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_lmv.h>
48 #include <lustre_mdc.h>
49 #include <linux/lustre_intent.h>
50 #include <linux/compat.h>
56 #ifndef VM_FAULT_RETRY
57 #define VM_FAULT_RETRY 0
60 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
61 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
62 #ifndef LOOKUP_CONTINUE
63 #define LOOKUP_CONTINUE LOOKUP_PARENT
66 /** Only used on client-side for indicating the tail of dir hash/offset. */
67 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
68 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
70 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
71 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
73 struct ll_dentry_data {
74 struct lookup_intent *lld_it;
75 unsigned int lld_sa_generation;
76 unsigned int lld_invalid:1;
77 struct rcu_head lld_rcu_head;
80 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
82 #define LLI_INODE_MAGIC 0x111d0de5
83 #define LLI_INODE_DEAD 0xdeadd00d
85 /* remote client permission cache */
86 #define REMOTE_PERM_HASHSIZE 16
88 struct ll_getname_data {
89 #ifdef HAVE_DIR_CONTEXT
90 struct dir_context ctx;
92 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
93 struct lu_fid lgd_fid; /* target fid we are looking for */
94 int lgd_found; /* inode matched? */
97 /* llite setxid/access permission for user on remote client */
98 struct ll_remote_perm {
99 struct hlist_node lrp_list;
104 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
105 is access permission with
106 lrp_fsuid/lrp_fsgid. */
110 /* MDS has an authority for the Size-on-MDS attributes. */
111 LLIF_MDS_SIZE_LOCK = (1 << 0),
112 /* Epoch close is postponed. */
113 LLIF_EPOCH_PENDING = (1 << 1),
114 /* DONE WRITING is allowed. */
115 LLIF_DONE_WRITING = (1 << 2),
116 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
118 LLIF_SOM_DIRTY = (1 << 3),
119 /* File data is modified. */
120 LLIF_DATA_MODIFIED = (1 << 4),
121 /* File is being restored */
122 LLIF_FILE_RESTORING = (1 << 5),
123 /* Xattr cache is attached to the file */
124 LLIF_XATTR_CACHE = (1 << 6),
127 struct ll_inode_info {
128 __u32 lli_inode_magic;
133 struct posix_acl *lli_posix_acl;
135 struct hlist_head *lli_remote_perms;
136 struct mutex lli_rmtperm_mutex;
138 /* identifying fields for both metadata and data stacks. */
139 struct lu_fid lli_fid;
140 /* Parent fid for accessing default stripe data on parent directory
141 * for allocating OST objects after a mknod() and later open-by-FID. */
142 struct lu_fid lli_pfid;
144 struct list_head lli_close_list;
145 struct list_head lli_oss_capas;
146 /* open count currently used by capability only, indicate whether
147 * capability needs renewal */
148 atomic_t lli_open_count;
149 struct obd_capa *lli_mds_capa;
150 cfs_time_t lli_rmtperm_time;
152 /* handle is to be sent to MDS later on done_writing and setattr.
153 * Open handle data are needed for the recovery to reconstruct
154 * the inode state on the MDS. XXX: recovery is not ready yet. */
155 struct obd_client_handle *lli_pending_och;
157 /* We need all three because every inode may be opened in different
159 struct obd_client_handle *lli_mds_read_och;
160 struct obd_client_handle *lli_mds_write_och;
161 struct obd_client_handle *lli_mds_exec_och;
162 __u64 lli_open_fd_read_count;
163 __u64 lli_open_fd_write_count;
164 __u64 lli_open_fd_exec_count;
165 /* Protects access to och pointers and their usage counters */
166 struct mutex lli_och_mutex;
168 struct inode lli_vfs_inode;
170 /* the most recent timestamps obtained from mds */
171 struct ost_lvb lli_lvb;
172 spinlock_t lli_agl_lock;
174 /* Try to make the d::member and f::member are aligned. Before using
175 * these members, make clear whether it is directory or not. */
179 /* serialize normal readdir and statahead-readdir. */
180 struct mutex d_readdir_mutex;
182 /* metadata statahead */
183 /* since parent-child threads can share the same @file
184 * struct, "opendir_key" is the token when dir close for
185 * case of parent exit before child -- it is me should
186 * cleanup the dir readahead. */
188 struct ll_statahead_info *d_sai;
189 /* protect statahead stuff. */
190 spinlock_t d_sa_lock;
191 /* "opendir_pid" is the token when lookup/revalid
192 * -- I am the owner of dir statahead. */
194 /* directory stripe information */
195 struct lmv_stripe_md *d_lsm_md;
196 /* striped directory size */
197 loff_t d_stripe_size;
198 /* striped directory nlink */
199 __u64 d_stripe_nlink;
202 #define lli_readdir_mutex u.d.d_readdir_mutex
203 #define lli_opendir_key u.d.d_opendir_key
204 #define lli_sai u.d.d_sai
205 #define lli_sa_lock u.d.d_sa_lock
206 #define lli_opendir_pid u.d.d_opendir_pid
207 #define lli_lsm_md u.d.d_lsm_md
208 #define lli_stripe_dir_size u.d.d_stripe_size
209 #define lli_stripe_dir_nlink u.d.d_stripe_nlink
211 /* for non-directory */
213 struct mutex f_size_mutex;
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 struct list_head f_agl_list;
231 /* for writepage() only to communicate to fsync */
235 * whenever a process try to read/write the file, the
236 * jobid of the process will be saved here, and it'll
237 * be packed into the write PRC when flush later.
