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, Whamcloud, Inc.
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 #ifndef DCACHE_LUSTRE_INVALID
72 #define DCACHE_LUSTRE_INVALID 0x4000000
75 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
76 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
78 struct ll_dentry_data {
81 struct obd_client_handle lld_cwd_och;
82 struct obd_client_handle lld_mnt_och;
83 struct lookup_intent *lld_it;
84 unsigned int lld_sa_generation;
87 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
89 extern struct file_operations ll_pgcache_seq_fops;
91 #define LLI_INODE_MAGIC 0x111d0de5
92 #define LLI_INODE_DEAD 0xdeadd00d
94 /* remote client permission cache */
95 #define REMOTE_PERM_HASHSIZE 16
97 struct ll_getname_data {
98 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
99 struct lu_fid lgd_fid; /* target fid we are looking for */
100 int lgd_found; /* inode matched? */
103 /* llite setxid/access permission for user on remote client */
104 struct ll_remote_perm {
105 cfs_hlist_node_t lrp_list;
110 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
111 is access permission with
112 lrp_fsuid/lrp_fsgid. */
116 /* MDS has an authority for the Size-on-MDS attributes. */
117 LLIF_MDS_SIZE_LOCK = (1 << 0),
118 /* Epoch close is postponed. */
119 LLIF_EPOCH_PENDING = (1 << 1),
120 /* DONE WRITING is allowed. */
121 LLIF_DONE_WRITING = (1 << 2),
122 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
124 LLIF_SOM_DIRTY = (1 << 3),
125 /* File is contented */
126 LLIF_CONTENDED = (1 << 4),
127 /* Truncate uses server lock for this file */
128 LLIF_SRVLOCK = (1 << 5),
132 struct ll_inode_info {
133 __u32 lli_inode_magic;
137 cfs_spinlock_t lli_lock;
138 struct posix_acl *lli_posix_acl;
140 cfs_hlist_head_t *lli_remote_perms;
141 cfs_mutex_t lli_rmtperm_mutex;
143 /* identifying fields for both metadata and data stacks. */
144 struct lu_fid lli_fid;
145 /* Parent fid for accessing default stripe data on parent directory
146 * for allocating OST objects after a mknod() and later open-by-FID. */
147 struct lu_fid lli_pfid;
149 cfs_list_t lli_close_list;
150 cfs_list_t lli_oss_capas;
151 /* open count currently used by capability only, indicate whether
152 * capability needs renewal */
153 cfs_atomic_t lli_open_count;
154 struct obd_capa *lli_mds_capa;
155 cfs_time_t lli_rmtperm_time;
157 /* handle is to be sent to MDS later on done_writing and setattr.
158 * Open handle data are needed for the recovery to reconstruct
159 * the inode state on the MDS. XXX: recovery is not ready yet. */
160 struct obd_client_handle *lli_pending_och;
162 /* We need all three because every inode may be opened in different
164 struct obd_client_handle *lli_mds_read_och;
165 struct obd_client_handle *lli_mds_write_och;
166 struct obd_client_handle *lli_mds_exec_och;
167 __u64 lli_open_fd_read_count;
168 __u64 lli_open_fd_write_count;
169 __u64 lli_open_fd_exec_count;
170 /* Protects access to och pointers and their usage counters, also
171 * atomicity of check-update of lli_has_smd */
172 cfs_mutex_t lli_och_mutex;
174 struct inode lli_vfs_inode;
176 /* the most recent timestamps obtained from mds */
177 struct ost_lvb lli_lvb;
178 cfs_spinlock_t lli_agl_lock;
180 /* Try to make the d::member and f::member are aligned. Before using
181 * these members, make clear whether it is directory or not. */
185 /* serialize normal readdir and statahead-readdir. */
186 cfs_mutex_t d_readdir_mutex;
188 /* metadata statahead */
189 /* since parent-child threads can share the same @file
190 * struct, "opendir_key" is the token when dir close for
191 * case of parent exit before child -- it is me should
192 * cleanup the dir readahead. */
194 struct ll_statahead_info *d_sai;
196 struct posix_acl *d_def_acl;
197 /* protect statahead stuff. */
198 cfs_spinlock_t d_sa_lock;
