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, 2012, 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
40 #include <lustre_acl.h>
42 #include <lustre_debug.h>
43 #include <lustre_ver.h>
44 #include <lustre_disk.h> /* for s2sbi */
45 #include <lustre_eacl.h>
47 /* for struct cl_lock_descr and struct cl_io */
48 #include <cl_object.h>
50 #include <lustre_mdc.h>
51 #include <linux/lustre_intent.h>
57 #ifndef VM_FAULT_RETRY
58 #define VM_FAULT_RETRY 0
61 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
62 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
63 #ifndef LOOKUP_CONTINUE
64 #define LOOKUP_CONTINUE LOOKUP_PARENT
67 /** Only used on client-side for indicating the tail of dir hash/offset. */
68 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
69 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
71 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
72 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
74 struct ll_dentry_data {
77 struct obd_client_handle lld_cwd_och;
78 struct obd_client_handle lld_mnt_och;
79 struct lookup_intent *lld_it;
80 unsigned int lld_sa_generation;
81 unsigned int lld_invalid:1;
82 struct rcu_head lld_rcu_head;
85 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
87 extern struct file_operations ll_pgcache_seq_fops;
89 #define LLI_INODE_MAGIC 0x111d0de5
90 #define LLI_INODE_DEAD 0xdeadd00d
92 /* remote client permission cache */
93 #define REMOTE_PERM_HASHSIZE 16
95 struct ll_getname_data {
96 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
97 struct lu_fid lgd_fid; /* target fid we are looking for */
98 int lgd_found; /* inode matched? */
101 /* llite setxid/access permission for user on remote client */
102 struct ll_remote_perm {
103 cfs_hlist_node_t lrp_list;
108 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
109 is access permission with
110 lrp_fsuid/lrp_fsgid. */
114 /* MDS has an authority for the Size-on-MDS attributes. */
115 LLIF_MDS_SIZE_LOCK = (1 << 0),
116 /* Epoch close is postponed. */
117 LLIF_EPOCH_PENDING = (1 << 1),
118 /* DONE WRITING is allowed. */
119 LLIF_DONE_WRITING = (1 << 2),
120 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
122 LLIF_SOM_DIRTY = (1 << 3),
123 /* File is contented */
124 LLIF_CONTENDED = (1 << 4),
125 /* Truncate uses server lock for this file */
126 LLIF_SRVLOCK = (1 << 5),
130 struct ll_inode_info {
131 __u32 lli_inode_magic;
136 struct posix_acl *lli_posix_acl;
138 cfs_hlist_head_t *lli_remote_perms;
139 struct mutex lli_rmtperm_mutex;
141 /* identifying fields for both metadata and data stacks. */
142 struct lu_fid lli_fid;
143 /* Parent fid for accessing default stripe data on parent directory
144 * for allocating OST objects after a mknod() and later open-by-FID. */
145 struct lu_fid lli_pfid;
147 cfs_list_t lli_close_list;
148 cfs_list_t lli_oss_capas;
149 /* open count currently used by capability only, indicate whether
150 * capability needs renewal */
151 cfs_atomic_t lli_open_count;
152 struct obd_capa *lli_mds_capa;
153 cfs_time_t lli_rmtperm_time;
155 /* handle is to be sent to MDS later on done_writing and setattr.
156 * Open handle data are needed for the recovery to reconstruct
157 * the inode state on the MDS. XXX: recovery is not ready yet. */
158 struct obd_client_handle *lli_pending_och;
160 /* We need all three because every inode may be opened in different
162 struct obd_client_handle *lli_mds_read_och;
163 struct obd_client_handle *lli_mds_write_och;
164 struct obd_client_handle *lli_mds_exec_och;
165 __u64 lli_open_fd_read_count;
166 __u64 lli_open_fd_write_count;
167 __u64 lli_open_fd_exec_count;
168 /* Protects access to och pointers and their usage counters */
169 struct mutex lli_och_mutex;
171 struct inode lli_vfs_inode;
173 /* the most recent timestamps obtained from mds */
174 struct ost_lvb lli_lvb;
175 spinlock_t lli_agl_lock;
177 /* Try to make the d::member and f::member are aligned. Before using
178 * these members, make clear whether it is directory or not. */
182 /* serialize normal readdir and statahead-readdir. */
183 struct mutex d_readdir_mutex;
185 /* metadata statahead */
186 /* since parent-child threads can share the same @file
187 * struct, "opendir_key" is the token when dir close for
188 * case of parent exit before child -- it is me should
189 * cleanup the dir readahead. */
191 struct ll_statahead_info *d_sai;
192 struct posix_acl *d_def_acl;
193 /* protect statahead stuff. */
194 spinlock_t d_sa_lock;
195 /* "opendir_pid" is the token when lookup/revalid
196 * -- I am the owner of dir statahead. */
200 #define lli_readdir_mutex u.d.d_readdir_mutex
201 #define lli_opendir_key u.d.d_opendir_key
202 #define lli_sai u.d.d_sai
203 #define lli_def_acl u.d.d_def_acl
204 #define lli_sa_lock u.d.d_sa_lock
205 #define lli_opendir_pid u.d.d_opendir_pid
207 /* for non-directory */
209 struct semaphore f_size_sem;
210 void *f_size_sem_owner;
211 char *f_symlink_name;
214 * struct rw_semaphore {
215 * signed long count; // align d.d_def_acl
216 * spinlock_t wait_lock; // align d.d_sa_lock
217 * struct list_head wait_list;
220 struct rw_semaphore f_trunc_sem;
221 struct mutex f_write_mutex;
223 struct rw_semaphore f_glimpse_sem;
224 cfs_time_t f_glimpse_time;
225 cfs_list_t f_agl_list;
228 /* for writepage() only to communicate to fsync */
232 * whenever a process try to read/write the file, the
233 * jobid of the process will be saved here, and it'll
234 * be packed into the write PRC when flush later.
