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>
54 #ifndef VM_FAULT_RETRY
55 #define VM_FAULT_RETRY 0
58 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
59 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
60 #ifndef LOOKUP_CONTINUE
61 #define LOOKUP_CONTINUE LOOKUP_PARENT
64 /** Only used on client-side for indicating the tail of dir hash/offset. */
65 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
66 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
68 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
69 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
71 struct ll_dentry_data {
74 struct obd_client_handle lld_cwd_och;
75 struct obd_client_handle lld_mnt_och;
76 struct lookup_intent *lld_it;
77 unsigned int lld_sa_generation;
78 unsigned int lld_invalid:1;
79 struct rcu_head lld_rcu_head;
82 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
84 extern struct file_operations ll_pgcache_seq_fops;
86 #define LLI_INODE_MAGIC 0x111d0de5
87 #define LLI_INODE_DEAD 0xdeadd00d
89 /* remote client permission cache */
90 #define REMOTE_PERM_HASHSIZE 16
92 struct ll_getname_data {
93 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
94 struct lu_fid lgd_fid; /* target fid we are looking for */
95 int lgd_found; /* inode matched? */
98 /* llite setxid/access permission for user on remote client */
99 struct ll_remote_perm {
100 cfs_hlist_node_t lrp_list;
105 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
106 is access permission with
107 lrp_fsuid/lrp_fsgid. */
111 /* MDS has an authority for the Size-on-MDS attributes. */
112 LLIF_MDS_SIZE_LOCK = (1 << 0),
113 /* Epoch close is postponed. */
114 LLIF_EPOCH_PENDING = (1 << 1),
115 /* DONE WRITING is allowed. */
116 LLIF_DONE_WRITING = (1 << 2),
117 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
119 LLIF_SOM_DIRTY = (1 << 3),
120 /* File is contented */
121 LLIF_CONTENDED = (1 << 4),
122 /* Truncate uses server lock for this file */
123 LLIF_SRVLOCK = (1 << 5),
124 /* File data is modified. */
125 LLIF_DATA_MODIFIED = (1 << 6),
128 struct ll_inode_info {
129 __u32 lli_inode_magic;
134 struct posix_acl *lli_posix_acl;
136 cfs_hlist_head_t *lli_remote_perms;
137 struct mutex lli_rmtperm_mutex;
139 /* identifying fields for both metadata and data stacks. */
140 struct lu_fid lli_fid;
141 /* Parent fid for accessing default stripe data on parent directory
142 * for allocating OST objects after a mknod() and later open-by-FID. */
143 struct lu_fid lli_pfid;
145 cfs_list_t lli_close_list;
146 cfs_list_t lli_oss_capas;
147 /* open count currently used by capability only, indicate whether
148 * capability needs renewal */
149 cfs_atomic_t lli_open_count;
150 struct obd_capa *lli_mds_capa;
151 cfs_time_t lli_rmtperm_time;
153 /* handle is to be sent to MDS later on done_writing and setattr.
154 * Open handle data are needed for the recovery to reconstruct
155 * the inode state on the MDS. XXX: recovery is not ready yet. */
156 struct obd_client_handle *lli_pending_och;
158 /* We need all three because every inode may be opened in different
160 struct obd_client_handle *lli_mds_read_och;
161 struct obd_client_handle *lli_mds_write_och;
162 struct obd_client_handle *lli_mds_exec_och;
163 __u64 lli_open_fd_read_count;
164 __u64 lli_open_fd_write_count;
165 __u64 lli_open_fd_exec_count;
166 /* Protects access to och pointers and their usage counters */
167 struct mutex lli_och_mutex;
169 struct inode lli_vfs_inode;
171 /* the most recent timestamps obtained from mds */
172 struct ost_lvb lli_lvb;
173 spinlock_t lli_agl_lock;
175 /* Try to make the d::member and f::member are aligned. Before using
176 * these members, make clear whether it is directory or not. */
180 /* serialize normal readdir and statahead-readdir. */
181 struct mutex d_readdir_mutex;
183 /* metadata statahead */
184 /* since parent-child threads can share the same @file
185 * struct, "opendir_key" is the token when dir close for
186 * case of parent exit before child -- it is me should
187 * cleanup the dir readahead. */
189 struct ll_statahead_info *d_sai;
190 struct posix_acl *d_def_acl;
191 /* protect statahead stuff. */
192 spinlock_t d_sa_lock;
193 /* "opendir_pid" is the token when lookup/revalid
194 * -- I am the owner of dir statahead. */
198 #define lli_readdir_mutex u.d.d_readdir_mutex
199 #define lli_opendir_key u.d.d_opendir_key
200 #define lli_sai u.d.d_sai
201 #define lli_def_acl u.d.d_def_acl
202 #define lli_sa_lock u.d.d_sa_lock
203 #define lli_opendir_pid u.d.d_opendir_pid
205 /* for non-directory */
207 struct semaphore f_size_sem;
208 void *f_size_sem_owner;
209 char *f_symlink_name;
212 * struct rw_semaphore {
213 * signed long count; // align d.d_def_acl
214 * spinlock_t wait_lock; // align d.d_sa_lock
215 * struct list_head wait_list;
218 struct rw_semaphore f_trunc_sem;
219 struct mutex f_write_mutex;
221 struct rw_semaphore f_glimpse_sem;
222 cfs_time_t f_glimpse_time;
223 cfs_list_t f_agl_list;
226 /* for writepage() only to communicate to fsync */
229 /* volatile file criteria is based on file name, this
230 * flag is used to keep the test result, so the strcmp
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_sem u.f.f_size_sem
246 #define lli_size_sem_owner u.f.f_size_sem_owner
247 #define lli_symlink_name u.f.f_symlink_name
248 #define lli_maxbytes u.f.f_maxbytes
249 #define lli_trunc_sem u.f.f_trunc_sem
250 #define lli_write_mutex u.f.f_write_mutex
251 #define lli_glimpse_sem u.f.f_glimpse_sem
252 #define lli_glimpse_time u.f.f_glimpse_time
253 #define lli_agl_list u.f.f_agl_list
254 #define lli_agl_index u.f.f_agl_index
255 #define lli_async_rc u.f.f_async_rc
256 #define lli_jobid u.f.f_jobid
257 #define lli_volatile u.f.f_volatile
261 /* XXX: For following frequent used members, although they maybe special
262 * used for non-directory object, it is some time-wasting to check
263 * whether the object is directory or not before using them. On the
264 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
265 * the "ll_inode_info" size even if moving those members into u.f.
