | /* mutex to request for layout lock exclusively. */ | |
280 | struct mutex lli_layout_mutex; | |
09aed8a5 | 281 | /* Layout version, protected by lli_layout_lock */ |
d7e09d03 | 282 | __u32 lli_layout_gen; |
09aed8a5 | 283 | spinlock_t lli_layout_lock; |
7fc1f831 AP |
284 | |
285 | struct rw_semaphore lli_xattrs_list_rwsem; | |
286 | struct mutex lli_xattrs_enq_lock; | |
287 | struct list_head lli_xattrs;/* ll_xattr_entry->xe_list */ | |
d7e09d03 PT |
288 | }; |
289 | ||
09aed8a5 JX |
290 | static inline __u32 ll_layout_version_get(struct ll_inode_info *lli) |
291 | { | |
292 | __u32 gen; | |
293 | ||
294 | spin_lock(&lli->lli_layout_lock); | |
295 | gen = lli->lli_layout_gen; | |
296 | spin_unlock(&lli->lli_layout_lock); | |
297 | ||
298 | return gen; | |
299 | } | |
300 | ||
301 | static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen) | |
302 | { | |
303 | spin_lock(&lli->lli_layout_lock); | |
304 | lli->lli_layout_gen = gen; | |
305 | spin_unlock(&lli->lli_layout_lock); | |
306 | } | |
307 | ||
7fc1f831 AP |
308 | int ll_xattr_cache_destroy(struct inode *inode); |
309 | ||
310 | int ll_xattr_cache_get(struct inode *inode, | |
311 | const char *name, | |
312 | char *buffer, | |
313 | size_t size, | |
314 | __u64 valid); | |
315 | ||
d7e09d03 PT |
316 | /* |
317 | * Locking to guarantee consistency of non-atomic updates to long long i_size, | |
318 | * consistency between file size and KMS. | |
319 | * | |
47a57bde | 320 | * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order. |
d7e09d03 PT |
321 | */ |
322 | ||
323 | void ll_inode_size_lock(struct inode *inode); | |
324 | void ll_inode_size_unlock(struct inode *inode); | |
325 | ||
326 | // FIXME: replace the name of this with LL_I to conform to kernel stuff | |
327 | // static inline struct ll_inode_info *LL_I(struct inode *inode) | |
328 | static inline struct ll_inode_info *ll_i2info(struct inode *inode) | |
329 | { | |
330 | return container_of(inode, struct ll_inode_info, lli_vfs_inode); | |
331 | } | |
332 | ||
333 | /* default to about 40meg of readahead on a given system. That much tied | |
334 | * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */ | |
335 | #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT)) | |
336 | ||
337 | /* default to read-ahead full files smaller than 2MB on the second read */ | |
338 | #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT)) | |
339 | ||
340 | enum ra_stat { | |
341 | RA_STAT_HIT = 0, | |
342 | RA_STAT_MISS, | |
343 | RA_STAT_DISTANT_READPAGE, | |
344 | RA_STAT_MISS_IN_WINDOW, | |
345 | RA_STAT_FAILED_GRAB_PAGE, | |
346 | RA_STAT_FAILED_MATCH, | |
347 | RA_STAT_DISCARDED, | |
348 | RA_STAT_ZERO_LEN, | |
349 | RA_STAT_ZERO_WINDOW, | |
350 | RA_STAT_EOF, | |
351 | RA_STAT_MAX_IN_FLIGHT, | |
352 | RA_STAT_WRONG_GRAB_PAGE, | |
353 | _NR_RA_STAT, | |
354 | }; | |
355 | ||
356 | struct ll_ra_info { | |
357 | atomic_t ra_cur_pages; | |
358 | unsigned long ra_max_pages; | |
359 | unsigned long ra_max_pages_per_file; | |
360 | unsigned long ra_max_read_ahead_whole_pages; | |
361 | }; | |
362 | ||
363 | /* ra_io_arg will be filled in the beginning of ll_readahead with | |
364 | * ras_lock, then the following ll_read_ahead_pages will read RA | |
365 | * pages according to this arg, all the items in this structure are | |
366 | * counted by page index. | |
367 | */ | |
368 | struct ra_io_arg { | |
369 | unsigned long ria_start; /* start offset of read-ahead*/ | |
370 | unsigned long ria_end; /* end offset of read-ahead*/ | |
371 | /* If stride read pattern is detected, ria_stoff means where | |
372 | * stride read is started. Note: for normal read-ahead, the | |
373 | * value here is meaningless, and also it will not be accessed*/ | |
374 | pgoff_t ria_stoff; | |
375 | /* ria_length and ria_pages are the length and pages length in the | |
376 | * stride I/O mode. And they will also be used to check whether | |
377 | * it is stride I/O read-ahead in the read-ahead pages*/ | |
378 | unsigned long ria_length; | |
379 | unsigned long ria_pages; | |
380 | }; | |
381 | ||
382 | /* LL_HIST_MAX=32 causes an overflow */ | |
383 | #define LL_HIST_MAX 28 | |
384 | #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */ | |
385 | #define LL_PROCESS_HIST_MAX 10 | |
386 | struct per_process_info { | |
387 | pid_t pid; | |
388 | struct obd_histogram pp_r_hist; | |
389 | struct obd_histogram pp_w_hist; | |
390 | }; | |
391 | ||
392 | /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */ | |
393 | struct ll_rw_extents_info { | |
394 | struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1]; | |
395 | }; | |
396 | ||
397 | #define LL_OFFSET_HIST_MAX 100 | |
398 | struct ll_rw_process_info { | |
399 | pid_t rw_pid; | |
400 | int rw_op; | |
401 | loff_t rw_range_start; | |
402 | loff_t rw_range_end; | |
403 | loff_t rw_last_file_pos; | |
404 | loff_t rw_offset; | |
405 | size_t rw_smallest_extent; | |
406 | size_t rw_largest_extent; | |
407 | struct ll_file_data *rw_last_file; | |
408 | }; | |
409 | ||
410 | enum stats_track_type { | |
411 | STATS_TRACK_ALL = 0, /* track all processes */ | |
412 | STATS_TRACK_PID, /* track process with this pid */ | |
413 | STATS_TRACK_PPID, /* track processes with this ppid */ | |
414 | STATS_TRACK_GID, /* track processes with this gid */ | |
415 | STATS_TRACK_LAST, | |
416 | }; | |
417 | ||
418 | /* flags for sbi->ll_flags */ | |
419 | #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */ | |
420 | #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */ | |
421 | #define LL_SBI_FLOCK 0x04 | |
422 | #define LL_SBI_USER_XATTR 0x08 /* support user xattr */ | |
423 | #define LL_SBI_ACL 0x10 /* support ACL */ | |
424 | #define LL_SBI_RMT_CLIENT 0x40 /* remote client */ | |
425 | #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */ | |
426 | #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */ | |
427 | #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */ | |
428 | #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */ | |
429 | #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */ | |
430 | #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */ | |
431 | #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */ | |
432 | #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */ | |
433 | #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */ | |
434 | #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */ | |
435 | #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */ | |
436 | #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */ | |
7fc1f831 | 437 | #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */ |
d7e09d03 PT |
438 | |
439 | #define LL_SBI_FLAGS { \ | |
440 | "nolck", \ | |
441 | "checksum", \ | |
442 | "flock", \ | |
443 | "xattr", \ | |
444 | "acl", \ | |
7fc1f831 | 445 | "???", \ |
d7e09d03 PT |
446 | "rmt_client", \ |
447 | "mds_capa", \ | |
448 | "oss_capa", \ | |
449 | "flock", \ | |
450 | "lru_resize", \ | |
451 | "lazy_statfs", \ | |
452 | "som", \ | |
453 | "32bit_api", \ | |
454 | "64bit_hash", \ | |
455 | "agl", \ | |
456 | "verbose", \ | |