236 * so the read/write statistics for jobid will not be
237 * accurate if the file is shared by different jobs.
239 char f_jobid[JOBSTATS_JOBID_SIZE];
242 #define lli_size_sem u.f.f_size_sem
243 #define lli_size_sem_owner u.f.f_size_sem_owner
244 #define lli_symlink_name u.f.f_symlink_name
245 #define lli_maxbytes u.f.f_maxbytes
246 #define lli_trunc_sem u.f.f_trunc_sem
247 #define lli_write_mutex u.f.f_write_mutex
248 #define lli_glimpse_sem u.f.f_glimpse_sem
249 #define lli_glimpse_time u.f.f_glimpse_time
250 #define lli_agl_list u.f.f_agl_list
251 #define lli_agl_index u.f.f_agl_index
252 #define lli_async_rc u.f.f_async_rc
253 #define lli_jobid u.f.f_jobid
257 /* XXX: For following frequent used members, although they maybe special
258 * used for non-directory object, it is some time-wasting to check
259 * whether the object is directory or not before using them. On the
260 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
261 * the "ll_inode_info" size even if moving those members into u.f.
262 * So keep them out side.
264 * In the future, if more members are added only for directory,
265 * some of the following members can be moved into u.f.
268 struct cl_object *lli_clob;
270 /* mutex to request for layout lock exclusively. */
271 struct mutex lli_layout_mutex;
272 /* valid only inside LAYOUT ibits lock, protected by lli_layout_mutex */
273 __u32 lli_layout_gen;
277 * Locking to guarantee consistency of non-atomic updates to long long i_size,
278 * consistency between file size and KMS.
280 * Implemented by ->lli_size_sem and ->lsm_lock, nested in that order.
283 void ll_inode_size_lock(struct inode *inode);
284 void ll_inode_size_unlock(struct inode *inode);
286 // FIXME: replace the name of this with LL_I to conform to kernel stuff
287 // static inline struct ll_inode_info *LL_I(struct inode *inode)
288 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
290 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
293 /* default to about 40meg of readahead on a given system. That much tied
294 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
295 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - CFS_PAGE_SHIFT))
297 /* default to read-ahead full files smaller than 2MB on the second read */
298 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - CFS_PAGE_SHIFT))
303 RA_STAT_DISTANT_READPAGE,
304 RA_STAT_MISS_IN_WINDOW,
305 RA_STAT_FAILED_GRAB_PAGE,
306 RA_STAT_FAILED_MATCH,
311 RA_STAT_MAX_IN_FLIGHT,
312 RA_STAT_WRONG_GRAB_PAGE,
317 cfs_atomic_t ra_cur_pages;
318 unsigned long ra_max_pages;
319 unsigned long ra_max_pages_per_file;
320 unsigned long ra_max_read_ahead_whole_pages;
323 /* ra_io_arg will be filled in the beginning of ll_readahead with
324 * ras_lock, then the following ll_read_ahead_pages will read RA
325 * pages according to this arg, all the items in this structure are
326 * counted by page index.
329 unsigned long ria_start; /* start offset of read-ahead*/
330 unsigned long ria_end; /* end offset of read-ahead*/
331 /* If stride read pattern is detected, ria_stoff means where
332 * stride read is started. Note: for normal read-ahead, the
333 * value here is meaningless, and also it will not be accessed*/
335 /* ria_length and ria_pages are the length and pages length in the
336 * stride I/O mode. And they will also be used to check whether
337 * it is stride I/O read-ahead in the read-ahead pages*/
338 unsigned long ria_length;
339 unsigned long ria_pages;
342 /* LL_HIST_MAX=32 causes an overflow */
343 #define LL_HIST_MAX 28
344 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
345 #define LL_PROCESS_HIST_MAX 10
346 struct per_process_info {
348 struct obd_histogram pp_r_hist;
349 struct obd_histogram pp_w_hist;
352 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
353 struct ll_rw_extents_info {
354 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
357 #define LL_OFFSET_HIST_MAX 100
358 struct ll_rw_process_info {
361 loff_t rw_range_start;
363 loff_t rw_last_file_pos;
365 size_t rw_smallest_extent;
366 size_t rw_largest_extent;
367 struct ll_file_data *rw_last_file;
370 enum stats_track_type {
371 STATS_TRACK_ALL = 0, /* track all processes */
372 STATS_TRACK_PID, /* track process with this pid */
373 STATS_TRACK_PPID, /* track processes with this ppid */
374 STATS_TRACK_GID, /* track processes with this gid */
378 /* flags for sbi->ll_flags */
379 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
380 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
381 #define LL_SBI_FLOCK 0x04
382 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
383 #define LL_SBI_ACL 0x10 /* support ACL */
384 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
385 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
386 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
387 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
388 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
389 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
390 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
391 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
392 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
393 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
