4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
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.
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).
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
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
27 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2013, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
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>
44 /* for struct cl_lock_descr and struct cl_io */
45 #include <cl_object.h>
47 #include <lustre_mdc.h>
48 #include <linux/lustre_intent.h>
49 #include <linux/compat.h>
55 #ifndef VM_FAULT_RETRY
56 #define VM_FAULT_RETRY 0
59 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
60 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
61 #ifndef LOOKUP_CONTINUE
62 #define LOOKUP_CONTINUE LOOKUP_PARENT
65 /** Only used on client-side for indicating the tail of dir hash/offset. */
66 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
67 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
69 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
70 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
72 struct ll_dentry_data {
75 struct obd_client_handle lld_cwd_och;
76 struct obd_client_handle lld_mnt_och;
77 struct lookup_intent *lld_it;
78 unsigned int lld_sa_generation;
79 unsigned int lld_invalid:1;
80 struct rcu_head lld_rcu_head;
83 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
85 extern struct file_operations ll_pgcache_seq_fops;
87 #define LLI_INODE_MAGIC 0x111d0de5
88 #define LLI_INODE_DEAD 0xdeadd00d
90 /* remote client permission cache */
91 #define REMOTE_PERM_HASHSIZE 16
93 struct ll_getname_data {
94 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
95 struct lu_fid lgd_fid; /* target fid we are looking for */
96 int lgd_found; /* inode matched? */
99 /* llite setxid/access permission for user on remote client */
100 struct ll_remote_perm {
101 cfs_hlist_node_t lrp_list;
106 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
107 is access permission with
108 lrp_fsuid/lrp_fsgid. */
112 /* MDS has an authority for the Size-on-MDS attributes. */
113 LLIF_MDS_SIZE_LOCK = (1 << 0),
114 /* Epoch close is postponed. */
115 LLIF_EPOCH_PENDING = (1 << 1),
116 /* DONE WRITING is allowed. */
117 LLIF_DONE_WRITING = (1 << 2),
118 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
120 LLIF_SOM_DIRTY = (1 << 3),
121 /* File is contented */
122 LLIF_CONTENDED = (1 << 4),
123 /* Truncate uses server lock for this file */
124 LLIF_SRVLOCK = (1 << 5),
125 /* File data is modified. */
126 LLIF_DATA_MODIFIED = (1 << 6),
127 /* File is being restored */
128 LLIF_FILE_RESTORING = (1 << 7),
129 /* Xattr cache is attached to the file */
130 LLIF_XATTR_CACHE = (1 << 8),
133 struct ll_inode_info {
134 __u32 lli_inode_magic;
139 struct posix_acl *lli_posix_acl;
141 cfs_hlist_head_t *lli_remote_perms;
142 struct mutex lli_rmtperm_mutex;
144 /* identifying fields for both metadata and data stacks. */
145 struct lu_fid lli_fid;
146 /* Parent fid for accessing default stripe data on parent directory
147 * for allocating OST objects after a mknod() and later open-by-FID. */
148 struct lu_fid lli_pfid;
150 cfs_list_t lli_close_list;
151 cfs_list_t lli_oss_capas;
152 /* open count currently used by capability only, indicate whether
153 * capability needs renewal */
154 cfs_atomic_t lli_open_count;
155 struct obd_capa *lli_mds_capa;
156 cfs_time_t lli_rmtperm_time;
158 /* handle is to be sent to MDS later on done_writing and setattr.
159 * Open handle data are needed for the recovery to reconstruct
160 * the inode state on the MDS. XXX: recovery is not ready yet. */
161 struct obd_client_handle *lli_pending_och;
163 /* We need all three because every inode may be opened in different
165 struct obd_client_handle *lli_mds_read_och;
166 struct obd_client_handle *lli_mds_write_och;
167 struct obd_client_handle *lli_mds_exec_och;
168 __u64 lli_open_fd_read_count;
169 __u64 lli_open_fd_write_count;
170 __u64 lli_open_fd_exec_count;
171 /* Protects access to och pointers and their usage counters */
172 struct mutex lli_och_mutex;
174 struct inode lli_vfs_inode;
176 /* the most recent timestamps obtained from mds */
177 struct ost_lvb lli_lvb;
178 spinlock_t lli_agl_lock;
180 /* Try to make the d::member and f::member are aligned. Before using
181 * these members, make clear whether it is directory or not. */
185 /* serialize normal readdir and statahead-readdir. */
186 struct mutex d_readdir_mutex;
188 /* metadata statahead */
189 /* since parent-child threads can share the same @file
190 * struct, "opendir_key" is the token when dir close for
191 * case of parent exit before child -- it is me should
192 * cleanup the dir readahead. */
194 struct ll_statahead_info *d_sai;
195 struct posix_acl *d_def_acl;
196 /* protect statahead stuff. */
197 spinlock_t d_sa_lock;
198 /* "opendir_pid" is the token when lookup/revalid
199 * -- I am the owner of dir statahead. */
203 #define lli_readdir_mutex u.d.d_readdir_mutex
204 #define lli_opendir_key u.d.d_opendir_key
205 #define lli_sai u.d.d_sai
206 #define lli_def_acl u.d.d_def_acl
207 #define lli_sa_lock u.d.d_sa_lock
208 #define lli_opendir_pid u.d.d_opendir_pid
210 /* for non-directory */
212 struct semaphore f_size_sem;
213 void *f_size_sem_owner;
214 char *f_symlink_name;
217 * struct rw_semaphore {
218 * signed long count; // align d.d_def_acl
219 * spinlock_t wait_lock; // align d.d_sa_lock
220 * struct list_head wait_list;
223 struct rw_semaphore f_trunc_sem;
224 struct mutex f_write_mutex;
226 struct rw_semaphore f_glimpse_sem;
227 cfs_time_t f_glimpse_time;
228 cfs_list_t f_agl_list;
231 /* for writepage() only to communicate to fsync */
234 /* volatile file criteria is based on file name, this
235 * flag is used to keep the test result, so the strcmp
240 * whenever a process try to read/write the file, the
241 * jobid of the process will be saved here, and it'll
242 * be packed into the write PRC when flush later.