457 | "layout", \ | |
7fc1f831 AP |
458 | "user_fid2path",\ |
459 | "xattr", \ | |
460 | } | |
d7e09d03 PT |
461 | |
462 | /* default value for ll_sb_info->contention_time */ | |
463 | #define SBI_DEFAULT_CONTENTION_SECONDS 60 | |
464 | /* default value for lockless_truncate_enable */ | |
465 | #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1 | |
466 | #define RCE_HASHES 32 | |
467 | ||
468 | struct rmtacl_ctl_entry { | |
469 | struct list_head rce_list; | |
470 | pid_t rce_key; /* hash key */ | |
471 | int rce_ops; /* acl operation type */ | |
472 | }; | |
473 | ||
474 | struct rmtacl_ctl_table { | |
475 | spinlock_t rct_lock; | |
476 | struct list_head rct_entries[RCE_HASHES]; | |
477 | }; | |
478 | ||
479 | #define EE_HASHES 32 | |
480 | ||
d7e09d03 PT |
481 | struct eacl_table { |
482 | spinlock_t et_lock; | |
483 | struct list_head et_entries[EE_HASHES]; | |
484 | }; | |
485 | ||
486 | struct ll_sb_info { | |
487 | struct list_head ll_list; | |
488 | /* this protects pglist and ra_info. It isn't safe to | |
489 | * grab from interrupt contexts */ | |
490 | spinlock_t ll_lock; | |
491 | spinlock_t ll_pp_extent_lock; /* pp_extent entry*/ | |
492 | spinlock_t ll_process_lock; /* ll_rw_process_info */ | |
493 | struct obd_uuid ll_sb_uuid; | |
494 | struct obd_export *ll_md_exp; | |
495 | struct obd_export *ll_dt_exp; | |
496 | struct proc_dir_entry* ll_proc_root; | |
497 | struct lu_fid ll_root_fid; /* root object fid */ | |
498 | ||
499 | int ll_flags; | |
7fc1f831 AP |
500 | unsigned int ll_umounting:1, |
501 | ll_xattr_cache_enabled:1; | |
d7e09d03 PT |
502 | struct list_head ll_conn_chain; /* per-conn chain of SBs */ |
503 | struct lustre_client_ocd ll_lco; | |
504 | ||
505 | struct list_head ll_orphan_dentry_list; /*please don't ask -p*/ | |
506 | struct ll_close_queue *ll_lcq; | |
507 | ||
508 | struct lprocfs_stats *ll_stats; /* lprocfs stats counter */ | |
509 | ||
d7e09d03 PT |
510 | struct cl_client_cache ll_cache; |
511 | ||
512 | struct lprocfs_stats *ll_ra_stats; | |
513 | ||
514 | struct ll_ra_info ll_ra_info; | |
515 | unsigned int ll_namelen; | |
516 | struct file_operations *ll_fop; | |
517 | ||
518 | /* =0 - hold lock over whole read/write | |
519 | * >0 - max. chunk to be read/written w/o lock re-acquiring */ | |
520 | unsigned long ll_max_rw_chunk; | |
521 | unsigned int ll_md_brw_size; /* used by readdir */ | |
522 | ||
523 | struct lu_site *ll_site; | |
524 | struct cl_device *ll_cl; | |
525 | /* Statistics */ | |
526 | struct ll_rw_extents_info ll_rw_extents_info; | |
527 | int ll_extent_process_count; | |
528 | struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX]; | |
529 | unsigned int ll_offset_process_count; | |
530 | struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX]; | |
531 | unsigned int ll_rw_offset_entry_count; | |
532 | int ll_stats_track_id; | |
533 | enum stats_track_type ll_stats_track_type; | |
534 | int ll_rw_stats_on; | |
535 | ||
536 | /* metadata stat-ahead */ | |
537 | unsigned int ll_sa_max; /* max statahead RPCs */ | |
538 | atomic_t ll_sa_total; /* statahead thread started | |
539 | * count */ | |
540 | atomic_t ll_sa_wrong; /* statahead thread stopped for | |
541 | * low hit ratio */ | |
542 | atomic_t ll_agl_total; /* AGL thread started count */ | |
543 | ||
544 | dev_t ll_sdev_orig; /* save s_dev before assign for | |
d0a0acc3 | 545 | * clustered nfs */ |
d7e09d03 PT |
546 | struct rmtacl_ctl_table ll_rct; |
547 | struct eacl_table ll_et; | |
bd994071 | 548 | __kernel_fsid_t ll_fsid; |
d7e09d03 PT |
549 | }; |
550 | ||
551 | #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024) | |
552 | ||
553 | struct ll_ra_read { | |
554 | pgoff_t lrr_start; | |
555 | pgoff_t lrr_count; | |
556 | struct task_struct *lrr_reader; | |
557 | struct list_head lrr_linkage; | |
558 | }; | |
559 | ||
560 | /* | |
561 | * per file-descriptor read-ahead data. | |
562 | */ | |
563 | struct ll_readahead_state { | |
564 | spinlock_t ras_lock; | |
565 | /* | |
566 | * index of the last page that read(2) needed and that wasn't in the | |
567 | * cache. Used by ras_update() to detect seeks. | |
568 | * | |
569 | * XXX nikita: if access seeks into cached region, Lustre doesn't see | |
570 | * this. | |
571 | */ | |
572 | unsigned long ras_last_readpage; | |
573 | /* | |
574 | * number of pages read after last read-ahead window reset. As window | |
575 | * is reset on each seek, this is effectively a number of consecutive | |
576 | * accesses. Maybe ->ras_accessed_in_window is better name. | |
577 | * | |
578 | * XXX nikita: window is also reset (by ras_update()) when Lustre | |
579 | * believes that memory pressure evicts read-ahead pages. In that | |
580 | * case, it probably doesn't make sense to expand window to | |
581 | * PTLRPC_MAX_BRW_PAGES on the third access. | |
582 | */ | |
583 | unsigned long ras_consecutive_pages; | |
584 | /* | |
585 | * number of read requests after the last read-ahead window reset | |
586 | * As window is reset on each seek, this is effectively the number | |
587 | * on consecutive read request and is used to trigger read-ahead. | |
588 | */ | |
589 | unsigned long ras_consecutive_requests; | |
590 | /* | |
591 | * Parameters of current read-ahead window. Handled by | |
592 | * ras_update(). On the initial access to the file or after a seek, | |
593 | * window is reset to 0. After 3 consecutive accesses, window is | |
594 | * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by | |
595 | * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages. | |
596 | */ | |
597 | unsigned long ras_window_start, ras_window_len; | |
598 | /* | |
599 | * Where next read-ahead should start at. This lies within read-ahead | |
600 | * window. Read-ahead window is read in pieces rather than at once | |
601 | * because: 1. lustre limits total number of pages under read-ahead by | |
602 | * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages | |
603 | * not covered by DLM lock. | |
604 | */ | |
605 | unsigned long ras_next_readahead; | |
606 | /* | |
607 | * Total number of ll_file_read requests issued, reads originating | |
608 | * due to mmap are not counted in this total. This value is used to | |
609 | * trigger full file read-ahead after multiple reads to a small file. | |
610 | */ | |
611 | unsigned long ras_requests; | |
612 | /* | |
613 | * Page index with respect to the current request, these value | |
614 | * will not be accurate when dealing with reads issued via mmap. | |
615 | */ | |
616 | unsigned long ras_request_index; | |
617 | /* | |
618 | * list of struct ll_ra_read's one per read(2) call current in | |
619 | * progress against this file descriptor. Used by read-ahead code, | |
620 | * protected by ->ras_lock. | |
621 | */ | |
622 | struct list_head ras_read_beads; | |
623 | /* | |
624 | * The following 3 items are used for detecting the stride I/O | |
625 | * mode. | |
626 | * In stride I/O mode, | |
627 | * ...............|-----data-----|****gap*****|--------|******|.... | |
628 | * offset |-stride_pages-|-stride_gap-| | |
