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 mutex f_size_mutex;
213 char *f_symlink_name;
216 * struct rw_semaphore {
217 * signed long count; // align d.d_def_acl
218 * spinlock_t wait_lock; // align d.d_sa_lock
219 * struct list_head wait_list;
222 struct rw_semaphore f_trunc_sem;
223 struct mutex f_write_mutex;
225 struct rw_semaphore f_glimpse_sem;
226 cfs_time_t f_glimpse_time;
227 cfs_list_t f_agl_list;
230 /* for writepage() only to communicate to fsync */
233 /* volatile file criteria is based on file name, this
234 * flag is used to keep the test result, so the strcmp
239 * whenever a process try to read/write the file, the
240 * jobid of the process will be saved here, and it'll
241 * be packed into the write PRC when flush later.
243 * so the read/write statistics for jobid will not be
244 * accurate if the file is shared by different jobs.
246 char f_jobid[JOBSTATS_JOBID_SIZE];
249 #define lli_size_mutex u.f.f_size_mutex
250 #define lli_symlink_name u.f.f_symlink_name
251 #define lli_maxbytes u.f.f_maxbytes
252 #define lli_trunc_sem u.f.f_trunc_sem
253 #define lli_write_mutex u.f.f_write_mutex
254 #define lli_glimpse_sem u.f.f_glimpse_sem
255 #define lli_glimpse_time u.f.f_glimpse_time
256 #define lli_agl_list u.f.f_agl_list
257 #define lli_agl_index u.f.f_agl_index
258 #define lli_async_rc u.f.f_async_rc
259 #define lli_jobid u.f.f_jobid
260 #define lli_volatile u.f.f_volatile
264 /* XXX: For following frequent used members, although they maybe special
265 * used for non-directory object, it is some time-wasting to check
266 * whether the object is directory or not before using them. On the
267 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
268 * the "ll_inode_info" size even if moving those members into u.f.
269 * So keep them out side.
271 * In the future, if more members are added only for directory,
272 * some of the following members can be moved into u.f.
275 struct cl_object *lli_clob;
277 /* mutex to request for layout lock exclusively. */
278 struct mutex lli_layout_mutex;
279 /* Layout version, protected by lli_layout_lock */
280 __u32 lli_layout_gen;
281 spinlock_t lli_layout_lock;
283 struct rw_semaphore lli_xattrs_list_rwsem;
284 struct mutex lli_xattrs_enq_lock;
285 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
288 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
292 spin_lock(&lli->lli_layout_lock);
293 gen = lli->lli_layout_gen;
294 spin_unlock(&lli->lli_layout_lock);
299 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
301 spin_lock(&lli->lli_layout_lock);
302 lli->lli_layout_gen = gen;
303 spin_unlock(&lli->lli_layout_lock);
306 int ll_xattr_cache_destroy(struct inode *inode);
308 int ll_xattr_cache_get(struct inode *inode,
315 * Locking to guarantee consistency of non-atomic updates to long long i_size,
316 * consistency between file size and KMS.
318 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
321 void ll_inode_size_lock(struct inode *inode);
322 void ll_inode_size_unlock(struct inode *inode);
324 // FIXME: replace the name of this with LL_I to conform to kernel stuff
325 // static inline struct ll_inode_info *LL_I(struct inode *inode)
326 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
328 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
331 /* default to about 40meg of readahead on a given system. That much tied
332 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
333 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT))
335 /* default to read-ahead full files smaller than 2MB on the second read */
336 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
341 RA_STAT_DISTANT_READPAGE,
342 RA_STAT_MISS_IN_WINDOW,
343 RA_STAT_FAILED_GRAB_PAGE,
344 RA_STAT_FAILED_MATCH,
349 RA_STAT_MAX_IN_FLIGHT,
350 RA_STAT_WRONG_GRAB_PAGE,
355 cfs_atomic_t ra_cur_pages;
356 unsigned long ra_max_pages;
357 unsigned long ra_max_pages_per_file;
358 unsigned long ra_max_read_ahead_whole_pages;
361 /* ra_io_arg will be filled in the beginning of ll_readahead with
362 * ras_lock, then the following ll_read_ahead_pages will read RA
363 * pages according to this arg, all the items in this structure are
364 * counted by page index.