244 * so the read/write statistics for jobid will not be
245 * accurate if the file is shared by different jobs.
247 char f_jobid[JOBSTATS_JOBID_SIZE];
250 #define lli_size_sem u.f.f_size_sem
251 #define lli_size_sem_owner u.f.f_size_sem_owner
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
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.
273 * In the future, if more members are added only for directory,
274 * some of the following members can be moved into u.f.
277 struct cl_object *lli_clob;
279 /* mutex to request for layout lock exclusively. */
280 struct mutex lli_layout_mutex;
281 /* valid only inside LAYOUT ibits lock, protected by lli_layout_mutex */
282 __u32 lli_layout_gen;
284 struct rw_semaphore lli_xattrs_list_rwsem;
285 struct mutex lli_xattrs_enq_lock;
286 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
289 int ll_xattr_cache_destroy(struct inode *inode);
291 int ll_xattr_cache_get(struct inode *inode,
297 int ll_xattr_cache_update(struct inode *inode,
305 * Locking to guarantee consistency of non-atomic updates to long long i_size,
306 * consistency between file size and KMS.
308 * Implemented by ->lli_size_sem and ->lsm_lock, nested in that order.
311 void ll_inode_size_lock(struct inode *inode);
312 void ll_inode_size_unlock(struct inode *inode);
314 // FIXME: replace the name of this with LL_I to conform to kernel stuff
315 // static inline struct ll_inode_info *LL_I(struct inode *inode)
316 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
318 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
321 /* default to about 40meg of readahead on a given system. That much tied
322 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
323 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT))
325 /* default to read-ahead full files smaller than 2MB on the second read */
326 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
331 RA_STAT_DISTANT_READPAGE,
332 RA_STAT_MISS_IN_WINDOW,
333 RA_STAT_FAILED_GRAB_PAGE,
334 RA_STAT_FAILED_MATCH,
339 RA_STAT_MAX_IN_FLIGHT,
340 RA_STAT_WRONG_GRAB_PAGE,
345 cfs_atomic_t ra_cur_pages;
346 unsigned long ra_max_pages;
347 unsigned long ra_max_pages_per_file;
348 unsigned long ra_max_read_ahead_whole_pages;
351 /* ra_io_arg will be filled in the beginning of ll_readahead with
352 * ras_lock, then the following ll_read_ahead_pages will read RA
353 * pages according to this arg, all the items in this structure are
354 * counted by page index.
357 unsigned long ria_start; /* start offset of read-ahead*/
358 unsigned long ria_end; /* end offset of read-ahead*/
359 /* If stride read pattern is detected, ria_stoff means where
360 * stride read is started. Note: for normal read-ahead, the
361 * value here is meaningless, and also it will not be accessed*/
363 /* ria_length and ria_pages are the length and pages length in the
364 * stride I/O mode. And they will also be used to check whether
365 * it is stride I/O read-ahead in the read-ahead pages*/
366 unsigned long ria_length;
367 unsigned long ria_pages;
370 /* LL_HIST_MAX=32 causes an overflow */
371 #define LL_HIST_MAX 28
372 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
373 #define LL_PROCESS_HIST_MAX 10
374 struct per_process_info {
376 struct obd_histogram pp_r_hist;
377 struct obd_histogram pp_w_hist;
380 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
381 struct ll_rw_extents_info {
382 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
385 #define LL_OFFSET_HIST_MAX 100
386 struct ll_rw_process_info {
389 loff_t rw_range_start;
391 loff_t rw_last_file_pos;
393 size_t rw_smallest_extent;
394 size_t rw_largest_extent;
395 struct ll_file_data *rw_last_file;
398 enum stats_track_type {
399 STATS_TRACK_ALL = 0, /* track all processes */
400 STATS_TRACK_PID, /* track process with this pid */
401 STATS_TRACK_PPID, /* track processes with this ppid */
402 STATS_TRACK_GID, /* track processes with this gid */
406 /* flags for sbi->ll_flags */
407 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
408 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
409 #define LL_SBI_FLOCK 0x04
410 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
411 #define LL_SBI_ACL 0x10 /* support ACL */
412 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
413 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
414 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
415 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
416 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
417 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
418 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
419 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
420 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
421 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
422 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
423 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
424 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
425 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
427 #define LL_SBI_FLAGS { \
450 /* default value for ll_sb_info->contention_time */
451 #define SBI_DEFAULT_CONTENTION_SECONDS 60
452 /* default value for lockless_truncate_enable */
453 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
454 #define RCE_HASHES 32
456 struct rmtacl_ctl_entry {