629 | * ras_stride_offset = offset; | |
630 | * ras_stride_length = stride_pages + stride_gap; | |
631 | * ras_stride_pages = stride_pages; | |
632 | * Note: all these three items are counted by pages. | |
633 | */ | |
634 | unsigned long ras_stride_length; | |
635 | unsigned long ras_stride_pages; | |
636 | pgoff_t ras_stride_offset; | |
637 | /* | |
638 | * number of consecutive stride request count, and it is similar as | |
639 | * ras_consecutive_requests, but used for stride I/O mode. | |
640 | * Note: only more than 2 consecutive stride request are detected, | |
641 | * stride read-ahead will be enable | |
642 | */ | |
643 | unsigned long ras_consecutive_stride_requests; | |
644 | }; | |
645 | ||
646 | extern struct kmem_cache *ll_file_data_slab; | |
647 | struct lustre_handle; | |
648 | struct ll_file_data { | |
649 | struct ll_readahead_state fd_ras; | |
d7e09d03 PT |
650 | struct ccc_grouplock fd_grouplock; |
651 | __u64 lfd_pos; | |
652 | __u32 fd_flags; | |
d3a8a4e2 JX |
653 | fmode_t fd_omode; |
654 | /* openhandle if lease exists for this file. | |
655 | * Borrow lli->lli_och_mutex to protect assignment */ | |
656 | struct obd_client_handle *fd_lease_och; | |
657 | struct obd_client_handle *fd_och; | |
d7e09d03 PT |
658 | struct file *fd_file; |
659 | /* Indicate whether need to report failure when close. | |
660 | * true: failure is known, not report again. | |
661 | * false: unknown failure, should report. */ | |
662 | bool fd_write_failed; | |
663 | }; | |
664 | ||
665 | struct lov_stripe_md; | |
666 | ||
667 | extern spinlock_t inode_lock; | |
668 | ||
669 | extern struct proc_dir_entry *proc_lustre_fs_root; | |
670 | ||
671 | static inline struct inode *ll_info2i(struct ll_inode_info *lli) | |
672 | { | |
673 | return &lli->lli_vfs_inode; | |
674 | } | |
675 | ||
676 | struct it_cb_data { | |
677 | struct inode *icbd_parent; | |
678 | struct dentry **icbd_childp; | |
679 | obd_id hash; | |
680 | }; | |
681 | ||
682 | __u32 ll_i2suppgid(struct inode *i); | |
683 | void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2); | |
684 | ||
685 | static inline int ll_need_32bit_api(struct ll_sb_info *sbi) | |
686 | { | |
687 | #if BITS_PER_LONG == 32 | |
688 | return 1; | |
689 | #else | |
56dba8cd DE |
690 | return unlikely( |
691 | #ifdef CONFIG_COMPAT | |
692 | is_compat_task() || | |
693 | #endif | |
694 | (sbi->ll_flags & LL_SBI_32BIT_API) | |
695 | ); | |
d7e09d03 PT |
696 | #endif |
697 | } | |
698 | ||
699 | #define LLAP_MAGIC 98764321 | |
700 | ||
701 | extern struct kmem_cache *ll_async_page_slab; | |
702 | extern size_t ll_async_page_slab_size; | |
703 | ||
704 | void ll_ra_read_in(struct file *f, struct ll_ra_read *rar); | |
705 | void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar); | |
706 | struct ll_ra_read *ll_ra_read_get(struct file *f); | |
707 | ||
708 | /* llite/lproc_llite.c */ | |
709 | #ifdef LPROCFS | |
710 | int lprocfs_register_mountpoint(struct proc_dir_entry *parent, | |
711 | struct super_block *sb, char *osc, char *mdc); | |
712 | void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi); | |
713 | void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count); | |
714 | void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars); | |
2c185ffa PT |
715 | void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid, |
716 | struct ll_file_data *file, loff_t pos, | |
717 | size_t count, int rw); | |
d7e09d03 PT |
718 | #else |
719 | static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent, | |
720 | struct super_block *sb, char *osc, char *mdc){return 0;} | |
721 | static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {} | |
2c185ffa PT |
722 | static inline |
723 | void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {} | |
724 | static inline void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars) | |
d7e09d03 PT |
725 | { |
726 | memset(lvars, 0, sizeof(*lvars)); | |
727 | } | |
2c185ffa PT |
728 | static inline void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid, |
729 | struct ll_file_data *file, loff_t pos, | |
730 | size_t count, int rw) {} | |
d7e09d03 PT |
731 | #endif |
732 | ||
733 | ||
734 | /* llite/dir.c */ | |
735 | void ll_release_page(struct page *page, int remove); | |
736 | extern struct file_operations ll_dir_operations; | |
737 | extern struct inode_operations ll_dir_inode_operations; | |
738 | struct page *ll_get_dir_page(struct inode *dir, __u64 hash, | |
739 | struct ll_dir_chain *chain); | |
0b09d381 | 740 | int ll_dir_read(struct inode *inode, struct dir_context *ctx); |
d7e09d03 PT |
741 | |
742 | int ll_get_mdt_idx(struct inode *inode); | |
743 | /* llite/namei.c */ | |
744 | int ll_objects_destroy(struct ptlrpc_request *request, | |
745 | struct inode *dir); | |
746 | struct inode *ll_iget(struct super_block *sb, ino_t hash, | |
747 | struct lustre_md *lic); | |
748 | int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *, | |
749 | void *data, int flag); | |
750 | struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de); | |
751 | int ll_rmdir_entry(struct inode *dir, char *name, int namelen); | |
752 | ||
753 | /* llite/rw.c */ | |
754 | int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to); | |
755 | int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to); | |
756 | int ll_writepage(struct page *page, struct writeback_control *wbc); | |
757 | int ll_writepages(struct address_space *, struct writeback_control *wbc); | |
758 | void ll_removepage(struct page *page); | |
759 | int ll_readpage(struct file *file, struct page *page); | |
760 | void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras); | |
761 | int ll_file_punch(struct inode *, loff_t, int); | |
762 | ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int); | |
763 | void ll_clear_file_contended(struct inode*); | |
764 | int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t); | |
765 | int ll_readahead(const struct lu_env *env, struct cl_io *io, | |
766 | struct ll_readahead_state *ras, struct address_space *mapping, | |
767 | struct cl_page_list *queue, int flags); | |
768 | ||
769 | /* llite/file.c */ | |
770 | extern struct file_operations ll_file_operations; | |
771 | extern struct file_operations ll_file_operations_flock; | |
772 | extern struct file_operations ll_file_operations_noflock; | |
773 | extern struct inode_operations ll_file_inode_operations; | |
774 | extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *, | |
775 | __u64); | |
776 | extern int ll_have_md_lock(struct inode *inode, __u64 *bits, | |
777 | ldlm_mode_t l_req_mode); | |
778 | extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits, | |
7fc1f831 AP |
779 | struct lustre_handle *lockh, __u64 flags, |
780 | ldlm_mode_t mode); | |
d7e09d03 PT |
781 | int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *, |
782 | __u64 bits); | |
783 | int ll_revalidate_nd(struct dentry *dentry, unsigned int flags); | |
784 | int ll_file_open(struct inode *inode, struct file *file); | |
785 | int ll_file_release(struct inode *inode, struct file *file); | |