367 unsigned long ria_start; /* start offset of read-ahead*/
368 unsigned long ria_end; /* end offset of read-ahead*/
369 /* If stride read pattern is detected, ria_stoff means where
370 * stride read is started. Note: for normal read-ahead, the
371 * value here is meaningless, and also it will not be accessed*/
373 /* ria_length and ria_pages are the length and pages length in the
374 * stride I/O mode. And they will also be used to check whether
375 * it is stride I/O read-ahead in the read-ahead pages*/
376 unsigned long ria_length;
377 unsigned long ria_pages;
380 /* LL_HIST_MAX=32 causes an overflow */
381 #define LL_HIST_MAX 28
382 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
383 #define LL_PROCESS_HIST_MAX 10
384 struct per_process_info {
386 struct obd_histogram pp_r_hist;
387 struct obd_histogram pp_w_hist;
390 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
391 struct ll_rw_extents_info {
392 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
395 #define LL_OFFSET_HIST_MAX 100
396 struct ll_rw_process_info {
399 loff_t rw_range_start;
401 loff_t rw_last_file_pos;
403 size_t rw_smallest_extent;
404 size_t rw_largest_extent;
405 struct ll_file_data *rw_last_file;
408 enum stats_track_type {
409 STATS_TRACK_ALL = 0, /* track all processes */
410 STATS_TRACK_PID, /* track process with this pid */
411 STATS_TRACK_PPID, /* track processes with this ppid */
412 STATS_TRACK_GID, /* track processes with this gid */
416 /* flags for sbi->ll_flags */
417 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
418 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
419 #define LL_SBI_FLOCK 0x04
420 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
421 #define LL_SBI_ACL 0x10 /* support ACL */
422 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
423 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
424 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
425 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
426 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
427 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
428 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
429 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
430 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
431 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
432 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
433 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
434 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
435 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
437 #define LL_SBI_FLAGS { \
460 /* default value for ll_sb_info->contention_time */
461 #define SBI_DEFAULT_CONTENTION_SECONDS 60
462 /* default value for lockless_truncate_enable */
463 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
464 #define RCE_HASHES 32
466 struct rmtacl_ctl_entry {
468 pid_t rce_key; /* hash key */
469 int rce_ops; /* acl operation type */
472 struct rmtacl_ctl_table {
474 cfs_list_t rct_entries[RCE_HASHES];
481 pid_t ee_key; /* hash key */
482 struct lu_fid ee_fid;
483 int ee_type; /* ACL type for ACCESS or DEFAULT */
484 ext_acl_xattr_header *ee_acl;
489 cfs_list_t et_entries[EE_HASHES];
494 /* this protects pglist and ra_info. It isn't safe to
495 * grab from interrupt contexts */
497 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
498 spinlock_t ll_process_lock; /* ll_rw_process_info */
499 struct obd_uuid ll_sb_uuid;
500 struct obd_export *ll_md_exp;
501 struct obd_export *ll_dt_exp;
502 struct proc_dir_entry* ll_proc_root;
503 struct lu_fid ll_root_fid; /* root object fid */
506 unsigned int ll_umounting:1,
507 ll_xattr_cache_enabled:1;
508 cfs_list_t ll_conn_chain; /* per-conn chain of SBs */
509 struct lustre_client_ocd ll_lco;
511 cfs_list_t ll_orphan_dentry_list; /*please don't ask -p*/
512 struct ll_close_queue *ll_lcq;
514 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
516 struct cl_client_cache ll_cache;
518 struct lprocfs_stats *ll_ra_stats;
520 struct ll_ra_info ll_ra_info;
521 unsigned int ll_namelen;
522 struct file_operations *ll_fop;
524 /* =0 - hold lock over whole read/write
525 * >0 - max. chunk to be read/written w/o lock re-acquiring */
526 unsigned long ll_max_rw_chunk;
527 unsigned int ll_md_brw_size; /* used by readdir */
529 struct lu_site *ll_site;
530 struct cl_device *ll_cl;
532 struct ll_rw_extents_info ll_rw_extents_info;
533 int ll_extent_process_count;
534 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
535 unsigned int ll_offset_process_count;
536 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
537 unsigned int ll_rw_offset_entry_count;
538 int ll_stats_track_id;
539 enum stats_track_type ll_stats_track_type;
542 /* metadata stat-ahead */
543 unsigned int ll_sa_max; /* max statahead RPCs */
544 atomic_t ll_sa_total; /* statahead thread started
546 atomic_t ll_sa_wrong; /* statahead thread stopped for
548 atomic_t ll_agl_total; /* AGL thread started count */
550 dev_t ll_sdev_orig; /* save s_dev before assign for
552 struct rmtacl_ctl_table ll_rct;
553 struct eacl_table ll_et;
556 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
561 struct task_struct *lrr_reader;
562 cfs_list_t lrr_linkage;
566 * per file-descriptor read-ahead data.
568 struct ll_readahead_state {
571 * index of the last page that read(2) needed and that wasn't in the
572 * cache. Used by ras_update() to detect seeks.
574 * XXX nikita: if access seeks into cached region, Lustre doesn't see
577 unsigned long ras_last_readpage;
579 * number of pages read after last read-ahead window reset. As window
580 * is reset on each seek, this is effectively a number of consecutive
581 * accesses. Maybe ->ras_accessed_in_window is better name.
583 * XXX nikita: window is also reset (by ras_update()) when Lustre
584 * believes that memory pressure evicts read-ahead pages. In that
585 * case, it probably doesn't make sense to expand window to
586 * PTLRPC_MAX_BRW_PAGES on the third access.