394 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
395 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
396 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
398 /* default value for ll_sb_info->contention_time */
399 #define SBI_DEFAULT_CONTENTION_SECONDS 60
400 /* default value for lockless_truncate_enable */
401 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
402 #define RCE_HASHES 32
404 struct rmtacl_ctl_entry {
406 pid_t rce_key; /* hash key */
407 int rce_ops; /* acl operation type */
410 struct rmtacl_ctl_table {
412 cfs_list_t rct_entries[RCE_HASHES];
419 pid_t ee_key; /* hash key */
420 struct lu_fid ee_fid;
421 int ee_type; /* ACL type for ACCESS or DEFAULT */
422 ext_acl_xattr_header *ee_acl;
427 cfs_list_t et_entries[EE_HASHES];
432 /* this protects pglist and ra_info. It isn't safe to
433 * grab from interrupt contexts */
435 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
436 spinlock_t ll_process_lock; /* ll_rw_process_info */
437 struct obd_uuid ll_sb_uuid;
438 struct obd_export *ll_md_exp;
439 struct obd_export *ll_dt_exp;
440 struct proc_dir_entry* ll_proc_root;
441 struct lu_fid ll_root_fid; /* root object fid */
444 cfs_list_t ll_conn_chain; /* per-conn chain of SBs */
445 struct lustre_client_ocd ll_lco;
447 cfs_list_t ll_orphan_dentry_list; /*please don't ask -p*/
448 struct ll_close_queue *ll_lcq;
450 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
452 struct cl_client_cache ll_cache;
454 struct lprocfs_stats *ll_ra_stats;
456 struct ll_ra_info ll_ra_info;
457 unsigned int ll_namelen;
458 struct file_operations *ll_fop;
460 /* =0 - hold lock over whole read/write
461 * >0 - max. chunk to be read/written w/o lock re-acquiring */
462 unsigned long ll_max_rw_chunk;
463 unsigned int ll_md_brw_size; /* used by readdir */
465 struct lu_site *ll_site;
466 struct cl_device *ll_cl;
468 struct ll_rw_extents_info ll_rw_extents_info;
469 int ll_extent_process_count;
470 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
471 unsigned int ll_offset_process_count;
472 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
473 unsigned int ll_rw_offset_entry_count;
474 int ll_stats_track_id;
475 enum stats_track_type ll_stats_track_type;
478 /* metadata stat-ahead */
479 unsigned int ll_sa_max; /* max statahead RPCs */
480 atomic_t ll_sa_total; /* statahead thread started
482 atomic_t ll_sa_wrong; /* statahead thread stopped for
484 atomic_t ll_agl_total; /* AGL thread started count */
486 dev_t ll_sdev_orig; /* save s_dev before assign for
488 struct rmtacl_ctl_table ll_rct;
489 struct eacl_table ll_et;
492 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
497 struct task_struct *lrr_reader;
498 cfs_list_t lrr_linkage;
502 * per file-descriptor read-ahead data.
504 struct ll_readahead_state {
507 * index of the last page that read(2) needed and that wasn't in the
508 * cache. Used by ras_update() to detect seeks.
510 * XXX nikita: if access seeks into cached region, Lustre doesn't see
513 unsigned long ras_last_readpage;
515 * number of pages read after last read-ahead window reset. As window
516 * is reset on each seek, this is effectively a number of consecutive
517 * accesses. Maybe ->ras_accessed_in_window is better name.
519 * XXX nikita: window is also reset (by ras_update()) when Lustre
520 * believes that memory pressure evicts read-ahead pages. In that
521 * case, it probably doesn't make sense to expand window to
522 * PTLRPC_MAX_BRW_PAGES on the third access.
524 unsigned long ras_consecutive_pages;
526 * number of read requests after the last read-ahead window reset
527 * As window is reset on each seek, this is effectively the number
528 * on consecutive read request and is used to trigger read-ahead.
530 unsigned long ras_consecutive_requests;
532 * Parameters of current read-ahead window. Handled by
533 * ras_update(). On the initial access to the file or after a seek,
534 * window is reset to 0. After 3 consecutive accesses, window is
535 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
536 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
538 unsigned long ras_window_start, ras_window_len;
540 * Where next read-ahead should start at. This lies within read-ahead
541 * window. Read-ahead window is read in pieces rather than at once
542 * because: 1. lustre limits total number of pages under read-ahead by
543 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
544 * not covered by DLM lock.
546 unsigned long ras_next_readahead;
548 * Total number of ll_file_read requests issued, reads originating
549 * due to mmap are not counted in this total. This value is used to
550 * trigger full file read-ahead after multiple reads to a small file.
552 unsigned long ras_requests;
554 * Page index with respect to the current request, these value
555 * will not be accurate when dealing with reads issued via mmap.
557 unsigned long ras_request_index;
559 * list of struct ll_ra_read's one per read(2) call current in
560 * progress against this file descriptor. Used by read-ahead code,
561 * protected by ->ras_lock.
563 cfs_list_t ras_read_beads;
565 * The following 3 items are used for detecting the stride I/O
567 * In stride I/O mode,
568 * ...............|-----data-----|****gap*****|--------|******|....
