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) 2003, 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 LLITE_INTERNAL_H | |
34 | #define LLITE_INTERNAL_H | |
67a235f5 GKH |
35 | #include "../include/lustre_debug.h" |
36 | #include "../include/lustre_ver.h" | |
37 | #include "../include/lustre_disk.h" /* for s2sbi */ | |
38 | #include "../include/lustre_eacl.h" | |
d7e09d03 PT |
39 | |
40 | /* for struct cl_lock_descr and struct cl_io */ | |
67a235f5 | 41 | #include "../include/cl_object.h" |
67a235f5 | 42 | #include "../include/lustre_mdc.h" |
00d65ec8 | 43 | #include "../include/lustre_intent.h" |
c7c99012 | 44 | #include <linux/compat.h> |
04eb2b7f | 45 | #include <linux/posix_acl_xattr.h> |
0d345656 | 46 | #include "vvp_internal.h" |
d7e09d03 PT |
47 | |
48 | #ifndef FMODE_EXEC | |
49 | #define FMODE_EXEC 0 | |
50 | #endif | |
51 | ||
52 | #ifndef VM_FAULT_RETRY | |
53 | #define VM_FAULT_RETRY 0 | |
54 | #endif | |
55 | ||
d7e09d03 PT |
56 | /** Only used on client-side for indicating the tail of dir hash/offset. */ |
57 | #define LL_DIR_END_OFF 0x7fffffffffffffffULL | |
58 | #define LL_DIR_END_OFF_32BIT 0x7fffffffUL | |
59 | ||
60 | #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0") | |
61 | #define LUSTRE_FPRIVATE(file) ((file)->private_data) | |
62 | ||
63 | struct ll_dentry_data { | |
d7e09d03 PT |
64 | struct lookup_intent *lld_it; |
65 | unsigned int lld_sa_generation; | |
66 | unsigned int lld_invalid:1; | |
67 | struct rcu_head lld_rcu_head; | |
68 | }; | |
69 | ||
b2952d62 | 70 | #define ll_d2d(de) ((struct ll_dentry_data *)((de)->d_fsdata)) |
d7e09d03 | 71 | |
d7e09d03 PT |
72 | #define LLI_INODE_MAGIC 0x111d0de5 |
73 | #define LLI_INODE_DEAD 0xdeadd00d | |
74 | ||
d7e09d03 | 75 | struct ll_getname_data { |
0b09d381 | 76 | struct dir_context ctx; |
d7e09d03 PT |
77 | char *lgd_name; /* points to a buffer with NAME_MAX+1 size */ |
78 | struct lu_fid lgd_fid; /* target fid we are looking for */ | |
79 | int lgd_found; /* inode matched? */ | |
80 | }; | |
81 | ||
98eae5e7 JH |
82 | struct ll_grouplock { |
83 | struct lu_env *lg_env; | |
84 | struct cl_io *lg_io; | |
85 | struct cl_lock *lg_lock; | |
86 | unsigned long lg_gid; | |
87 | }; | |
88 | ||
d7e09d03 PT |
89 | enum lli_flags { |
90 | /* MDS has an authority for the Size-on-MDS attributes. */ | |
91 | LLIF_MDS_SIZE_LOCK = (1 << 0), | |
92 | /* Epoch close is postponed. */ | |
93 | LLIF_EPOCH_PENDING = (1 << 1), | |
94 | /* DONE WRITING is allowed. */ | |
95 | LLIF_DONE_WRITING = (1 << 2), | |
96 | /* Sizeon-on-MDS attributes are changed. An attribute update needs to | |
c0894c6c OD |
97 | * be sent to MDS. |
98 | */ | |
d7e09d03 | 99 | LLIF_SOM_DIRTY = (1 << 3), |
d7e09d03 | 100 | /* File data is modified. */ |
2d95f10e | 101 | LLIF_DATA_MODIFIED = (1 << 4), |
5ea17d6c | 102 | /* File is being restored */ |
2d95f10e | 103 | LLIF_FILE_RESTORING = (1 << 5), |
7fc1f831 | 104 | /* Xattr cache is attached to the file */ |
2d95f10e | 105 | LLIF_XATTR_CACHE = (1 << 6), |
d7e09d03 PT |
106 | }; |
107 | ||
108 | struct ll_inode_info { | |
109 | __u32 lli_inode_magic; | |
110 | __u32 lli_flags; | |
111 | __u64 lli_ioepoch; | |
112 | ||
113 | spinlock_t lli_lock; | |
114 | struct posix_acl *lli_posix_acl; | |
115 | ||
d7e09d03 PT |
116 | /* identifying fields for both metadata and data stacks. */ |
117 | struct lu_fid lli_fid; | |
118 | /* Parent fid for accessing default stripe data on parent directory | |
c0894c6c OD |
119 | * for allocating OST objects after a mknod() and later open-by-FID. |
120 | */ | |
d7e09d03 PT |
121 | struct lu_fid lli_pfid; |
122 | ||
82bd9200 OD |
123 | struct list_head lli_close_list; |
124 | ||
d7e09d03 PT |
125 | /* handle is to be sent to MDS later on done_writing and setattr. |
126 | * Open handle data are needed for the recovery to reconstruct | |
c0894c6c OD |
127 | * the inode state on the MDS. XXX: recovery is not ready yet. |
128 | */ | |
d7e09d03 PT |
129 | struct obd_client_handle *lli_pending_och; |
130 | ||
131 | /* We need all three because every inode may be opened in different | |
c0894c6c OD |
132 | * modes |
133 | */ | |
d7e09d03 PT |
134 | struct obd_client_handle *lli_mds_read_och; |
135 | struct obd_client_handle *lli_mds_write_och; | |
136 | struct obd_client_handle *lli_mds_exec_och; | |
137 | __u64 lli_open_fd_read_count; | |
138 | __u64 lli_open_fd_write_count; | |
139 | __u64 lli_open_fd_exec_count; | |
140 | /* Protects access to och pointers and their usage counters */ | |
141 | struct mutex lli_och_mutex; | |
142 | ||
143 | struct inode lli_vfs_inode; | |
144 | ||
145 | /* the most recent timestamps obtained from mds */ | |
d2995737 JH |
146 | s64 lli_atime; |
147 | s64 lli_mtime; | |
148 | s64 lli_ctime; | |
d7e09d03 PT |
149 | spinlock_t lli_agl_lock; |
150 | ||
151 | /* Try to make the d::member and f::member are aligned. Before using | |
c0894c6c OD |
152 | * these members, make clear whether it is directory or not. |
153 | */ | |
d7e09d03 PT |
154 | union { |
155 | /* for directory */ | |
156 | struct { | |
157 | /* serialize normal readdir and statahead-readdir. */ | |
158 | struct mutex d_readdir_mutex; | |
159 | ||
160 | /* metadata statahead */ | |
161 | /* since parent-child threads can share the same @file | |
162 | * struct, "opendir_key" is the token when dir close for | |
163 | * case of parent exit before child -- it is me should | |
c0894c6c OD |
164 | * cleanup the dir readahead. |
165 | */ | |
d7e09d03 PT |
166 | void *d_opendir_key; |
167 | struct ll_statahead_info *d_sai; | |
d7e09d03 PT |
168 | /* protect statahead stuff. */ |
169 | spinlock_t d_sa_lock; | |
c0894c6c OD |
170 | /* "opendir_pid" is the token when lookup/revalidate |
171 | * -- I am the owner of dir statahead. | |
172 | */ | |
d7e09d03 PT |
173 | pid_t d_opendir_pid; |
174 | } d; | |
175 | ||
176 | #define lli_readdir_mutex u.d.d_readdir_mutex | |
177 | #define lli_opendir_key u.d.d_opendir_key | |
178 | #define lli_sai u.d.d_sai | |
d7e09d03 PT |
179 | #define lli_sa_lock u.d.d_sa_lock |
180 | #define lli_opendir_pid u.d.d_opendir_pid | |
181 | ||
182 | /* for non-directory */ | |
183 | struct { | |
47a57bde | 184 | struct mutex f_size_mutex; |
d7e09d03 PT |
185 | char *f_symlink_name; |
186 | __u64 f_maxbytes; | |
187 | /* | |
188 | * struct rw_semaphore { | |
189 | * signed long count; // align d.d_def_acl | |
190 | * spinlock_t wait_lock; // align d.d_sa_lock | |
191 | * struct list_head wait_list; | |
192 | * } | |
193 | */ | |
194 | struct rw_semaphore f_trunc_sem; | |
195 | struct mutex f_write_mutex; | |
196 | ||
197 | struct rw_semaphore f_glimpse_sem; | |
a649ad1d | 198 | unsigned long f_glimpse_time; |
d7e09d03 PT |
199 | struct list_head f_agl_list; |
200 | __u64 f_agl_index; | |
201 | ||
202 | /* for writepage() only to communicate to fsync */ | |
203 | int f_async_rc; | |
204 | ||
d7e09d03 PT |
205 | /* |
206 | * whenever a process try to read/write the file, the | |
207 | * jobid of the process will be saved here, and it'll | |