588 unsigned long ras_consecutive_pages;
590 * number of read requests after the last read-ahead window reset
591 * As window is reset on each seek, this is effectively the number
592 * on consecutive read request and is used to trigger read-ahead.
594 unsigned long ras_consecutive_requests;
596 * Parameters of current read-ahead window. Handled by
597 * ras_update(). On the initial access to the file or after a seek,
598 * window is reset to 0. After 3 consecutive accesses, window is
599 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
600 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
602 unsigned long ras_window_start, ras_window_len;
604 * Where next read-ahead should start at. This lies within read-ahead
605 * window. Read-ahead window is read in pieces rather than at once
606 * because: 1. lustre limits total number of pages under read-ahead by
607 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
608 * not covered by DLM lock.
610 unsigned long ras_next_readahead;
612 * Total number of ll_file_read requests issued, reads originating
613 * due to mmap are not counted in this total. This value is used to
614 * trigger full file read-ahead after multiple reads to a small file.
616 unsigned long ras_requests;
618 * Page index with respect to the current request, these value
619 * will not be accurate when dealing with reads issued via mmap.
621 unsigned long ras_request_index;
623 * list of struct ll_ra_read's one per read(2) call current in
624 * progress against this file descriptor. Used by read-ahead code,
625 * protected by ->ras_lock.
627 cfs_list_t ras_read_beads;
629 * The following 3 items are used for detecting the stride I/O
631 * In stride I/O mode,
632 * ...............|-----data-----|****gap*****|--------|******|....
633 * offset |-stride_pages-|-stride_gap-|
634 * ras_stride_offset = offset;
635 * ras_stride_length = stride_pages + stride_gap;
636 * ras_stride_pages = stride_pages;
637 * Note: all these three items are counted by pages.
639 unsigned long ras_stride_length;
640 unsigned long ras_stride_pages;
641 pgoff_t ras_stride_offset;
643 * number of consecutive stride request count, and it is similar as
644 * ras_consecutive_requests, but used for stride I/O mode.
645 * Note: only more than 2 consecutive stride request are detected,
646 * stride read-ahead will be enable
648 unsigned long ras_consecutive_stride_requests;
651 extern struct kmem_cache *ll_file_data_slab;
652 struct lustre_handle;
653 struct ll_file_data {
654 struct ll_readahead_state fd_ras;
655 struct ccc_grouplock fd_grouplock;
659 /* openhandle if lease exists for this file.
660 * Borrow lli->lli_och_mutex to protect assignment */
661 struct obd_client_handle *fd_lease_och;
662 struct obd_client_handle *fd_och;
663 struct file *fd_file;
664 /* Indicate whether need to report failure when close.
665 * true: failure is known, not report again.
666 * false: unknown failure, should report. */
667 bool fd_write_failed;
670 struct lov_stripe_md;
672 extern spinlock_t inode_lock;
674 extern struct proc_dir_entry *proc_lustre_fs_root;
676 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
678 return &lli->lli_vfs_inode;
682 struct inode *icbd_parent;
683 struct dentry **icbd_childp;
687 __u32 ll_i2suppgid(struct inode *i);
688 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
690 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
692 #if BITS_PER_LONG == 32
699 (sbi->ll_flags & LL_SBI_32BIT_API)
704 #define LLAP_MAGIC 98764321
706 extern struct kmem_cache *ll_async_page_slab;
707 extern size_t ll_async_page_slab_size;
709 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
710 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
711 struct ll_ra_read *ll_ra_read_get(struct file *f);
713 /* llite/lproc_llite.c */
715 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
716 struct super_block *sb, char *osc, char *mdc);
717 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
718 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
719 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
721 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
722 struct super_block *sb, char *osc, char *mdc){return 0;}
723 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
724 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
725 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
727 memset(lvars, 0, sizeof(*lvars));
733 void ll_release_page(struct page *page, int remove);
734 extern struct file_operations ll_dir_operations;
735 extern struct inode_operations ll_dir_inode_operations;
736 struct page *ll_get_dir_page(struct inode *dir, __u64 hash,
737 struct ll_dir_chain *chain);
738 int ll_dir_read(struct inode *inode, __u64 *_pos, void *cookie,
741 int ll_get_mdt_idx(struct inode *inode);
743 int ll_objects_destroy(struct ptlrpc_request *request,
745 struct inode *ll_iget(struct super_block *sb, ino_t hash,
746 struct lustre_md *lic);
747 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
748 void *data, int flag);
749 #ifndef HAVE_IOP_ATOMIC_OPEN
750 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
753 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
754 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
757 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
758 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
759 int ll_writepage(struct page *page, struct writeback_control *wbc);
760 int ll_writepages(struct address_space *, struct writeback_control *wbc);
761 void ll_removepage(struct page *page);