569 * offset |-stride_pages-|-stride_gap-|
570 * ras_stride_offset = offset;
571 * ras_stride_length = stride_pages + stride_gap;
572 * ras_stride_pages = stride_pages;
573 * Note: all these three items are counted by pages.
575 unsigned long ras_stride_length;
576 unsigned long ras_stride_pages;
577 pgoff_t ras_stride_offset;
579 * number of consecutive stride request count, and it is similar as
580 * ras_consecutive_requests, but used for stride I/O mode.
581 * Note: only more than 2 consecutive stride request are detected,
582 * stride read-ahead will be enable
584 unsigned long ras_consecutive_stride_requests;
587 extern cfs_mem_cache_t *ll_file_data_slab;
588 struct lustre_handle;
589 struct ll_file_data {
590 struct ll_readahead_state fd_ras;
592 struct ccc_grouplock fd_grouplock;
595 struct file *fd_file;
596 /* Indicate whether need to report failure when close.
597 * true: failure is known, not report again.
598 * false: unknown failure, should report. */
599 bool fd_write_failed;
602 struct lov_stripe_md;
604 extern spinlock_t inode_lock;
606 extern struct proc_dir_entry *proc_lustre_fs_root;
608 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
610 return &lli->lli_vfs_inode;
614 struct inode *icbd_parent;
615 struct dentry **icbd_childp;
619 __u32 ll_i2suppgid(struct inode *i);
620 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
622 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
624 #if BITS_PER_LONG == 32
627 return unlikely(cfs_curproc_is_32bit() || (sbi->ll_flags & LL_SBI_32BIT_API));
631 #define LLAP_MAGIC 98764321
633 extern cfs_mem_cache_t *ll_async_page_slab;
634 extern size_t ll_async_page_slab_size;
636 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
637 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
638 struct ll_ra_read *ll_ra_read_get(struct file *f);
640 /* llite/lproc_llite.c */
642 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
643 struct super_block *sb, char *osc, char *mdc);
644 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
645 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
646 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
648 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
649 struct super_block *sb, char *osc, char *mdc){return 0;}
650 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
651 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
652 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
654 memset(lvars, 0, sizeof(*lvars));
660 void ll_release_page(struct page *page, int remove);
661 extern struct file_operations ll_dir_operations;
662 extern struct inode_operations ll_dir_inode_operations;
663 struct page *ll_get_dir_page(struct inode *dir, __u64 hash,
664 struct ll_dir_chain *chain);
665 int ll_dir_read(struct inode *inode, __u64 *_pos, void *cookie,
668 int ll_get_mdt_idx(struct inode *inode);
670 int ll_objects_destroy(struct ptlrpc_request *request,
672 struct inode *ll_iget(struct super_block *sb, ino_t hash,
673 struct lustre_md *lic);
674 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
675 void *data, int flag);
676 #ifndef HAVE_IOP_ATOMIC_OPEN
677 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
680 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
683 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
684 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
685 int ll_writepage(struct page *page, struct writeback_control *wbc);
686 int ll_writepages(struct address_space *, struct writeback_control *wbc);
687 void ll_removepage(struct page *page);
688 int ll_readpage(struct file *file, struct page *page);
689 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
690 int ll_file_punch(struct inode *, loff_t, int);
691 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
692 void ll_clear_file_contended(struct inode*);
693 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
694 int ll_readahead(const struct lu_env *env, struct cl_io *io,
695 struct ll_readahead_state *ras, struct address_space *mapping,
696 struct cl_page_list *queue, int flags);
699 extern struct file_operations ll_file_operations;
700 extern struct file_operations ll_file_operations_flock;
701 extern struct file_operations ll_file_operations_noflock;
702 extern struct inode_operations ll_file_inode_operations;
703 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
705 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
706 ldlm_mode_t l_req_mode);
707 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
708 struct lustre_handle *lockh, __u64 flags);
709 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
711 #ifdef HAVE_IOP_ATOMIC_OPEN
712 int ll_revalidate_nd(struct dentry *dentry, unsigned int flags);
714 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
716 int ll_file_open(struct inode *inode, struct file *file);
717 int ll_file_release(struct inode *inode, struct file *file);
718 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
719 struct lov_stripe_md *lsm, lstat_t *st);
720 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
721 int ll_local_open(struct file *file,
722 struct lookup_intent *it, struct ll_file_data *fd,
723 struct obd_client_handle *och);
724 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
725 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
727 int ll_md_real_close(struct inode *inode, int flags);
728 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
729 struct obd_client_handle **och, unsigned long flags);
730 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
731 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
732 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
733 __u64 ioepoch, int sync);
734 int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data,
735 struct md_open_data **mod);
736 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
737 struct lustre_handle *fh);
738 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
739 struct ll_file_data *file, loff_t pos,
740 size_t count, int rw);
741 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
742 struct lookup_intent *it, struct kstat *stat);
743 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
744 struct ll_file_data *ll_file_data_get(void);