208 | * be packed into the write PRC when flush later. | |
209 | * | |
210 | * so the read/write statistics for jobid will not be | |
211 | * accurate if the file is shared by different jobs. | |
212 | */ | |
213 | char f_jobid[JOBSTATS_JOBID_SIZE]; | |
214 | } f; | |
215 | ||
47a57bde | 216 | #define lli_size_mutex u.f.f_size_mutex |
d7e09d03 PT |
217 | #define lli_symlink_name u.f.f_symlink_name |
218 | #define lli_maxbytes u.f.f_maxbytes | |
219 | #define lli_trunc_sem u.f.f_trunc_sem | |
220 | #define lli_write_mutex u.f.f_write_mutex | |
221 | #define lli_glimpse_sem u.f.f_glimpse_sem | |
222 | #define lli_glimpse_time u.f.f_glimpse_time | |
223 | #define lli_agl_list u.f.f_agl_list | |
224 | #define lli_agl_index u.f.f_agl_index | |
225 | #define lli_async_rc u.f.f_async_rc | |
226 | #define lli_jobid u.f.f_jobid | |
d7e09d03 PT |
227 | |
228 | } u; | |
229 | ||
230 | /* XXX: For following frequent used members, although they maybe special | |
231 | * used for non-directory object, it is some time-wasting to check | |
232 | * whether the object is directory or not before using them. On the | |
233 | * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce | |
234 | * the "ll_inode_info" size even if moving those members into u.f. | |
235 | * So keep them out side. | |
236 | * | |
237 | * In the future, if more members are added only for directory, | |
238 | * some of the following members can be moved into u.f. | |
239 | */ | |
240 | bool lli_has_smd; | |
241 | struct cl_object *lli_clob; | |
242 | ||
243 | /* mutex to request for layout lock exclusively. */ | |
244 | struct mutex lli_layout_mutex; | |
09aed8a5 | 245 | /* Layout version, protected by lli_layout_lock */ |
d7e09d03 | 246 | __u32 lli_layout_gen; |
09aed8a5 | 247 | spinlock_t lli_layout_lock; |
7fc1f831 AP |
248 | |
249 | struct rw_semaphore lli_xattrs_list_rwsem; | |
250 | struct mutex lli_xattrs_enq_lock; | |
251 | struct list_head lli_xattrs;/* ll_xattr_entry->xe_list */ | |
d7e09d03 PT |
252 | }; |
253 | ||
09aed8a5 JX |
254 | static inline __u32 ll_layout_version_get(struct ll_inode_info *lli) |
255 | { | |
256 | __u32 gen; | |
257 | ||
258 | spin_lock(&lli->lli_layout_lock); | |
259 | gen = lli->lli_layout_gen; | |
260 | spin_unlock(&lli->lli_layout_lock); | |
261 | ||
262 | return gen; | |
263 | } | |
264 | ||
265 | static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen) | |
266 | { | |
267 | spin_lock(&lli->lli_layout_lock); | |
268 | lli->lli_layout_gen = gen; | |
269 | spin_unlock(&lli->lli_layout_lock); | |
270 | } | |
271 | ||
7fc1f831 AP |
272 | int ll_xattr_cache_destroy(struct inode *inode); |
273 | ||
e15ba45d OD |
274 | int ll_xattr_cache_get(struct inode *inode, const char *name, |
275 | char *buffer, size_t size, __u64 valid); | |
7fc1f831 | 276 | |
d7e09d03 PT |
277 | /* |
278 | * Locking to guarantee consistency of non-atomic updates to long long i_size, | |
279 | * consistency between file size and KMS. | |
280 | * | |
47a57bde | 281 | * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order. |
d7e09d03 PT |
282 | */ |
283 | ||
284 | void ll_inode_size_lock(struct inode *inode); | |
285 | void ll_inode_size_unlock(struct inode *inode); | |
286 | ||
995c8b4a GD |
287 | /* FIXME: replace the name of this with LL_I to conform to kernel stuff */ |
288 | /* static inline struct ll_inode_info *LL_I(struct inode *inode) */ | |
d7e09d03 PT |
289 | static inline struct ll_inode_info *ll_i2info(struct inode *inode) |
290 | { | |
291 | return container_of(inode, struct ll_inode_info, lli_vfs_inode); | |
292 | } | |
293 | ||
294 | /* default to about 40meg of readahead on a given system. That much tied | |
c0894c6c OD |
295 | * up in 512k readahead requests serviced at 40ms each is about 1GB/s. |
296 | */ | |
09cbfeaf | 297 | #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_SHIFT)) |
d7e09d03 PT |
298 | |
299 | /* default to read-ahead full files smaller than 2MB on the second read */ | |
09cbfeaf | 300 | #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_SHIFT)) |
d7e09d03 PT |
301 | |
302 | enum ra_stat { | |
303 | RA_STAT_HIT = 0, | |
304 | RA_STAT_MISS, | |
305 | RA_STAT_DISTANT_READPAGE, | |
306 | RA_STAT_MISS_IN_WINDOW, | |
307 | RA_STAT_FAILED_GRAB_PAGE, | |
308 | RA_STAT_FAILED_MATCH, | |
309 | RA_STAT_DISCARDED, | |
310 | RA_STAT_ZERO_LEN, | |
311 | RA_STAT_ZERO_WINDOW, | |
312 | RA_STAT_EOF, | |
313 | RA_STAT_MAX_IN_FLIGHT, | |
314 | RA_STAT_WRONG_GRAB_PAGE, | |
fd7444fe | 315 | RA_STAT_FAILED_REACH_END, |
d7e09d03 PT |
316 | _NR_RA_STAT, |
317 | }; | |
318 | ||
319 | struct ll_ra_info { | |
320 | atomic_t ra_cur_pages; | |
321 | unsigned long ra_max_pages; | |
322 | unsigned long ra_max_pages_per_file; | |
323 | unsigned long ra_max_read_ahead_whole_pages; | |
324 | }; | |
325 | ||
326 | /* ra_io_arg will be filled in the beginning of ll_readahead with | |
327 | * ras_lock, then the following ll_read_ahead_pages will read RA | |
328 | * pages according to this arg, all the items in this structure are | |
329 | * counted by page index. | |
330 | */ | |
331 | struct ra_io_arg { | |
332 | unsigned long ria_start; /* start offset of read-ahead*/ | |
333 | unsigned long ria_end; /* end offset of read-ahead*/ | |
334 | /* If stride read pattern is detected, ria_stoff means where | |
335 | * stride read is started. Note: for normal read-ahead, the | |
c0894c6c OD |
336 | * value here is meaningless, and also it will not be accessed |
337 | */ | |
d7e09d03 PT |
338 | pgoff_t ria_stoff; |
339 | /* ria_length and ria_pages are the length and pages length in the | |
340 | * stride I/O mode. And they will also be used to check whether | |
c0894c6c OD |
341 | * it is stride I/O read-ahead in the read-ahead pages |
342 | */ | |
d7e09d03 PT |
343 | unsigned long ria_length; |
344 | unsigned long ria_pages; | |
345 | }; | |
346 | ||
347 | /* LL_HIST_MAX=32 causes an overflow */ | |
348 | #define LL_HIST_MAX 28 | |
349 | #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */ | |
350 | #define LL_PROCESS_HIST_MAX 10 | |
351 | struct per_process_info { | |
352 | pid_t pid; | |
353 | struct obd_histogram pp_r_hist; | |
354 | struct obd_histogram pp_w_hist; | |
355 | }; | |
356 | ||
357 | /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */ | |
358 | struct ll_rw_extents_info { | |
359 | struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1]; | |
360 | }; | |
361 | ||
362 | #define LL_OFFSET_HIST_MAX 100 | |
363 | struct ll_rw_process_info { | |
364 | pid_t rw_pid; | |
365 | int rw_op; | |
366 | loff_t rw_range_start; | |
367 | loff_t rw_range_end; | |
368 | loff_t rw_last_file_pos; | |
369 | loff_t rw_offset; | |
370 | size_t rw_smallest_extent; | |
371 | size_t rw_largest_extent; | |
372 | struct ll_file_data *rw_last_file; | |
373 | }; | |
374 | ||
375 | enum stats_track_type { | |
376 | STATS_TRACK_ALL = 0, /* track all processes */ | |
377 | STATS_TRACK_PID, /* track process with this pid */ | |