762 int ll_readpage(struct file *file, struct page *page);
763 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
764 int ll_file_punch(struct inode *, loff_t, int);
765 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
766 void ll_clear_file_contended(struct inode*);
767 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
768 int ll_readahead(const struct lu_env *env, struct cl_io *io,
769 struct ll_readahead_state *ras, struct address_space *mapping,
770 struct cl_page_list *queue, int flags);
773 extern struct file_operations ll_file_operations;
774 extern struct file_operations ll_file_operations_flock;
775 extern struct file_operations ll_file_operations_noflock;
776 extern struct inode_operations ll_file_inode_operations;
777 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
779 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
780 ldlm_mode_t l_req_mode);
781 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
782 struct lustre_handle *lockh, __u64 flags,
784 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
786 #ifdef HAVE_IOP_ATOMIC_OPEN
787 int ll_revalidate_nd(struct dentry *dentry, unsigned int flags);
789 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
791 int ll_file_open(struct inode *inode, struct file *file);
792 int ll_file_release(struct inode *inode, struct file *file);
793 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
794 struct lov_stripe_md *lsm, lstat_t *st);
795 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
796 int ll_local_open(struct file *file,
797 struct lookup_intent *it, struct ll_file_data *fd,
798 struct obd_client_handle *och);
799 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
800 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
802 int ll_md_real_close(struct inode *inode, fmode_t fmode);
803 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
804 struct obd_client_handle **och, unsigned long flags);
805 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
806 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
807 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
808 __u64 ioepoch, int sync);
809 int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data,
810 struct md_open_data **mod);
811 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
812 struct lustre_handle *fh);
813 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
814 struct ll_file_data *file, loff_t pos,
815 size_t count, int rw);
816 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
817 struct lookup_intent *it, struct kstat *stat);
818 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
819 struct ll_file_data *ll_file_data_get(void);
820 struct posix_acl * ll_get_acl(struct inode *inode, int type);
822 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
823 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
825 # ifndef HAVE_INODE_PERMISION_2ARGS
826 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
828 int ll_inode_permission(struct inode *inode, int mask);
832 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
833 __u64 flags, struct lov_user_md *lum,
835 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
836 struct lov_mds_md **lmm, int *lmm_size,
837 struct ptlrpc_request **request);
838 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
840 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmmp,
841 int *lmm_size, struct ptlrpc_request **request);
842 #ifdef HAVE_FILE_FSYNC_4ARGS
843 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
844 #elif defined(HAVE_FILE_FSYNC_2ARGS)
845 int ll_fsync(struct file *file, int data);
847 int ll_fsync(struct file *file, struct dentry *dentry, int data);
849 int ll_merge_lvb(const struct lu_env *env, struct inode *inode);
850 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
851 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
852 int ll_fid2path(struct inode *inode, void *arg);
853 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
854 int ll_hsm_release(struct inode *inode);
856 struct obd_client_handle *ll_lease_open(struct inode *inode, struct file *file,
857 fmode_t mode, __u64 flags);
858 int ll_lease_close(struct obd_client_handle *och, struct inode *inode,
863 int ll_d_init(struct dentry *de);
864 extern struct dentry_operations ll_d_ops;
865 void ll_intent_drop_lock(struct lookup_intent *);
866 void ll_intent_release(struct lookup_intent *);
867 void ll_invalidate_aliases(struct inode *);
868 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
869 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
870 #ifdef HAVE_D_COMPARE_7ARGS
871 int ll_dcompare(const struct dentry *parent, const struct inode *pinode,
872 const struct dentry *dentry, const struct inode *inode,
873 unsigned int len, const char *str, const struct qstr *d_name);
875 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
877 int ll_revalidate_it_finish(struct ptlrpc_request *request,
878 struct lookup_intent *it, struct dentry *de);
880 /* llite/llite_lib.c */
881 extern struct super_operations lustre_super_operations;
883 char *ll_read_opt(const char *opt, char *data);
884 void ll_lli_init(struct ll_inode_info *lli);
885 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
886 void ll_put_super(struct super_block *sb);
887 void ll_kill_super(struct super_block *sb);
888 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
889 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
890 void ll_clear_inode(struct inode *inode);
891 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import);
892 int ll_setattr(struct dentry *de, struct iattr *attr);