199 /* "opendir_pid" is the token when lookup/revalid
200 * -- I am the owner of dir statahead. */
204 #define lli_readdir_mutex u.d.d_readdir_mutex
205 #define lli_opendir_key u.d.d_opendir_key
206 #define lli_sai u.d.d_sai
207 #define lli_sa_pos u.d.d_sa_pos
208 #define lli_def_acl u.d.d_def_acl
209 #define lli_sa_lock u.d.d_sa_lock
210 #define lli_opendir_pid u.d.d_opendir_pid
212 /* for non-directory */
214 cfs_semaphore_t f_size_sem;
215 void *f_size_sem_owner;
216 char *f_symlink_name;
219 * cfs_rw_semaphore_t {
220 * signed long count; // align u.d.d_def_acl
221 * cfs_spinlock_t wait_lock; // align u.d.d_sa_lock
222 * struct list_head wait_list;
225 cfs_rw_semaphore_t f_trunc_sem;
226 cfs_mutex_t f_write_mutex;
228 cfs_rw_semaphore_t f_glimpse_sem;
229 cfs_time_t f_glimpse_time;
230 cfs_list_t f_agl_list;
233 /* for writepage() only to communicate to fsync */
237 * whenever a process try to read/write the file, the
238 * jobid of the process will be saved here, and it'll
239 * be packed into the write PRC when flush later.
241 * so the read/write statistics for jobid will not be
242 * accurate if the file is shared by different jobs.
244 char f_jobid[JOBSTATS_JOBID_SIZE];
247 #define lli_size_sem u.f.f_size_sem
248 #define lli_size_sem_owner u.f.f_size_sem_owner
249 #define lli_symlink_name u.f.f_symlink_name
250 #define lli_maxbytes u.f.f_maxbytes
251 #define lli_trunc_sem u.f.f_trunc_sem
252 #define lli_write_mutex u.f.f_write_mutex
253 #define lli_glimpse_sem u.f.f_glimpse_sem
254 #define lli_glimpse_time u.f.f_glimpse_time
255 #define lli_agl_list u.f.f_agl_list
256 #define lli_agl_index u.f.f_agl_index
257 #define lli_async_rc u.f.f_async_rc
258 #define lli_jobid u.f.f_jobid
262 /* XXX: For following frequent used members, although they maybe special
263 * used for non-directory object, it is some time-wasting to check
264 * whether the object is directory or not before using them. On the
265 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
266 * the "ll_inode_info" size even if moving those members into u.f.
267 * So keep them out side.
269 * In the future, if more members are added only for directory,
270 * some of the following members can be moved into u.f.
273 struct cl_object *lli_clob;
275 /* mutex to request for layout lock exclusively. */
276 cfs_mutex_t lli_layout_mutex;
280 * Locking to guarantee consistency of non-atomic updates to long long i_size,
281 * consistency between file size and KMS.
283 * Implemented by ->lli_size_sem and ->lsm_lock, nested in that order.
286 void ll_inode_size_lock(struct inode *inode);
287 void ll_inode_size_unlock(struct inode *inode);
289 // FIXME: replace the name of this with LL_I to conform to kernel stuff
290 // static inline struct ll_inode_info *LL_I(struct inode *inode)
291 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
293 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
296 /* default to about 40meg of readahead on a given system. That much tied
297 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
298 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - CFS_PAGE_SHIFT))
300 /* default to read-ahead full files smaller than 2MB on the second read */
301 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - CFS_PAGE_SHIFT))
306 RA_STAT_DISTANT_READPAGE,
307 RA_STAT_MISS_IN_WINDOW,
308 RA_STAT_FAILED_GRAB_PAGE,
309 RA_STAT_FAILED_MATCH,
314 RA_STAT_MAX_IN_FLIGHT,
315 RA_STAT_WRONG_GRAB_PAGE,
320 cfs_atomic_t ra_cur_pages;
321 unsigned long ra_max_pages;
322 unsigned long ra_max_pages_per_file;
323 unsigned long ra_max_read_ahead_whole_pages;
326 /* ra_io_arg will be filled in the beginning of ll_readahead with
327 * ras_lock, then the following ll_read_ahead_pages will read RA
328 * pages according to this arg, all the items in this structure are
329 * counted by page index.