786 | int ll_glimpse_ioctl(struct ll_sb_info *sbi, | |
787 | struct lov_stripe_md *lsm, lstat_t *st); | |
788 | void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch); | |
789 | int ll_local_open(struct file *file, | |
790 | struct lookup_intent *it, struct ll_file_data *fd, | |
791 | struct obd_client_handle *och); | |
792 | int ll_release_openhandle(struct dentry *, struct lookup_intent *); | |
793 | int ll_md_close(struct obd_export *md_exp, struct inode *inode, | |
794 | struct file *file); | |
45b2a010 | 795 | int ll_md_real_close(struct inode *inode, fmode_t fmode); |
d7e09d03 PT |
796 | void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data, |
797 | struct obd_client_handle **och, unsigned long flags); | |
798 | void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data); | |
799 | int ll_som_update(struct inode *inode, struct md_op_data *op_data); | |
800 | int ll_inode_getattr(struct inode *inode, struct obdo *obdo, | |
801 | __u64 ioepoch, int sync); | |
802 | int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data, | |
803 | struct md_open_data **mod); | |
804 | void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data, | |
805 | struct lustre_handle *fh); | |
d7e09d03 PT |
806 | int ll_getattr_it(struct vfsmount *mnt, struct dentry *de, |
807 | struct lookup_intent *it, struct kstat *stat); | |
808 | int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat); | |
809 | struct ll_file_data *ll_file_data_get(void); | |
810 | struct posix_acl * ll_get_acl(struct inode *inode, int type); | |
811 | ||
812 | int ll_inode_permission(struct inode *inode, int mask); | |
813 | ||
814 | int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file, | |
815 | int flags, struct lov_user_md *lum, | |
816 | int lum_size); | |
817 | int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename, | |
818 | struct lov_mds_md **lmm, int *lmm_size, | |
819 | struct ptlrpc_request **request); | |
820 | int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump, | |
821 | int set_default); | |
822 | int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmmp, | |
823 | int *lmm_size, struct ptlrpc_request **request); | |
824 | int ll_fsync(struct file *file, loff_t start, loff_t end, int data); | |
825 | int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap, | |
826 | int num_bytes); | |
827 | int ll_merge_lvb(const struct lu_env *env, struct inode *inode); | |
828 | int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg); | |
829 | int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg); | |
830 | int ll_fid2path(struct inode *inode, void *arg); | |
831 | int ll_data_version(struct inode *inode, __u64 *data_version, int extent_lock); | |
48d23e61 | 832 | int ll_hsm_release(struct inode *inode); |
d7e09d03 | 833 | |
d3a8a4e2 | 834 | struct obd_client_handle *ll_lease_open(struct inode *inode, struct file *file, |
48d23e61 | 835 | fmode_t mode, __u64 flags); |
d3a8a4e2 JX |
836 | int ll_lease_close(struct obd_client_handle *och, struct inode *inode, |
837 | bool *lease_broken); | |
838 | ||
d7e09d03 PT |
839 | /* llite/dcache.c */ |
840 | ||
3ea8f3bc | 841 | int ll_d_init(struct dentry *de); |
d7e09d03 PT |
842 | extern struct dentry_operations ll_d_ops; |
843 | void ll_intent_drop_lock(struct lookup_intent *); | |
844 | void ll_intent_release(struct lookup_intent *); | |
845 | void ll_invalidate_aliases(struct inode *); | |
846 | void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft); | |
847 | void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry); | |
90181e1d | 848 | int ll_dcompare(const struct dentry *parent, const struct dentry *dentry, |
d7e09d03 PT |
849 | unsigned int len, const char *str, const struct qstr *d_name); |
850 | int ll_revalidate_it_finish(struct ptlrpc_request *request, | |
851 | struct lookup_intent *it, struct dentry *de); | |
852 | ||
853 | /* llite/llite_lib.c */ | |
854 | extern struct super_operations lustre_super_operations; | |
855 | ||
856 | char *ll_read_opt(const char *opt, char *data); | |
857 | void ll_lli_init(struct ll_inode_info *lli); | |
858 | int ll_fill_super(struct super_block *sb, struct vfsmount *mnt); | |
859 | void ll_put_super(struct super_block *sb); | |
860 | void ll_kill_super(struct super_block *sb); | |
861 | struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock); | |
862 | struct inode *ll_inode_from_lock(struct ldlm_lock *lock); | |
863 | void ll_clear_inode(struct inode *inode); | |
a720b790 | 864 | int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import); |
d7e09d03 PT |
865 | int ll_setattr(struct dentry *de, struct iattr *attr); |
866 | int ll_statfs(struct dentry *de, struct kstatfs *sfs); | |
867 | int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs, | |
868 | __u64 max_age, __u32 flags); | |
869 | void ll_update_inode(struct inode *inode, struct lustre_md *md); | |
870 | void ll_read_inode2(struct inode *inode, void *opaque); | |
871 | void ll_delete_inode(struct inode *inode); | |
872 | int ll_iocontrol(struct inode *inode, struct file *file, | |
873 | unsigned int cmd, unsigned long arg); | |
874 | int ll_flush_ctx(struct inode *inode); | |
875 | void ll_umount_begin(struct super_block *sb); | |
876 | int ll_remount_fs(struct super_block *sb, int *flags, char *data); | |
877 | int ll_show_options(struct seq_file *seq, struct dentry *dentry); | |
878 | void ll_dirty_page_discard_warn(struct page *page, int ioret); | |
879 | int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req, | |
880 | struct super_block *, struct lookup_intent *); | |
881 | void lustre_dump_dentry(struct dentry *, int recur); | |
882 | void lustre_dump_inode(struct inode *); | |
883 | int ll_obd_statfs(struct inode *inode, void *arg); | |
884 | int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize); | |
44779340 BB |
885 | int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize); |
886 | int ll_get_max_cookiesize(struct ll_sb_info *sbi, int *max_cookiesize); | |
887 | int ll_get_default_cookiesize(struct ll_sb_info *sbi, int *default_cookiesize); | |
d7e09d03 PT |
888 | int ll_process_config(struct lustre_cfg *lcfg); |
889 | struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data, | |
890 | struct inode *i1, struct inode *i2, | |
891 | const char *name, int namelen, | |
892 | int mode, __u32 opc, void *data); | |
893 | void ll_finish_md_op_data(struct md_op_data *op_data); | |
894 | int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg); | |
895 | char *ll_get_fsname(struct super_block *sb, char *buf, int buflen); | |
896 | ||
897 | /* llite/llite_nfs.c */ | |
898 | extern struct export_operations lustre_export_operations; | |
899 | __u32 get_uuid2int(const char *name, int len); | |
bd994071 | 900 | void get_uuid2fsid(const char *name, int len, __kernel_fsid_t *fsid); |
d7e09d03 PT |
901 | struct inode *search_inode_for_lustre(struct super_block *sb, |
902 | const struct lu_fid *fid); | |
903 | ||
904 | /* llite/special.c */ | |
905 | extern struct inode_operations ll_special_inode_operations; | |
906 | extern struct file_operations ll_special_chr_inode_fops; | |