239 * so the read/write statistics for jobid will not be
240 * accurate if the file is shared by different jobs.
242 char f_jobid[JOBSTATS_JOBID_SIZE];
245 #define lli_size_mutex u.f.f_size_mutex
246 #define lli_symlink_name u.f.f_symlink_name
247 #define lli_maxbytes u.f.f_maxbytes
248 #define lli_trunc_sem u.f.f_trunc_sem
249 #define lli_write_mutex u.f.f_write_mutex
250 #define lli_glimpse_sem u.f.f_glimpse_sem
251 #define lli_glimpse_time u.f.f_glimpse_time
252 #define lli_agl_list u.f.f_agl_list
253 #define lli_agl_index u.f.f_agl_index
254 #define lli_async_rc u.f.f_async_rc
255 #define lli_jobid u.f.f_jobid
259 /* XXX: For following frequent used members, although they maybe special
260 * used for non-directory object, it is some time-wasting to check
261 * whether the object is directory or not before using them. On the
262 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
263 * the "ll_inode_info" size even if moving those members into u.f.
264 * So keep them out side.
266 * In the future, if more members are added only for directory,
267 * some of the following members can be moved into u.f.
270 struct cl_object *lli_clob;
272 /* mutex to request for layout lock exclusively. */
273 struct mutex lli_layout_mutex;
274 /* Layout version, protected by lli_layout_lock */
275 __u32 lli_layout_gen;
276 spinlock_t lli_layout_lock;
278 struct rw_semaphore lli_xattrs_list_rwsem;
279 struct mutex lli_xattrs_enq_lock;
280 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
283 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
287 spin_lock(&lli->lli_layout_lock);
288 gen = lli->lli_layout_gen;
289 spin_unlock(&lli->lli_layout_lock);
294 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
296 spin_lock(&lli->lli_layout_lock);
297 lli->lli_layout_gen = gen;
298 spin_unlock(&lli->lli_layout_lock);
301 int ll_xattr_cache_destroy(struct inode *inode);
303 int ll_xattr_cache_get(struct inode *inode,
310 * Locking to guarantee consistency of non-atomic updates to long long i_size,
311 * consistency between file size and KMS.
313 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
316 void ll_inode_size_lock(struct inode *inode);
317 void ll_inode_size_unlock(struct inode *inode);
319 // FIXME: replace the name of this with LL_I to conform to kernel stuff
320 // static inline struct ll_inode_info *LL_I(struct inode *inode)
321 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
323 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
326 /* default to about 40meg of readahead on a given system. That much tied
327 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
328 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT))
330 /* default to read-ahead full files smaller than 2MB on the second read */
331 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
336 RA_STAT_DISTANT_READPAGE,
337 RA_STAT_MISS_IN_WINDOW,
338 RA_STAT_FAILED_GRAB_PAGE,
339 RA_STAT_FAILED_MATCH,
344 RA_STAT_MAX_IN_FLIGHT,
345 RA_STAT_WRONG_GRAB_PAGE,
346 RA_STAT_FAILED_REACH_END,
351 atomic_t ra_cur_pages;
352 unsigned long ra_max_pages;
353 unsigned long ra_max_pages_per_file;
354 unsigned long ra_max_read_ahead_whole_pages;
357 /* ra_io_arg will be filled in the beginning of ll_readahead with
358 * ras_lock, then the following ll_read_ahead_pages will read RA
359 * pages according to this arg, all the items in this structure are
360 * counted by page index.