332 unsigned long ria_start; /* start offset of read-ahead*/
333 unsigned long ria_end; /* end offset of read-ahead*/
334 /* If stride read pattern is detected, ria_stoff means where
335 * stride read is started. Note: for normal read-ahead, the
336 * value here is meaningless, and also it will not be accessed*/
338 /* ria_length and ria_pages are the length and pages length in the
339 * stride I/O mode. And they will also be used to check whether
340 * it is stride I/O read-ahead in the read-ahead pages*/
341 unsigned long ria_length;
342 unsigned long ria_pages;
345 /* LL_HIST_MAX=32 causes an overflow */
346 #define LL_HIST_MAX 28
347 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
348 #define LL_PROCESS_HIST_MAX 10
349 struct per_process_info {
351 struct obd_histogram pp_r_hist;
352 struct obd_histogram pp_w_hist;
355 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
356 struct ll_rw_extents_info {
357 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
360 #define LL_OFFSET_HIST_MAX 100
361 struct ll_rw_process_info {
364 loff_t rw_range_start;
366 loff_t rw_last_file_pos;
368 size_t rw_smallest_extent;
369 size_t rw_largest_extent;
370 struct ll_file_data *rw_last_file;
373 enum stats_track_type {
374 STATS_TRACK_ALL = 0, /* track all processes */
375 STATS_TRACK_PID, /* track process with this pid */
376 STATS_TRACK_PPID, /* track processes with this ppid */
377 STATS_TRACK_GID, /* track processes with this gid */
381 /* flags for sbi->ll_flags */
382 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
383 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
384 #define LL_SBI_FLOCK 0x04
385 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
386 #define LL_SBI_ACL 0x10 /* support ACL */
387 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
388 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
389 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
390 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
391 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
392 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
393 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
394 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
395 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
396 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
397 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
398 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
399 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
401 /* default value for ll_sb_info->contention_time */
402 #define SBI_DEFAULT_CONTENTION_SECONDS 60
403 /* default value for lockless_truncate_enable */
404 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
405 #define RCE_HASHES 32
407 struct rmtacl_ctl_entry {
409 pid_t rce_key; /* hash key */
410 int rce_ops; /* acl operation type */
413 struct rmtacl_ctl_table {
414 cfs_spinlock_t rct_lock;
415 cfs_list_t rct_entries[RCE_HASHES];
422 pid_t ee_key; /* hash key */
423 struct lu_fid ee_fid;
424 int ee_type; /* ACL type for ACCESS or DEFAULT */
425 ext_acl_xattr_header *ee_acl;
429 cfs_spinlock_t et_lock;
430 cfs_list_t et_entries[EE_HASHES];
435 /* this protects pglist and ra_info. It isn't safe to
436 * grab from interrupt contexts */
437 cfs_spinlock_t ll_lock;
438 cfs_spinlock_t ll_pp_extent_lock; /* Lock for pp_extent entries */
439 cfs_spinlock_t ll_process_lock; /* Lock for ll_rw_process_info */
440 struct obd_uuid ll_sb_uuid;
441 struct obd_export *ll_md_exp;
442 struct obd_export *ll_dt_exp;
443 struct proc_dir_entry* ll_proc_root;
444 struct lu_fid ll_root_fid; /* root object fid */
447 cfs_list_t ll_conn_chain; /* per-conn chain of SBs */
448 struct lustre_client_ocd ll_lco;
450 cfs_list_t ll_orphan_dentry_list; /*please don't ask -p*/
451 struct ll_close_queue *ll_lcq;
453 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
455 unsigned long ll_async_page_max;
456 unsigned long ll_async_page_count;
458 struct lprocfs_stats *ll_ra_stats;
460 struct ll_ra_info ll_ra_info;
461 unsigned int ll_namelen;
462 struct file_operations *ll_fop;
464 /* =0 - hold lock over whole read/write
465 * >0 - max. chunk to be read/written w/o lock re-acquiring */
466 unsigned long ll_max_rw_chunk;
467 unsigned int ll_md_brw_size; /* used by readdir */
469 struct lu_site *ll_site;
470 struct cl_device *ll_cl;
472 struct ll_rw_extents_info ll_rw_extents_info;
473 int ll_extent_process_count;
474 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
475 unsigned int ll_offset_process_count;
476 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
477 unsigned int ll_rw_offset_entry_count;
478 int ll_stats_track_id;
479 enum stats_track_type ll_stats_track_type;
482 /* metadata stat-ahead */
483 unsigned int ll_sa_max; /* max statahead RPCs */
484 atomic_t ll_sa_total; /* statahead thread started
486 atomic_t ll_sa_wrong; /* statahead thread stopped for
488 atomic_t ll_agl_total; /* AGL thread started count */
490 dev_t ll_sdev_orig; /* save s_dev before assign for
492 struct rmtacl_ctl_table ll_rct;
493 struct eacl_table ll_et;
494 struct vfsmount *ll_mnt;
497 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
502 struct task_struct *lrr_reader;
503 cfs_list_t lrr_linkage;
507 * per file-descriptor read-ahead data.