907 | extern struct file_operations ll_special_chr_file_fops; | |
908 | extern struct file_operations ll_special_blk_inode_fops; | |
909 | extern struct file_operations ll_special_fifo_inode_fops; | |
910 | extern struct file_operations ll_special_fifo_file_fops; | |
911 | extern struct file_operations ll_special_sock_inode_fops; | |
912 | ||
913 | /* llite/symlink.c */ | |
914 | extern struct inode_operations ll_fast_symlink_inode_operations; | |
915 | ||
916 | /* llite/llite_close.c */ | |
917 | struct ll_close_queue { | |
918 | spinlock_t lcq_lock; | |
919 | struct list_head lcq_head; | |
920 | wait_queue_head_t lcq_waitq; | |
921 | struct completion lcq_comp; | |
922 | atomic_t lcq_stop; | |
923 | }; | |
924 | ||
925 | struct ccc_object *cl_inode2ccc(struct inode *inode); | |
926 | ||
927 | ||
928 | void vvp_write_pending (struct ccc_object *club, struct ccc_page *page); | |
929 | void vvp_write_complete(struct ccc_object *club, struct ccc_page *page); | |
930 | ||
d0a0acc3 | 931 | /* specific architecture can implement only part of this list */ |
d7e09d03 PT |
932 | enum vvp_io_subtype { |
933 | /** normal IO */ | |
934 | IO_NORMAL, | |
d7e09d03 PT |
935 | /** io started from splice_{read|write} */ |
936 | IO_SPLICE | |
937 | }; | |
938 | ||
939 | /* IO subtypes */ | |
940 | struct vvp_io { | |
941 | /** io subtype */ | |
942 | enum vvp_io_subtype cui_io_subtype; | |
943 | ||
944 | union { | |
d7e09d03 PT |
945 | struct { |
946 | struct pipe_inode_info *cui_pipe; | |
947 | unsigned int cui_flags; | |
948 | } splice; | |
949 | struct vvp_fault_io { | |
950 | /** | |
951 | * Inode modification time that is checked across DLM | |
952 | * lock request. | |
953 | */ | |
954 | time_t ft_mtime; | |
955 | struct vm_area_struct *ft_vma; | |
956 | /** | |
957 | * locked page returned from vvp_io | |
958 | */ | |
959 | struct page *ft_vmpage; | |
960 | struct vm_fault_api { | |
961 | /** | |
962 | * kernel fault info | |
963 | */ | |
964 | struct vm_fault *ft_vmf; | |
965 | /** | |
966 | * fault API used bitflags for return code. | |
967 | */ | |
968 | unsigned int ft_flags; | |
969 | } fault; | |
970 | } fault; | |
971 | } u; | |
972 | /** | |
973 | * Read-ahead state used by read and page-fault IO contexts. | |
974 | */ | |
975 | struct ll_ra_read cui_bead; | |
976 | /** | |
977 | * Set when cui_bead has been initialized. | |
978 | */ | |
979 | int cui_ra_window_set; | |
980 | /** | |
981 | * Partially truncated page, that vvp_io_trunc_start() keeps locked | |
982 | * across truncate. | |
983 | */ | |
984 | struct cl_page *cui_partpage; | |
985 | }; | |
986 | ||
987 | /** | |
988 | * IO arguments for various VFS I/O interfaces. | |
989 | */ | |
990 | struct vvp_io_args { | |
74c0da19 | 991 | /** normal/splice */ |
d7e09d03 PT |
992 | enum vvp_io_subtype via_io_subtype; |
993 | ||
994 | union { | |
995 | struct { | |
996 | struct kiocb *via_iocb; | |
997 | struct iovec *via_iov; | |
998 | unsigned long via_nrsegs; | |
999 | } normal; | |
d7e09d03 PT |
1000 | struct { |
1001 | struct pipe_inode_info *via_pipe; | |
1002 | unsigned int via_flags; | |
1003 | } splice; | |
1004 | } u; | |
1005 | }; | |
1006 | ||
1007 | struct ll_cl_context { | |
1008 | void *lcc_cookie; | |
1009 | struct cl_io *lcc_io; | |
1010 | struct cl_page *lcc_page; | |
1011 | struct lu_env *lcc_env; | |
1012 | int lcc_refcheck; | |
1013 | int lcc_created; | |
1014 | }; | |
1015 | ||
1016 | struct vvp_thread_info { | |
1017 | struct ost_lvb vti_lvb; | |
1018 | struct cl_2queue vti_queue; | |
1019 | struct iovec vti_local_iov; | |
1020 | struct vvp_io_args vti_args; | |
1021 | struct ra_io_arg vti_ria; | |
1022 | struct kiocb vti_kiocb; | |
1023 | struct ll_cl_context vti_io_ctx; | |
1024 | }; | |
1025 | ||
1026 | static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env) | |
1027 | { | |
1028 | extern struct lu_context_key vvp_key; | |
1029 | struct vvp_thread_info *info; | |
1030 | ||
1031 | info = lu_context_key_get(&env->le_ctx, &vvp_key); | |
1032 | LASSERT(info != NULL); | |
1033 | return info; | |
1034 | } | |
1035 | ||
1036 | static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env, | |
1037 | enum vvp_io_subtype type) | |
1038 | { | |
1039 | struct vvp_io_args *ret = &vvp_env_info(env)->vti_args; | |
1040 | ||
1041 | ret->via_io_subtype = type; | |
1042 | ||
1043 | return ret; | |
1044 | } | |
1045 | ||
1046 | struct vvp_session { | |
1047 | struct vvp_io vs_ios; | |
1048 | }; | |
1049 | ||
1050 | static inline struct vvp_session *vvp_env_session(const struct lu_env *env) | |
1051 | { | |
1052 | extern struct lu_context_key vvp_session_key; | |
1053 | struct vvp_session *ses; | |
1054 | ||
1055 | ses = lu_context_key_get(env->le_ses, &vvp_session_key); | |
1056 | LASSERT(ses != NULL); | |
1057 | return ses; | |
1058 | } | |
1059 | ||
1060 | static inline struct vvp_io *vvp_env_io(const struct lu_env *env) | |
1061 | { | |
1062 | return &vvp_env_session(env)->vs_ios; | |
1063 | } | |
1064 | ||
1065 | void ll_queue_done_writing(struct inode *inode, unsigned long flags); | |
1066 | void ll_close_thread_shutdown(struct ll_close_queue *lcq); | |
1067 | int ll_close_thread_start(struct ll_close_queue **lcq_ret); | |
1068 | ||
1069 | /* llite/llite_mmap.c */ | |
1070 | typedef struct rb_root rb_root_t; | |
1071 | typedef struct rb_node rb_node_t; | |
1072 | ||
1073 | struct ll_lock_tree_node; | |
1074 | struct ll_lock_tree { | |
1075 | rb_root_t lt_root; | |
1076 | struct list_head lt_locked_list; | |
1077 | struct ll_file_data *lt_fd; | |
1078 | }; | |
1079 | ||
1080 | int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last); | |
1081 | int ll_file_mmap(struct file * file, struct vm_area_struct * vma); | |
1082 | struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start, | |
1083 | __u64 end, ldlm_mode_t mode); | |
1084 | void policy_from_vma(ldlm_policy_data_t *policy, | |
1085 | struct vm_area_struct *vma, unsigned long addr, size_t count); | |
1086 | struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr, | |
1087 | size_t count); | |
1088 | ||
1089 | static inline void ll_invalidate_page(struct page *vmpage) | |
1090 | { | |
1091 | struct address_space *mapping = vmpage->mapping; | |
1092 | loff_t offset = vmpage->index << PAGE_CACHE_SHIFT; | |
1093 | ||
1094 | LASSERT(PageLocked(vmpage)); | |
1095 | if (mapping == NULL) | |
1096 | return; | |
1097 | ||
1098 | ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE); | |
1099 | truncate_complete_page(mapping, vmpage); | |
1100 | } | |
1101 | ||
1102 | #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi) | |
1103 | ||
1104 | /* don't need an addref as the sb_info should be holding one */ | |
1105 | static inline struct obd_export *ll_s2dtexp(struct super_block *sb) | |
1106 | { | |
1107 | return ll_s2sbi(sb)->ll_dt_exp; | |
1108 | } | |
1109 | ||
1110 | /* don't need an addref as the sb_info should be holding one */ | |
1111 | static inline struct obd_export *ll_s2mdexp(struct super_block *sb) | |
1112 | { | |
1113 | return ll_s2sbi(sb)->ll_md_exp; | |
1114 | } | |