363 unsigned long ria_start; /* start offset of read-ahead*/
364 unsigned long ria_end; /* end offset of read-ahead*/
365 /* If stride read pattern is detected, ria_stoff means where
366 * stride read is started. Note: for normal read-ahead, the
367 * value here is meaningless, and also it will not be accessed*/
369 /* ria_length and ria_pages are the length and pages length in the
370 * stride I/O mode. And they will also be used to check whether
371 * it is stride I/O read-ahead in the read-ahead pages*/
372 unsigned long ria_length;
373 unsigned long ria_pages;
376 /* LL_HIST_MAX=32 causes an overflow */
377 #define LL_HIST_MAX 28
378 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
379 #define LL_PROCESS_HIST_MAX 10
380 struct per_process_info {
382 struct obd_histogram pp_r_hist;
383 struct obd_histogram pp_w_hist;
386 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
387 struct ll_rw_extents_info {
388 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
391 #define LL_OFFSET_HIST_MAX 100
392 struct ll_rw_process_info {
395 loff_t rw_range_start;
397 loff_t rw_last_file_pos;
399 size_t rw_smallest_extent;
400 size_t rw_largest_extent;
401 struct ll_file_data *rw_last_file;
404 enum stats_track_type {
405 STATS_TRACK_ALL = 0, /* track all processes */
406 STATS_TRACK_PID, /* track process with this pid */
407 STATS_TRACK_PPID, /* track processes with this ppid */
408 STATS_TRACK_GID, /* track processes with this gid */
412 /* flags for sbi->ll_flags */
413 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
414 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
415 #define LL_SBI_FLOCK 0x04
416 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
417 #define LL_SBI_ACL 0x10 /* support ACL */
418 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
419 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
420 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
421 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
422 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
423 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
424 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
425 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
426 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
427 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
428 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
429 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
430 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
431 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
433 #define LL_SBI_FLAGS { \
456 #define RCE_HASHES 32
458 struct rmtacl_ctl_entry {
459 struct list_head rce_list;
460 pid_t rce_key; /* hash key */
461 int rce_ops; /* acl operation type */
464 struct rmtacl_ctl_table {
466 struct list_head rct_entries[RCE_HASHES];
472 struct list_head ee_list;
473 pid_t ee_key; /* hash key */
474 struct lu_fid ee_fid;
475 int ee_type; /* ACL type for ACCESS or DEFAULT */
476 ext_acl_xattr_header *ee_acl;
481 struct list_head et_entries[EE_HASHES];
485 struct list_head ll_list;
486 /* this protects pglist and ra_info. It isn't safe to
487 * grab from interrupt contexts */
489 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
490 spinlock_t ll_process_lock; /* ll_rw_process_info */
491 struct obd_uuid ll_sb_uuid;
492 struct obd_export *ll_md_exp;
493 struct obd_export *ll_dt_exp;
494 struct proc_dir_entry* ll_proc_root;
495 struct lu_fid ll_root_fid; /* root object fid */
498 unsigned int ll_umounting:1,
499 ll_xattr_cache_enabled:1;
500 /* per-conn chain of SBs */
501 struct list_head ll_conn_chain;
502 struct lustre_client_ocd ll_lco;
504 /*please don't ask -p*/
505 struct list_head ll_orphan_dentry_list;
506 struct ll_close_queue *ll_lcq;
508 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
510 /* Used to track "unstable" pages on a client, and maintain a
511 * LRU list of clean pages. An "unstable" page is defined as
512 * any page which is sent to a server as part of a bulk request,
513 * but is uncommitted to stable storage. */
514 struct cl_client_cache ll_cache;
516 struct lprocfs_stats *ll_ra_stats;
518 struct ll_ra_info ll_ra_info;
519 unsigned int ll_namelen;
520 struct file_operations *ll_fop;
522 /* =0 - hold lock over whole read/write
523 * >0 - max. chunk to be read/written w/o lock re-acquiring */
524 unsigned long ll_max_rw_chunk;
525 unsigned int ll_md_brw_size; /* used by readdir */
527 struct lu_site *ll_site;
528 struct cl_device *ll_cl;
530 struct ll_rw_extents_info ll_rw_extents_info;
531 int ll_extent_process_count;
532 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
533 unsigned int ll_offset_process_count;
534 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
535 unsigned int ll_rw_offset_entry_count;
536 int ll_stats_track_id;
537 enum stats_track_type ll_stats_track_type;
540 /* metadata stat-ahead */
541 unsigned int ll_sa_max; /* max statahead RPCs */
542 atomic_t ll_sa_total; /* statahead thread started
544 atomic_t ll_sa_wrong; /* statahead thread stopped for
546 atomic_t ll_agl_total; /* AGL thread started count */
548 dev_t ll_sdev_orig; /* save s_dev before assign for
550 struct rmtacl_ctl_table ll_rct;
551 struct eacl_table ll_et;
554 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
559 struct task_struct *lrr_reader;
560 struct list_head lrr_linkage;
564 * per file-descriptor read-ahead data.
566 struct ll_readahead_state {
569 * index of the last page that read(2) needed and that wasn't in the
570 * cache. Used by ras_update() to detect seeks.
572 * XXX nikita: if access seeks into cached region, Lustre doesn't see
575 unsigned long ras_last_readpage;
577 * number of pages read after last read-ahead window reset. As window
578 * is reset on each seek, this is effectively a number of consecutive
579 * accesses. Maybe ->ras_accessed_in_window is better name.
581 * XXX nikita: window is also reset (by ras_update()) when Lustre
582 * believes that memory pressure evicts read-ahead pages. In that
583 * case, it probably doesn't make sense to expand window to
584 * PTLRPC_MAX_BRW_PAGES on the third access.
586 unsigned long ras_consecutive_pages;
588 * number of read requests after the last read-ahead window reset
589 * As window is reset on each seek, this is effectively the number
590 * on consecutive read request and is used to trigger read-ahead.
592 unsigned long ras_consecutive_requests;
594 * Parameters of current read-ahead window. Handled by
595 * ras_update(). On the initial access to the file or after a seek,
596 * window is reset to 0. After 3 consecutive accesses, window is
597 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
598 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
600 unsigned long ras_window_start, ras_window_len;
602 * Where next read-ahead should start at. This lies within read-ahead
603 * window. Read-ahead window is read in pieces rather than at once
604 * because: 1. lustre limits total number of pages under read-ahead by
605 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
606 * not covered by DLM lock.
608 unsigned long ras_next_readahead;
610 * Total number of ll_file_read requests issued, reads originating
611 * due to mmap are not counted in this total. This value is used to
612 * trigger full file read-ahead after multiple reads to a small file.