266 * So keep them out side.
268 * In the future, if more members are added only for directory,
269 * some of the following members can be moved into u.f.
272 struct cl_object *lli_clob;
274 /* mutex to request for layout lock exclusively. */
275 struct mutex lli_layout_mutex;
276 /* valid only inside LAYOUT ibits lock, protected by lli_layout_mutex */
277 __u32 lli_layout_gen;
281 * Locking to guarantee consistency of non-atomic updates to long long i_size,
282 * consistency between file size and KMS.
284 * Implemented by ->lli_size_sem and ->lsm_lock, nested in that order.
287 void ll_inode_size_lock(struct inode *inode);
288 void ll_inode_size_unlock(struct inode *inode);
290 // FIXME: replace the name of this with LL_I to conform to kernel stuff
291 // static inline struct ll_inode_info *LL_I(struct inode *inode)
292 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
294 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
297 /* default to about 40meg of readahead on a given system. That much tied
298 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
299 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT))
301 /* default to read-ahead full files smaller than 2MB on the second read */
302 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
307 RA_STAT_DISTANT_READPAGE,
308 RA_STAT_MISS_IN_WINDOW,
309 RA_STAT_FAILED_GRAB_PAGE,
310 RA_STAT_FAILED_MATCH,
315 RA_STAT_MAX_IN_FLIGHT,
316 RA_STAT_WRONG_GRAB_PAGE,
321 cfs_atomic_t ra_cur_pages;
322 unsigned long ra_max_pages;
323 unsigned long ra_max_pages_per_file;
324 unsigned long ra_max_read_ahead_whole_pages;
327 /* ra_io_arg will be filled in the beginning of ll_readahead with
328 * ras_lock, then the following ll_read_ahead_pages will read RA
329 * pages according to this arg, all the items in this structure are
330 * counted by page index.
333 unsigned long ria_start; /* start offset of read-ahead*/
334 unsigned long ria_end; /* end offset of read-ahead*/
335 /* If stride read pattern is detected, ria_stoff means where
336 * stride read is started. Note: for normal read-ahead, the
337 * value here is meaningless, and also it will not be accessed*/
339 /* ria_length and ria_pages are the length and pages length in the
340 * stride I/O mode. And they will also be used to check whether
341 * it is stride I/O read-ahead in the read-ahead pages*/
342 unsigned long ria_length;
343 unsigned long ria_pages;
346 /* LL_HIST_MAX=32 causes an overflow */
347 #define LL_HIST_MAX 28
348 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
349 #define LL_PROCESS_HIST_MAX 10
350 struct per_process_info {
352 struct obd_histogram pp_r_hist;
353 struct obd_histogram pp_w_hist;
356 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
357 struct ll_rw_extents_info {
358 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
361 #define LL_OFFSET_HIST_MAX 100
362 struct ll_rw_process_info {
365 loff_t rw_range_start;
367 loff_t rw_last_file_pos;
369 size_t rw_smallest_extent;
370 size_t rw_largest_extent;
371 struct ll_file_data *rw_last_file;
374 enum stats_track_type {
375 STATS_TRACK_ALL = 0, /* track all processes */
376 STATS_TRACK_PID, /* track process with this pid */
377 STATS_TRACK_PPID, /* track processes with this ppid */
378 STATS_TRACK_GID, /* track processes with this gid */
382 /* flags for sbi->ll_flags */
383 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
384 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
385 #define LL_SBI_FLOCK 0x04
386 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
387 #define LL_SBI_ACL 0x10 /* support ACL */
388 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
389 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
390 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
391 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
392 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
393 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
394 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
395 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
396 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
397 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
398 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
399 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
400 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
402 #define LL_SBI_FLAGS { \
422 /* default value for ll_sb_info->contention_time */
423 #define SBI_DEFAULT_CONTENTION_SECONDS 60
424 /* default value for lockless_truncate_enable */
425 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
426 #define RCE_HASHES 32
428 struct rmtacl_ctl_entry {
430 pid_t rce_key; /* hash key */
431 int rce_ops; /* acl operation type */
434 struct rmtacl_ctl_table {
436 cfs_list_t rct_entries[RCE_HASHES];
443 pid_t ee_key; /* hash key */
444 struct lu_fid ee_fid;
445 int ee_type; /* ACL type for ACCESS or DEFAULT */
446 ext_acl_xattr_header *ee_acl;
451 cfs_list_t et_entries[EE_HASHES];