745 struct posix_acl * ll_get_acl(struct inode *inode, int type);
747 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
748 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
750 # ifndef HAVE_INODE_PERMISION_2ARGS
751 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
753 int ll_inode_permission(struct inode *inode, int mask);
757 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
758 int flags, struct lov_user_md *lum,
760 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
761 struct lov_mds_md **lmm, int *lmm_size,
762 struct ptlrpc_request **request);
763 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
765 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmm,
766 int *lmm_size, struct ptlrpc_request **request);
767 #ifdef HAVE_FILE_FSYNC_4ARGS
768 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
769 #elif defined(HAVE_FILE_FSYNC_2ARGS)
770 int ll_fsync(struct file *file, int data);
772 int ll_fsync(struct file *file, struct dentry *dentry, int data);
774 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
776 int ll_merge_lvb(struct inode *inode);
777 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
778 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
779 int ll_fid2path(struct inode *inode, void *arg);
783 int ll_dops_init(struct dentry *de, int block, int init_sa);
784 extern struct dentry_operations ll_d_ops;
785 void ll_intent_drop_lock(struct lookup_intent *);
786 void ll_intent_release(struct lookup_intent *);
787 void ll_invalidate_aliases(struct inode *);
788 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
789 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
790 #ifdef HAVE_D_COMPARE_7ARGS
791 int ll_dcompare(const struct dentry *parent, const struct inode *pinode,
792 const struct dentry *dentry, const struct inode *inode,
793 unsigned int len, const char *str, const struct qstr *d_name);
795 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
797 int ll_revalidate_it_finish(struct ptlrpc_request *request,
798 struct lookup_intent *it, struct dentry *de);
800 /* llite/llite_lib.c */
801 extern struct super_operations lustre_super_operations;
803 char *ll_read_opt(const char *opt, char *data);
804 void ll_lli_init(struct ll_inode_info *lli);
805 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
806 void ll_put_super(struct super_block *sb);
807 void ll_kill_super(struct super_block *sb);
808 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
809 void ll_clear_inode(struct inode *inode);
810 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr);
811 int ll_setattr(struct dentry *de, struct iattr *attr);
812 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
813 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
814 __u64 max_age, __u32 flags);
815 void ll_update_inode(struct inode *inode, struct lustre_md *md);
816 void ll_read_inode2(struct inode *inode, void *opaque);
817 void ll_delete_inode(struct inode *inode);
818 int ll_iocontrol(struct inode *inode, struct file *file,
819 unsigned int cmd, unsigned long arg);
820 int ll_flush_ctx(struct inode *inode);
821 #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT
822 void ll_umount_begin(struct vfsmount *vfsmnt, int flags);
824 void ll_umount_begin(struct super_block *sb);
826 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
827 #ifdef HAVE_SUPEROPS_USE_DENTRY
828 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
830 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
832 void ll_dirty_page_discard_warn(cfs_page_t *page, int ioret);
833 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
834 struct super_block *);
835 void lustre_dump_dentry(struct dentry *, int recur);
836 void lustre_dump_inode(struct inode *);
837 int ll_obd_statfs(struct inode *inode, void *arg);
838 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
839 int ll_process_config(struct lustre_cfg *lcfg);
840 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
841 struct inode *i1, struct inode *i2,
842 const char *name, int namelen,
843 int mode, __u32 opc, void *data);
844 void ll_finish_md_op_data(struct md_op_data *op_data);
845 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
846 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
848 /* llite/llite_nfs.c */
849 extern struct export_operations lustre_export_operations;
850 __u32 get_uuid2int(const char *name, int len);
852 /* llite/special.c */
853 extern struct inode_operations ll_special_inode_operations;
854 extern struct file_operations ll_special_chr_inode_fops;
855 extern struct file_operations ll_special_chr_file_fops;
856 extern struct file_operations ll_special_blk_inode_fops;
857 extern struct file_operations ll_special_fifo_inode_fops;
858 extern struct file_operations ll_special_fifo_file_fops;
859 extern struct file_operations ll_special_sock_inode_fops;
861 /* llite/symlink.c */
862 extern struct inode_operations ll_fast_symlink_inode_operations;
864 /* llite/llite_close.c */
865 struct ll_close_queue {
868 cfs_waitq_t lcq_waitq;
869 struct completion lcq_comp;
870 cfs_atomic_t lcq_stop;
873 struct ccc_object *cl_inode2ccc(struct inode *inode);
876 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
877 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
879 /* specific achitecture can implement only part of this list */
880 enum vvp_io_subtype {
883 /** io called from .sendfile */
885 /** io started from splice_{read|write} */
892 enum vvp_io_subtype cui_io_subtype;
896 read_actor_t cui_actor;
900 struct pipe_inode_info *cui_pipe;
901 unsigned int cui_flags;
903 struct vvp_fault_io {
905 * Inode modification time that is checked across DLM
909 struct vm_area_struct *ft_vma;
911 * locked page returned from vvp_io
913 cfs_page_t *ft_vmpage;
914 #ifndef HAVE_VM_OP_FAULT
915 struct vm_nopage_api {
917 * Virtual address at which fault occurred.
919 unsigned long ft_address;
921 * Fault type, as to be supplied to
927 struct vm_fault_api {
931 struct vm_fault *ft_vmf;
933 * fault API used bitflags for return code.
935 unsigned int ft_flags;
941 * Read-ahead state used by read and page-fault IO contexts.
943 struct ll_ra_read cui_bead;
945 * Set when cui_bead has been initialized.