614 unsigned long ras_requests;
616 * Page index with respect to the current request, these value
617 * will not be accurate when dealing with reads issued via mmap.
619 unsigned long ras_request_index;
621 * list of struct ll_ra_read's one per read(2) call current in
622 * progress against this file descriptor. Used by read-ahead code,
623 * protected by ->ras_lock.
625 struct list_head ras_read_beads;
627 * The following 3 items are used for detecting the stride I/O
629 * In stride I/O mode,
630 * ...............|-----data-----|****gap*****|--------|******|....
631 * offset |-stride_pages-|-stride_gap-|
632 * ras_stride_offset = offset;
633 * ras_stride_length = stride_pages + stride_gap;
634 * ras_stride_pages = stride_pages;
635 * Note: all these three items are counted by pages.
637 unsigned long ras_stride_length;
638 unsigned long ras_stride_pages;
639 pgoff_t ras_stride_offset;
641 * number of consecutive stride request count, and it is similar as
642 * ras_consecutive_requests, but used for stride I/O mode.
643 * Note: only more than 2 consecutive stride request are detected,
644 * stride read-ahead will be enable
646 unsigned long ras_consecutive_stride_requests;
649 extern struct kmem_cache *ll_file_data_slab;
650 struct lustre_handle;
651 struct ll_file_data {
652 struct ll_readahead_state fd_ras;
653 struct ccc_grouplock fd_grouplock;
657 /* openhandle if lease exists for this file.
658 * Borrow lli->lli_och_mutex to protect assignment */
659 struct obd_client_handle *fd_lease_och;
660 struct obd_client_handle *fd_och;
661 struct file *fd_file;
662 /* Indicate whether need to report failure when close.
663 * true: failure is known, not report again.
664 * false: unknown failure, should report. */
665 bool fd_write_failed;
668 struct lov_stripe_md;
670 extern spinlock_t inode_lock;
672 extern struct proc_dir_entry *proc_lustre_fs_root;
674 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
676 return &lli->lli_vfs_inode;
679 __u32 ll_i2suppgid(struct inode *i);
680 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
682 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
684 #if BITS_PER_LONG == 32
686 #elif defined(CONFIG_COMPAT)
687 return unlikely(is_compat_task() || (sbi->ll_flags & LL_SBI_32BIT_API));
689 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
693 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
694 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
695 struct ll_ra_read *ll_ra_read_get(struct file *f);
697 /* llite/lproc_llite.c */
699 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
700 struct super_block *sb, char *osc, char *mdc);
701 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
702 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
703 extern struct lprocfs_seq_vars lprocfs_llite_obd_vars[];
705 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
706 struct super_block *sb, char *osc, char *mdc){return 0;}
707 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
708 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
713 extern const struct file_operations ll_dir_operations;
714 extern const struct inode_operations ll_dir_inode_operations;
715 #ifdef HAVE_DIR_CONTEXT
716 int ll_dir_read(struct inode *inode, struct md_op_data *op_data,
717 struct dir_context *ctx);
719 int ll_dir_read(struct inode *inode, struct md_op_data *op_data,
720 void *cookie, filldir_t filldir);
722 int ll_get_mdt_idx(struct inode *inode);
723 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
725 struct lu_dirent *ll_dir_entry_start(struct inode *dir,
726 struct md_op_data *op_data,
727 struct page **ppage);
729 struct lu_dirent *ll_dir_entry_next(struct inode *dir,
730 struct md_op_data *op_data,
731 struct lu_dirent *ent,
732 struct page **ppage);
734 extern const struct inode_operations ll_special_inode_operations;
736 int ll_objects_destroy(struct ptlrpc_request *request,
738 struct inode *ll_iget(struct super_block *sb, ino_t hash,
739 struct lustre_md *lic);
740 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
741 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
742 void *data, int flag);
743 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
744 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
745 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
748 int ll_writepage(struct page *page, struct writeback_control *wbc);
749 int ll_writepages(struct address_space *, struct writeback_control *wbc);
750 int ll_readpage(struct file *file, struct page *page);
751 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
752 int ll_readahead(const struct lu_env *env, struct cl_io *io,
753 struct cl_page_list *queue, struct ll_readahead_state *ras,
755 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
756 struct ll_cl_context *ll_cl_init(struct file *file, struct page *vmpage);
757 void ll_cl_fini(struct ll_cl_context *lcc);
759 #ifndef MS_HAS_NEW_AOPS
760 extern const struct address_space_operations ll_aops;
762 extern const struct address_space_operations_ext ll_aops;
766 extern struct file_operations ll_file_operations;
767 extern struct file_operations ll_file_operations_flock;
768 extern struct file_operations ll_file_operations_noflock;
769 extern struct inode_operations ll_file_inode_operations;
770 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
771 ldlm_mode_t l_req_mode);
772 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
773 struct lustre_handle *lockh, __u64 flags,
776 int ll_file_open(struct inode *inode, struct file *file);
777 int ll_file_release(struct inode *inode, struct file *file);
778 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
779 struct lov_stripe_md *lsm, lstat_t *st);
780 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
781 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
782 int ll_md_real_close(struct inode *inode, fmode_t fmode);
783 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
784 struct obd_client_handle **och, unsigned long flags);