456 /* this protects pglist and ra_info. It isn't safe to
457 * grab from interrupt contexts */
459 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
460 spinlock_t ll_process_lock; /* ll_rw_process_info */
461 struct obd_uuid ll_sb_uuid;
462 struct obd_export *ll_md_exp;
463 struct obd_export *ll_dt_exp;
464 struct proc_dir_entry* ll_proc_root;
465 struct lu_fid ll_root_fid; /* root object fid */
469 cfs_list_t ll_conn_chain; /* per-conn chain of SBs */
470 struct lustre_client_ocd ll_lco;
472 cfs_list_t ll_orphan_dentry_list; /*please don't ask -p*/
473 struct ll_close_queue *ll_lcq;
475 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
477 struct cl_client_cache ll_cache;
479 struct lprocfs_stats *ll_ra_stats;
481 struct ll_ra_info ll_ra_info;
482 unsigned int ll_namelen;
483 struct file_operations *ll_fop;
485 /* =0 - hold lock over whole read/write
486 * >0 - max. chunk to be read/written w/o lock re-acquiring */
487 unsigned long ll_max_rw_chunk;
488 unsigned int ll_md_brw_size; /* used by readdir */
490 struct lu_site *ll_site;
491 struct cl_device *ll_cl;
493 struct ll_rw_extents_info ll_rw_extents_info;
494 int ll_extent_process_count;
495 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
496 unsigned int ll_offset_process_count;
497 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
498 unsigned int ll_rw_offset_entry_count;
499 int ll_stats_track_id;
500 enum stats_track_type ll_stats_track_type;
503 /* metadata stat-ahead */
504 unsigned int ll_sa_max; /* max statahead RPCs */
505 atomic_t ll_sa_total; /* statahead thread started
507 atomic_t ll_sa_wrong; /* statahead thread stopped for
509 atomic_t ll_agl_total; /* AGL thread started count */
511 dev_t ll_sdev_orig; /* save s_dev before assign for
513 struct rmtacl_ctl_table ll_rct;
514 struct eacl_table ll_et;
515 __kernel_fsid_t ll_fsid;
518 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
523 struct task_struct *lrr_reader;
524 cfs_list_t lrr_linkage;
528 * per file-descriptor read-ahead data.
530 struct ll_readahead_state {
533 * index of the last page that read(2) needed and that wasn't in the
534 * cache. Used by ras_update() to detect seeks.
536 * XXX nikita: if access seeks into cached region, Lustre doesn't see
539 unsigned long ras_last_readpage;
541 * number of pages read after last read-ahead window reset. As window
542 * is reset on each seek, this is effectively a number of consecutive
543 * accesses. Maybe ->ras_accessed_in_window is better name.
545 * XXX nikita: window is also reset (by ras_update()) when Lustre
546 * believes that memory pressure evicts read-ahead pages. In that
547 * case, it probably doesn't make sense to expand window to
548 * PTLRPC_MAX_BRW_PAGES on the third access.
550 unsigned long ras_consecutive_pages;
552 * number of read requests after the last read-ahead window reset
553 * As window is reset on each seek, this is effectively the number
554 * on consecutive read request and is used to trigger read-ahead.
556 unsigned long ras_consecutive_requests;
558 * Parameters of current read-ahead window. Handled by
559 * ras_update(). On the initial access to the file or after a seek,
560 * window is reset to 0. After 3 consecutive accesses, window is
561 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
562 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
564 unsigned long ras_window_start, ras_window_len;
566 * Where next read-ahead should start at. This lies within read-ahead
567 * window. Read-ahead window is read in pieces rather than at once
568 * because: 1. lustre limits total number of pages under read-ahead by
569 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
570 * not covered by DLM lock.
572 unsigned long ras_next_readahead;
574 * Total number of ll_file_read requests issued, reads originating
575 * due to mmap are not counted in this total. This value is used to
576 * trigger full file read-ahead after multiple reads to a small file.
578 unsigned long ras_requests;
580 * Page index with respect to the current request, these value
581 * will not be accurate when dealing with reads issued via mmap.
583 unsigned long ras_request_index;
585 * list of struct ll_ra_read's one per read(2) call current in
586 * progress against this file descriptor. Used by read-ahead code,
587 * protected by ->ras_lock.
589 cfs_list_t ras_read_beads;
591 * The following 3 items are used for detecting the stride I/O
593 * In stride I/O mode,
594 * ...............|-----data-----|****gap*****|--------|******|....
595 * offset |-stride_pages-|-stride_gap-|
596 * ras_stride_offset = offset;
597 * ras_stride_length = stride_pages + stride_gap;
598 * ras_stride_pages = stride_pages;
599 * Note: all these three items are counted by pages.
601 unsigned long ras_stride_length;
602 unsigned long ras_stride_pages;
603 pgoff_t ras_stride_offset;
605 * number of consecutive stride request count, and it is similar as
606 * ras_consecutive_requests, but used for stride I/O mode.