378 | STATS_TRACK_PPID, /* track processes with this ppid */ | |
379 | STATS_TRACK_GID, /* track processes with this gid */ | |
380 | STATS_TRACK_LAST, | |
381 | }; | |
382 | ||
383 | /* flags for sbi->ll_flags */ | |
384 | #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */ | |
385 | #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */ | |
386 | #define LL_SBI_FLOCK 0x04 | |
387 | #define LL_SBI_USER_XATTR 0x08 /* support user xattr */ | |
388 | #define LL_SBI_ACL 0x10 /* support ACL */ | |
341f1f0a | 389 | /* LL_SBI_RMT_CLIENT 0x40 remote client */ |
ef2e0f55 OD |
390 | #define LL_SBI_MDS_CAPA 0x80 /* support mds capa, obsolete */ |
391 | #define LL_SBI_OSS_CAPA 0x100 /* support oss capa, obsolete */ | |
d7e09d03 PT |
392 | #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */ |
393 | #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */ | |
394 | #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */ | |
395 | #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */ | |
396 | #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */ | |
397 | #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */ | |
398 | #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */ | |
399 | #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */ | |
400 | #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */ | |
401 | #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */ | |
7fc1f831 | 402 | #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */ |
d7e09d03 PT |
403 | |
404 | #define LL_SBI_FLAGS { \ | |
405 | "nolck", \ | |
406 | "checksum", \ | |
407 | "flock", \ | |
408 | "xattr", \ | |
409 | "acl", \ | |
7fc1f831 | 410 | "???", \ |
341f1f0a | 411 | "???", \ |
d7e09d03 PT |
412 | "mds_capa", \ |
413 | "oss_capa", \ | |
414 | "flock", \ | |
415 | "lru_resize", \ | |
416 | "lazy_statfs", \ | |
417 | "som", \ | |
418 | "32bit_api", \ | |
419 | "64bit_hash", \ | |
420 | "agl", \ | |
421 | "verbose", \ | |
422 | "layout", \ | |
7fc1f831 AP |
423 | "user_fid2path",\ |
424 | "xattr", \ | |
425 | } | |
d7e09d03 | 426 | |
d7e09d03 | 427 | struct ll_sb_info { |
d7e09d03 | 428 | /* this protects pglist and ra_info. It isn't safe to |
c0894c6c OD |
429 | * grab from interrupt contexts |
430 | */ | |
d7e09d03 PT |
431 | spinlock_t ll_lock; |
432 | spinlock_t ll_pp_extent_lock; /* pp_extent entry*/ | |
433 | spinlock_t ll_process_lock; /* ll_rw_process_info */ | |
434 | struct obd_uuid ll_sb_uuid; | |
435 | struct obd_export *ll_md_exp; | |
436 | struct obd_export *ll_dt_exp; | |
ae7c0f48 | 437 | struct dentry *ll_debugfs_entry; |
d7e09d03 PT |
438 | struct lu_fid ll_root_fid; /* root object fid */ |
439 | ||
440 | int ll_flags; | |
7fc1f831 AP |
441 | unsigned int ll_umounting:1, |
442 | ll_xattr_cache_enabled:1; | |
d7e09d03 PT |
443 | struct lustre_client_ocd ll_lco; |
444 | ||
d7e09d03 PT |
445 | struct ll_close_queue *ll_lcq; |
446 | ||
447 | struct lprocfs_stats *ll_stats; /* lprocfs stats counter */ | |
448 | ||
ac5b1481 PS |
449 | /* |
450 | * Used to track "unstable" pages on a client, and maintain a | |
451 | * LRU list of clean pages. An "unstable" page is defined as | |
452 | * any page which is sent to a server as part of a bulk request, | |
453 | * but is uncommitted to stable storage. | |
454 | */ | |
1b02bde3 | 455 | struct cl_client_cache *ll_cache; |
d7e09d03 PT |
456 | |
457 | struct lprocfs_stats *ll_ra_stats; | |
458 | ||
459 | struct ll_ra_info ll_ra_info; | |
460 | unsigned int ll_namelen; | |
461 | struct file_operations *ll_fop; | |
462 | ||
d7e09d03 PT |
463 | unsigned int ll_md_brw_size; /* used by readdir */ |
464 | ||
465 | struct lu_site *ll_site; | |
466 | struct cl_device *ll_cl; | |
467 | /* Statistics */ | |
468 | struct ll_rw_extents_info ll_rw_extents_info; | |
469 | int ll_extent_process_count; | |
470 | struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX]; | |
471 | unsigned int ll_offset_process_count; | |
472 | struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX]; | |
473 | unsigned int ll_rw_offset_entry_count; | |
474 | int ll_stats_track_id; | |
475 | enum stats_track_type ll_stats_track_type; | |
476 | int ll_rw_stats_on; | |
477 | ||
478 | /* metadata stat-ahead */ | |
479 | unsigned int ll_sa_max; /* max statahead RPCs */ | |
480 | atomic_t ll_sa_total; /* statahead thread started | |
c0894c6c OD |
481 | * count |
482 | */ | |
d7e09d03 | 483 | atomic_t ll_sa_wrong; /* statahead thread stopped for |
c0894c6c OD |
484 | * low hit ratio |
485 | */ | |
d7e09d03 PT |
486 | atomic_t ll_agl_total; /* AGL thread started count */ |
487 | ||
c0894c6c OD |
488 | dev_t ll_sdev_orig; /* save s_dev before assign for |
489 | * clustered nfs | |
490 | */ | |
bd994071 | 491 | __kernel_fsid_t ll_fsid; |
fd0d04ba OD |
492 | struct kobject ll_kobj; /* sysfs object */ |
493 | struct super_block *ll_sb; /* struct super_block (for sysfs code)*/ | |
494 | struct completion ll_kobj_unregister; | |
d7e09d03 PT |
495 | }; |
496 | ||
d7e09d03 PT |
497 | /* |
498 | * per file-descriptor read-ahead data. | |
499 | */ | |
500 | struct ll_readahead_state { | |
501 | spinlock_t ras_lock; | |
502 | /* | |
503 | * index of the last page that read(2) needed and that wasn't in the | |
504 | * cache. Used by ras_update() to detect seeks. | |
505 | * | |
506 | * XXX nikita: if access seeks into cached region, Lustre doesn't see | |
507 | * this. | |
508 | */ | |
509 | unsigned long ras_last_readpage; | |
510 | /* | |
511 | * number of pages read after last read-ahead window reset. As window | |
512 | * is reset on each seek, this is effectively a number of consecutive | |
513 | * accesses. Maybe ->ras_accessed_in_window is better name. | |
514 | * | |
515 | * XXX nikita: window is also reset (by ras_update()) when Lustre | |
516 | * believes that memory pressure evicts read-ahead pages. In that | |
517 | * case, it probably doesn't make sense to expand window to | |
518 | * PTLRPC_MAX_BRW_PAGES on the third access. | |
519 | */ | |
520 | unsigned long ras_consecutive_pages; | |
521 | /* | |
522 | * number of read requests after the last read-ahead window reset | |
523 | * As window is reset on each seek, this is effectively the number | |
524 | * on consecutive read request and is used to trigger read-ahead. | |
525 | */ | |
526 | unsigned long ras_consecutive_requests; | |
527 | /* | |
528 | * Parameters of current read-ahead window. Handled by | |
529 | * ras_update(). On the initial access to the file or after a seek, | |
530 | * window is reset to 0. After 3 consecutive accesses, window is | |
531 | * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by | |
532 | * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages. | |
533 | */ | |
534 | unsigned long ras_window_start, ras_window_len; | |
535 | /* | |
536 | * Where next read-ahead should start at. This lies within read-ahead | |
537 | * window. Read-ahead window is read in pieces rather than at once | |
538 | * because: 1. lustre limits total number of pages under read-ahead by | |