785 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
786 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
787 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
788 __u64 ioepoch, int sync);
789 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
790 struct lustre_handle *fh);
791 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
792 struct ll_file_data *file, loff_t pos,
793 size_t count, int rw);
794 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
795 struct posix_acl *ll_get_acl(struct inode *inode, int type);
796 int ll_migrate(struct inode *parent, struct file *file, int mdtidx,
797 const char *name, int namelen);
798 int ll_get_fid_by_name(struct inode *parent, const char *name,
799 int namelen, struct lu_fid *fid);
800 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
801 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
803 # ifndef HAVE_INODE_PERMISION_2ARGS
804 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
806 int ll_inode_permission(struct inode *inode, int mask);
810 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
811 __u64 flags, struct lov_user_md *lum,
813 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
814 struct lov_mds_md **lmm, int *lmm_size,
815 struct ptlrpc_request **request);
816 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
818 int ll_dir_getstripe(struct inode *inode, void **lmmp,
819 int *lmm_size, struct ptlrpc_request **request,
821 #ifdef HAVE_FILE_FSYNC_4ARGS
822 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
823 #elif defined(HAVE_FILE_FSYNC_2ARGS)
824 int ll_fsync(struct file *file, int data);
826 int ll_fsync(struct file *file, struct dentry *dentry, int data);
828 int ll_merge_lvb(const struct lu_env *env, struct inode *inode);
829 int ll_fid2path(struct inode *inode, void *arg);
830 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
831 int ll_hsm_release(struct inode *inode);
835 int ll_d_init(struct dentry *de);
836 extern const struct dentry_operations ll_d_ops;
837 void ll_intent_drop_lock(struct lookup_intent *);
838 void ll_intent_release(struct lookup_intent *);
839 void ll_invalidate_aliases(struct inode *);
840 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
841 int ll_revalidate_it_finish(struct ptlrpc_request *request,
842 struct lookup_intent *it, struct dentry *de);
844 /* llite/llite_lib.c */
845 extern struct super_operations lustre_super_operations;
847 void ll_lli_init(struct ll_inode_info *lli);
848 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
849 void ll_put_super(struct super_block *sb);
850 void ll_kill_super(struct super_block *sb);
851 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
852 void ll_clear_inode(struct inode *inode);
853 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import);
854 int ll_setattr(struct dentry *de, struct iattr *attr);
855 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
856 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
857 __u64 max_age, __u32 flags);
858 int ll_update_inode(struct inode *inode, struct lustre_md *md);
859 int ll_read_inode2(struct inode *inode, void *opaque);
860 void ll_delete_inode(struct inode *inode);
861 int ll_iocontrol(struct inode *inode, struct file *file,
862 unsigned int cmd, unsigned long arg);
863 int ll_flush_ctx(struct inode *inode);
864 void ll_umount_begin(struct super_block *sb);
865 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
866 #ifdef HAVE_SUPEROPS_USE_DENTRY
867 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
869 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
871 void ll_dirty_page_discard_warn(struct page *page, int ioret);
872 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
873 struct super_block *, struct lookup_intent *);
874 void lustre_dump_dentry(struct dentry *, int recur);
875 int ll_obd_statfs(struct inode *inode, void *arg);
876 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
877 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
878 int ll_get_max_cookiesize(struct ll_sb_info *sbi, int *max_cookiesize);
879 int ll_get_default_cookiesize(struct ll_sb_info *sbi, int *default_cookiesize);
880 int ll_process_config(struct lustre_cfg *lcfg);
881 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
882 struct inode *i1, struct inode *i2,
883 const char *name, int namelen,
884 int mode, __u32 opc, void *data);
885 void ll_finish_md_op_data(struct md_op_data *op_data);
886 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
887 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
889 /* llite/llite_nfs.c */
890 extern struct export_operations lustre_export_operations;
891 __u32 get_uuid2int(const char *name, int len);
892 struct inode *search_inode_for_lustre(struct super_block *sb,
893 const struct lu_fid *fid);
895 /* llite/symlink.c */
896 extern struct inode_operations ll_fast_symlink_inode_operations;
898 /* llite/llite_close.c */
899 struct ll_close_queue {
901 struct list_head lcq_head;
902 wait_queue_head_t lcq_waitq;
903 struct completion lcq_comp;
907 struct ccc_object *cl_inode2ccc(struct inode *inode);
910 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
911 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
913 /* specific achitecture can implement only part of this list */
914 enum vvp_io_subtype {
917 /** io started from splice_{read|write} */
924 enum vvp_io_subtype cui_io_subtype;
928 struct pipe_inode_info *cui_pipe;
929 unsigned int cui_flags;
931 struct vvp_fault_io {
933 * Inode modification time that is checked across DLM
937 struct vm_area_struct *ft_vma;
939 * locked page returned from vvp_io
941 struct page *ft_vmpage;
942 struct vm_fault_api {
946 struct vm_fault *ft_vmf;
948 * fault API used bitflags for return code.
950 unsigned int ft_flags;
955 * Read-ahead state used by read and page-fault IO contexts.
957 struct ll_ra_read cui_bead;
959 * Set when cui_bead has been initialized.
961 int cui_ra_window_set;
965 * IO arguments for various VFS I/O interfaces.