509 struct ll_readahead_state {
510 cfs_spinlock_t ras_lock;
512 * index of the last page that read(2) needed and that wasn't in the
513 * cache. Used by ras_update() to detect seeks.
515 * XXX nikita: if access seeks into cached region, Lustre doesn't see
518 unsigned long ras_last_readpage;
520 * number of pages read after last read-ahead window reset. As window
521 * is reset on each seek, this is effectively a number of consecutive
522 * accesses. Maybe ->ras_accessed_in_window is better name.
524 * XXX nikita: window is also reset (by ras_update()) when Lustre
525 * believes that memory pressure evicts read-ahead pages. In that
526 * case, it probably doesn't make sense to expand window to
527 * PTLRPC_MAX_BRW_PAGES on the third access.
529 unsigned long ras_consecutive_pages;
531 * number of read requests after the last read-ahead window reset
532 * As window is reset on each seek, this is effectively the number
533 * on consecutive read request and is used to trigger read-ahead.
535 unsigned long ras_consecutive_requests;
537 * Parameters of current read-ahead window. Handled by
538 * ras_update(). On the initial access to the file or after a seek,
539 * window is reset to 0. After 3 consecutive accesses, window is
540 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
541 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
543 unsigned long ras_window_start, ras_window_len;
545 * Where next read-ahead should start at. This lies within read-ahead
546 * window. Read-ahead window is read in pieces rather than at once
547 * because: 1. lustre limits total number of pages under read-ahead by
548 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
549 * not covered by DLM lock.
551 unsigned long ras_next_readahead;
553 * Total number of ll_file_read requests issued, reads originating
554 * due to mmap are not counted in this total. This value is used to
555 * trigger full file read-ahead after multiple reads to a small file.
557 unsigned long ras_requests;
559 * Page index with respect to the current request, these value
560 * will not be accurate when dealing with reads issued via mmap.
562 unsigned long ras_request_index;
564 * list of struct ll_ra_read's one per read(2) call current in
565 * progress against this file descriptor. Used by read-ahead code,
566 * protected by ->ras_lock.
568 cfs_list_t ras_read_beads;
570 * The following 3 items are used for detecting the stride I/O
572 * In stride I/O mode,
573 * ...............|-----data-----|****gap*****|--------|******|....
574 * offset |-stride_pages-|-stride_gap-|
575 * ras_stride_offset = offset;
576 * ras_stride_length = stride_pages + stride_gap;
577 * ras_stride_pages = stride_pages;
578 * Note: all these three items are counted by pages.
580 unsigned long ras_stride_length;
581 unsigned long ras_stride_pages;
582 pgoff_t ras_stride_offset;
584 * number of consecutive stride request count, and it is similar as
585 * ras_consecutive_requests, but used for stride I/O mode.
586 * Note: only more than 2 consecutive stride request are detected,
587 * stride read-ahead will be enable
589 unsigned long ras_consecutive_stride_requests;
597 extern cfs_mem_cache_t *ll_file_data_slab;
598 struct lustre_handle;
599 struct ll_file_data {
600 struct ll_readahead_state fd_ras;
602 struct ccc_grouplock fd_grouplock;
603 struct ll_file_dir fd_dir;
605 struct file *fd_file;
606 /* Indicate whether need to report failure when close.