607 * Note: only more than 2 consecutive stride request are detected,
608 * stride read-ahead will be enable
610 unsigned long ras_consecutive_stride_requests;
613 extern struct kmem_cache *ll_file_data_slab;
614 struct lustre_handle;
615 struct ll_file_data {
616 struct ll_readahead_state fd_ras;
618 struct ccc_grouplock fd_grouplock;
621 struct file *fd_file;
622 /* Indicate whether need to report failure when close.
623 * true: failure is known, not report again.
624 * false: unknown failure, should report. */
625 bool fd_write_failed;
628 struct lov_stripe_md;
630 extern spinlock_t inode_lock;
632 extern struct proc_dir_entry *proc_lustre_fs_root;
634 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
636 return &lli->lli_vfs_inode;
640 struct inode *icbd_parent;
641 struct dentry **icbd_childp;
645 __u32 ll_i2suppgid(struct inode *i);
646 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
648 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
650 #if BITS_PER_LONG == 32
653 return unlikely(cfs_curproc_is_32bit() || (sbi->ll_flags & LL_SBI_32BIT_API));
657 #define LLAP_MAGIC 98764321
659 extern struct kmem_cache *ll_async_page_slab;
660 extern size_t ll_async_page_slab_size;
662 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
663 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
664 struct ll_ra_read *ll_ra_read_get(struct file *f);
666 /* llite/lproc_llite.c */
668 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
669 struct super_block *sb, char *osc, char *mdc);
670 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
671 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
672 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
674 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
675 struct super_block *sb, char *osc, char *mdc){return 0;}
676 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
677 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
678 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
680 memset(lvars, 0, sizeof(*lvars));
686 void ll_release_page(struct page *page, int remove);
687 extern struct file_operations ll_dir_operations;
688 extern struct inode_operations ll_dir_inode_operations;
689 struct page *ll_get_dir_page(struct inode *dir, __u64 hash,
690 struct ll_dir_chain *chain);
691 int ll_dir_read(struct inode *inode, __u64 *_pos, void *cookie,
694 int ll_get_mdt_idx(struct inode *inode);
696 int ll_objects_destroy(struct ptlrpc_request *request,
698 struct inode *ll_iget(struct super_block *sb, ino_t hash,
699 struct lustre_md *lic);
700 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
701 void *data, int flag);
702 #ifndef HAVE_IOP_ATOMIC_OPEN
703 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
706 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
707 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
710 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
711 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
712 int ll_writepage(struct page *page, struct writeback_control *wbc);
713 int ll_writepages(struct address_space *, struct writeback_control *wbc);
714 void ll_removepage(struct page *page);
715 int ll_readpage(struct file *file, struct page *page);
716 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
717 int ll_file_punch(struct inode *, loff_t, int);
718 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
719 void ll_clear_file_contended(struct inode*);
720 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
721 int ll_readahead(const struct lu_env *env, struct cl_io *io,
722 struct ll_readahead_state *ras, struct address_space *mapping,
723 struct cl_page_list *queue, int flags);
726 extern struct file_operations ll_file_operations;
727 extern struct file_operations ll_file_operations_flock;
728 extern struct file_operations ll_file_operations_noflock;
729 extern struct inode_operations ll_file_inode_operations;
730 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
732 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
733 ldlm_mode_t l_req_mode);
734 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
735 struct lustre_handle *lockh, __u64 flags);
736 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
738 #ifdef HAVE_IOP_ATOMIC_OPEN
739 int ll_revalidate_nd(struct dentry *dentry, unsigned int flags);
741 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
743 int ll_file_open(struct inode *inode, struct file *file);
744 int ll_file_release(struct inode *inode, struct file *file);
745 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
746 struct lov_stripe_md *lsm, lstat_t *st);
747 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
748 int ll_local_open(struct file *file,
749 struct lookup_intent *it, struct ll_file_data *fd,
750 struct obd_client_handle *och);
751 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
752 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
754 int ll_md_real_close(struct inode *inode, int flags);
755 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
756 struct obd_client_handle **och, unsigned long flags);
757 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
758 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
759 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
760 __u64 ioepoch, int sync);
761 int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data,
762 struct md_open_data **mod);
763 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
764 struct lustre_handle *fh);
765 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
766 struct ll_file_data *file, loff_t pos,
767 size_t count, int rw);
768 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
769 struct lookup_intent *it, struct kstat *stat);
770 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
771 struct ll_file_data *ll_file_data_get(void);
772 struct posix_acl * ll_get_acl(struct inode *inode, int type);