458 pid_t rce_key; /* hash key */
459 int rce_ops; /* acl operation type */
462 struct rmtacl_ctl_table {
464 cfs_list_t rct_entries[RCE_HASHES];
471 pid_t ee_key; /* hash key */
472 struct lu_fid ee_fid;
473 int ee_type; /* ACL type for ACCESS or DEFAULT */
474 ext_acl_xattr_header *ee_acl;
479 cfs_list_t et_entries[EE_HASHES];
484 /* this protects pglist and ra_info. It isn't safe to
485 * grab from interrupt contexts */
487 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
488 spinlock_t ll_process_lock; /* ll_rw_process_info */
489 struct obd_uuid ll_sb_uuid;
490 struct obd_export *ll_md_exp;
491 struct obd_export *ll_dt_exp;
492 struct proc_dir_entry* ll_proc_root;
493 struct lu_fid ll_root_fid; /* root object fid */
496 unsigned int ll_umounting:1,
497 ll_xattr_cache_enabled:1;
498 cfs_list_t ll_conn_chain; /* per-conn chain of SBs */
499 struct lustre_client_ocd ll_lco;
501 cfs_list_t ll_orphan_dentry_list; /*please don't ask -p*/
502 struct ll_close_queue *ll_lcq;
504 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
506 struct cl_client_cache ll_cache;
508 struct lprocfs_stats *ll_ra_stats;
510 struct ll_ra_info ll_ra_info;
511 unsigned int ll_namelen;
512 struct file_operations *ll_fop;
514 /* =0 - hold lock over whole read/write
515 * >0 - max. chunk to be read/written w/o lock re-acquiring */
516 unsigned long ll_max_rw_chunk;
517 unsigned int ll_md_brw_size; /* used by readdir */
519 struct lu_site *ll_site;
520 struct cl_device *ll_cl;
522 struct ll_rw_extents_info ll_rw_extents_info;
523 int ll_extent_process_count;
524 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
525 unsigned int ll_offset_process_count;
526 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
527 unsigned int ll_rw_offset_entry_count;
528 int ll_stats_track_id;
529 enum stats_track_type ll_stats_track_type;
532 /* metadata stat-ahead */
533 unsigned int ll_sa_max; /* max statahead RPCs */
534 atomic_t ll_sa_total; /* statahead thread started
536 atomic_t ll_sa_wrong; /* statahead thread stopped for
538 atomic_t ll_agl_total; /* AGL thread started count */
540 dev_t ll_sdev_orig; /* save s_dev before assign for
542 struct rmtacl_ctl_table ll_rct;
543 struct eacl_table ll_et;
546 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
551 struct task_struct *lrr_reader;
552 cfs_list_t lrr_linkage;
556 * per file-descriptor read-ahead data.
558 struct ll_readahead_state {
561 * index of the last page that read(2) needed and that wasn't in the
562 * cache. Used by ras_update() to detect seeks.
564 * XXX nikita: if access seeks into cached region, Lustre doesn't see
567 unsigned long ras_last_readpage;
569 * number of pages read after last read-ahead window reset. As window
570 * is reset on each seek, this is effectively a number of consecutive
571 * accesses. Maybe ->ras_accessed_in_window is better name.
573 * XXX nikita: window is also reset (by ras_update()) when Lustre
574 * believes that memory pressure evicts read-ahead pages. In that
575 * case, it probably doesn't make sense to expand window to
576 * PTLRPC_MAX_BRW_PAGES on the third access.
578 unsigned long ras_consecutive_pages;
580 * number of read requests after the last read-ahead window reset
581 * As window is reset on each seek, this is effectively the number
582 * on consecutive read request and is used to trigger read-ahead.
584 unsigned long ras_consecutive_requests;
586 * Parameters of current read-ahead window. Handled by
587 * ras_update(). On the initial access to the file or after a seek,
588 * window is reset to 0. After 3 consecutive accesses, window is
589 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
590 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
592 unsigned long ras_window_start, ras_window_len;
594 * Where next read-ahead should start at. This lies within read-ahead
595 * window. Read-ahead window is read in pieces rather than at once
596 * because: 1. lustre limits total number of pages under read-ahead by
597 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
598 * not covered by DLM lock.
600 unsigned long ras_next_readahead;
602 * Total number of ll_file_read requests issued, reads originating
603 * due to mmap are not counted in this total. This value is used to
604 * trigger full file read-ahead after multiple reads to a small file.
606 unsigned long ras_requests;
608 * Page index with respect to the current request, these value
609 * will not be accurate when dealing with reads issued via mmap.
611 unsigned long ras_request_index;
613 * list of struct ll_ra_read's one per read(2) call current in
614 * progress against this file descriptor. Used by read-ahead code,
615 * protected by ->ras_lock.
617 cfs_list_t ras_read_beads;
619 * The following 3 items are used for detecting the stride I/O
621 * In stride I/O mode,
622 * ...............|-----data-----|****gap*****|--------|******|....
623 * offset |-stride_pages-|-stride_gap-|
624 * ras_stride_offset = offset;
625 * ras_stride_length = stride_pages + stride_gap;
626 * ras_stride_pages = stride_pages;
627 * Note: all these three items are counted by pages.
629 unsigned long ras_stride_length;
630 unsigned long ras_stride_pages;
631 pgoff_t ras_stride_offset;
633 * number of consecutive stride request count, and it is similar as
634 * ras_consecutive_requests, but used for stride I/O mode.