1115 | ||
1116 | static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi) | |
1117 | { | |
1118 | struct obd_device *obd = sbi->ll_md_exp->exp_obd; | |
1119 | if (obd == NULL) | |
1120 | LBUG(); | |
1121 | return &obd->u.cli; | |
1122 | } | |
1123 | ||
1124 | // FIXME: replace the name of this with LL_SB to conform to kernel stuff | |
1125 | static inline struct ll_sb_info *ll_i2sbi(struct inode *inode) | |
1126 | { | |
1127 | return ll_s2sbi(inode->i_sb); | |
1128 | } | |
1129 | ||
1130 | static inline struct obd_export *ll_i2dtexp(struct inode *inode) | |
1131 | { | |
1132 | return ll_s2dtexp(inode->i_sb); | |
1133 | } | |
1134 | ||
1135 | static inline struct obd_export *ll_i2mdexp(struct inode *inode) | |
1136 | { | |
1137 | return ll_s2mdexp(inode->i_sb); | |
1138 | } | |
1139 | ||
1140 | static inline struct lu_fid *ll_inode2fid(struct inode *inode) | |
1141 | { | |
1142 | struct lu_fid *fid; | |
1143 | ||
1144 | LASSERT(inode != NULL); | |
1145 | fid = &ll_i2info(inode)->lli_fid; | |
1146 | ||
1147 | return fid; | |
1148 | } | |
1149 | ||
d7e09d03 PT |
1150 | static inline __u64 ll_file_maxbytes(struct inode *inode) |
1151 | { | |
1152 | return ll_i2info(inode)->lli_maxbytes; | |
1153 | } | |
1154 | ||
1155 | /* llite/xattr.c */ | |
1156 | int ll_setxattr(struct dentry *dentry, const char *name, | |
1157 | const void *value, size_t size, int flags); | |
1158 | ssize_t ll_getxattr(struct dentry *dentry, const char *name, | |
1159 | void *buffer, size_t size); | |
1160 | ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size); | |
1161 | int ll_removexattr(struct dentry *dentry, const char *name); | |
1162 | ||
1163 | /* llite/remote_perm.c */ | |
1164 | extern struct kmem_cache *ll_remote_perm_cachep; | |
1165 | extern struct kmem_cache *ll_rmtperm_hash_cachep; | |
1166 | ||
1167 | struct hlist_head *alloc_rmtperm_hash(void); | |
1168 | void free_rmtperm_hash(struct hlist_head *hash); | |
1169 | int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm); | |
1170 | int lustre_check_remote_perm(struct inode *inode, int mask); | |
1171 | ||
1172 | /* llite/llite_capa.c */ | |
54319351 | 1173 | extern struct timer_list ll_capa_timer; |
d7e09d03 PT |
1174 | |
1175 | int ll_capa_thread_start(void); | |
1176 | void ll_capa_thread_stop(void); | |
1177 | void ll_capa_timer_callback(unsigned long unused); | |
1178 | ||
1179 | struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa); | |
1180 | int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa); | |
1181 | ||
1182 | void ll_capa_open(struct inode *inode); | |
1183 | void ll_capa_close(struct inode *inode); | |
1184 | ||
1185 | struct obd_capa *ll_mdscapa_get(struct inode *inode); | |
1186 | struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc); | |
1187 | ||
1188 | void ll_truncate_free_capa(struct obd_capa *ocapa); | |
1189 | void ll_clear_inode_capas(struct inode *inode); | |
1190 | void ll_print_capa_stat(struct ll_sb_info *sbi); | |
1191 | ||
1192 | /* llite/llite_cl.c */ | |
1193 | extern struct lu_device_type vvp_device_type; | |
1194 | ||
1195 | /** | |
1196 | * Common IO arguments for various VFS I/O interfaces. | |
1197 | */ | |
1198 | int cl_sb_init(struct super_block *sb); | |
1199 | int cl_sb_fini(struct super_block *sb); | |
1200 | enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma); | |
1201 | void ll_io_init(struct cl_io *io, const struct file *file, int write); | |
1202 | ||
1203 | void ras_update(struct ll_sb_info *sbi, struct inode *inode, | |
1204 | struct ll_readahead_state *ras, unsigned long index, | |
1205 | unsigned hit); | |
1206 | void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len); | |
1207 | int ll_is_file_contended(struct file *file); | |
1208 | void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which); | |
1209 | ||
1210 | /* llite/llite_rmtacl.c */ | |
1211 | #ifdef CONFIG_FS_POSIX_ACL | |
9147dc8d PT |
1212 | struct eacl_entry { |
1213 | struct list_head ee_list; | |
1214 | pid_t ee_key; /* hash key */ | |
1215 | struct lu_fid ee_fid; | |
1216 | int ee_type; /* ACL type for ACCESS or DEFAULT */ | |
1217 | ext_acl_xattr_header *ee_acl; | |
1218 | }; | |
1219 | ||
d7e09d03 PT |
1220 | obd_valid rce_ops2valid(int ops); |
1221 | struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key); | |
1222 | int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops); | |
1223 | int rct_del(struct rmtacl_ctl_table *rct, pid_t key); | |
1224 | void rct_init(struct rmtacl_ctl_table *rct); | |
1225 | void rct_fini(struct rmtacl_ctl_table *rct); | |
1226 | ||
1227 | void ee_free(struct eacl_entry *ee); | |
1228 | int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type, | |
1229 | ext_acl_xattr_header *header); | |
1230 | struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key, | |
1231 | struct lu_fid *fid, int type); | |
1232 | void et_search_free(struct eacl_table *et, pid_t key); | |
1233 | void et_init(struct eacl_table *et); | |
1234 | void et_fini(struct eacl_table *et); | |
9147dc8d PT |
1235 | #else |
1236 | static inline obd_valid rce_ops2valid(int ops) | |
1237 | { | |
1238 | return 0; | |
1239 | } | |
d7e09d03 PT |
1240 | #endif |
1241 | ||
1242 | /* statahead.c */ | |
1243 | ||
1244 | #define LL_SA_RPC_MIN 2 | |
1245 | #define LL_SA_RPC_DEF 32 | |
1246 | #define LL_SA_RPC_MAX 8192 | |
1247 | ||
1248 | #define LL_SA_CACHE_BIT 5 | |
1249 | #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT) | |
1250 | #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1) | |
1251 | ||
1252 | /* per inode struct, for dir only */ | |
1253 | struct ll_statahead_info { | |
1254 | struct inode *sai_inode; | |
1255 | atomic_t sai_refcount; /* when access this struct, hold | |
1256 | * refcount */ | |
1257 | unsigned int sai_generation; /* generation for statahead */ | |
1258 | unsigned int sai_max; /* max ahead of lookup */ | |
1259 | __u64 sai_sent; /* stat requests sent count */ | |
1260 | __u64 sai_replied; /* stat requests which received | |
1261 | * reply */ | |
1262 | __u64 sai_index; /* index of statahead entry */ | |
1263 | __u64 sai_index_wait; /* index of entry which is the | |
1264 | * caller is waiting for */ | |
1265 | __u64 sai_hit; /* hit count */ | |
1266 | __u64 sai_miss; /* miss count: | |
1267 | * for "ls -al" case, it includes | |
1268 | * hidden dentry miss; | |
1269 | * for "ls -l" case, it does not | |
1270 | * include hidden dentry miss. | |
1271 | * "sai_miss_hidden" is used for | |
1272 | * the later case. | |
1273 | */ | |
1274 | unsigned int sai_consecutive_miss; /* consecutive miss */ | |
1275 | unsigned int sai_miss_hidden;/* "ls -al", but first dentry | |
1276 | * is not a hidden one */ | |
1277 | unsigned int sai_skip_hidden;/* skipped hidden dentry count */ | |
1278 | unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for | |
1279 | * hidden entries */ | |
1280 | sai_in_readpage:1,/* statahead is in readdir()*/ | |
1281 | sai_agl_valid:1;/* AGL is valid for the dir */ | |
1282 | wait_queue_head_t sai_waitq; /* stat-ahead wait queue */ | |