774 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
775 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
777 # ifndef HAVE_INODE_PERMISION_2ARGS
778 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
780 int ll_inode_permission(struct inode *inode, int mask);
784 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
785 int flags, struct lov_user_md *lum,
787 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
788 struct lov_mds_md **lmm, int *lmm_size,
789 struct ptlrpc_request **request);
790 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
792 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmmp,
793 int *lmm_size, struct ptlrpc_request **request);
794 #ifdef HAVE_FILE_FSYNC_4ARGS
795 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
796 #elif defined(HAVE_FILE_FSYNC_2ARGS)
797 int ll_fsync(struct file *file, int data);
799 int ll_fsync(struct file *file, struct dentry *dentry, int data);
801 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
803 int ll_merge_lvb(const struct lu_env *env, struct inode *inode);
804 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
805 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
806 int ll_fid2path(struct inode *inode, void *arg);
807 int ll_data_version(struct inode *inode, __u64 *data_version, int extent_lock);
811 int ll_dops_init(struct dentry *de, int block, int init_sa);
812 extern struct dentry_operations ll_d_ops;
813 void ll_intent_drop_lock(struct lookup_intent *);
814 void ll_intent_release(struct lookup_intent *);
815 void ll_invalidate_aliases(struct inode *);
816 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
817 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
818 #ifdef HAVE_D_COMPARE_7ARGS
819 int ll_dcompare(const struct dentry *parent, const struct inode *pinode,
820 const struct dentry *dentry, const struct inode *inode,
821 unsigned int len, const char *str, const struct qstr *d_name);
823 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
825 int ll_revalidate_it_finish(struct ptlrpc_request *request,
826 struct lookup_intent *it, struct dentry *de);
828 /* llite/llite_lib.c */
829 extern struct super_operations lustre_super_operations;
831 char *ll_read_opt(const char *opt, char *data);
832 void ll_lli_init(struct ll_inode_info *lli);
833 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
834 void ll_put_super(struct super_block *sb);
835 void ll_kill_super(struct super_block *sb);
836 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
837 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
838 void ll_clear_inode(struct inode *inode);
839 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr);
840 int ll_setattr(struct dentry *de, struct iattr *attr);
841 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
842 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
843 __u64 max_age, __u32 flags);
844 void ll_update_inode(struct inode *inode, struct lustre_md *md);
845 void ll_read_inode2(struct inode *inode, void *opaque);
846 void ll_delete_inode(struct inode *inode);
847 int ll_iocontrol(struct inode *inode, struct file *file,
848 unsigned int cmd, unsigned long arg);
849 int ll_flush_ctx(struct inode *inode);
850 void ll_umount_begin(struct super_block *sb);
851 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
852 #ifdef HAVE_SUPEROPS_USE_DENTRY
853 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
855 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
857 void ll_dirty_page_discard_warn(struct page *page, int ioret);
858 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
859 struct super_block *, struct lookup_intent *);
860 void lustre_dump_dentry(struct dentry *, int recur);
861 void lustre_dump_inode(struct inode *);
862 int ll_obd_statfs(struct inode *inode, void *arg);
863 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
864 int ll_process_config(struct lustre_cfg *lcfg);
865 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
866 struct inode *i1, struct inode *i2,
867 const char *name, int namelen,
868 int mode, __u32 opc, void *data);
869 void ll_finish_md_op_data(struct md_op_data *op_data);
870 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
871 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
873 /* llite/llite_nfs.c */
874 extern struct export_operations lustre_export_operations;
875 __u32 get_uuid2int(const char *name, int len);
876 void get_uuid2fsid(const char *name, int len, __kernel_fsid_t *fsid);
877 struct inode *search_inode_for_lustre(struct super_block *sb,
878 const struct lu_fid *fid);
880 /* llite/special.c */
881 extern struct inode_operations ll_special_inode_operations;
882 extern struct file_operations ll_special_chr_inode_fops;
883 extern struct file_operations ll_special_chr_file_fops;
884 extern struct file_operations ll_special_blk_inode_fops;
885 extern struct file_operations ll_special_fifo_inode_fops;
886 extern struct file_operations ll_special_fifo_file_fops;
887 extern struct file_operations ll_special_sock_inode_fops;
889 /* llite/symlink.c */
890 extern struct inode_operations ll_fast_symlink_inode_operations;
892 /* llite/llite_close.c */
893 struct ll_close_queue {
896 cfs_waitq_t lcq_waitq;
897 struct completion lcq_comp;
898 cfs_atomic_t lcq_stop;
901 struct ccc_object *cl_inode2ccc(struct inode *inode);
904 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
905 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
907 /* specific achitecture can implement only part of this list */
908 enum vvp_io_subtype {
911 /** io called from .sendfile */
913 /** io started from splice_{read|write} */
920 enum vvp_io_subtype cui_io_subtype;
924 read_actor_t cui_actor;
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;
963 * Partially truncated page, that vvp_io_trunc_start() keeps locked
966 struct cl_page *cui_partpage;
970 * IO arguments for various VFS I/O interfaces.