635 * Note: only more than 2 consecutive stride request are detected,
636 * stride read-ahead will be enable
638 unsigned long ras_consecutive_stride_requests;
641 extern struct kmem_cache *ll_file_data_slab;
642 struct lustre_handle;
643 struct ll_file_data {
644 struct ll_readahead_state fd_ras;
645 struct ccc_grouplock fd_grouplock;
649 /* openhandle if lease exists for this file.
650 * Borrow lli->lli_och_mutex to protect assignment */
651 struct obd_client_handle *fd_lease_och;
652 struct obd_client_handle *fd_och;
653 struct file *fd_file;
654 /* Indicate whether need to report failure when close.
655 * true: failure is known, not report again.
656 * false: unknown failure, should report. */
657 bool fd_write_failed;
660 struct lov_stripe_md;
662 extern spinlock_t inode_lock;
664 extern struct proc_dir_entry *proc_lustre_fs_root;
666 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
668 return &lli->lli_vfs_inode;
672 struct inode *icbd_parent;
673 struct dentry **icbd_childp;
677 __u32 ll_i2suppgid(struct inode *i);
678 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
680 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
682 #if BITS_PER_LONG == 32
689 (sbi->ll_flags & LL_SBI_32BIT_API)
694 #define LLAP_MAGIC 98764321
696 extern struct kmem_cache *ll_async_page_slab;
697 extern size_t ll_async_page_slab_size;
699 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
700 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
701 struct ll_ra_read *ll_ra_read_get(struct file *f);
703 /* llite/lproc_llite.c */
705 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
706 struct super_block *sb, char *osc, char *mdc);
707 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
708 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
709 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
711 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
712 struct super_block *sb, char *osc, char *mdc){return 0;}
713 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
714 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
715 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
717 memset(lvars, 0, sizeof(*lvars));
723 void ll_release_page(struct page *page, int remove);
724 extern struct file_operations ll_dir_operations;
725 extern struct inode_operations ll_dir_inode_operations;
726 struct page *ll_get_dir_page(struct inode *dir, __u64 hash,
727 struct ll_dir_chain *chain);
728 int ll_dir_read(struct inode *inode, __u64 *_pos, void *cookie,
731 int ll_get_mdt_idx(struct inode *inode);
733 int ll_objects_destroy(struct ptlrpc_request *request,
735 struct inode *ll_iget(struct super_block *sb, ino_t hash,
736 struct lustre_md *lic);
737 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
738 void *data, int flag);
739 #ifndef HAVE_IOP_ATOMIC_OPEN
740 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
743 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
744 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
747 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
748 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
749 int ll_writepage(struct page *page, struct writeback_control *wbc);
750 int ll_writepages(struct address_space *, struct writeback_control *wbc);
751 void ll_removepage(struct page *page);
752 int ll_readpage(struct file *file, struct page *page);
753 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
754 int ll_file_punch(struct inode *, loff_t, int);
755 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
756 void ll_clear_file_contended(struct inode*);
757 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
758 int ll_readahead(const struct lu_env *env, struct cl_io *io,
759 struct ll_readahead_state *ras, struct address_space *mapping,
760 struct cl_page_list *queue, int flags);
763 extern struct file_operations ll_file_operations;
764 extern struct file_operations ll_file_operations_flock;
765 extern struct file_operations ll_file_operations_noflock;
766 extern struct inode_operations ll_file_inode_operations;
767 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
769 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
770 ldlm_mode_t l_req_mode);
771 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
772 struct lustre_handle *lockh, __u64 flags,
774 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
776 #ifdef HAVE_IOP_ATOMIC_OPEN
777 int ll_revalidate_nd(struct dentry *dentry, unsigned int flags);
779 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
781 int ll_file_open(struct inode *inode, struct file *file);
782 int ll_file_release(struct inode *inode, struct file *file);
783 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
784 struct lov_stripe_md *lsm, lstat_t *st);
785 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
786 int ll_local_open(struct file *file,
787 struct lookup_intent *it, struct ll_file_data *fd,
788 struct obd_client_handle *och);
789 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
790 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
792 int ll_md_real_close(struct inode *inode, int flags);
793 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
794 struct obd_client_handle **och, unsigned long flags);
795 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
796 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
797 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
798 __u64 ioepoch, int sync);
799 int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data,
800 struct md_open_data **mod);
801 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
802 struct lustre_handle *fh);
803 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
804 struct ll_file_data *file, loff_t pos,
805 size_t count, int rw);
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);
812 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
813 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
815 # ifndef HAVE_INODE_PERMISION_2ARGS
816 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
818 int ll_inode_permission(struct inode *inode, int mask);
822 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