968 /** normal/sendfile/splice */
969 enum vvp_io_subtype via_io_subtype;
973 struct kiocb *via_iocb;
974 struct iovec *via_iov;
975 unsigned long via_nrsegs;
978 struct pipe_inode_info *via_pipe;
979 unsigned int via_flags;
984 struct ll_cl_context {
986 struct cl_io *lcc_io;
987 struct cl_page *lcc_page;
988 struct lu_env *lcc_env;
992 struct vvp_thread_info {
993 struct iovec vti_local_iov;
994 struct vvp_io_args vti_args;
995 struct ra_io_arg vti_ria;
996 struct kiocb vti_kiocb;
997 struct ll_cl_context vti_io_ctx;
1000 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
1002 extern struct lu_context_key vvp_key;
1003 struct vvp_thread_info *info;
1005 info = lu_context_key_get(&env->le_ctx, &vvp_key);
1006 LASSERT(info != NULL);
1010 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
1011 enum vvp_io_subtype type)
1013 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
1015 ret->via_io_subtype = type;
1020 struct vvp_session {
1021 struct vvp_io vs_ios;
1024 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1026 extern struct lu_context_key vvp_session_key;
1027 struct vvp_session *ses;
1029 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1030 LASSERT(ses != NULL);
1034 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1036 return &vvp_env_session(env)->vs_ios;
1039 int vvp_global_init(void);
1040 void vvp_global_fini(void);
1042 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1043 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1044 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1046 /* llite/llite_mmap.c */
1048 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1049 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1050 void policy_from_vma(ldlm_policy_data_t *policy,
1051 struct vm_area_struct *vma, unsigned long addr, size_t count);
1052 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1055 static inline void ll_invalidate_page(struct page *vmpage)
1057 struct address_space *mapping = vmpage->mapping;
1058 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1060 LASSERT(PageLocked(vmpage));
1061 if (mapping == NULL)
1065 * truncate_complete_page() calls
1066 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
1068 ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE);
1069 truncate_complete_page(mapping, vmpage);
1072 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1074 /* don't need an addref as the sb_info should be holding one */
1075 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1077 return ll_s2sbi(sb)->ll_dt_exp;
1080 /* don't need an addref as the sb_info should be holding one */
1081 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1083 return ll_s2sbi(sb)->ll_md_exp;
1086 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1088 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1094 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1095 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1097 return ll_s2sbi(inode->i_sb);
1100 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1102 return ll_s2dtexp(inode->i_sb);
1105 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1107 return ll_s2mdexp(inode->i_sb);
1110 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1114 LASSERT(inode != NULL);
1115 fid = &ll_i2info(inode)->lli_fid;
1120 static inline __u64 ll_file_maxbytes(struct inode *inode)
1122 return ll_i2info(inode)->lli_maxbytes;
1126 int ll_setxattr(struct dentry *dentry, const char *name,
1127 const void *value, size_t size, int flags);
1128 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1129 void *buffer, size_t size);
1130 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1131 int ll_removexattr(struct dentry *dentry, const char *name);
1133 /* llite/remote_perm.c */
1134 extern struct kmem_cache *ll_remote_perm_cachep;
1135 extern struct kmem_cache *ll_rmtperm_hash_cachep;
1137 void free_rmtperm_hash(struct hlist_head *hash);
1138 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1139 int lustre_check_remote_perm(struct inode *inode, int mask);
1141 /* llite/llite_capa.c */
1142 extern struct timer_list ll_capa_timer;
1144 int ll_capa_thread_start(void);
1145 void ll_capa_thread_stop(void);
1146 void ll_capa_timer_callback(unsigned long unused);
1148 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1150 void ll_capa_open(struct inode *inode);
1151 void ll_capa_close(struct inode *inode);
1153 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1154 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1156 void ll_truncate_free_capa(struct obd_capa *ocapa);
1157 void ll_clear_inode_capas(struct inode *inode);
1158 void ll_print_capa_stat(struct ll_sb_info *sbi);
1160 /* llite/llite_cl.c */
1161 extern struct lu_device_type vvp_device_type;
1164 * Common IO arguments for various VFS I/O interfaces.
1166 int cl_sb_init(struct super_block *sb);
1167 int cl_sb_fini(struct super_block *sb);
1169 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1170 struct ll_readahead_state *ras, unsigned long index,
1172 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1173 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1175 /* llite/llite_rmtacl.c */
1176 #ifdef CONFIG_FS_POSIX_ACL
1177 obd_valid rce_ops2valid(int ops);
1178 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1179 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1180 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1181 void rct_init(struct rmtacl_ctl_table *rct);
1182 void rct_fini(struct rmtacl_ctl_table *rct);
1184 void ee_free(struct eacl_entry *ee);
1185 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1186 ext_acl_xattr_header *header);
1187 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1188 struct lu_fid *fid, int type);
1189 void et_search_free(struct eacl_table *et, pid_t key);
1190 void et_init(struct eacl_table *et);
1191 void et_fini(struct eacl_table *et);
1196 #define LL_SA_RPC_MIN 2
1197 #define LL_SA_RPC_DEF 32
1198 #define LL_SA_RPC_MAX 8192
1200 #define LL_SA_CACHE_BIT 5
1201 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1202 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1204 /* per inode struct, for dir only */
1205 struct ll_statahead_info {
1206 struct inode *sai_inode;
1207 atomic_t sai_refcount; /* when access this struct, hold
1209 unsigned int sai_generation; /* generation for statahead */
1210 unsigned int sai_max; /* max ahead of lookup */
1211 __u64 sai_sent; /* stat requests sent count */
1212 __u64 sai_replied; /* stat requests which received
1214 __u64 sai_index; /* index of statahead entry */
1215 __u64 sai_index_wait; /* index of entry which is the
1216 * caller is waiting for */
1217 __u64 sai_hit; /* hit count */
1218 __u64 sai_miss; /* miss count:
1219 * for "ls -al" case, it
1220 * includes hidden dentry miss;
1221 * for "ls -l" case, it does not
1222 * include hidden dentry miss.