539 | * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages | |
540 | * not covered by DLM lock. | |
541 | */ | |
542 | unsigned long ras_next_readahead; | |
543 | /* | |
544 | * Total number of ll_file_read requests issued, reads originating | |
545 | * due to mmap are not counted in this total. This value is used to | |
546 | * trigger full file read-ahead after multiple reads to a small file. | |
547 | */ | |
548 | unsigned long ras_requests; | |
549 | /* | |
550 | * Page index with respect to the current request, these value | |
551 | * will not be accurate when dealing with reads issued via mmap. | |
552 | */ | |
553 | unsigned long ras_request_index; | |
d7e09d03 PT |
554 | /* |
555 | * The following 3 items are used for detecting the stride I/O | |
556 | * mode. | |
557 | * In stride I/O mode, | |
558 | * ...............|-----data-----|****gap*****|--------|******|.... | |
559 | * offset |-stride_pages-|-stride_gap-| | |
560 | * ras_stride_offset = offset; | |
561 | * ras_stride_length = stride_pages + stride_gap; | |
562 | * ras_stride_pages = stride_pages; | |
563 | * Note: all these three items are counted by pages. | |
564 | */ | |
565 | unsigned long ras_stride_length; | |
566 | unsigned long ras_stride_pages; | |
567 | pgoff_t ras_stride_offset; | |
568 | /* | |
569 | * number of consecutive stride request count, and it is similar as | |
570 | * ras_consecutive_requests, but used for stride I/O mode. | |
571 | * Note: only more than 2 consecutive stride request are detected, | |
572 | * stride read-ahead will be enable | |
573 | */ | |
574 | unsigned long ras_consecutive_stride_requests; | |
575 | }; | |
576 | ||
577 | extern struct kmem_cache *ll_file_data_slab; | |
578 | struct lustre_handle; | |
579 | struct ll_file_data { | |
580 | struct ll_readahead_state fd_ras; | |
98eae5e7 | 581 | struct ll_grouplock fd_grouplock; |
d7e09d03 PT |
582 | __u64 lfd_pos; |
583 | __u32 fd_flags; | |
d3a8a4e2 JX |
584 | fmode_t fd_omode; |
585 | /* openhandle if lease exists for this file. | |
c0894c6c OD |
586 | * Borrow lli->lli_och_mutex to protect assignment |
587 | */ | |
d3a8a4e2 JX |
588 | struct obd_client_handle *fd_lease_och; |
589 | struct obd_client_handle *fd_och; | |
d7e09d03 PT |
590 | struct file *fd_file; |
591 | /* Indicate whether need to report failure when close. | |
592 | * true: failure is known, not report again. | |
c0894c6c OD |
593 | * false: unknown failure, should report. |
594 | */ | |
d7e09d03 | 595 | bool fd_write_failed; |
966c4a8f JX |
596 | rwlock_t fd_lock; /* protect lcc list */ |
597 | struct list_head fd_lccs; /* list of ll_cl_context */ | |
d7e09d03 PT |
598 | }; |
599 | ||
600 | struct lov_stripe_md; | |
601 | ||
ae7c0f48 | 602 | extern struct dentry *llite_root; |
fd0d04ba | 603 | extern struct kset *llite_kset; |
d7e09d03 PT |
604 | |
605 | static inline struct inode *ll_info2i(struct ll_inode_info *lli) | |
606 | { | |
607 | return &lli->lli_vfs_inode; | |
608 | } | |
609 | ||
d7e09d03 | 610 | __u32 ll_i2suppgid(struct inode *i); |
1d8cb70c | 611 | void ll_i2gids(__u32 *suppgids, struct inode *i1, struct inode *i2); |
d7e09d03 PT |
612 | |
613 | static inline int ll_need_32bit_api(struct ll_sb_info *sbi) | |
614 | { | |
615 | #if BITS_PER_LONG == 32 | |
616 | return 1; | |
2d95f10e | 617 | #elif defined(CONFIG_COMPAT) |
6d8bedff | 618 | return unlikely(in_compat_syscall() || (sbi->ll_flags & LL_SBI_32BIT_API)); |
d7e09d03 | 619 | #else |
2d95f10e | 620 | return unlikely(sbi->ll_flags & LL_SBI_32BIT_API); |
d7e09d03 PT |
621 | #endif |
622 | } | |
623 | ||
bc4320a9 | 624 | void ll_ras_enter(struct file *f); |
d7e09d03 | 625 | |
98eae5e7 | 626 | /* llite/lcommon_misc.c */ |
5c5af0fc JH |
627 | int cl_init_ea_size(struct obd_export *md_exp, struct obd_export *dt_exp); |
628 | int cl_ocd_update(struct obd_device *host, | |
629 | struct obd_device *watched, | |
630 | enum obd_notify_event ev, void *owner, void *data); | |
98eae5e7 JH |
631 | int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock, |
632 | struct ll_grouplock *cg); | |
633 | void cl_put_grouplock(struct ll_grouplock *cg); | |
d7e09d03 PT |
634 | |
635 | /* llite/lproc_llite.c */ | |
ae7c0f48 OD |
636 | int ldebugfs_register_mountpoint(struct dentry *parent, |
637 | struct super_block *sb, char *osc, char *mdc); | |
638 | void ldebugfs_unregister_mountpoint(struct ll_sb_info *sbi); | |
d7e09d03 PT |
639 | void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count); |
640 | void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars); | |
2c185ffa PT |
641 | void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid, |
642 | struct ll_file_data *file, loff_t pos, | |
643 | size_t count, int rw); | |
d7e09d03 PT |
644 | |
645 | /* llite/dir.c */ | |
646 | void ll_release_page(struct page *page, int remove); | |
2d95f10e JH |
647 | extern const struct file_operations ll_dir_operations; |
648 | extern const struct inode_operations ll_dir_inode_operations; | |
d7e09d03 PT |
649 | struct page *ll_get_dir_page(struct inode *dir, __u64 hash, |
650 | struct ll_dir_chain *chain); | |
0b09d381 | 651 | int ll_dir_read(struct inode *inode, struct dir_context *ctx); |
d7e09d03 PT |
652 | |
653 | int ll_get_mdt_idx(struct inode *inode); | |
654 | /* llite/namei.c */ | |
2d95f10e JH |
655 | extern const struct inode_operations ll_special_inode_operations; |
656 | ||
d7e09d03 PT |
657 | int ll_objects_destroy(struct ptlrpc_request *request, |
658 | struct inode *dir); | |
659 | struct inode *ll_iget(struct super_block *sb, ino_t hash, | |
660 | struct lustre_md *lic); | |
661 | int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *, | |
662 | void *data, int flag); | |
663 | struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de); | |
d7e09d03 PT |
664 | |
665 | /* llite/rw.c */ | |
d7e09d03 PT |
666 | int ll_writepage(struct page *page, struct writeback_control *wbc); |
667 | int ll_writepages(struct address_space *, struct writeback_control *wbc); | |
d7e09d03 PT |
668 | int ll_readpage(struct file *file, struct page *page); |
669 | void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras); | |
d7e09d03 | 670 | int ll_readahead(const struct lu_env *env, struct cl_io *io, |
fd7444fe JX |
671 | struct cl_page_list *queue, struct ll_readahead_state *ras, |
672 | bool hit); | |
966c4a8f JX |
673 | struct ll_cl_context *ll_cl_find(struct file *file); |
674 | void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io); | |
675 | void ll_cl_remove(struct file *file, const struct lu_env *env); | |
d7e09d03 | 676 | |
2d95f10e | 677 | extern const struct address_space_operations ll_aops; |
2d95f10e | 678 | |
d7e09d03 PT |
679 | /* llite/file.c */ |
680 | extern struct file_operations ll_file_operations; | |
681 | extern struct file_operations ll_file_operations_flock; | |
682 | extern struct file_operations ll_file_operations_noflock; | |
d2d32738 | 683 | extern const struct inode_operations ll_file_inode_operations; |