947 int cui_ra_window_set;
949 * Partially truncated page, that vvp_io_trunc_start() keeps locked
952 struct cl_page *cui_partpage;
956 * IO arguments for various VFS I/O interfaces.
959 /** normal/sendfile/splice */
960 enum vvp_io_subtype via_io_subtype;
964 #ifndef HAVE_FILE_WRITEV
965 struct kiocb *via_iocb;
967 struct iovec *via_iov;
968 unsigned long via_nrsegs;
971 read_actor_t via_actor;
975 struct pipe_inode_info *via_pipe;
976 unsigned int via_flags;
981 struct ll_cl_context {
983 struct cl_io *lcc_io;
984 struct cl_page *lcc_page;
985 struct lu_env *lcc_env;
990 struct vvp_thread_info {
991 struct ost_lvb vti_lvb;
992 struct cl_2queue vti_queue;
993 struct iovec vti_local_iov;
994 struct vvp_io_args vti_args;
995 struct ra_io_arg vti_ria;
996 struct kiocb vti_kiocb;
997 struct ll_cl_context vti_io_ctx;
1000 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
1002 extern struct lu_context_key vvp_key;
1003 struct vvp_thread_info *info;
1005 info = lu_context_key_get(&env->le_ctx, &vvp_key);
1006 LASSERT(info != NULL);
1010 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
1011 enum vvp_io_subtype type)
1013 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
1015 ret->via_io_subtype = type;
1020 struct vvp_session {
1021 struct vvp_io vs_ios;
1024 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1026 extern struct lu_context_key vvp_session_key;
1027 struct vvp_session *ses;
1029 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1030 LASSERT(ses != NULL);
1034 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1036 return &vvp_env_session(env)->vs_ios;
1039 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1040 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1041 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1043 /* llite/llite_mmap.c */
1044 typedef struct rb_root rb_root_t;
1045 typedef struct rb_node rb_node_t;
1047 struct ll_lock_tree_node;
1048 struct ll_lock_tree {
1050 cfs_list_t lt_locked_list;
1051 struct ll_file_data *lt_fd;
1054 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1055 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1056 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
1057 __u64 end, ldlm_mode_t mode);
1058 void policy_from_vma(ldlm_policy_data_t *policy,
1059 struct vm_area_struct *vma, unsigned long addr, size_t count);
1060 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1063 static inline void ll_invalidate_page(struct page *vmpage)
1065 struct address_space *mapping = vmpage->mapping;
1066 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1068 LASSERT(PageLocked(vmpage));
1069 if (mapping == NULL)
1072 ll_teardown_mmaps(mapping, offset, offset + CFS_PAGE_SIZE);
1073 truncate_complete_page(mapping, vmpage);
1076 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1078 /* don't need an addref as the sb_info should be holding one */
1079 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1081 return ll_s2sbi(sb)->ll_dt_exp;
1084 /* don't need an addref as the sb_info should be holding one */
1085 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1087 return ll_s2sbi(sb)->ll_md_exp;
1090 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1092 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1098 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1099 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1101 return ll_s2sbi(inode->i_sb);
1104 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1106 return ll_s2dtexp(inode->i_sb);
1109 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1111 return ll_s2mdexp(inode->i_sb);
1114 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1118 LASSERT(inode != NULL);
1119 fid = &ll_i2info(inode)->lli_fid;
1124 static inline int ll_mds_max_easize(struct super_block *sb)
1126 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1129 static inline __u64 ll_file_maxbytes(struct inode *inode)
1131 return ll_i2info(inode)->lli_maxbytes;
1135 int ll_setxattr(struct dentry *dentry, const char *name,
1136 const void *value, size_t size, int flags);
1137 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1138 void *buffer, size_t size);
1139 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1140 int ll_removexattr(struct dentry *dentry, const char *name);
1142 /* llite/remote_perm.c */
1143 extern cfs_mem_cache_t *ll_remote_perm_cachep;
1144 extern cfs_mem_cache_t *ll_rmtperm_hash_cachep;
1146 cfs_hlist_head_t *alloc_rmtperm_hash(void);
1147 void free_rmtperm_hash(cfs_hlist_head_t *hash);
1148 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1149 int lustre_check_remote_perm(struct inode *inode, int mask);
1151 /* llite/llite_capa.c */
1152 extern cfs_timer_t ll_capa_timer;
1154 int ll_capa_thread_start(void);
1155 void ll_capa_thread_stop(void);
1156 void ll_capa_timer_callback(unsigned long unused);
1158 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1159 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1161 void ll_capa_open(struct inode *inode);
1162 void ll_capa_close(struct inode *inode);
1164 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1165 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1167 void ll_truncate_free_capa(struct obd_capa *ocapa);
1168 void ll_clear_inode_capas(struct inode *inode);
1169 void ll_print_capa_stat(struct ll_sb_info *sbi);
1171 /* llite/llite_cl.c */
1172 extern struct lu_device_type vvp_device_type;
1175 * Common IO arguments for various VFS I/O interfaces.