1283 | struct ptlrpc_thread sai_thread; /* stat-ahead thread */ | |
1284 | struct ptlrpc_thread sai_agl_thread; /* AGL thread */ | |
1285 | struct list_head sai_entries; /* entry list */ | |
1286 | struct list_head sai_entries_received; /* entries returned */ | |
1287 | struct list_head sai_entries_stated; /* entries stated */ | |
1288 | struct list_head sai_entries_agl; /* AGL entries to be sent */ | |
1289 | struct list_head sai_cache[LL_SA_CACHE_SIZE]; | |
1290 | spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE]; | |
1291 | atomic_t sai_cache_count; /* entry count in cache */ | |
1292 | }; | |
1293 | ||
1294 | int do_statahead_enter(struct inode *dir, struct dentry **dentry, | |
1295 | int only_unplug); | |
1296 | void ll_stop_statahead(struct inode *dir, void *key); | |
1297 | ||
1298 | static inline int ll_glimpse_size(struct inode *inode) | |
1299 | { | |
1300 | struct ll_inode_info *lli = ll_i2info(inode); | |
1301 | int rc; | |
1302 | ||
1303 | down_read(&lli->lli_glimpse_sem); | |
1304 | rc = cl_glimpse_size(inode); | |
1305 | lli->lli_glimpse_time = cfs_time_current(); | |
1306 | up_read(&lli->lli_glimpse_sem); | |
1307 | return rc; | |
1308 | } | |
1309 | ||
1310 | static inline void | |
1311 | ll_statahead_mark(struct inode *dir, struct dentry *dentry) | |
1312 | { | |
1313 | struct ll_inode_info *lli = ll_i2info(dir); | |
1314 | struct ll_statahead_info *sai = lli->lli_sai; | |
1315 | struct ll_dentry_data *ldd = ll_d2d(dentry); | |
1316 | ||
1317 | /* not the same process, don't mark */ | |
1318 | if (lli->lli_opendir_pid != current_pid()) | |
1319 | return; | |
1320 | ||
3ea8f3bc LS |
1321 | LASSERT(ldd != NULL); |
1322 | if (sai != NULL) | |
d7e09d03 PT |
1323 | ldd->lld_sa_generation = sai->sai_generation; |
1324 | } | |
1325 | ||
1326 | static inline int | |
f236f69b | 1327 | d_need_statahead(struct inode *dir, struct dentry *dentryp) |
d7e09d03 PT |
1328 | { |
1329 | struct ll_inode_info *lli; | |
1330 | struct ll_dentry_data *ldd; | |
1331 | ||
1332 | if (ll_i2sbi(dir)->ll_sa_max == 0) | |
1333 | return -EAGAIN; | |
1334 | ||
1335 | lli = ll_i2info(dir); | |
1336 | /* not the same process, don't statahead */ | |
1337 | if (lli->lli_opendir_pid != current_pid()) | |
1338 | return -EAGAIN; | |
1339 | ||
1340 | /* statahead has been stopped */ | |
1341 | if (lli->lli_opendir_key == NULL) | |
1342 | return -EAGAIN; | |
1343 | ||
1344 | ldd = ll_d2d(dentryp); | |
1345 | /* | |
1346 | * When stats a dentry, the system trigger more than once "revalidate" | |
1347 | * or "lookup", for "getattr", for "getxattr", and maybe for others. | |
1348 | * Under patchless client mode, the operation intent is not accurate, | |
1349 | * which maybe misguide the statahead thread. For example: | |
1350 | * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe | |
1351 | * have the same operation intent -- "IT_GETATTR". | |
1352 | * In fact, one dentry should has only one chance to interact with the | |
1353 | * statahead thread, otherwise the statahead windows will be confused. | |
1354 | * The solution is as following: | |
1355 | * Assign "lld_sa_generation" with "sai_generation" when a dentry | |
1356 | * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR" | |
1357 | * will bypass interacting with statahead thread for checking: | |
1358 | * "lld_sa_generation == lli_sai->sai_generation" | |
1359 | */ | |
1360 | if (ldd && lli->lli_sai && | |
1361 | ldd->lld_sa_generation == lli->lli_sai->sai_generation) | |
1362 | return -EAGAIN; | |
1363 | ||
1364 | return 1; | |
1365 | } | |
1366 | ||
1367 | static inline int | |
1368 | ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug) | |
1369 | { | |
1370 | int ret; | |
1371 | ||
f236f69b | 1372 | ret = d_need_statahead(dir, *dentryp); |
d7e09d03 PT |
1373 | if (ret <= 0) |
1374 | return ret; | |
1375 | ||
1376 | return do_statahead_enter(dir, dentryp, only_unplug); | |
1377 | } | |
1378 | ||
d0a0acc3 | 1379 | /* llite ioctl register support routine */ |
d7e09d03 PT |
1380 | enum llioc_iter { |
1381 | LLIOC_CONT = 0, | |
1382 | LLIOC_STOP | |
1383 | }; | |
1384 | ||
1385 | #define LLIOC_MAX_CMD 256 | |
1386 | ||
1387 | /* | |
1388 | * Rules to write a callback function: | |
1389 | * | |
1390 | * Parameters: | |
d0a0acc3 | 1391 | * @magic: Dynamic ioctl call routine will feed this value with the pointer |
d7e09d03 PT |
1392 | * returned to ll_iocontrol_register. Callback functions should use this |
1393 | * data to check the potential collasion of ioctl cmd. If collasion is | |
1394 | * found, callback function should return LLIOC_CONT. | |
1395 | * @rcp: The result of ioctl command. | |
1396 | * | |
1397 | * Return values: | |
1398 | * If @magic matches the pointer returned by ll_iocontrol_data, the | |
1399 | * callback should return LLIOC_STOP; return LLIOC_STOP otherwise. | |
1400 | */ | |
1401 | typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode, | |
1402 | struct file *file, unsigned int cmd, unsigned long arg, | |
1403 | void *magic, int *rcp); | |
1404 | ||
1405 | enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file, | |
1406 | unsigned int cmd, unsigned long arg, int *rcp); | |
1407 | ||
1408 | /* export functions */ | |
1409 | /* Register ioctl block dynamatically for a regular file. | |
1410 | * | |
1411 | * @cmd: the array of ioctl command set | |
1412 | * @count: number of commands in the @cmd | |
1413 | * @cb: callback function, it will be called if an ioctl command is found to | |
1414 | * belong to the command list @cmd. | |
1415 | * | |
d0a0acc3 | 1416 | * Return value: |
d7e09d03 PT |
1417 | * A magic pointer will be returned if success; |
1418 | * otherwise, NULL will be returned. | |
1419 | * */ | |
1420 | void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd); | |
1421 | void ll_iocontrol_unregister(void *magic); | |
1422 | ||
1423 | ||
1424 | /* lclient compat stuff */ | |
1425 | #define cl_inode_info ll_inode_info | |
1426 | #define cl_i2info(info) ll_i2info(info) | |
1427 | #define cl_inode_mode(inode) ((inode)->i_mode) | |
1428 | #define cl_i2sbi ll_i2sbi | |
1429 | ||
1430 | static inline struct ll_file_data *cl_iattr2fd(struct inode *inode, | |
1431 | const struct iattr *attr) | |
1432 | { | |
1433 | LASSERT(attr->ia_valid & ATTR_FILE); | |
1434 | return LUSTRE_FPRIVATE(attr->ia_file); | |
1435 | } | |
1436 | ||
1437 | static inline void cl_isize_lock(struct inode *inode) | |
1438 | { | |
1439 | ll_inode_size_lock(inode); | |
1440 | } | |
1441 | ||
1442 | static inline void cl_isize_unlock(struct inode *inode) | |
1443 | { | |
1444 | ll_inode_size_unlock(inode); | |
1445 | } | |
1446 | ||
1447 | static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms) | |
1448 | { | |
47a57bde | 1449 | LASSERT(mutex_is_locked(&ll_i2info(inode)->lli_size_mutex)); |
d7e09d03 PT |
1450 | i_size_write(inode, kms); |
1451 | } | |
1452 | ||
1453 | static inline void cl_isize_write(struct inode *inode, loff_t kms) | |
1454 | { | |
1455 | ll_inode_size_lock(inode); | |
1456 | i_size_write(inode, kms); | |
1457 | ll_inode_size_unlock(inode); | |
1458 | } | |
1459 | ||