893 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
894 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
895 __u64 max_age, __u32 flags);
896 void ll_update_inode(struct inode *inode, struct lustre_md *md);
897 void ll_read_inode2(struct inode *inode, void *opaque);
898 void ll_delete_inode(struct inode *inode);
899 int ll_iocontrol(struct inode *inode, struct file *file,
900 unsigned int cmd, unsigned long arg);
901 int ll_flush_ctx(struct inode *inode);
902 void ll_umount_begin(struct super_block *sb);
903 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
904 #ifdef HAVE_SUPEROPS_USE_DENTRY
905 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
907 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
909 void ll_dirty_page_discard_warn(struct page *page, int ioret);
910 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
911 struct super_block *, struct lookup_intent *);
912 void lustre_dump_dentry(struct dentry *, int recur);
913 void lustre_dump_inode(struct inode *);
914 int ll_obd_statfs(struct inode *inode, void *arg);
915 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
916 int ll_process_config(struct lustre_cfg *lcfg);
917 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
918 struct inode *i1, struct inode *i2,
919 const char *name, int namelen,
920 int mode, __u32 opc, void *data);
921 void ll_finish_md_op_data(struct md_op_data *op_data);
922 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
923 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
925 /* llite/llite_nfs.c */
926 extern struct export_operations lustre_export_operations;
927 __u32 get_uuid2int(const char *name, int len);
928 struct inode *search_inode_for_lustre(struct super_block *sb,
929 const struct lu_fid *fid);
931 /* llite/special.c */
932 extern struct inode_operations ll_special_inode_operations;
933 extern struct file_operations ll_special_chr_inode_fops;
934 extern struct file_operations ll_special_chr_file_fops;
935 extern struct file_operations ll_special_blk_inode_fops;
936 extern struct file_operations ll_special_fifo_inode_fops;
937 extern struct file_operations ll_special_fifo_file_fops;
938 extern struct file_operations ll_special_sock_inode_fops;
940 /* llite/symlink.c */
941 extern struct inode_operations ll_fast_symlink_inode_operations;
943 /* llite/llite_close.c */
944 struct ll_close_queue {
947 wait_queue_head_t lcq_waitq;
948 struct completion lcq_comp;
949 cfs_atomic_t lcq_stop;
952 struct ccc_object *cl_inode2ccc(struct inode *inode);
955 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
956 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
958 /* specific achitecture can implement only part of this list */
959 enum vvp_io_subtype {
962 /** io called from .sendfile */
964 /** io started from splice_{read|write} */
971 enum vvp_io_subtype cui_io_subtype;
975 read_actor_t cui_actor;
979 struct pipe_inode_info *cui_pipe;
980 unsigned int cui_flags;
982 struct vvp_fault_io {
984 * Inode modification time that is checked across DLM
988 struct vm_area_struct *ft_vma;
990 * locked page returned from vvp_io
992 struct page *ft_vmpage;
993 struct vm_fault_api {
997 struct vm_fault *ft_vmf;
999 * fault API used bitflags for return code.
1001 unsigned int ft_flags;
1006 * Read-ahead state used by read and page-fault IO contexts.
1008 struct ll_ra_read cui_bead;
1010 * Set when cui_bead has been initialized.
1012 int cui_ra_window_set;
1014 * Partially truncated page, that vvp_io_trunc_start() keeps locked
1017 struct cl_page *cui_partpage;
1021 * IO arguments for various VFS I/O interfaces.
1023 struct vvp_io_args {
1024 /** normal/sendfile/splice */
1025 enum vvp_io_subtype via_io_subtype;
1029 struct kiocb *via_iocb;
1030 struct iovec *via_iov;
1031 unsigned long via_nrsegs;
1034 read_actor_t via_actor;
1038 struct pipe_inode_info *via_pipe;
1039 unsigned int via_flags;
1044 struct ll_cl_context {
1046 struct cl_io *lcc_io;
1047 struct cl_page *lcc_page;
1048 struct lu_env *lcc_env;
1053 struct vvp_thread_info {
1054 struct ost_lvb vti_lvb;
1055 struct cl_2queue vti_queue;
1056 struct iovec vti_local_iov;
1057 struct vvp_io_args vti_args;
1058 struct ra_io_arg vti_ria;
1059 struct kiocb vti_kiocb;
1060 struct ll_cl_context vti_io_ctx;
1063 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
1065 extern struct lu_context_key vvp_key;
1066 struct vvp_thread_info *info;
1068 info = lu_context_key_get(&env->le_ctx, &vvp_key);
1069 LASSERT(info != NULL);
1073 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
1074 enum vvp_io_subtype type)
1076 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
1078 ret->via_io_subtype = type;
1083 struct vvp_session {
1084 struct vvp_io vs_ios;
1087 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1089 extern struct lu_context_key vvp_session_key;
1090 struct vvp_session *ses;
1092 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1093 LASSERT(ses != NULL);
1097 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1099 return &vvp_env_session(env)->vs_ios;
1102 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1103 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1104 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1106 /* llite/llite_mmap.c */
1107 typedef struct rb_root rb_root_t;
1108 typedef struct rb_node rb_node_t;
1110 struct ll_lock_tree_node;
1111 struct ll_lock_tree {
1113 cfs_list_t lt_locked_list;
1114 struct ll_file_data *lt_fd;
1117 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1118 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1119 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