823 __u64 flags, struct lov_user_md *lum,
825 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
826 struct lov_mds_md **lmm, int *lmm_size,
827 struct ptlrpc_request **request);
828 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
830 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmmp,
831 int *lmm_size, struct ptlrpc_request **request);
832 #ifdef HAVE_FILE_FSYNC_4ARGS
833 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
834 #elif defined(HAVE_FILE_FSYNC_2ARGS)
835 int ll_fsync(struct file *file, int data);
837 int ll_fsync(struct file *file, struct dentry *dentry, int data);
839 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
841 int ll_merge_lvb(const struct lu_env *env, struct inode *inode);
842 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
843 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
844 int ll_fid2path(struct inode *inode, void *arg);
845 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
846 int ll_hsm_release(struct inode *inode);
848 struct obd_client_handle *ll_lease_open(struct inode *inode, struct file *file,
849 fmode_t mode, __u64 flags);
850 int ll_lease_close(struct obd_client_handle *och, struct inode *inode,
855 int ll_d_init(struct dentry *de);
856 extern struct dentry_operations ll_d_ops;
857 void ll_intent_drop_lock(struct lookup_intent *);
858 void ll_intent_release(struct lookup_intent *);
859 void ll_invalidate_aliases(struct inode *);
860 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
861 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
862 #ifdef HAVE_D_COMPARE_7ARGS
863 int ll_dcompare(const struct dentry *parent, const struct inode *pinode,
864 const struct dentry *dentry, const struct inode *inode,
865 unsigned int len, const char *str, const struct qstr *d_name);
867 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
869 int ll_revalidate_it_finish(struct ptlrpc_request *request,
870 struct lookup_intent *it, struct dentry *de);
872 /* llite/llite_lib.c */
873 extern struct super_operations lustre_super_operations;
875 char *ll_read_opt(const char *opt, char *data);
876 void ll_lli_init(struct ll_inode_info *lli);
877 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
878 void ll_put_super(struct super_block *sb);
879 void ll_kill_super(struct super_block *sb);
880 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
881 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
882 void ll_clear_inode(struct inode *inode);
883 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import);
884 int ll_setattr(struct dentry *de, struct iattr *attr);
885 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
886 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
887 __u64 max_age, __u32 flags);
888 void ll_update_inode(struct inode *inode, struct lustre_md *md);
889 void ll_read_inode2(struct inode *inode, void *opaque);
890 void ll_delete_inode(struct inode *inode);
891 int ll_iocontrol(struct inode *inode, struct file *file,
892 unsigned int cmd, unsigned long arg);
893 int ll_flush_ctx(struct inode *inode);
894 void ll_umount_begin(struct super_block *sb);
895 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
896 #ifdef HAVE_SUPEROPS_USE_DENTRY
897 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
899 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
901 void ll_dirty_page_discard_warn(struct page *page, int ioret);
902 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
903 struct super_block *, struct lookup_intent *);
904 void lustre_dump_dentry(struct dentry *, int recur);
905 void lustre_dump_inode(struct inode *);
906 int ll_obd_statfs(struct inode *inode, void *arg);
907 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
908 int ll_process_config(struct lustre_cfg *lcfg);
909 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
910 struct inode *i1, struct inode *i2,
911 const char *name, int namelen,
912 int mode, __u32 opc, void *data);
913 void ll_finish_md_op_data(struct md_op_data *op_data);
914 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
915 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
917 /* llite/llite_nfs.c */
918 extern struct export_operations lustre_export_operations;
919 __u32 get_uuid2int(const char *name, int len);
920 struct inode *search_inode_for_lustre(struct super_block *sb,
921 const struct lu_fid *fid);
923 /* llite/special.c */
924 extern struct inode_operations ll_special_inode_operations;
925 extern struct file_operations ll_special_chr_inode_fops;
926 extern struct file_operations ll_special_chr_file_fops;
927 extern struct file_operations ll_special_blk_inode_fops;
928 extern struct file_operations ll_special_fifo_inode_fops;
929 extern struct file_operations ll_special_fifo_file_fops;
930 extern struct file_operations ll_special_sock_inode_fops;
932 /* llite/symlink.c */
933 extern struct inode_operations ll_fast_symlink_inode_operations;
935 /* llite/llite_close.c */
936 struct ll_close_queue {
939 cfs_waitq_t lcq_waitq;
940 struct completion lcq_comp;
941 cfs_atomic_t lcq_stop;
944 struct ccc_object *cl_inode2ccc(struct inode *inode);
947 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
948 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
950 /* specific achitecture can implement only part of this list */
951 enum vvp_io_subtype {
954 /** io called from .sendfile */
956 /** io started from splice_{read|write} */
963 enum vvp_io_subtype cui_io_subtype;
967 read_actor_t cui_actor;
971 struct pipe_inode_info *cui_pipe;
972 unsigned int cui_flags;
974 struct vvp_fault_io {
976 * Inode modification time that is checked across DLM
980 struct vm_area_struct *ft_vma;
982 * locked page returned from vvp_io
984 struct page *ft_vmpage;
985 struct vm_fault_api {
989 struct vm_fault *ft_vmf;
991 * fault API used bitflags for return code.
993 unsigned int ft_flags;
998 * Read-ahead state used by read and page-fault IO contexts.
1000 struct ll_ra_read cui_bead;
1002 * Set when cui_bead has been initialized.
1004 int cui_ra_window_set;
1006 * Partially truncated page, that vvp_io_trunc_start() keeps locked
1009 struct cl_page *cui_partpage;
1013 * IO arguments for various VFS I/O interfaces.