973 /** normal/sendfile/splice */
974 enum vvp_io_subtype via_io_subtype;
978 #ifndef HAVE_FILE_WRITEV
979 struct kiocb *via_iocb;
981 struct iovec *via_iov;
982 unsigned long via_nrsegs;
985 read_actor_t via_actor;
989 struct pipe_inode_info *via_pipe;
990 unsigned int via_flags;
995 struct ll_cl_context {
997 struct cl_io *lcc_io;
998 struct cl_page *lcc_page;
999 struct lu_env *lcc_env;
1004 struct vvp_thread_info {
1005 struct ost_lvb vti_lvb;
1006 struct cl_2queue vti_queue;
1007 struct iovec vti_local_iov;
1008 struct vvp_io_args vti_args;
1009 struct ra_io_arg vti_ria;
1010 struct kiocb vti_kiocb;
1011 struct ll_cl_context vti_io_ctx;
1014 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
1016 extern struct lu_context_key vvp_key;
1017 struct vvp_thread_info *info;
1019 info = lu_context_key_get(&env->le_ctx, &vvp_key);
1020 LASSERT(info != NULL);
1024 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
1025 enum vvp_io_subtype type)
1027 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
1029 ret->via_io_subtype = type;
1034 struct vvp_session {
1035 struct vvp_io vs_ios;
1038 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1040 extern struct lu_context_key vvp_session_key;
1041 struct vvp_session *ses;
1043 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1044 LASSERT(ses != NULL);
1048 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1050 return &vvp_env_session(env)->vs_ios;
1053 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1054 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1055 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1057 /* llite/llite_mmap.c */
1058 typedef struct rb_root rb_root_t;
1059 typedef struct rb_node rb_node_t;
1061 struct ll_lock_tree_node;
1062 struct ll_lock_tree {
1064 cfs_list_t lt_locked_list;
1065 struct ll_file_data *lt_fd;
1068 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1069 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1070 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
1071 __u64 end, ldlm_mode_t mode);
1072 void policy_from_vma(ldlm_policy_data_t *policy,
1073 struct vm_area_struct *vma, unsigned long addr, size_t count);
1074 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1077 static inline void ll_invalidate_page(struct page *vmpage)
1079 struct address_space *mapping = vmpage->mapping;
1080 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1082 LASSERT(PageLocked(vmpage));
1083 if (mapping == NULL)
1086 ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE);
1087 truncate_complete_page(mapping, vmpage);
1090 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1092 /* don't need an addref as the sb_info should be holding one */
1093 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1095 return ll_s2sbi(sb)->ll_dt_exp;
1098 /* don't need an addref as the sb_info should be holding one */
1099 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1101 return ll_s2sbi(sb)->ll_md_exp;
1104 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1106 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1112 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1113 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1115 return ll_s2sbi(inode->i_sb);
1118 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1120 return ll_s2dtexp(inode->i_sb);
1123 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1125 return ll_s2mdexp(inode->i_sb);
1128 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1132 LASSERT(inode != NULL);
1133 fid = &ll_i2info(inode)->lli_fid;
1138 static inline int ll_mds_max_easize(struct super_block *sb)
1140 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1143 static inline __u64 ll_file_maxbytes(struct inode *inode)
1145 return ll_i2info(inode)->lli_maxbytes;
1149 int ll_setxattr(struct dentry *dentry, const char *name,
1150 const void *value, size_t size, int flags);
1151 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1152 void *buffer, size_t size);
1153 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1154 int ll_removexattr(struct dentry *dentry, const char *name);
1156 /* llite/remote_perm.c */
1157 extern struct kmem_cache *ll_remote_perm_cachep;
1158 extern struct kmem_cache *ll_rmtperm_hash_cachep;
1160 cfs_hlist_head_t *alloc_rmtperm_hash(void);
1161 void free_rmtperm_hash(cfs_hlist_head_t *hash);
1162 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1163 int lustre_check_remote_perm(struct inode *inode, int mask);
1165 /* llite/llite_capa.c */
1166 extern cfs_timer_t ll_capa_timer;
1168 int ll_capa_thread_start(void);
1169 void ll_capa_thread_stop(void);
1170 void ll_capa_timer_callback(unsigned long unused);
1172 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1173 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1175 void ll_capa_open(struct inode *inode);
1176 void ll_capa_close(struct inode *inode);
1178 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1179 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1181 void ll_truncate_free_capa(struct obd_capa *ocapa);
1182 void ll_clear_inode_capas(struct inode *inode);
1183 void ll_print_capa_stat(struct ll_sb_info *sbi);
1185 /* llite/llite_cl.c */
1186 extern struct lu_device_type vvp_device_type;
1189 * Common IO arguments for various VFS I/O interfaces.
1191 int cl_sb_init(struct super_block *sb);
1192 int cl_sb_fini(struct super_block *sb);
1193 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1194 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1196 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1197 struct ll_readahead_state *ras, unsigned long index,
1199 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1200 int ll_is_file_contended(struct file *file);
1201 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1203 /* llite/llite_rmtacl.c */
1204 #ifdef CONFIG_FS_POSIX_ACL
1205 obd_valid rce_ops2valid(int ops);
1206 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1207 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1208 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1209 void rct_init(struct rmtacl_ctl_table *rct);
1210 void rct_fini(struct rmtacl_ctl_table *rct);
1212 void ee_free(struct eacl_entry *ee);
1213 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1214 ext_acl_xattr_header *header);
1215 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1216 struct lu_fid *fid, int type);
1217 void et_search_free(struct eacl_table *et, pid_t key);
1218 void et_init(struct eacl_table *et);
1219 void et_fini(struct eacl_table *et);
1224 #define LL_SA_RPC_MIN 2
1225 #define LL_SA_RPC_DEF 32
1226 #define LL_SA_RPC_MAX 8192
1228 #define LL_SA_CACHE_BIT 5
1229 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1230 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1232 /* per inode struct, for dir only */
1233 struct ll_statahead_info {
1234 struct inode *sai_inode;
1235 cfs_atomic_t sai_refcount; /* when access this struct, hold
1237 unsigned int sai_generation; /* generation for statahead */
1238 unsigned int sai_max; /* max ahead of lookup */
1239 __u64 sai_sent; /* stat requests sent count */
1240 __u64 sai_replied; /* stat requests which received
1242 __u64 sai_index; /* index of statahead entry */
1243 __u64 sai_index_wait; /* index of entry which is the
1244 * caller is waiting for */
1245 __u64 sai_hit; /* hit count */
1246 __u64 sai_miss; /* miss count:
1247 * for "ls -al" case, it includes
1248 * hidden dentry miss;
1249 * for "ls -l" case, it does not
1250 * include hidden dentry miss.