1120 __u64 end, ldlm_mode_t mode);
1121 void policy_from_vma(ldlm_policy_data_t *policy,
1122 struct vm_area_struct *vma, unsigned long addr, size_t count);
1123 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1126 static inline void ll_invalidate_page(struct page *vmpage)
1128 struct address_space *mapping = vmpage->mapping;
1129 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1131 LASSERT(PageLocked(vmpage));
1132 if (mapping == NULL)
1135 ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE);
1136 truncate_complete_page(mapping, vmpage);
1139 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1141 /* don't need an addref as the sb_info should be holding one */
1142 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1144 return ll_s2sbi(sb)->ll_dt_exp;
1147 /* don't need an addref as the sb_info should be holding one */
1148 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1150 return ll_s2sbi(sb)->ll_md_exp;
1153 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1155 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1161 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1162 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1164 return ll_s2sbi(inode->i_sb);
1167 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1169 return ll_s2dtexp(inode->i_sb);
1172 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1174 return ll_s2mdexp(inode->i_sb);
1177 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1181 LASSERT(inode != NULL);
1182 fid = &ll_i2info(inode)->lli_fid;
1187 static inline int ll_mds_max_easize(struct super_block *sb)
1189 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1192 static inline __u64 ll_file_maxbytes(struct inode *inode)
1194 return ll_i2info(inode)->lli_maxbytes;
1198 int ll_setxattr(struct dentry *dentry, const char *name,
1199 const void *value, size_t size, int flags);
1200 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1201 void *buffer, size_t size);
1202 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1203 int ll_removexattr(struct dentry *dentry, const char *name);
1205 /* llite/remote_perm.c */
1206 extern struct kmem_cache *ll_remote_perm_cachep;
1207 extern struct kmem_cache *ll_rmtperm_hash_cachep;
1209 cfs_hlist_head_t *alloc_rmtperm_hash(void);
1210 void free_rmtperm_hash(cfs_hlist_head_t *hash);
1211 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1212 int lustre_check_remote_perm(struct inode *inode, int mask);
1214 /* llite/llite_capa.c */
1215 extern struct timer_list ll_capa_timer;
1217 int ll_capa_thread_start(void);
1218 void ll_capa_thread_stop(void);
1219 void ll_capa_timer_callback(unsigned long unused);
1221 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1222 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1224 void ll_capa_open(struct inode *inode);
1225 void ll_capa_close(struct inode *inode);
1227 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1228 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1230 void ll_truncate_free_capa(struct obd_capa *ocapa);
1231 void ll_clear_inode_capas(struct inode *inode);
1232 void ll_print_capa_stat(struct ll_sb_info *sbi);
1234 /* llite/llite_cl.c */
1235 extern struct lu_device_type vvp_device_type;
1238 * Common IO arguments for various VFS I/O interfaces.
1240 int cl_sb_init(struct super_block *sb);
1241 int cl_sb_fini(struct super_block *sb);
1242 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1243 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1245 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1246 struct ll_readahead_state *ras, unsigned long index,
1248 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1249 int ll_is_file_contended(struct file *file);
1250 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1252 /* llite/llite_rmtacl.c */
1253 #ifdef CONFIG_FS_POSIX_ACL
1254 obd_valid rce_ops2valid(int ops);
1255 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1256 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1257 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1258 void rct_init(struct rmtacl_ctl_table *rct);
1259 void rct_fini(struct rmtacl_ctl_table *rct);
1261 void ee_free(struct eacl_entry *ee);
1262 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1263 ext_acl_xattr_header *header);
1264 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1265 struct lu_fid *fid, int type);
1266 void et_search_free(struct eacl_table *et, pid_t key);
1267 void et_init(struct eacl_table *et);
1268 void et_fini(struct eacl_table *et);
1273 #define LL_SA_RPC_MIN 2
1274 #define LL_SA_RPC_DEF 32
1275 #define LL_SA_RPC_MAX 8192
1277 #define LL_SA_CACHE_BIT 5
1278 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1279 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1281 /* per inode struct, for dir only */
1282 struct ll_statahead_info {
1283 struct inode *sai_inode;
1284 cfs_atomic_t sai_refcount; /* when access this struct, hold
1286 unsigned int sai_generation; /* generation for statahead */
1287 unsigned int sai_max; /* max ahead of lookup */
1288 __u64 sai_sent; /* stat requests sent count */
1289 __u64 sai_replied; /* stat requests which received
1291 __u64 sai_index; /* index of statahead entry */
1292 __u64 sai_index_wait; /* index of entry which is the
1293 * caller is waiting for */
1294 __u64 sai_hit; /* hit count */
1295 __u64 sai_miss; /* miss count:
1296 * for "ls -al" case, it includes
1297 * hidden dentry miss;
1298 * for "ls -l" case, it does not
1299 * include hidden dentry miss.