1015 struct vvp_io_args {
1016 /** normal/sendfile/splice */
1017 enum vvp_io_subtype via_io_subtype;
1021 struct kiocb *via_iocb;
1022 struct iovec *via_iov;
1023 unsigned long via_nrsegs;
1026 read_actor_t via_actor;
1030 struct pipe_inode_info *via_pipe;
1031 unsigned int via_flags;
1036 struct ll_cl_context {
1038 struct cl_io *lcc_io;
1039 struct cl_page *lcc_page;
1040 struct lu_env *lcc_env;
1045 struct vvp_thread_info {
1046 struct ost_lvb vti_lvb;
1047 struct cl_2queue vti_queue;
1048 struct iovec vti_local_iov;
1049 struct vvp_io_args vti_args;
1050 struct ra_io_arg vti_ria;
1051 struct kiocb vti_kiocb;
1052 struct ll_cl_context vti_io_ctx;
1055 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
1057 extern struct lu_context_key vvp_key;
1058 struct vvp_thread_info *info;
1060 info = lu_context_key_get(&env->le_ctx, &vvp_key);
1061 LASSERT(info != NULL);
1065 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
1066 enum vvp_io_subtype type)
1068 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
1070 ret->via_io_subtype = type;
1075 struct vvp_session {
1076 struct vvp_io vs_ios;
1079 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1081 extern struct lu_context_key vvp_session_key;
1082 struct vvp_session *ses;
1084 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1085 LASSERT(ses != NULL);
1089 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1091 return &vvp_env_session(env)->vs_ios;
1094 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1095 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1096 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1098 /* llite/llite_mmap.c */
1099 typedef struct rb_root rb_root_t;
1100 typedef struct rb_node rb_node_t;
1102 struct ll_lock_tree_node;
1103 struct ll_lock_tree {
1105 cfs_list_t lt_locked_list;
1106 struct ll_file_data *lt_fd;
1109 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1110 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1111 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
1112 __u64 end, ldlm_mode_t mode);
1113 void policy_from_vma(ldlm_policy_data_t *policy,
1114 struct vm_area_struct *vma, unsigned long addr, size_t count);
1115 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1118 static inline void ll_invalidate_page(struct page *vmpage)
1120 struct address_space *mapping = vmpage->mapping;
1121 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1123 LASSERT(PageLocked(vmpage));
1124 if (mapping == NULL)
1127 ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE);
1128 truncate_complete_page(mapping, vmpage);
1131 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1133 /* don't need an addref as the sb_info should be holding one */
1134 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1136 return ll_s2sbi(sb)->ll_dt_exp;
1139 /* don't need an addref as the sb_info should be holding one */
1140 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1142 return ll_s2sbi(sb)->ll_md_exp;
1145 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1147 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1153 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1154 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1156 return ll_s2sbi(inode->i_sb);
1159 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1161 return ll_s2dtexp(inode->i_sb);
1164 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1166 return ll_s2mdexp(inode->i_sb);
1169 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1173 LASSERT(inode != NULL);
1174 fid = &ll_i2info(inode)->lli_fid;
1179 static inline int ll_mds_max_easize(struct super_block *sb)
1181 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1184 static inline __u64 ll_file_maxbytes(struct inode *inode)
1186 return ll_i2info(inode)->lli_maxbytes;
1190 int ll_setxattr(struct dentry *dentry, const char *name,
1191 const void *value, size_t size, int flags);
1192 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1193 void *buffer, size_t size);
1194 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1195 int ll_removexattr(struct dentry *dentry, const char *name);
1197 /* llite/remote_perm.c */
1198 extern struct kmem_cache *ll_remote_perm_cachep;
1199 extern struct kmem_cache *ll_rmtperm_hash_cachep;
1201 cfs_hlist_head_t *alloc_rmtperm_hash(void);
1202 void free_rmtperm_hash(cfs_hlist_head_t *hash);
1203 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1204 int lustre_check_remote_perm(struct inode *inode, int mask);
1206 /* llite/llite_capa.c */
1207 extern cfs_timer_t ll_capa_timer;
1209 int ll_capa_thread_start(void);
1210 void ll_capa_thread_stop(void);
1211 void ll_capa_timer_callback(unsigned long unused);
1213 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1214 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1216 void ll_capa_open(struct inode *inode);
1217 void ll_capa_close(struct inode *inode);
1219 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1220 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1222 void ll_truncate_free_capa(struct obd_capa *ocapa);
1223 void ll_clear_inode_capas(struct inode *inode);
1224 void ll_print_capa_stat(struct ll_sb_info *sbi);
1226 /* llite/llite_cl.c */
1227 extern struct lu_device_type vvp_device_type;
1230 * Common IO arguments for various VFS I/O interfaces.
1232 int cl_sb_init(struct super_block *sb);
1233 int cl_sb_fini(struct super_block *sb);
1234 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1235 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1237 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1238 struct ll_readahead_state *ras, unsigned long index,
1240 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1241 int ll_is_file_contended(struct file *file);
1242 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1244 /* llite/llite_rmtacl.c */
1245 #ifdef CONFIG_FS_POSIX_ACL
1246 obd_valid rce_ops2valid(int ops);
1247 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1248 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1249 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1250 void rct_init(struct rmtacl_ctl_table *rct);
1251 void rct_fini(struct rmtacl_ctl_table *rct);
1253 void ee_free(struct eacl_entry *ee);
1254 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1255 ext_acl_xattr_header *header);
1256 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1257 struct lu_fid *fid, int type);
1258 void et_search_free(struct eacl_table *et, pid_t key);
1259 void et_init(struct eacl_table *et);
1260 void et_fini(struct eacl_table *et);
1265 #define LL_SA_RPC_MIN 2
1266 #define LL_SA_RPC_DEF 32
1267 #define LL_SA_RPC_MAX 8192
1269 #define LL_SA_CACHE_BIT 5
1270 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1271 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1273 /* per inode struct, for dir only */
1274 struct ll_statahead_info {
1275 struct inode *sai_inode;
1276 cfs_atomic_t sai_refcount; /* when access this struct, hold
1278 unsigned int sai_generation; /* generation for statahead */
1279 unsigned int sai_max; /* max ahead of lookup */
1280 __u64 sai_sent; /* stat requests sent count */
1281 __u64 sai_replied; /* stat requests which received
1283 __u64 sai_index; /* index of statahead entry */
1284 __u64 sai_index_wait; /* index of entry which is the
1285 * caller is waiting for */
1286 __u64 sai_hit; /* hit count */
1287 __u64 sai_miss; /* miss count:
1288 * for "ls -al" case, it includes
1289 * hidden dentry miss;
1290 * for "ls -l" case, it does not
1291 * include hidden dentry miss.