607 * true: failure is known, not report again.
608 * false: unknown failure, should report. */
609 bool fd_write_failed;
612 struct lov_stripe_md;
614 extern cfs_spinlock_t inode_lock;
616 extern struct proc_dir_entry *proc_lustre_fs_root;
618 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
620 return &lli->lli_vfs_inode;
624 struct inode *icbd_parent;
625 struct dentry **icbd_childp;
629 __u32 ll_i2suppgid(struct inode *i);
630 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
632 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
634 #if BITS_PER_LONG == 32
637 return unlikely(cfs_curproc_is_32bit() || (sbi->ll_flags & LL_SBI_32BIT_API));
641 #define LLAP_MAGIC 98764321
643 extern cfs_mem_cache_t *ll_async_page_slab;
644 extern size_t ll_async_page_slab_size;
646 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
647 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
648 struct ll_ra_read *ll_ra_read_get(struct file *f);
650 /* llite/lproc_llite.c */
652 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
653 struct super_block *sb, char *osc, char *mdc);
654 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
655 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
656 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
658 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
659 struct super_block *sb, char *osc, char *mdc){return 0;}
660 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
661 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
662 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
664 memset(lvars, 0, sizeof(*lvars));
670 void ll_release_page(struct page *page, int remove);
671 extern struct file_operations ll_dir_operations;
672 extern struct inode_operations ll_dir_inode_operations;
673 struct page *ll_get_dir_page(struct file *filp, struct inode *dir, __u64 hash,
674 struct ll_dir_chain *chain);
675 int ll_readdir(struct file *filp, void *cookie, filldir_t filldir);
677 int ll_get_mdt_idx(struct inode *inode);
678 char *ll_get_fsname(struct inode *inode);
680 int ll_objects_destroy(struct ptlrpc_request *request,
682 struct inode *ll_iget(struct super_block *sb, ino_t hash,
683 struct lustre_md *lic);
684 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
685 void *data, int flag);
686 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
688 int ll_lookup_it_finish(struct ptlrpc_request *request,
689 struct lookup_intent *it, void *data);
690 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
693 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
694 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
695 int ll_writepage(struct page *page, struct writeback_control *wbc);
696 int ll_writepages(struct address_space *, struct writeback_control *wbc);
697 void ll_removepage(struct page *page);
698 int ll_readpage(struct file *file, struct page *page);
699 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
700 void ll_truncate(struct inode *inode);
701 int ll_file_punch(struct inode *, loff_t, int);
702 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
703 void ll_clear_file_contended(struct inode*);
704 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
705 int ll_readahead(const struct lu_env *env, struct cl_io *io,
706 struct ll_readahead_state *ras, struct address_space *mapping,
707 struct cl_page_list *queue, int flags);
710 extern struct file_operations ll_file_operations;
711 extern struct file_operations ll_file_operations_flock;
712 extern struct file_operations ll_file_operations_noflock;
713 extern struct inode_operations ll_file_inode_operations;
714 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
716 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
717 ldlm_mode_t l_req_mode);
718 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
719 struct lustre_handle *lockh);
720 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
722 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
723 int ll_file_open(struct inode *inode, struct file *file);
724 int ll_file_release(struct inode *inode, struct file *file);
725 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
726 struct lov_stripe_md *lsm, lstat_t *st);
727 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
728 int ll_local_open(struct file *file,
729 struct lookup_intent *it, struct ll_file_data *fd,
730 struct obd_client_handle *och);
731 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
732 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
734 int ll_md_real_close(struct inode *inode, int flags);
735 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
736 struct obd_client_handle **och, unsigned long flags);
737 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
738 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
739 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
740 __u64 ioepoch, int sync);
741 int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data,
742 struct md_open_data **mod);
743 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
744 struct lustre_handle *fh);
745 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
746 struct ll_file_data *file, loff_t pos,
747 size_t count, int rw);
748 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
749 struct lookup_intent *it, struct kstat *stat);
750 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
751 struct ll_file_data *ll_file_data_get(void);
752 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
753 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