1300 * "sai_miss_hidden" is used for
1303 unsigned int sai_consecutive_miss; /* consecutive miss */
1304 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1305 * is not a hidden one */
1306 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1307 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1309 sai_in_readpage:1,/* statahead is in readdir()*/
1310 sai_agl_valid:1;/* AGL is valid for the dir */
1311 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1312 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1313 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1314 cfs_list_t sai_entries; /* entry list */
1315 cfs_list_t sai_entries_received; /* entries returned */
1316 cfs_list_t sai_entries_stated; /* entries stated */
1317 cfs_list_t sai_entries_agl; /* AGL entries to be sent */
1318 cfs_list_t sai_cache[LL_SA_CACHE_SIZE];
1319 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1320 cfs_atomic_t sai_cache_count; /* entry count in cache */
1323 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1325 void ll_stop_statahead(struct inode *dir, void *key);
1327 static inline int ll_glimpse_size(struct inode *inode)
1329 struct ll_inode_info *lli = ll_i2info(inode);
1332 down_read(&lli->lli_glimpse_sem);
1333 rc = cl_glimpse_size(inode);
1334 lli->lli_glimpse_time = cfs_time_current();
1335 up_read(&lli->lli_glimpse_sem);
1340 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1342 struct ll_inode_info *lli = ll_i2info(dir);
1343 struct ll_statahead_info *sai = lli->lli_sai;
1344 struct ll_dentry_data *ldd = ll_d2d(dentry);
1346 /* not the same process, don't mark */
1347 if (lli->lli_opendir_pid != current_pid())
1350 LASSERT(ldd != NULL);
1352 ldd->lld_sa_generation = sai->sai_generation;
1356 ll_need_statahead(struct inode *dir, struct dentry *dentryp)
1358 struct ll_inode_info *lli;
1359 struct ll_dentry_data *ldd;
1361 if (ll_i2sbi(dir)->ll_sa_max == 0)
1364 lli = ll_i2info(dir);
1365 /* not the same process, don't statahead */
1366 if (lli->lli_opendir_pid != current_pid())
1369 /* statahead has been stopped */
1370 if (lli->lli_opendir_key == NULL)
1373 ldd = ll_d2d(dentryp);
1375 * When stats a dentry, the system trigger more than once "revalidate"
1376 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1377 * Under patchless client mode, the operation intent is not accurate,
1378 * which maybe misguide the statahead thread. For example:
1379 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1380 * have the same operation intent -- "IT_GETATTR".
1381 * In fact, one dentry should has only one chance to interact with the
1382 * statahead thread, otherwise the statahead windows will be confused.
1383 * The solution is as following:
1384 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1385 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1386 * will bypass interacting with statahead thread for checking:
1387 * "lld_sa_generation == lli_sai->sai_generation"
1389 if (ldd && lli->lli_sai &&
1390 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1397 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1401 ret = ll_need_statahead(dir, *dentryp);
1405 return do_statahead_enter(dir, dentryp, only_unplug);
1408 /* llite ioctl register support rountine */
1415 #define LLIOC_MAX_CMD 256
1418 * Rules to write a callback function:
1421 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1422 * returned to ll_iocontrol_register. Callback functions should use this
1423 * data to check the potential collasion of ioctl cmd. If collasion is
1424 * found, callback function should return LLIOC_CONT.
1425 * @rcp: The result of ioctl command.
1428 * If @magic matches the pointer returned by ll_iocontrol_data, the
1429 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1431 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1432 struct file *file, unsigned int cmd, unsigned long arg,
1433 void *magic, int *rcp);
1435 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1436 unsigned int cmd, unsigned long arg, int *rcp);
1438 /* export functions */
1439 /* Register ioctl block dynamatically for a regular file.
1441 * @cmd: the array of ioctl command set
1442 * @count: number of commands in the @cmd
1443 * @cb: callback function, it will be called if an ioctl command is found to
1444 * belong to the command list @cmd.
1447 * A magic pointer will be returned if success;
1448 * otherwise, NULL will be returned.