1460 | #define cl_isize_read(inode) i_size_read(inode) | |
1461 | ||
1462 | static inline int cl_merge_lvb(const struct lu_env *env, struct inode *inode) | |
1463 | { | |
1464 | return ll_merge_lvb(env, inode); | |
1465 | } | |
1466 | ||
1467 | #define cl_inode_atime(inode) LTIME_S((inode)->i_atime) | |
1468 | #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime) | |
1469 | #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime) | |
1470 | ||
1471 | struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt); | |
1472 | ||
1473 | int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end, | |
65fb55d1 | 1474 | enum cl_fsync_mode mode, int ignore_layout); |
d7e09d03 PT |
1475 | |
1476 | /** direct write pages */ | |
1477 | struct ll_dio_pages { | |
1478 | /** page array to be written. we don't support | |
1479 | * partial pages except the last one. */ | |
1480 | struct page **ldp_pages; | |
1481 | /* offset of each page */ | |
1482 | loff_t *ldp_offsets; | |
1483 | /** if ldp_offsets is NULL, it means a sequential | |
1484 | * pages to be written, then this is the file offset | |
1485 | * of the * first page. */ | |
1486 | loff_t ldp_start_offset; | |
1487 | /** how many bytes are to be written. */ | |
1488 | size_t ldp_size; | |
1489 | /** # of pages in the array. */ | |
1490 | int ldp_nr; | |
1491 | }; | |
1492 | ||
1493 | static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt, | |
1494 | int rc) | |
1495 | { | |
1496 | int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ : | |
1497 | LPROC_LL_OSC_WRITE; | |
1498 | ||
1499 | ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc); | |
1500 | } | |
1501 | ||
1502 | extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io, | |
1503 | int rw, struct inode *inode, | |
1504 | struct ll_dio_pages *pv); | |
1505 | ||
1506 | static inline int ll_file_nolock(const struct file *file) | |
1507 | { | |
1508 | struct ll_file_data *fd = LUSTRE_FPRIVATE(file); | |
1509 | struct inode *inode = file->f_dentry->d_inode; | |
1510 | ||
1511 | LASSERT(fd != NULL); | |
1512 | return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) || | |
1513 | (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK)); | |
1514 | } | |
1515 | ||
1516 | static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode, | |
1517 | struct lookup_intent *it, __u64 *bits) | |
1518 | { | |
1519 | if (!it->d.lustre.it_lock_set) { | |
1520 | struct lustre_handle handle; | |
1521 | ||
1522 | /* If this inode is a remote object, it will get two | |
1523 | * separate locks in different namespaces, Master MDT, | |
1524 | * where the name entry is, will grant LOOKUP lock, | |
1525 | * remote MDT, where the object is, will grant | |
d0a0acc3 | 1526 | * UPDATE|PERM lock. The inode will be attached to both |
d7e09d03 PT |
1527 | * LOOKUP and PERM locks, so revoking either locks will |
1528 | * case the dcache being cleared */ | |
1529 | if (it->d.lustre.it_remote_lock_mode) { | |
1530 | handle.cookie = it->d.lustre.it_remote_lock_handle; | |
1531 | CDEBUG(D_DLMTRACE, "setting l_data to inode %p" | |
1532 | "(%lu/%u) for remote lock "LPX64"\n", inode, | |
1533 | inode->i_ino, inode->i_generation, | |
1534 | handle.cookie); | |
1535 | md_set_lock_data(exp, &handle.cookie, inode, NULL); | |
1536 | } | |
1537 | ||
1538 | handle.cookie = it->d.lustre.it_lock_handle; | |
1539 | ||
1540 | CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)" | |
1541 | " for lock "LPX64"\n", inode, inode->i_ino, | |
1542 | inode->i_generation, handle.cookie); | |
1543 | ||
1544 | md_set_lock_data(exp, &handle.cookie, inode, | |
1545 | &it->d.lustre.it_lock_bits); | |
1546 | it->d.lustre.it_lock_set = 1; | |
1547 | } | |
1548 | ||
1549 | if (bits != NULL) | |
1550 | *bits = it->d.lustre.it_lock_bits; | |
1551 | } | |
1552 | ||
1553 | static inline void ll_lock_dcache(struct inode *inode) | |
1554 | { | |
1555 | spin_lock(&inode->i_lock); | |
1556 | } | |
1557 | ||
1558 | static inline void ll_unlock_dcache(struct inode *inode) | |
1559 | { | |
1560 | spin_unlock(&inode->i_lock); | |
1561 | } | |
1562 | ||
1563 | static inline int d_lustre_invalid(const struct dentry *dentry) | |
1564 | { | |
1565 | struct ll_dentry_data *lld = ll_d2d(dentry); | |
1566 | ||
1567 | return (lld == NULL) || lld->lld_invalid; | |
1568 | } | |
1569 | ||
1570 | static inline void __d_lustre_invalidate(struct dentry *dentry) | |
1571 | { | |
1572 | struct ll_dentry_data *lld = ll_d2d(dentry); | |
1573 | ||
1574 | if (lld != NULL) | |
1575 | lld->lld_invalid = 1; | |
1576 | } | |
1577 | ||
1578 | /* | |
1579 | * Mark dentry INVALID, if dentry refcount is zero (this is normally case for | |
1580 | * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later; | |
1581 | * else dput() of the last refcount will unhash this dentry and kill it. | |
1582 | */ | |
b1d2a127 | 1583 | static inline void d_lustre_invalidate(struct dentry *dentry, int nested) |
d7e09d03 PT |
1584 | { |
1585 | CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p " | |
1586 | "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry, | |
193deee1 | 1587 | dentry->d_parent, dentry->d_inode, d_count(dentry)); |
d7e09d03 | 1588 | |
b1d2a127 PT |
1589 | spin_lock_nested(&dentry->d_lock, |
1590 | nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL); | |
d7e09d03 | 1591 | __d_lustre_invalidate(dentry); |
193deee1 | 1592 | if (d_count(dentry) == 0) |
d7e09d03 PT |
1593 | __d_drop(dentry); |
1594 | spin_unlock(&dentry->d_lock); | |
1595 | } | |
1596 | ||
1597 | static inline void d_lustre_revalidate(struct dentry *dentry) | |
1598 | { | |
1599 | spin_lock(&dentry->d_lock); | |
1600 | LASSERT(ll_d2d(dentry) != NULL); | |
1601 | ll_d2d(dentry)->lld_invalid = 0; | |
1602 | spin_unlock(&dentry->d_lock); | |
1603 | } | |
1604 | ||
1605 | #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) | |
1606 | /* Compatibility for old (1.8) compiled userspace quota code */ | |
1607 | struct if_quotactl_18 { | |
1608 | __u32 qc_cmd; | |
1609 | __u32 qc_type; | |
1610 | __u32 qc_id; | |
1611 | __u32 qc_stat; | |
1612 | struct obd_dqinfo qc_dqinfo; | |
1613 | struct obd_dqblk qc_dqblk; | |
1614 | char obd_type[16]; | |
1615 | struct obd_uuid obd_uuid; | |
1616 | }; | |
1617 | #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *) | |
1618 | /* End compatibility for old (1.8) compiled userspace quota code */ | |
1619 | #else | |
1620 | #warning "remove old LL_IOC_QUOTACTL_18 compatibility code" | |
1621 | #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */ | |
1622 | ||
1623 | enum { | |
1624 | LL_LAYOUT_GEN_NONE = ((__u32)-2), /* layout lock was cancelled */ | |
1625 | LL_LAYOUT_GEN_EMPTY = ((__u32)-1) /* for empty layout */ | |
1626 | }; | |
1627 | ||
1628 | int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf); | |
1629 | int ll_layout_refresh(struct inode *inode, __u32 *gen); | |
5ea17d6c | 1630 | int ll_layout_restore(struct inode *inode); |
d7e09d03 | 1631 | |
7fc1f831 AP |
1632 | int ll_xattr_init(void); |
1633 | void ll_xattr_fini(void); | |
1634 | ||
d7e09d03 | 1635 | #endif /* LLITE_INTERNAL_H */ |