1223 * "sai_miss_hidden" is used for
1226 unsigned int sai_consecutive_miss; /* consecutive miss */
1227 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1228 * is not a hidden one */
1229 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1230 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1232 sai_agl_valid:1;/* AGL is valid for the dir */
1233 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1234 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1235 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1236 struct list_head sai_entries; /* entry list */
1237 struct list_head sai_entries_received; /* entries returned */
1238 struct list_head sai_entries_stated; /* entries stated */
1239 struct list_head sai_entries_agl; /* AGL entries to be sent */
1240 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1241 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1242 atomic_t sai_cache_count; /* entry count in cache */
1245 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1247 void ll_stop_statahead(struct inode *dir, void *key);
1249 static inline int ll_glimpse_size(struct inode *inode)
1251 struct ll_inode_info *lli = ll_i2info(inode);
1254 down_read(&lli->lli_glimpse_sem);
1255 rc = cl_glimpse_size(inode);
1256 lli->lli_glimpse_time = cfs_time_current();
1257 up_read(&lli->lli_glimpse_sem);
1262 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1264 struct ll_inode_info *lli = ll_i2info(dir);
1265 struct ll_statahead_info *sai = lli->lli_sai;
1266 struct ll_dentry_data *ldd = ll_d2d(dentry);
1268 /* not the same process, don't mark */
1269 if (lli->lli_opendir_pid != current_pid())
1272 LASSERT(ldd != NULL);
1274 ldd->lld_sa_generation = sai->sai_generation;
1278 d_need_statahead(struct inode *dir, struct dentry *dentryp)
1280 struct ll_inode_info *lli;
1281 struct ll_dentry_data *ldd;
1283 if (ll_i2sbi(dir)->ll_sa_max == 0)
1286 lli = ll_i2info(dir);
1287 /* not the same process, don't statahead */
1288 if (lli->lli_opendir_pid != current_pid())
1291 /* statahead has been stopped */
1292 if (lli->lli_opendir_key == NULL)
1295 ldd = ll_d2d(dentryp);
1297 * When stats a dentry, the system trigger more than once "revalidate"
1298 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1299 * Under patchless client mode, the operation intent is not accurate,
1300 * which maybe misguide the statahead thread. For example:
1301 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1302 * have the same operation intent -- "IT_GETATTR".
1303 * In fact, one dentry should has only one chance to interact with the
1304 * statahead thread, otherwise the statahead windows will be confused.
1305 * The solution is as following:
1306 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1307 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1308 * will bypass interacting with statahead thread for checking:
1309 * "lld_sa_generation == lli_sai->sai_generation"
1311 if (ldd && lli->lli_sai &&
1312 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1319 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1323 ret = d_need_statahead(dir, *dentryp);
1327 return do_statahead_enter(dir, dentryp, only_unplug);
1330 /* llite ioctl register support rountine */
1337 #define LLIOC_MAX_CMD 256
1340 * Rules to write a callback function:
1343 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1344 * returned to ll_iocontrol_register. Callback functions should use this
1345 * data to check the potential collasion of ioctl cmd. If collasion is
1346 * found, callback function should return LLIOC_CONT.
1347 * @rcp: The result of ioctl command.
1350 * If @magic matches the pointer returned by ll_iocontrol_data, the
1351 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1353 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1354 struct file *file, unsigned int cmd, unsigned long arg,
1355 void *magic, int *rcp);
1357 /* export functions */
1358 /* Register ioctl block dynamatically for a regular file.
1360 * @cmd: the array of ioctl command set
1361 * @count: number of commands in the @cmd
1362 * @cb: callback function, it will be called if an ioctl command is found to
1363 * belong to the command list @cmd.
1366 * A magic pointer will be returned if success;
1367 * otherwise, NULL will be returned.