755 # ifndef HAVE_INODE_PERMISION_2ARGS
756 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
758 int ll_inode_permission(struct inode *inode, int mask);
761 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
762 int flags, struct lov_user_md *lum,
764 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
765 struct lov_mds_md **lmm, int *lmm_size,
766 struct ptlrpc_request **request);
767 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
769 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmm,
770 int *lmm_size, struct ptlrpc_request **request);
771 #ifdef HAVE_FILE_FSYNC_4ARGS
772 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
773 #elif defined(HAVE_FILE_FSYNC_2ARGS)
774 int ll_fsync(struct file *file, int data);
776 int ll_fsync(struct file *file, struct dentry *dentry, int data);
778 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
780 int ll_merge_lvb(struct inode *inode);
781 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
782 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
783 int ll_fid2path(struct inode *inode, void *arg);
787 int ll_dops_init(struct dentry *de, int block, int init_sa);
788 extern struct dentry_operations ll_d_ops;
789 void ll_intent_drop_lock(struct lookup_intent *);
790 void ll_intent_release(struct lookup_intent *);
791 void ll_invalidate_aliases(struct inode *);
792 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
793 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
794 #ifdef HAVE_D_COMPARE_7ARGS
795 int ll_dcompare(const struct dentry *parent, const struct inode *pinode,
796 const struct dentry *dentry, const struct inode *inode,
797 unsigned int len, const char *str, const struct qstr *d_name);
799 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
801 int ll_revalidate_it_finish(struct ptlrpc_request *request,
802 struct lookup_intent *it, struct dentry *de);
804 /* llite/llite_lib.c */
805 extern struct super_operations lustre_super_operations;
807 char *ll_read_opt(const char *opt, char *data);
808 void ll_lli_init(struct ll_inode_info *lli);
809 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
810 void ll_put_super(struct super_block *sb);
811 void ll_kill_super(struct super_block *sb);
812 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
813 void ll_clear_inode(struct inode *inode);
814 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr);
815 int ll_setattr(struct dentry *de, struct iattr *attr);
816 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
817 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
818 __u64 max_age, __u32 flags);
819 void ll_update_inode(struct inode *inode, struct lustre_md *md);
820 void ll_read_inode2(struct inode *inode, void *opaque);
821 void ll_delete_inode(struct inode *inode);
822 int ll_iocontrol(struct inode *inode, struct file *file,
823 unsigned int cmd, unsigned long arg);
824 int ll_flush_ctx(struct inode *inode);
825 #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT
826 void ll_umount_begin(struct vfsmount *vfsmnt, int flags);
828 void ll_umount_begin(struct super_block *sb);
830 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
831 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
832 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
833 struct super_block *);
834 void lustre_dump_dentry(struct dentry *, int recur);
835 void lustre_dump_inode(struct inode *);
836 int ll_obd_statfs(struct inode *inode, void *arg);
837 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
838 int ll_process_config(struct lustre_cfg *lcfg);
839 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
840 struct inode *i1, struct inode *i2,
841 const char *name, int namelen,
842 int mode, __u32 opc, void *data);
843 void ll_finish_md_op_data(struct md_op_data *op_data);
844 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
846 /* llite/llite_nfs.c */
847 extern struct export_operations lustre_export_operations;
848 __u32 get_uuid2int(const char *name, int len);
850 /* llite/special.c */
851 extern struct inode_operations ll_special_inode_operations;
852 extern struct file_operations ll_special_chr_inode_fops;
853 extern struct file_operations ll_special_chr_file_fops;
854 extern struct file_operations ll_special_blk_inode_fops;
855 extern struct file_operations ll_special_fifo_inode_fops;
856 extern struct file_operations ll_special_fifo_file_fops;
857 extern struct file_operations ll_special_sock_inode_fops;
859 /* llite/symlink.c */
860 extern struct inode_operations ll_fast_symlink_inode_operations;
862 /* llite/llite_close.c */
863 struct ll_close_queue {
864 cfs_spinlock_t lcq_lock;
866 cfs_waitq_t lcq_waitq;
867 cfs_completion_t lcq_comp;
868 cfs_atomic_t lcq_stop;
871 struct ccc_object *cl_inode2ccc(struct inode *inode);
874 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
875 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
877 /* specific achitecture can implement only part of this list */
878 enum vvp_io_subtype {
881 /** io called from .sendfile */
883 /** io started from splice_{read|write} */
890 enum vvp_io_subtype cui_io_subtype;
894 read_actor_t cui_actor;
898 struct pipe_inode_info *cui_pipe;
899 unsigned int cui_flags;
901 struct vvp_fault_io {
903 * Inode modification time that is checked across DLM
907 struct vm_area_struct *ft_vma;
909 * locked page returned from vvp_io
911 cfs_page_t *ft_vmpage;
912 #ifndef HAVE_VM_OP_FAULT
913 struct vm_nopage_api {
915 * Virtual address at which fault occurred.