1292 * "sai_miss_hidden" is used for
1295 unsigned int sai_consecutive_miss; /* consecutive miss */
1296 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1297 * is not a hidden one */
1298 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1299 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1301 sai_in_readpage:1,/* statahead is in readdir()*/
1302 sai_agl_valid:1;/* AGL is valid for the dir */
1303 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
1304 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1305 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1306 cfs_list_t sai_entries; /* entry list */
1307 cfs_list_t sai_entries_received; /* entries returned */
1308 cfs_list_t sai_entries_stated; /* entries stated */
1309 cfs_list_t sai_entries_agl; /* AGL entries to be sent */
1310 cfs_list_t sai_cache[LL_SA_CACHE_SIZE];
1311 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1312 cfs_atomic_t sai_cache_count; /* entry count in cache */
1315 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1317 void ll_stop_statahead(struct inode *dir, void *key);
1319 static inline int ll_glimpse_size(struct inode *inode)
1321 struct ll_inode_info *lli = ll_i2info(inode);
1324 down_read(&lli->lli_glimpse_sem);
1325 rc = cl_glimpse_size(inode);
1326 lli->lli_glimpse_time = cfs_time_current();
1327 up_read(&lli->lli_glimpse_sem);
1332 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1334 struct ll_inode_info *lli = ll_i2info(dir);
1335 struct ll_statahead_info *sai = lli->lli_sai;
1336 struct ll_dentry_data *ldd = ll_d2d(dentry);
1338 /* not the same process, don't mark */
1339 if (lli->lli_opendir_pid != current_pid())
1342 LASSERT(ldd != NULL);
1344 ldd->lld_sa_generation = sai->sai_generation;
1348 ll_need_statahead(struct inode *dir, struct dentry *dentryp)
1350 struct ll_inode_info *lli;
1351 struct ll_dentry_data *ldd;
1353 if (ll_i2sbi(dir)->ll_sa_max == 0)
1356 lli = ll_i2info(dir);
1357 /* not the same process, don't statahead */
1358 if (lli->lli_opendir_pid != current_pid())
1361 /* statahead has been stopped */
1362 if (lli->lli_opendir_key == NULL)
1365 ldd = ll_d2d(dentryp);
1367 * When stats a dentry, the system trigger more than once "revalidate"
1368 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1369 * Under patchless client mode, the operation intent is not accurate,
1370 * which maybe misguide the statahead thread. For example:
1371 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1372 * have the same operation intent -- "IT_GETATTR".
1373 * In fact, one dentry should has only one chance to interact with the
1374 * statahead thread, otherwise the statahead windows will be confused.
1375 * The solution is as following:
1376 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1377 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1378 * will bypass interacting with statahead thread for checking:
1379 * "lld_sa_generation == lli_sai->sai_generation"
1381 if (ldd && lli->lli_sai &&
1382 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1389 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1393 ret = ll_need_statahead(dir, *dentryp);
1397 return do_statahead_enter(dir, dentryp, only_unplug);
1400 /* llite ioctl register support rountine */
1407 #define LLIOC_MAX_CMD 256
1410 * Rules to write a callback function:
1413 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1414 * returned to ll_iocontrol_register. Callback functions should use this
1415 * data to check the potential collasion of ioctl cmd. If collasion is
1416 * found, callback function should return LLIOC_CONT.
1417 * @rcp: The result of ioctl command.
1420 * If @magic matches the pointer returned by ll_iocontrol_data, the
1421 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1423 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1424 struct file *file, unsigned int cmd, unsigned long arg,
1425 void *magic, int *rcp);
1427 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1428 unsigned int cmd, unsigned long arg, int *rcp);
1430 /* export functions */
1431 /* Register ioctl block dynamatically for a regular file.
1433 * @cmd: the array of ioctl command set
1434 * @count: number of commands in the @cmd
1435 * @cb: callback function, it will be called if an ioctl command is found to
1436 * belong to the command list @cmd.
1439 * A magic pointer will be returned if success;
1440 * otherwise, NULL will be returned.