8150a97f | 684 | int ll_have_md_lock(struct inode *inode, __u64 *bits, |
52ee0d20 OD |
685 | enum ldlm_mode l_req_mode); |
686 | enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits, | |
687 | struct lustre_handle *lockh, __u64 flags, | |
688 | enum ldlm_mode mode); | |
d7e09d03 PT |
689 | int ll_file_open(struct inode *inode, struct file *file); |
690 | int ll_file_release(struct inode *inode, struct file *file); | |
691 | int ll_glimpse_ioctl(struct ll_sb_info *sbi, | |
692 | struct lov_stripe_md *lsm, lstat_t *st); | |
693 | void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch); | |
e22fdcc8 | 694 | int ll_release_openhandle(struct inode *, struct lookup_intent *); |
45b2a010 | 695 | int ll_md_real_close(struct inode *inode, fmode_t fmode); |
d7e09d03 PT |
696 | void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data, |
697 | struct obd_client_handle **och, unsigned long flags); | |
698 | void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data); | |
699 | int ll_som_update(struct inode *inode, struct md_op_data *op_data); | |
700 | int ll_inode_getattr(struct inode *inode, struct obdo *obdo, | |
701 | __u64 ioepoch, int sync); | |
d7e09d03 PT |
702 | void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data, |
703 | struct lustre_handle *fh); | |
d7e09d03 | 704 | int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat); |
2d95f10e | 705 | struct posix_acl *ll_get_acl(struct inode *inode, int type); |
d7e09d03 PT |
706 | |
707 | int ll_inode_permission(struct inode *inode, int mask); | |
708 | ||
c139f3ce | 709 | int ll_lov_setstripe_ea_info(struct inode *inode, struct dentry *dentry, |
d467220e | 710 | __u64 flags, struct lov_user_md *lum, |
d7e09d03 PT |
711 | int lum_size); |
712 | int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename, | |
713 | struct lov_mds_md **lmm, int *lmm_size, | |
714 | struct ptlrpc_request **request); | |
715 | int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump, | |
716 | int set_default); | |
717 | int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmmp, | |
718 | int *lmm_size, struct ptlrpc_request **request); | |
719 | int ll_fsync(struct file *file, loff_t start, loff_t end, int data); | |
d2995737 | 720 | int ll_merge_attr(const struct lu_env *env, struct inode *inode); |
2b358b4e | 721 | int ll_fid2path(struct inode *inode, void __user *arg); |
e1798006 | 722 | int ll_data_version(struct inode *inode, __u64 *data_version, int flags); |
48d23e61 | 723 | int ll_hsm_release(struct inode *inode); |
d7e09d03 PT |
724 | |
725 | /* llite/dcache.c */ | |
726 | ||
3ea8f3bc | 727 | int ll_d_init(struct dentry *de); |
2d95f10e | 728 | extern const struct dentry_operations ll_d_ops; |
d7e09d03 PT |
729 | void ll_intent_drop_lock(struct lookup_intent *); |
730 | void ll_intent_release(struct lookup_intent *); | |
731 | void ll_invalidate_aliases(struct inode *); | |
dbca51dd | 732 | void ll_lookup_finish_locks(struct lookup_intent *it, struct inode *inode); |
d7e09d03 | 733 | int ll_revalidate_it_finish(struct ptlrpc_request *request, |
dbca51dd | 734 | struct lookup_intent *it, struct inode *inode); |
d7e09d03 PT |
735 | |
736 | /* llite/llite_lib.c */ | |
737 | extern struct super_operations lustre_super_operations; | |
738 | ||
d7e09d03 PT |
739 | void ll_lli_init(struct ll_inode_info *lli); |
740 | int ll_fill_super(struct super_block *sb, struct vfsmount *mnt); | |
741 | void ll_put_super(struct super_block *sb); | |
742 | void ll_kill_super(struct super_block *sb); | |
743 | struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock); | |
d7e09d03 | 744 | void ll_clear_inode(struct inode *inode); |
a720b790 | 745 | int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import); |
d7e09d03 PT |
746 | int ll_setattr(struct dentry *de, struct iattr *attr); |
747 | int ll_statfs(struct dentry *de, struct kstatfs *sfs); | |
748 | int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs, | |
749 | __u64 max_age, __u32 flags); | |
750 | void ll_update_inode(struct inode *inode, struct lustre_md *md); | |
751 | void ll_read_inode2(struct inode *inode, void *opaque); | |
752 | void ll_delete_inode(struct inode *inode); | |
753 | int ll_iocontrol(struct inode *inode, struct file *file, | |
754 | unsigned int cmd, unsigned long arg); | |
755 | int ll_flush_ctx(struct inode *inode); | |
756 | void ll_umount_begin(struct super_block *sb); | |
757 | int ll_remount_fs(struct super_block *sb, int *flags, char *data); | |
758 | int ll_show_options(struct seq_file *seq, struct dentry *dentry); | |
759 | void ll_dirty_page_discard_warn(struct page *page, int ioret); | |
760 | int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req, | |
761 | struct super_block *, struct lookup_intent *); | |
4c6243ec | 762 | int ll_obd_statfs(struct inode *inode, void __user *arg); |
d7e09d03 | 763 | int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize); |
44779340 | 764 | int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize); |
d7e09d03 PT |
765 | int ll_process_config(struct lustre_cfg *lcfg); |
766 | struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data, | |
767 | struct inode *i1, struct inode *i2, | |
768 | const char *name, int namelen, | |
769 | int mode, __u32 opc, void *data); | |
770 | void ll_finish_md_op_data(struct md_op_data *op_data); | |
771 | int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg); | |
772 | char *ll_get_fsname(struct super_block *sb, char *buf, int buflen); | |
44ecac68 | 773 | void ll_open_cleanup(struct super_block *sb, struct ptlrpc_request *open_req); |
d7e09d03 PT |
774 | |
775 | /* llite/llite_nfs.c */ | |
98aa7661 | 776 | extern const struct export_operations lustre_export_operations; |
d7e09d03 | 777 | __u32 get_uuid2int(const char *name, int len); |
bd994071 | 778 | void get_uuid2fsid(const char *name, int len, __kernel_fsid_t *fsid); |
d7e09d03 PT |
779 | struct inode *search_inode_for_lustre(struct super_block *sb, |
780 | const struct lu_fid *fid); | |
781 | ||
d7e09d03 | 782 | /* llite/symlink.c */ |
d2d32738 | 783 | extern const struct inode_operations ll_fast_symlink_inode_operations; |
d7e09d03 PT |
784 | |
785 | /* llite/llite_close.c */ | |
786 | struct ll_close_queue { | |
787 | spinlock_t lcq_lock; | |
788 | struct list_head lcq_head; | |
789 | wait_queue_head_t lcq_waitq; | |
790 | struct completion lcq_comp; | |
791 | atomic_t lcq_stop; | |
792 | }; | |
793 | ||
3a52f803 JH |
794 | void vvp_write_pending(struct vvp_object *club, struct vvp_page *page); |
795 | void vvp_write_complete(struct vvp_object *club, struct vvp_page *page); | |
d7e09d03 PT |
796 | |
797 | /** | |
798 | * IO arguments for various VFS I/O interfaces. | |
799 | */ | |
800 | struct vvp_io_args { | |
74c0da19 | 801 | /** normal/splice */ |
d7e09d03 PT |
802 | enum vvp_io_subtype via_io_subtype; |
803 | ||
804 | union { | |
805 | struct { | |
806 | struct kiocb *via_iocb; | |
b42b15fd | 807 | struct iov_iter *via_iter; |
d7e09d03 | 808 | } normal; |
d7e09d03 PT |
809 | struct { |