1177 int cl_sb_init(struct super_block *sb);
1178 int cl_sb_fini(struct super_block *sb);
1179 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1180 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1182 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1183 struct ll_readahead_state *ras, unsigned long index,
1185 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1186 int ll_is_file_contended(struct file *file);
1187 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1189 /* llite/llite_rmtacl.c */
1190 #ifdef CONFIG_FS_POSIX_ACL
1191 obd_valid rce_ops2valid(int ops);
1192 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1193 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1194 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1195 void rct_init(struct rmtacl_ctl_table *rct);
1196 void rct_fini(struct rmtacl_ctl_table *rct);
1198 void ee_free(struct eacl_entry *ee);
1199 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1200 ext_acl_xattr_header *header);
1201 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1202 struct lu_fid *fid, int type);
1203 void et_search_free(struct eacl_table *et, pid_t key);
1204 void et_init(struct eacl_table *et);
1205 void et_fini(struct eacl_table *et);
1210 #define LL_SA_RPC_MIN 2
1211 #define LL_SA_RPC_DEF 32
1212 #define LL_SA_RPC_MAX 8192
1214 #define LL_SA_CACHE_BIT 5
1215 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1216 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1218 /* per inode struct, for dir only */
1219 struct ll_statahead_info {
1220 struct inode *sai_inode;
1221 cfs_atomic_t sai_refcount; /* when access this struct, hold
1223 unsigned int sai_generation; /* generation for statahead */
1224 unsigned int sai_max; /* max ahead of lookup */
1225 __u64 sai_sent; /* stat requests sent count */
1226 __u64 sai_replied; /* stat requests which received
1228 __u64 sai_index; /* index of statahead entry */
1229 __u64 sai_index_wait; /* index of entry which is the
1230 * caller is waiting for */
1231 __u64 sai_hit; /* hit count */
1232 __u64 sai_miss; /* miss count:
1233 * for "ls -al" case, it includes
1234 * hidden dentry miss;
1235 * for "ls -l" case, it does not
1236 * include hidden dentry miss.
1237 * "sai_miss_hidden" is used for
1240 unsigned int sai_consecutive_miss; /* consecutive miss */
1241 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1242 * is not a hidden one */
1243 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1244 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1246 sai_in_readpage:1,/* statahead is in readdir()*/
1247 sai_agl_valid:1;/* AGL is valid for the dir */
1248 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
1249 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1250 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1251 cfs_list_t sai_entries_sent; /* entries sent out */
1252 cfs_list_t sai_entries_received; /* entries returned */
1253 cfs_list_t sai_entries_stated; /* entries stated */
1254 cfs_list_t sai_entries_agl; /* AGL entries to be sent */
1255 cfs_list_t sai_cache[LL_SA_CACHE_SIZE];
1256 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1257 cfs_atomic_t sai_cache_count; /* entry count in cache */
1260 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1262 void ll_stop_statahead(struct inode *dir, void *key);
1264 static inline int ll_glimpse_size(struct inode *inode)
1266 struct ll_inode_info *lli = ll_i2info(inode);
1269 down_read(&lli->lli_glimpse_sem);
1270 rc = cl_glimpse_size(inode);
1271 lli->lli_glimpse_time = cfs_time_current();
1272 up_read(&lli->lli_glimpse_sem);
1277 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1279 struct ll_inode_info *lli = ll_i2info(dir);
1280 struct ll_statahead_info *sai = lli->lli_sai;
1281 struct ll_dentry_data *ldd = ll_d2d(dentry);
1283 /* not the same process, don't mark */
1284 if (lli->lli_opendir_pid != cfs_curproc_pid())
1287 if (sai != NULL && ldd != NULL)
1288 ldd->lld_sa_generation = sai->sai_generation;
1292 ll_need_statahead(struct inode *dir, struct dentry *dentryp)
1294 struct ll_inode_info *lli;
1295 struct ll_dentry_data *ldd;
1297 if (ll_i2sbi(dir)->ll_sa_max == 0)
1300 lli = ll_i2info(dir);
1301 /* not the same process, don't statahead */
1302 if (lli->lli_opendir_pid != cfs_curproc_pid())
1305 /* statahead has been stopped */
1306 if (lli->lli_opendir_key == NULL)
1309 ldd = ll_d2d(dentryp);
1311 * When stats a dentry, the system trigger more than once "revalidate"
1312 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1313 * Under patchless client mode, the operation intent is not accurate,
1314 * which maybe misguide the statahead thread. For example:
1315 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1316 * have the same operation intent -- "IT_GETATTR".
1317 * In fact, one dentry should has only one chance to interact with the
1318 * statahead thread, otherwise the statahead windows will be confused.
1319 * The solution is as following:
1320 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1321 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1322 * will bypass interacting with statahead thread for checking:
1323 * "lld_sa_generation == lli_sai->sai_generation"
1325 if (ldd && lli->lli_sai &&
1326 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1333 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1337 ret = ll_need_statahead(dir, *dentryp);
1341 return do_statahead_enter(dir, dentryp, only_unplug);
1344 /* llite ioctl register support rountine */
1351 #define LLIOC_MAX_CMD 256
1354 * Rules to write a callback function:
1357 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1358 * returned to ll_iocontrol_register. Callback functions should use this
1359 * data to check the potential collasion of ioctl cmd. If collasion is
1360 * found, callback function should return LLIOC_CONT.
1361 * @rcp: The result of ioctl command.