1442 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1443 void ll_iocontrol_unregister(void *magic);
1447 /* lclient compat stuff */
1448 #define cl_inode_info ll_inode_info
1449 #define cl_i2info(info) ll_i2info(info)
1450 #define cl_inode_mode(inode) ((inode)->i_mode)
1451 #define cl_i2sbi ll_i2sbi
1453 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1454 const struct iattr *attr)
1456 LASSERT(attr->ia_valid & ATTR_FILE);
1457 return LUSTRE_FPRIVATE(attr->ia_file);
1460 static inline void cl_isize_lock(struct inode *inode)
1462 ll_inode_size_lock(inode);
1465 static inline void cl_isize_unlock(struct inode *inode)
1467 ll_inode_size_unlock(inode);
1470 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1472 LASSERT(down_trylock(&ll_i2info(inode)->lli_size_sem) != 0);
1473 i_size_write(inode, kms);
1476 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1478 ll_inode_size_lock(inode);
1479 i_size_write(inode, kms);
1480 ll_inode_size_unlock(inode);
1483 #define cl_isize_read(inode) i_size_read(inode)
1485 static inline int cl_merge_lvb(const struct lu_env *env, struct inode *inode)
1487 return ll_merge_lvb(env, inode);
1490 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1491 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1492 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1494 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1496 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1497 enum cl_fsync_mode mode, int ignore_layout);
1499 /** direct write pages */
1500 struct ll_dio_pages {
1501 /** page array to be written. we don't support
1502 * partial pages except the last one. */
1503 struct page **ldp_pages;
1504 /* offset of each page */
1505 loff_t *ldp_offsets;
1506 /** if ldp_offsets is NULL, it means a sequential
1507 * pages to be written, then this is the file offset
1508 * of the * first page. */
1509 loff_t ldp_start_offset;
1510 /** how many bytes are to be written. */
1512 /** # of pages in the array. */
1516 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1519 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1522 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1525 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1526 int rw, struct inode *inode,
1527 struct ll_dio_pages *pv);
1529 static inline int ll_file_nolock(const struct file *file)
1531 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1532 struct inode *inode = file->f_dentry->d_inode;
1534 LASSERT(fd != NULL);
1535 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1536 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1539 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1540 struct lookup_intent *it, __u64 *bits)
1542 if (!it->d.lustre.it_lock_set) {
1543 struct lustre_handle handle;
1545 /* If this inode is a remote object, it will get two
1546 * separate locks in different namespaces, Master MDT,
1547 * where the name entry is, will grant LOOKUP lock,
1548 * remote MDT, where the object is, will grant
1549 * UPDATE|PERM lock. The inode will be attched to both
1550 * LOOKUP and PERM locks, so revoking either locks will
1551 * case the dcache being cleared */
1552 if (it->d.lustre.it_remote_lock_mode) {
1553 handle.cookie = it->d.lustre.it_remote_lock_handle;
1554 CDEBUG(D_DLMTRACE, "setting l_data to inode %p"
1555 "(%lu/%u) for remote lock "LPX64"\n", inode,
1556 inode->i_ino, inode->i_generation,
1558 md_set_lock_data(exp, &handle.cookie, inode, NULL);
1561 handle.cookie = it->d.lustre.it_lock_handle;
1563 CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)"
1564 " for lock "LPX64"\n", inode, inode->i_ino,
1565 inode->i_generation, handle.cookie);
1567 md_set_lock_data(exp, &handle.cookie, inode,
1568 &it->d.lustre.it_lock_bits);
1569 it->d.lustre.it_lock_set = 1;
1573 *bits = it->d.lustre.it_lock_bits;
1576 static inline void ll_lock_dcache(struct inode *inode)
1578 #ifdef HAVE_DCACHE_LOCK
1579 spin_lock(&dcache_lock);
1581 spin_lock(&inode->i_lock);
1585 static inline void ll_unlock_dcache(struct inode *inode)
1587 #ifdef HAVE_DCACHE_LOCK
1588 spin_unlock(&dcache_lock);
1590 spin_unlock(&inode->i_lock);
1594 static inline int d_lustre_invalid(const struct dentry *dentry)
1596 struct ll_dentry_data *lld = ll_d2d(dentry);
1598 return (lld == NULL) || lld->lld_invalid;
1601 static inline void __d_lustre_invalidate(struct dentry *dentry)
1603 struct ll_dentry_data *lld = ll_d2d(dentry);
1606 lld->lld_invalid = 1;
1610 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1611 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1612 * else dput() of the last refcount will unhash this dentry and kill it.
1614 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1616 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1617 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1618 dentry->d_parent, dentry->d_inode, d_refcount(dentry));
1620 spin_lock_nested(&dentry->d_lock,
1621 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1622 __d_lustre_invalidate(dentry);
1623 if (d_refcount(dentry) == 0)
1625 spin_unlock(&dentry->d_lock);
1628 static inline void d_lustre_revalidate(struct dentry *dentry)
1630 spin_lock(&dentry->d_lock);
1631 LASSERT(ll_d2d(dentry) != NULL);
1632 ll_d2d(dentry)->lld_invalid = 0;
1633 spin_unlock(&dentry->d_lock);
1636 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0)
1637 /* Compatibility for old (1.8) compiled userspace quota code */
1638 struct if_quotactl_18 {
1643 struct obd_dqinfo qc_dqinfo;
1644 struct obd_dqblk qc_dqblk;
1646 struct obd_uuid obd_uuid;
1648 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1649 /* End compatibility for old (1.8) compiled userspace quota code */
1651 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1652 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */
1655 LL_LAYOUT_GEN_NONE = ((__u32)-2), /* layout lock was cancelled */
1656 LL_LAYOUT_GEN_EMPTY = ((__u32)-1) /* for empty layout */
1659 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1660 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1661 int ll_layout_restore(struct inode *inode);
1663 int ll_xattr_init(void);
1664 void ll_xattr_fini(void);
1666 #endif /* LLITE_INTERNAL_H */