810 | struct pipe_inode_info *via_pipe; | |
811 | unsigned int via_flags; | |
812 | } splice; | |
813 | } u; | |
814 | }; | |
815 | ||
816 | struct ll_cl_context { | |
966c4a8f | 817 | struct list_head lcc_list; |
d7e09d03 | 818 | void *lcc_cookie; |
966c4a8f | 819 | const struct lu_env *lcc_env; |
d7e09d03 PT |
820 | struct cl_io *lcc_io; |
821 | struct cl_page *lcc_page; | |
d7e09d03 PT |
822 | }; |
823 | ||
9989a58e JH |
824 | struct ll_thread_info { |
825 | struct vvp_io_args lti_args; | |
826 | struct ra_io_arg lti_ria; | |
827 | struct ll_cl_context lti_io_ctx; | |
d7e09d03 PT |
828 | }; |
829 | ||
9989a58e JH |
830 | extern struct lu_context_key ll_thread_key; |
831 | static inline struct ll_thread_info *ll_env_info(const struct lu_env *env) | |
d7e09d03 | 832 | { |
9989a58e | 833 | struct ll_thread_info *lti; |
d7e09d03 | 834 | |
9989a58e JH |
835 | lti = lu_context_key_get(&env->le_ctx, &ll_thread_key); |
836 | LASSERT(lti); | |
837 | return lti; | |
d7e09d03 PT |
838 | } |
839 | ||
9989a58e JH |
840 | static inline struct vvp_io_args *ll_env_args(const struct lu_env *env, |
841 | enum vvp_io_subtype type) | |
d7e09d03 | 842 | { |
9989a58e | 843 | struct vvp_io_args *via = &ll_env_info(env)->lti_args; |
d7e09d03 | 844 | |
9989a58e | 845 | via->via_io_subtype = type; |
d7e09d03 | 846 | |
9989a58e | 847 | return via; |
d7e09d03 PT |
848 | } |
849 | ||
850 | void ll_queue_done_writing(struct inode *inode, unsigned long flags); | |
851 | void ll_close_thread_shutdown(struct ll_close_queue *lcq); | |
852 | int ll_close_thread_start(struct ll_close_queue **lcq_ret); | |
853 | ||
854 | /* llite/llite_mmap.c */ | |
d7e09d03 PT |
855 | |
856 | int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last); | |
aff9d8e8 | 857 | int ll_file_mmap(struct file *file, struct vm_area_struct *vma); |
e15ba45d OD |
858 | void policy_from_vma(ldlm_policy_data_t *policy, struct vm_area_struct *vma, |
859 | unsigned long addr, size_t count); | |
d7e09d03 PT |
860 | struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr, |
861 | size_t count); | |
862 | ||
863 | static inline void ll_invalidate_page(struct page *vmpage) | |
864 | { | |
865 | struct address_space *mapping = vmpage->mapping; | |
09cbfeaf | 866 | loff_t offset = vmpage->index << PAGE_SHIFT; |
d7e09d03 PT |
867 | |
868 | LASSERT(PageLocked(vmpage)); | |
6e16818b | 869 | if (!mapping) |
d7e09d03 PT |
870 | return; |
871 | ||
7addf402 JX |
872 | /* |
873 | * truncate_complete_page() calls | |
874 | * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete(). | |
875 | */ | |
09cbfeaf | 876 | ll_teardown_mmaps(mapping, offset, offset + PAGE_SIZE); |
d7e09d03 PT |
877 | truncate_complete_page(mapping, vmpage); |
878 | } | |
879 | ||
880 | #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi) | |
881 | ||
882 | /* don't need an addref as the sb_info should be holding one */ | |
883 | static inline struct obd_export *ll_s2dtexp(struct super_block *sb) | |
884 | { | |
885 | return ll_s2sbi(sb)->ll_dt_exp; | |
886 | } | |
887 | ||
888 | /* don't need an addref as the sb_info should be holding one */ | |
889 | static inline struct obd_export *ll_s2mdexp(struct super_block *sb) | |
890 | { | |
891 | return ll_s2sbi(sb)->ll_md_exp; | |
892 | } | |
893 | ||
894 | static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi) | |
895 | { | |
896 | struct obd_device *obd = sbi->ll_md_exp->exp_obd; | |
50ffcb7e | 897 | |
6e16818b | 898 | if (!obd) |
d7e09d03 PT |
899 | LBUG(); |
900 | return &obd->u.cli; | |
901 | } | |
902 | ||
995c8b4a | 903 | /* FIXME: replace the name of this with LL_SB to conform to kernel stuff */ |
d7e09d03 PT |
904 | static inline struct ll_sb_info *ll_i2sbi(struct inode *inode) |
905 | { | |
906 | return ll_s2sbi(inode->i_sb); | |
907 | } | |
908 | ||
909 | static inline struct obd_export *ll_i2dtexp(struct inode *inode) | |
910 | { | |
911 | return ll_s2dtexp(inode->i_sb); | |
912 | } | |
913 | ||
914 | static inline struct obd_export *ll_i2mdexp(struct inode *inode) | |
915 | { | |
916 | return ll_s2mdexp(inode->i_sb); | |
917 | } | |
918 | ||
919 | static inline struct lu_fid *ll_inode2fid(struct inode *inode) | |
920 | { | |
921 | struct lu_fid *fid; | |
922 | ||
6e16818b | 923 | LASSERT(inode); |
d7e09d03 PT |
924 | fid = &ll_i2info(inode)->lli_fid; |
925 | ||
926 | return fid; | |
927 | } | |
928 | ||
d7e09d03 PT |
929 | static inline __u64 ll_file_maxbytes(struct inode *inode) |
930 | { | |
931 | return ll_i2info(inode)->lli_maxbytes; | |
932 | } | |
933 | ||
934 | /* llite/xattr.c */ | |
3767e255 AV |
935 | int ll_setxattr(struct dentry *dentry, struct inode *inode, |
936 | const char *name, const void *value, size_t size, int flags); | |
ce23e640 AV |
937 | ssize_t ll_getxattr(struct dentry *dentry, struct inode *inode, |
938 | const char *name, void *buffer, size_t size); | |
d7e09d03 PT |
939 | ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size); |
940 | int ll_removexattr(struct dentry *dentry, const char *name); | |
941 | ||
d7e09d03 PT |
942 | /** |
943 | * Common IO arguments for various VFS I/O interfaces. | |
944 | */ | |
945 | int cl_sb_init(struct super_block *sb); | |
946 | int cl_sb_fini(struct super_block *sb); | |
d7e09d03 PT |
947 | void ll_io_init(struct cl_io *io, const struct file *file, int write); |
948 | ||
949 | void ras_update(struct ll_sb_info *sbi, struct inode *inode, | |
950 | struct ll_readahead_state *ras, unsigned long index, | |
951 | unsigned hit); | |
952 | void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len); | |
fd7444fe | 953 | void ll_ra_stats_inc(struct inode *inode, enum ra_stat which); |
d7e09d03 | 954 | |
d7e09d03 | 955 | /* statahead.c */ |
d7e09d03 PT |
956 | #define LL_SA_RPC_MIN 2 |
957 | #define LL_SA_RPC_DEF 32 | |
958 | #define LL_SA_RPC_MAX 8192 | |
959 | ||
960 | #define LL_SA_CACHE_BIT 5 | |
961 | #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT) | |
962 | #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1) | |
963 | ||
964 | /* per inode struct, for dir only */ | |
965 | struct ll_statahead_info { | |
966 | struct inode *sai_inode; | |
967 | atomic_t sai_refcount; /* when access this struct, hold | |
c0894c6c OD |
968 | * refcount |
969 | */ | |
d7e09d03 PT |
970 | unsigned int sai_generation; /* generation for statahead */ |
971 | unsigned int sai_max; /* max ahead of lookup */ | |
972 | __u64 sai_sent; /* stat requests sent count */ | |
973 | __u64 sai_replied; /* stat requests which received | |
c0894c6c OD |
974 | * reply |
975 | */ | |
d7e09d03 PT |
976 | __u64 sai_index; /* index of statahead entry */ |
977 | __u64 sai_index_wait; /* index of entry which is the | |
c0894c6c OD |
978 | * caller is waiting for |
979 | */ | |
d7e09d03 PT |
980 | __u64 sai_hit; /* hit count */ |
981 | __u64 sai_miss; /* miss count: | |
c0894c6c OD |
982 | * for "ls -al" case, it includes |
983 | * hidden dentry miss; | |
984 | * for "ls -l" case, it does not | |
985 | * include hidden dentry miss. | |