1251 * "sai_miss_hidden" is used for
1254 unsigned int sai_consecutive_miss; /* consecutive miss */
1255 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1256 * is not a hidden one */
1257 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1258 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1260 sai_in_readpage:1,/* statahead is in readdir()*/
1261 sai_agl_valid:1;/* AGL is valid for the dir */
1262 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
1263 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1264 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1265 cfs_list_t sai_entries; /* entry list */
1266 cfs_list_t sai_entries_received; /* entries returned */
1267 cfs_list_t sai_entries_stated; /* entries stated */
1268 cfs_list_t sai_entries_agl; /* AGL entries to be sent */
1269 cfs_list_t sai_cache[LL_SA_CACHE_SIZE];
1270 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1271 cfs_atomic_t sai_cache_count; /* entry count in cache */
1274 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1276 void ll_stop_statahead(struct inode *dir, void *key);
1278 static inline int ll_glimpse_size(struct inode *inode)
1280 struct ll_inode_info *lli = ll_i2info(inode);
1283 down_read(&lli->lli_glimpse_sem);
1284 rc = cl_glimpse_size(inode);
1285 lli->lli_glimpse_time = cfs_time_current();
1286 up_read(&lli->lli_glimpse_sem);
1291 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1293 struct ll_inode_info *lli = ll_i2info(dir);
1294 struct ll_statahead_info *sai = lli->lli_sai;
1295 struct ll_dentry_data *ldd = ll_d2d(dentry);
1297 /* not the same process, don't mark */
1298 if (lli->lli_opendir_pid != cfs_curproc_pid())
1301 if (sai != NULL && ldd != NULL)
1302 ldd->lld_sa_generation = sai->sai_generation;
1306 ll_need_statahead(struct inode *dir, struct dentry *dentryp)
1308 struct ll_inode_info *lli;
1309 struct ll_dentry_data *ldd;
1311 if (ll_i2sbi(dir)->ll_sa_max == 0)
1314 lli = ll_i2info(dir);
1315 /* not the same process, don't statahead */
1316 if (lli->lli_opendir_pid != cfs_curproc_pid())
1319 /* statahead has been stopped */
1320 if (lli->lli_opendir_key == NULL)
1323 ldd = ll_d2d(dentryp);
1325 * When stats a dentry, the system trigger more than once "revalidate"
1326 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1327 * Under patchless client mode, the operation intent is not accurate,
1328 * which maybe misguide the statahead thread. For example:
1329 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1330 * have the same operation intent -- "IT_GETATTR".
1331 * In fact, one dentry should has only one chance to interact with the
1332 * statahead thread, otherwise the statahead windows will be confused.
1333 * The solution is as following:
1334 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1335 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1336 * will bypass interacting with statahead thread for checking:
1337 * "lld_sa_generation == lli_sai->sai_generation"
1339 if (ldd && lli->lli_sai &&
1340 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1347 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1351 ret = ll_need_statahead(dir, *dentryp);
1355 return do_statahead_enter(dir, dentryp, only_unplug);
1358 /* llite ioctl register support rountine */
1365 #define LLIOC_MAX_CMD 256
1368 * Rules to write a callback function:
1371 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1372 * returned to ll_iocontrol_register. Callback functions should use this
1373 * data to check the potential collasion of ioctl cmd. If collasion is
1374 * found, callback function should return LLIOC_CONT.
1375 * @rcp: The result of ioctl command.
1378 * If @magic matches the pointer returned by ll_iocontrol_data, the
1379 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1381 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1382 struct file *file, unsigned int cmd, unsigned long arg,
1383 void *magic, int *rcp);
1385 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1386 unsigned int cmd, unsigned long arg, int *rcp);
1388 /* export functions */
1389 /* Register ioctl block dynamatically for a regular file.
1391 * @cmd: the array of ioctl command set
1392 * @count: number of commands in the @cmd
1393 * @cb: callback function, it will be called if an ioctl command is found to
1394 * belong to the command list @cmd.
1397 * A magic pointer will be returned if success;
1398 * otherwise, NULL will be returned.