1369 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1370 void ll_iocontrol_unregister(void *magic);
1374 /* lclient compat stuff */
1375 #define cl_inode_info ll_inode_info
1376 #define cl_i2info(info) ll_i2info(info)
1377 #define cl_inode_mode(inode) ((inode)->i_mode)
1378 #define cl_i2sbi ll_i2sbi
1380 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1381 const struct iattr *attr)
1383 LASSERT(attr->ia_valid & ATTR_FILE);
1384 return LUSTRE_FPRIVATE(attr->ia_file);
1387 static inline void cl_isize_lock(struct inode *inode)
1389 ll_inode_size_lock(inode);
1392 static inline void cl_isize_unlock(struct inode *inode)
1394 ll_inode_size_unlock(inode);
1397 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1399 LASSERT(mutex_is_locked(&ll_i2info(inode)->lli_size_mutex));
1400 i_size_write(inode, kms);
1403 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1405 ll_inode_size_lock(inode);
1406 i_size_write(inode, kms);
1407 ll_inode_size_unlock(inode);
1410 #define cl_isize_read(inode) i_size_read(inode)
1412 static inline int cl_merge_lvb(const struct lu_env *env, struct inode *inode)
1414 return ll_merge_lvb(env, inode);
1417 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1418 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1419 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1421 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1423 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1424 enum cl_fsync_mode mode, int ignore_layout);
1426 /** direct write pages */
1427 struct ll_dio_pages {
1428 /** page array to be written. we don't support
1429 * partial pages except the last one. */
1430 struct page **ldp_pages;
1431 /* offset of each page */
1432 loff_t *ldp_offsets;
1433 /** if ldp_offsets is NULL, it means a sequential
1434 * pages to be written, then this is the file offset
1435 * of the * first page. */
1436 loff_t ldp_start_offset;
1437 /** how many bytes are to be written. */
1439 /** # of pages in the array. */
1443 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1446 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1449 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1452 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1453 int rw, struct inode *inode,
1454 struct ll_dio_pages *pv);
1456 static inline int ll_file_nolock(const struct file *file)
1458 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1459 struct inode *inode = file->f_dentry->d_inode;
1461 LASSERT(fd != NULL);
1462 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1463 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1466 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1467 struct lookup_intent *it, __u64 *bits)
1469 if (!it->d.lustre.it_lock_set) {
1470 struct lustre_handle handle;
1472 /* If this inode is a remote object, it will get two
1473 * separate locks in different namespaces, Master MDT,
1474 * where the name entry is, will grant LOOKUP lock,
1475 * remote MDT, where the object is, will grant
1476 * UPDATE|PERM lock. The inode will be attched to both
1477 * LOOKUP and PERM locks, so revoking either locks will
1478 * case the dcache being cleared */
1479 if (it->d.lustre.it_remote_lock_mode) {
1480 handle.cookie = it->d.lustre.it_remote_lock_handle;
1481 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1482 "(%p) for remote lock "LPX64"\n",
1483 PFID(ll_inode2fid(inode)), inode,
1485 md_set_lock_data(exp, &handle.cookie, inode, NULL);
1488 handle.cookie = it->d.lustre.it_lock_handle;
1490 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1491 " for lock "LPX64"\n",
1492 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1494 md_set_lock_data(exp, &handle.cookie, inode,
1495 &it->d.lustre.it_lock_bits);
1496 it->d.lustre.it_lock_set = 1;
1500 *bits = it->d.lustre.it_lock_bits;
1503 static inline void ll_lock_dcache(struct inode *inode)
1505 #ifdef HAVE_DCACHE_LOCK
1506 spin_lock(&dcache_lock);
1508 spin_lock(&inode->i_lock);
1512 static inline void ll_unlock_dcache(struct inode *inode)
1514 #ifdef HAVE_DCACHE_LOCK
1515 spin_unlock(&dcache_lock);
1517 spin_unlock(&inode->i_lock);
1521 static inline int d_lustre_invalid(const struct dentry *dentry)
1523 struct ll_dentry_data *lld = ll_d2d(dentry);
1525 return (lld == NULL) || lld->lld_invalid;
1528 static inline void __d_lustre_invalidate(struct dentry *dentry)
1530 struct ll_dentry_data *lld = ll_d2d(dentry);
1533 lld->lld_invalid = 1;
1537 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1538 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1539 * else dput() of the last refcount will unhash this dentry and kill it.
1541 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1543 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1544 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1545 dentry->d_parent, dentry->d_inode, d_count(dentry));
1547 spin_lock_nested(&dentry->d_lock,
1548 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1549 __d_lustre_invalidate(dentry);
1550 if (d_count(dentry) == 0)
1552 spin_unlock(&dentry->d_lock);
1555 static inline void d_lustre_revalidate(struct dentry *dentry)
1557 spin_lock(&dentry->d_lock);
1558 LASSERT(ll_d2d(dentry) != NULL);
1559 ll_d2d(dentry)->lld_invalid = 0;
1560 spin_unlock(&dentry->d_lock);
1563 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0)
1564 /* Compatibility for old (1.8) compiled userspace quota code */
1565 struct if_quotactl_18 {
1570 struct obd_dqinfo qc_dqinfo;
1571 struct obd_dqblk qc_dqblk;
1573 struct obd_uuid obd_uuid;
1575 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1576 /* End compatibility for old (1.8) compiled userspace quota code */
1578 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1579 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */
1582 LL_LAYOUT_GEN_NONE = ((__u32)-2), /* layout lock was cancelled */
1583 LL_LAYOUT_GEN_EMPTY = ((__u32)-1) /* for empty layout */
1586 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1587 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1588 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1590 int ll_xattr_init(void);
1591 void ll_xattr_fini(void);
1593 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1594 struct cl_page *page, enum cl_req_type crt);
1596 #endif /* LLITE_INTERNAL_H */