986 | * "sai_miss_hidden" is used for | |
987 | * the later case. | |
988 | */ | |
d7e09d03 PT |
989 | unsigned int sai_consecutive_miss; /* consecutive miss */ |
990 | unsigned int sai_miss_hidden;/* "ls -al", but first dentry | |
c0894c6c OD |
991 | * is not a hidden one |
992 | */ | |
d7e09d03 PT |
993 | unsigned int sai_skip_hidden;/* skipped hidden dentry count */ |
994 | unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for | |
c0894c6c OD |
995 | * hidden entries |
996 | */ | |
d7e09d03 | 997 | sai_agl_valid:1;/* AGL is valid for the dir */ |
c0894c6c | 998 | wait_queue_head_t sai_waitq; /* stat-ahead wait queue */ |
d7e09d03 PT |
999 | struct ptlrpc_thread sai_thread; /* stat-ahead thread */ |
1000 | struct ptlrpc_thread sai_agl_thread; /* AGL thread */ | |
c0894c6c OD |
1001 | struct list_head sai_entries; /* entry list */ |
1002 | struct list_head sai_entries_received; /* entries returned */ | |
1003 | struct list_head sai_entries_stated; /* entries stated */ | |
1004 | struct list_head sai_entries_agl; /* AGL entries to be sent */ | |
1005 | struct list_head sai_cache[LL_SA_CACHE_SIZE]; | |
d7e09d03 PT |
1006 | spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE]; |
1007 | atomic_t sai_cache_count; /* entry count in cache */ | |
1008 | }; | |
1009 | ||
1010 | int do_statahead_enter(struct inode *dir, struct dentry **dentry, | |
1011 | int only_unplug); | |
1012 | void ll_stop_statahead(struct inode *dir, void *key); | |
1013 | ||
5c5af0fc JH |
1014 | blkcnt_t dirty_cnt(struct inode *inode); |
1015 | ||
1016 | int cl_glimpse_size0(struct inode *inode, int agl); | |
1017 | int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io, | |
1018 | struct inode *inode, struct cl_object *clob, int agl); | |
1019 | ||
1020 | static inline int cl_glimpse_size(struct inode *inode) | |
1021 | { | |
1022 | return cl_glimpse_size0(inode, 0); | |
1023 | } | |
1024 | ||
1025 | static inline int cl_agl(struct inode *inode) | |
1026 | { | |
1027 | return cl_glimpse_size0(inode, 1); | |
1028 | } | |
1029 | ||
d7e09d03 PT |
1030 | static inline int ll_glimpse_size(struct inode *inode) |
1031 | { | |
1032 | struct ll_inode_info *lli = ll_i2info(inode); | |
1033 | int rc; | |
1034 | ||
1035 | down_read(&lli->lli_glimpse_sem); | |
1036 | rc = cl_glimpse_size(inode); | |
1037 | lli->lli_glimpse_time = cfs_time_current(); | |
1038 | up_read(&lli->lli_glimpse_sem); | |
1039 | return rc; | |
1040 | } | |
1041 | ||
1042 | static inline void | |
1043 | ll_statahead_mark(struct inode *dir, struct dentry *dentry) | |
1044 | { | |
1045 | struct ll_inode_info *lli = ll_i2info(dir); | |
1046 | struct ll_statahead_info *sai = lli->lli_sai; | |
1047 | struct ll_dentry_data *ldd = ll_d2d(dentry); | |
1048 | ||
1049 | /* not the same process, don't mark */ | |
1050 | if (lli->lli_opendir_pid != current_pid()) | |
1051 | return; | |
1052 | ||
6e16818b OD |
1053 | LASSERT(ldd); |
1054 | if (sai) | |
d7e09d03 PT |
1055 | ldd->lld_sa_generation = sai->sai_generation; |
1056 | } | |
1057 | ||
1058 | static inline int | |
f236f69b | 1059 | d_need_statahead(struct inode *dir, struct dentry *dentryp) |
d7e09d03 PT |
1060 | { |
1061 | struct ll_inode_info *lli; | |
1062 | struct ll_dentry_data *ldd; | |
1063 | ||
1064 | if (ll_i2sbi(dir)->ll_sa_max == 0) | |
1065 | return -EAGAIN; | |
1066 | ||
1067 | lli = ll_i2info(dir); | |
1068 | /* not the same process, don't statahead */ | |
1069 | if (lli->lli_opendir_pid != current_pid()) | |
1070 | return -EAGAIN; | |
1071 | ||
1072 | /* statahead has been stopped */ | |
6e16818b | 1073 | if (!lli->lli_opendir_key) |
d7e09d03 PT |
1074 | return -EAGAIN; |
1075 | ||
1076 | ldd = ll_d2d(dentryp); | |
1077 | /* | |
1078 | * When stats a dentry, the system trigger more than once "revalidate" | |
1079 | * or "lookup", for "getattr", for "getxattr", and maybe for others. | |
1080 | * Under patchless client mode, the operation intent is not accurate, | |
1081 | * which maybe misguide the statahead thread. For example: | |
1082 | * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe | |
1083 | * have the same operation intent -- "IT_GETATTR". | |
1084 | * In fact, one dentry should has only one chance to interact with the | |
1085 | * statahead thread, otherwise the statahead windows will be confused. | |
1086 | * The solution is as following: | |
1087 | * Assign "lld_sa_generation" with "sai_generation" when a dentry | |
1088 | * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR" | |
1089 | * will bypass interacting with statahead thread for checking: | |
1090 | * "lld_sa_generation == lli_sai->sai_generation" | |
1091 | */ | |
1092 | if (ldd && lli->lli_sai && | |
1093 | ldd->lld_sa_generation == lli->lli_sai->sai_generation) | |
1094 | return -EAGAIN; | |
1095 | ||
1096 | return 1; | |
1097 | } | |
1098 | ||
1099 | static inline int | |
1100 | ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug) | |
1101 | { | |
1102 | int ret; | |
1103 | ||
f236f69b | 1104 | ret = d_need_statahead(dir, *dentryp); |
d7e09d03 PT |
1105 | if (ret <= 0) |
1106 | return ret; | |
1107 | ||
1108 | return do_statahead_enter(dir, dentryp, only_unplug); | |
1109 | } | |
1110 | ||
d0a0acc3 | 1111 | /* llite ioctl register support routine */ |
d7e09d03 PT |
1112 | enum llioc_iter { |
1113 | LLIOC_CONT = 0, | |
1114 | LLIOC_STOP | |
1115 | }; | |
1116 | ||
1117 | #define LLIOC_MAX_CMD 256 | |
1118 | ||
1119 | /* | |
1120 | * Rules to write a callback function: | |
1121 | * | |
1122 | * Parameters: | |
d0a0acc3 | 1123 | * @magic: Dynamic ioctl call routine will feed this value with the pointer |
d7e09d03 PT |
1124 | * returned to ll_iocontrol_register. Callback functions should use this |
1125 | * data to check the potential collasion of ioctl cmd. If collasion is | |
1126 | * found, callback function should return LLIOC_CONT. | |
1127 | * @rcp: The result of ioctl command. | |
1128 | * | |
1129 | * Return values: | |
1130 | * If @magic matches the pointer returned by ll_iocontrol_data, the | |
1131 | * callback should return LLIOC_STOP; return LLIOC_STOP otherwise. | |
1132 | */ | |
1133 | typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode, | |
1134 | struct file *file, unsigned int cmd, unsigned long arg, | |
1135 | void *magic, int *rcp); | |
1136 | ||
d7e09d03 PT |
1137 | /* export functions */ |
1138 | /* Register ioctl block dynamatically for a regular file. | |
1139 | * | |
1140 | * @cmd: the array of ioctl command set | |
1141 | * @count: number of commands in the @cmd | |
1142 | * @cb: callback function, it will be called if an ioctl command is found to | |
1143 | * belong to the command list @cmd. | |
1144 | * | |
d0a0acc3 | 1145 | * Return value: |
d7e09d03 PT |
1146 | * A magic pointer will be returned if success; |
1147 | * otherwise, NULL will be returned. | |
1148 | * */ | |
1149 | void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd); | |
1150 | void ll_iocontrol_unregister(void *magic); | |
1151 | ||