1400 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1401 void ll_iocontrol_unregister(void *magic);
1405 /* lclient compat stuff */
1406 #define cl_inode_info ll_inode_info
1407 #define cl_i2info(info) ll_i2info(info)
1408 #define cl_inode_mode(inode) ((inode)->i_mode)
1409 #define cl_i2sbi ll_i2sbi
1411 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1412 const struct iattr *attr)
1414 LASSERT(attr->ia_valid & ATTR_FILE);
1415 return LUSTRE_FPRIVATE(attr->ia_file);
1418 static inline void cl_isize_lock(struct inode *inode)
1420 ll_inode_size_lock(inode);
1423 static inline void cl_isize_unlock(struct inode *inode)
1425 ll_inode_size_unlock(inode);
1428 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1430 LASSERT_SEM_LOCKED(&ll_i2info(inode)->lli_size_sem);
1431 i_size_write(inode, kms);
1434 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1436 ll_inode_size_lock(inode);
1437 i_size_write(inode, kms);
1438 ll_inode_size_unlock(inode);
1441 #define cl_isize_read(inode) i_size_read(inode)
1443 static inline int cl_merge_lvb(const struct lu_env *env, struct inode *inode)
1445 return ll_merge_lvb(env, inode);
1448 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1449 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1450 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1452 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1454 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1455 enum cl_fsync_mode mode, int ignore_layout);
1457 /** direct write pages */
1458 struct ll_dio_pages {
1459 /** page array to be written. we don't support
1460 * partial pages except the last one. */
1461 struct page **ldp_pages;
1462 /* offset of each page */
1463 loff_t *ldp_offsets;
1464 /** if ldp_offsets is NULL, it means a sequential
1465 * pages to be written, then this is the file offset
1466 * of the * first page. */
1467 loff_t ldp_start_offset;
1468 /** how many bytes are to be written. */
1470 /** # of pages in the array. */
1474 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1477 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1480 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1483 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1484 int rw, struct inode *inode,
1485 struct ll_dio_pages *pv);
1487 static inline int ll_file_nolock(const struct file *file)
1489 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1490 struct inode *inode = file->f_dentry->d_inode;
1492 LASSERT(fd != NULL);
1493 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1494 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1497 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1498 struct lookup_intent *it, __u64 *bits)
1500 if (!it->d.lustre.it_lock_set) {
1501 struct lustre_handle handle;
1503 /* If this inode is a remote object, it will get two
1504 * separate locks in different namespaces, Master MDT,
1505 * where the name entry is, will grant LOOKUP lock,
1506 * remote MDT, where the object is, will grant
1507 * UPDATE|PERM lock. The inode will be attched to both
1508 * LOOKUP and PERM locks, so revoking either locks will
1509 * case the dcache being cleared */
1510 if (it->d.lustre.it_remote_lock_mode) {
1511 handle.cookie = it->d.lustre.it_remote_lock_handle;
1512 CDEBUG(D_DLMTRACE, "setting l_data to inode %p"
1513 "(%lu/%u) for remote lock "LPX64"\n", inode,
1514 inode->i_ino, inode->i_generation,
1516 md_set_lock_data(exp, &handle.cookie, inode, NULL);
1519 handle.cookie = it->d.lustre.it_lock_handle;
1521 CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)"
1522 " for lock "LPX64"\n", inode, inode->i_ino,
1523 inode->i_generation, handle.cookie);
1525 md_set_lock_data(exp, &handle.cookie, inode,
1526 &it->d.lustre.it_lock_bits);
1527 it->d.lustre.it_lock_set = 1;
1531 *bits = it->d.lustre.it_lock_bits;
1534 static inline void ll_lock_dcache(struct inode *inode)
1536 #ifdef HAVE_DCACHE_LOCK
1537 spin_lock(&dcache_lock);
1539 spin_lock(&inode->i_lock);
1543 static inline void ll_unlock_dcache(struct inode *inode)
1545 #ifdef HAVE_DCACHE_LOCK
1546 spin_unlock(&dcache_lock);
1548 spin_unlock(&inode->i_lock);
1552 static inline int d_lustre_invalid(const struct dentry *dentry)
1554 struct ll_dentry_data *lld = ll_d2d(dentry);
1556 return (lld == NULL) || lld->lld_invalid;
1559 static inline void __d_lustre_invalidate(struct dentry *dentry)
1561 struct ll_dentry_data *lld = ll_d2d(dentry);
1564 lld->lld_invalid = 1;
1568 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1569 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1570 * else dput() of the last refcount will unhash this dentry and kill it.
1572 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1574 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1575 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1576 dentry->d_parent, dentry->d_inode, d_refcount(dentry));
1578 spin_lock_nested(&dentry->d_lock,
1579 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1580 __d_lustre_invalidate(dentry);
1581 if (d_refcount(dentry) == 0)
1583 spin_unlock(&dentry->d_lock);
1586 static inline void d_lustre_revalidate(struct dentry *dentry)
1588 spin_lock(&dentry->d_lock);
1589 LASSERT(ll_d2d(dentry) != NULL);
1590 ll_d2d(dentry)->lld_invalid = 0;
1591 spin_unlock(&dentry->d_lock);
1594 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0)
1595 /* Compatibility for old (1.8) compiled userspace quota code */
1596 struct if_quotactl_18 {
1601 struct obd_dqinfo qc_dqinfo;
1602 struct obd_dqblk qc_dqblk;
1604 struct obd_uuid obd_uuid;
1606 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1607 /* End compatibility for old (1.8) compiled userspace quota code */
1609 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1610 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */
1613 LL_LAYOUT_GEN_NONE = ((__u32)-2), /* layout lock was cancelled */
1614 LL_LAYOUT_GEN_EMPTY = ((__u32)-1) /* for empty layout */
1617 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1618 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1620 #endif /* LLITE_INTERNAL_H */