1364 * If @magic matches the pointer returned by ll_iocontrol_data, the
1365 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1367 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1368 struct file *file, unsigned int cmd, unsigned long arg,
1369 void *magic, int *rcp);
1371 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1372 unsigned int cmd, unsigned long arg, int *rcp);
1374 /* export functions */
1375 /* Register ioctl block dynamatically for a regular file.
1377 * @cmd: the array of ioctl command set
1378 * @count: number of commands in the @cmd
1379 * @cb: callback function, it will be called if an ioctl command is found to
1380 * belong to the command list @cmd.
1383 * A magic pointer will be returned if success;
1384 * otherwise, NULL will be returned.
1386 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1387 void ll_iocontrol_unregister(void *magic);
1391 /* lclient compat stuff */
1392 #define cl_inode_info ll_inode_info
1393 #define cl_i2info(info) ll_i2info(info)
1394 #define cl_inode_mode(inode) ((inode)->i_mode)
1395 #define cl_i2sbi ll_i2sbi
1397 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1398 const struct iattr *attr)
1400 LASSERT(attr->ia_valid & ATTR_FILE);
1401 return LUSTRE_FPRIVATE(attr->ia_file);
1404 static inline void cl_isize_lock(struct inode *inode)
1406 ll_inode_size_lock(inode);
1409 static inline void cl_isize_unlock(struct inode *inode)
1411 ll_inode_size_unlock(inode);
1414 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1416 LASSERT_SEM_LOCKED(&ll_i2info(inode)->lli_size_sem);
1417 i_size_write(inode, kms);
1420 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1422 ll_inode_size_lock(inode);
1423 i_size_write(inode, kms);
1424 ll_inode_size_unlock(inode);
1427 #define cl_isize_read(inode) i_size_read(inode)
1429 static inline int cl_merge_lvb(struct inode *inode)
1431 return ll_merge_lvb(inode);
1434 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1435 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1436 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1438 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1440 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1441 enum cl_fsync_mode mode);
1443 /** direct write pages */
1444 struct ll_dio_pages {
1445 /** page array to be written. we don't support
1446 * partial pages except the last one. */
1447 struct page **ldp_pages;
1448 /* offset of each page */
1449 loff_t *ldp_offsets;
1450 /** if ldp_offsets is NULL, it means a sequential
1451 * pages to be written, then this is the file offset
1452 * of the * first page. */
1453 loff_t ldp_start_offset;
1454 /** how many bytes are to be written. */
1456 /** # of pages in the array. */
1460 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1463 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1466 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1469 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1470 int rw, struct inode *inode,
1471 struct ll_dio_pages *pv);
1473 static inline int ll_file_nolock(const struct file *file)
1475 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1476 struct inode *inode = file->f_dentry->d_inode;
1478 LASSERT(fd != NULL);
1479 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1480 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1483 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1484 struct lookup_intent *it, __u64 *bits)
1486 if (!it->d.lustre.it_lock_set) {
1487 CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)\n",
1488 inode, inode->i_ino, inode->i_generation);
1489 md_set_lock_data(exp, &it->d.lustre.it_lock_handle,
1490 inode, &it->d.lustre.it_lock_bits);
1491 it->d.lustre.it_lock_set = 1;
1495 *bits = it->d.lustre.it_lock_bits;
1498 static inline void ll_lock_dcache(struct inode *inode)
1500 #ifdef HAVE_DCACHE_LOCK
1501 spin_lock(&dcache_lock);
1503 spin_lock(&inode->i_lock);
1507 static inline void ll_unlock_dcache(struct inode *inode)
1509 #ifdef HAVE_DCACHE_LOCK
1510 spin_unlock(&dcache_lock);
1512 spin_unlock(&inode->i_lock);
1516 static inline int d_lustre_invalid(const struct dentry *dentry)
1518 struct ll_dentry_data *lld = ll_d2d(dentry);
1520 return (lld == NULL) || lld->lld_invalid;
1523 static inline void __d_lustre_invalidate(struct dentry *dentry)
1525 struct ll_dentry_data *lld = ll_d2d(dentry);
1528 lld->lld_invalid = 1;
1532 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1533 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1534 * else dput() of the last refcount will unhash this dentry and kill it.
1536 static inline void d_lustre_invalidate(struct dentry *dentry)
1538 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1539 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1540 dentry->d_parent, dentry->d_inode, d_refcount(dentry));
1542 spin_lock(&dentry->d_lock);
1543 __d_lustre_invalidate(dentry);
1544 if (d_refcount(dentry) == 0)
1546 spin_unlock(&dentry->d_lock);
1549 static inline void d_lustre_revalidate(struct dentry *dentry)
1551 spin_lock(&dentry->d_lock);
1552 LASSERT(ll_d2d(dentry) != NULL);
1553 ll_d2d(dentry)->lld_invalid = 0;
1554 spin_unlock(&dentry->d_lock);
1557 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0)
1558 /* Compatibility for old (1.8) compiled userspace quota code */
1559 struct if_quotactl_18 {
1564 struct obd_dqinfo qc_dqinfo;
1565 struct obd_dqblk qc_dqblk;
1567 struct obd_uuid obd_uuid;
1569 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1570 /* End compatibility for old (1.8) compiled userspace quota code */
1572 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1573 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */
1575 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1576 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1578 #endif /* LLITE_INTERNAL_H */