d7e09d03 | 1152 | int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end, |
65fb55d1 | 1153 | enum cl_fsync_mode mode, int ignore_layout); |
d7e09d03 PT |
1154 | |
1155 | /** direct write pages */ | |
1156 | struct ll_dio_pages { | |
1157 | /** page array to be written. we don't support | |
c0894c6c OD |
1158 | * partial pages except the last one. |
1159 | */ | |
d7e09d03 PT |
1160 | struct page **ldp_pages; |
1161 | /* offset of each page */ | |
1162 | loff_t *ldp_offsets; | |
1163 | /** if ldp_offsets is NULL, it means a sequential | |
1164 | * pages to be written, then this is the file offset | |
c0894c6c OD |
1165 | * of the first page. |
1166 | */ | |
d7e09d03 PT |
1167 | loff_t ldp_start_offset; |
1168 | /** how many bytes are to be written. */ | |
1169 | size_t ldp_size; | |
1170 | /** # of pages in the array. */ | |
1171 | int ldp_nr; | |
1172 | }; | |
1173 | ||
1174 | static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt, | |
1175 | int rc) | |
1176 | { | |
1177 | int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ : | |
1178 | LPROC_LL_OSC_WRITE; | |
1179 | ||
3c95b839 | 1180 | ll_stats_ops_tally(ll_s2sbi(cl2vvp_dev(dev)->vdv_sb), opc, rc); |
d7e09d03 PT |
1181 | } |
1182 | ||
8150a97f JP |
1183 | ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io, |
1184 | int rw, struct inode *inode, | |
1185 | struct ll_dio_pages *pv); | |
d7e09d03 PT |
1186 | |
1187 | static inline int ll_file_nolock(const struct file *file) | |
1188 | { | |
1189 | struct ll_file_data *fd = LUSTRE_FPRIVATE(file); | |
2a8a3597 | 1190 | struct inode *inode = file_inode(file); |
d7e09d03 | 1191 | |
d7e09d03 PT |
1192 | return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) || |
1193 | (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK)); | |
1194 | } | |
1195 | ||
1196 | static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode, | |
1197 | struct lookup_intent *it, __u64 *bits) | |
1198 | { | |
1199 | if (!it->d.lustre.it_lock_set) { | |
1200 | struct lustre_handle handle; | |
1201 | ||
1202 | /* If this inode is a remote object, it will get two | |
1203 | * separate locks in different namespaces, Master MDT, | |
1204 | * where the name entry is, will grant LOOKUP lock, | |
1205 | * remote MDT, where the object is, will grant | |
d0a0acc3 | 1206 | * UPDATE|PERM lock. The inode will be attached to both |
d7e09d03 | 1207 | * LOOKUP and PERM locks, so revoking either locks will |
c0894c6c OD |
1208 | * case the dcache being cleared |
1209 | */ | |
d7e09d03 PT |
1210 | if (it->d.lustre.it_remote_lock_mode) { |
1211 | handle.cookie = it->d.lustre.it_remote_lock_handle; | |
97a075cd JN |
1212 | CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"%p for remote lock %#llx\n", |
1213 | PFID(ll_inode2fid(inode)), inode, | |
d7e09d03 PT |
1214 | handle.cookie); |
1215 | md_set_lock_data(exp, &handle.cookie, inode, NULL); | |
1216 | } | |
1217 | ||
1218 | handle.cookie = it->d.lustre.it_lock_handle; | |
1219 | ||
97a075cd JN |
1220 | CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"%p for lock %#llx\n", |
1221 | PFID(ll_inode2fid(inode)), inode, handle.cookie); | |
d7e09d03 PT |
1222 | |
1223 | md_set_lock_data(exp, &handle.cookie, inode, | |
1224 | &it->d.lustre.it_lock_bits); | |
1225 | it->d.lustre.it_lock_set = 1; | |
1226 | } | |
1227 | ||
6e16818b | 1228 | if (bits) |
d7e09d03 PT |
1229 | *bits = it->d.lustre.it_lock_bits; |
1230 | } | |
1231 | ||
d7e09d03 PT |
1232 | static inline int d_lustre_invalid(const struct dentry *dentry) |
1233 | { | |
1234 | struct ll_dentry_data *lld = ll_d2d(dentry); | |
1235 | ||
6e16818b | 1236 | return !lld || lld->lld_invalid; |
d7e09d03 PT |
1237 | } |
1238 | ||
1239 | static inline void __d_lustre_invalidate(struct dentry *dentry) | |
1240 | { | |
1241 | struct ll_dentry_data *lld = ll_d2d(dentry); | |
1242 | ||
6e16818b | 1243 | if (lld) |
d7e09d03 PT |
1244 | lld->lld_invalid = 1; |
1245 | } | |
1246 | ||
1247 | /* | |
1248 | * Mark dentry INVALID, if dentry refcount is zero (this is normally case for | |
1249 | * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later; | |
1250 | * else dput() of the last refcount will unhash this dentry and kill it. | |
1251 | */ | |
b1d2a127 | 1252 | static inline void d_lustre_invalidate(struct dentry *dentry, int nested) |
d7e09d03 | 1253 | { |
dab363f9 LT |
1254 | CDEBUG(D_DENTRY, "invalidate dentry %pd (%p) parent %p inode %p refc %d\n", |
1255 | dentry, dentry, | |
2b0143b5 | 1256 | dentry->d_parent, d_inode(dentry), d_count(dentry)); |
d7e09d03 | 1257 | |
b1d2a127 PT |
1258 | spin_lock_nested(&dentry->d_lock, |
1259 | nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL); | |
d7e09d03 | 1260 | __d_lustre_invalidate(dentry); |
7f830d8d AP |
1261 | /* |
1262 | * We should be careful about dentries created by d_obtain_alias(). | |
1263 | * These dentries are not put in the dentry tree, instead they are | |
1264 | * linked to sb->s_anon through dentry->d_hash. | |
1265 | * shrink_dcache_for_umount() shrinks the tree and sb->s_anon list. | |
1266 | * If we unhashed such a dentry, unmount would not be able to find | |
1267 | * it and busy inodes would be reported. | |
1268 | */ | |
1269 | if (d_count(dentry) == 0 && !(dentry->d_flags & DCACHE_DISCONNECTED)) | |
d7e09d03 PT |
1270 | __d_drop(dentry); |
1271 | spin_unlock(&dentry->d_lock); | |
1272 | } | |
1273 | ||
1274 | static inline void d_lustre_revalidate(struct dentry *dentry) | |
1275 | { | |
1276 | spin_lock(&dentry->d_lock); | |
6e16818b | 1277 | LASSERT(ll_d2d(dentry)); |
d7e09d03 PT |
1278 | ll_d2d(dentry)->lld_invalid = 0; |
1279 | spin_unlock(&dentry->d_lock); | |
1280 | } | |
1281 | ||
d7e09d03 PT |
1282 | enum { |
1283 | LL_LAYOUT_GEN_NONE = ((__u32)-2), /* layout lock was cancelled */ | |
1284 | LL_LAYOUT_GEN_EMPTY = ((__u32)-1) /* for empty layout */ | |
1285 | }; | |
1286 | ||
1287 | int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf); | |
1288 | int ll_layout_refresh(struct inode *inode, __u32 *gen); | |
1b1594da | 1289 | int ll_layout_restore(struct inode *inode, loff_t start, __u64 length); |
d7e09d03 | 1290 | |
7fc1f831 AP |
1291 | int ll_xattr_init(void); |
1292 | void ll_xattr_fini(void); | |
1293 | ||
77605e41 JX |
1294 | int ll_page_sync_io(const struct lu_env *env, struct cl_io *io, |
1295 | struct cl_page *page, enum cl_req_type crt); | |
1296 | ||
a37bec74 | 1297 | /* lcommon_cl.c */ |
5c5af0fc JH |
1298 | int cl_setattr_ost(struct inode *inode, const struct iattr *attr); |
1299 | ||
a37bec74 JH |
1300 | extern struct lu_env *cl_inode_fini_env; |
1301 | extern int cl_inode_fini_refcheck; | |
1302 | ||
5c5af0fc JH |
1303 | int cl_file_inode_init(struct inode *inode, struct lustre_md *md); |
1304 | void cl_inode_fini(struct inode *inode); | |
1305 | int cl_local_size(struct inode *inode); | |
1306 | ||
1307 | __u64 cl_fid_build_ino(const struct lu_fid *fid, int api32); | |
1308 | __u32 cl_fid_build_gen(const struct lu_fid *fid); | |
1309 | ||
d7e09d03 | 1310 | #endif /* LLITE_INTERNAL_H */ |