1450 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1451 void ll_iocontrol_unregister(void *magic);
1455 /* lclient compat stuff */
1456 #define cl_inode_info ll_inode_info
1457 #define cl_i2info(info) ll_i2info(info)
1458 #define cl_inode_mode(inode) ((inode)->i_mode)
1459 #define cl_i2sbi ll_i2sbi
1461 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1462 const struct iattr *attr)
1464 LASSERT(attr->ia_valid & ATTR_FILE);
1465 return LUSTRE_FPRIVATE(attr->ia_file);
1468 static inline void cl_isize_lock(struct inode *inode)
1470 ll_inode_size_lock(inode);
1473 static inline void cl_isize_unlock(struct inode *inode)
1475 ll_inode_size_unlock(inode);
1478 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1480 LASSERT(mutex_is_locked(&ll_i2info(inode)->lli_size_mutex));
1481 i_size_write(inode, kms);
1484 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1486 ll_inode_size_lock(inode);
1487 i_size_write(inode, kms);
1488 ll_inode_size_unlock(inode);
1491 #define cl_isize_read(inode) i_size_read(inode)
1493 static inline int cl_merge_lvb(const struct lu_env *env, struct inode *inode)
1495 return ll_merge_lvb(env, inode);
1498 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1499 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1500 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1502 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1504 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1505 enum cl_fsync_mode mode, int ignore_layout);
1507 /** direct write pages */
1508 struct ll_dio_pages {
1509 /** page array to be written. we don't support
1510 * partial pages except the last one. */
1511 struct page **ldp_pages;
1512 /* offset of each page */
1513 loff_t *ldp_offsets;
1514 /** if ldp_offsets is NULL, it means a sequential
1515 * pages to be written, then this is the file offset
1516 * of the * first page. */
1517 loff_t ldp_start_offset;
1518 /** how many bytes are to be written. */
1520 /** # of pages in the array. */
1524 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1527 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1530 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1533 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1534 int rw, struct inode *inode,
1535 struct ll_dio_pages *pv);
1537 static inline int ll_file_nolock(const struct file *file)
1539 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1540 struct inode *inode = file->f_dentry->d_inode;
1542 LASSERT(fd != NULL);
1543 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1544 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1547 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1548 struct lookup_intent *it, __u64 *bits)
1550 if (!it->d.lustre.it_lock_set) {
1551 struct lustre_handle handle;
1553 /* If this inode is a remote object, it will get two
1554 * separate locks in different namespaces, Master MDT,
1555 * where the name entry is, will grant LOOKUP lock,
1556 * remote MDT, where the object is, will grant
1557 * UPDATE|PERM lock. The inode will be attched to both
1558 * LOOKUP and PERM locks, so revoking either locks will
1559 * case the dcache being cleared */
1560 if (it->d.lustre.it_remote_lock_mode) {
1561 handle.cookie = it->d.lustre.it_remote_lock_handle;
1562 CDEBUG(D_DLMTRACE, "setting l_data to inode %p"
1563 "(%lu/%u) for remote lock "LPX64"\n", inode,
1564 inode->i_ino, inode->i_generation,
1566 md_set_lock_data(exp, &handle.cookie, inode, NULL);
1569 handle.cookie = it->d.lustre.it_lock_handle;
1571 CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)"
1572 " for lock "LPX64"\n", inode, inode->i_ino,
1573 inode->i_generation, handle.cookie);
1575 md_set_lock_data(exp, &handle.cookie, inode,
1576 &it->d.lustre.it_lock_bits);
1577 it->d.lustre.it_lock_set = 1;
1581 *bits = it->d.lustre.it_lock_bits;
1584 static inline void ll_lock_dcache(struct inode *inode)
1586 #ifdef HAVE_DCACHE_LOCK
1587 spin_lock(&dcache_lock);
1589 spin_lock(&inode->i_lock);
1593 static inline void ll_unlock_dcache(struct inode *inode)
1595 #ifdef HAVE_DCACHE_LOCK
1596 spin_unlock(&dcache_lock);
1598 spin_unlock(&inode->i_lock);
1602 static inline int d_lustre_invalid(const struct dentry *dentry)
1604 struct ll_dentry_data *lld = ll_d2d(dentry);
1606 return (lld == NULL) || lld->lld_invalid;
1609 static inline void __d_lustre_invalidate(struct dentry *dentry)
1611 struct ll_dentry_data *lld = ll_d2d(dentry);
1614 lld->lld_invalid = 1;
1618 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1619 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1620 * else dput() of the last refcount will unhash this dentry and kill it.
1622 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1624 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1625 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1626 dentry->d_parent, dentry->d_inode, d_refcount(dentry));
1628 spin_lock_nested(&dentry->d_lock,
1629 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1630 __d_lustre_invalidate(dentry);
1631 if (d_refcount(dentry) == 0)
1633 spin_unlock(&dentry->d_lock);
1636 static inline void d_lustre_revalidate(struct dentry *dentry)
1638 spin_lock(&dentry->d_lock);
1639 LASSERT(ll_d2d(dentry) != NULL);
1640 ll_d2d(dentry)->lld_invalid = 0;
1641 spin_unlock(&dentry->d_lock);
1644 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0)
1645 /* Compatibility for old (1.8) compiled userspace quota code */
1646 struct if_quotactl_18 {
1651 struct obd_dqinfo qc_dqinfo;
1652 struct obd_dqblk qc_dqblk;
1654 struct obd_uuid obd_uuid;
1656 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1657 /* End compatibility for old (1.8) compiled userspace quota code */
1659 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1660 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */
1663 LL_LAYOUT_GEN_NONE = ((__u32)-2), /* layout lock was cancelled */
1664 LL_LAYOUT_GEN_EMPTY = ((__u32)-1) /* for empty layout */
1667 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1668 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1669 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1671 int ll_xattr_init(void);
1672 void ll_xattr_fini(void);
1674 #endif /* LLITE_INTERNAL_H */