917 unsigned long ft_address;
919 * Fault type, as to be supplied to
925 struct vm_fault_api {
929 struct vm_fault *ft_vmf;
931 * fault API used bitflags for return code.
933 unsigned int ft_flags;
939 * Read-ahead state used by read and page-fault IO contexts.
941 struct ll_ra_read cui_bead;
943 * Set when cui_bead has been initialized.
945 int cui_ra_window_set;
947 * Partially truncated page, that vvp_io_trunc_start() keeps locked
950 struct cl_page *cui_partpage;
954 * IO arguments for various VFS I/O interfaces.
957 /** normal/sendfile/splice */
958 enum vvp_io_subtype via_io_subtype;
962 #ifndef HAVE_FILE_WRITEV
963 struct kiocb *via_iocb;
965 struct iovec *via_iov;
966 unsigned long via_nrsegs;
969 read_actor_t via_actor;
973 struct pipe_inode_info *via_pipe;
974 unsigned int via_flags;
979 struct ll_cl_context {
981 struct cl_io *lcc_io;
982 struct cl_page *lcc_page;
983 struct lu_env *lcc_env;
988 struct vvp_thread_info {
989 struct ost_lvb vti_lvb;
990 struct cl_2queue vti_queue;
991 struct iovec vti_local_iov;
992 struct vvp_io_args vti_args;
993 struct ra_io_arg vti_ria;
994 struct kiocb vti_kiocb;
995 struct ll_cl_context vti_io_ctx;
998 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
1000 extern struct lu_context_key vvp_key;
1001 struct vvp_thread_info *info;
1003 info = lu_context_key_get(&env->le_ctx, &vvp_key);
1004 LASSERT(info != NULL);
1008 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
1009 enum vvp_io_subtype type)
1011 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
1013 ret->via_io_subtype = type;
1018 struct vvp_session {
1019 struct vvp_io vs_ios;
1022 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1024 extern struct lu_context_key vvp_session_key;
1025 struct vvp_session *ses;
1027 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1028 LASSERT(ses != NULL);
1032 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1034 return &vvp_env_session(env)->vs_ios;
1037 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1038 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1039 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1041 /* llite/llite_mmap.c */
1042 typedef struct rb_root rb_root_t;
1043 typedef struct rb_node rb_node_t;
1045 struct ll_lock_tree_node;
1046 struct ll_lock_tree {
1048 cfs_list_t lt_locked_list;
1049 struct ll_file_data *lt_fd;
1052 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1053 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1054 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
1055 __u64 end, ldlm_mode_t mode);
1056 void policy_from_vma(ldlm_policy_data_t *policy,
1057 struct vm_area_struct *vma, unsigned long addr, size_t count);
1058 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1061 static inline void ll_invalidate_page(struct page *vmpage)
1063 struct address_space *mapping = vmpage->mapping;
1064 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1066 LASSERT(PageLocked(vmpage));
1067 if (mapping == NULL)
1070 ll_teardown_mmaps(mapping, offset, offset + CFS_PAGE_SIZE);
1071 truncate_complete_page(mapping, vmpage);
1074 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1076 /* don't need an addref as the sb_info should be holding one */
1077 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1079 return ll_s2sbi(sb)->ll_dt_exp;
1082 /* don't need an addref as the sb_info should be holding one */
1083 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1085 return ll_s2sbi(sb)->ll_md_exp;
1088 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1090 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1096 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1097 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1099 return ll_s2sbi(inode->i_sb);
1102 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1104 return ll_s2dtexp(inode->i_sb);
1107 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1109 return ll_s2mdexp(inode->i_sb);
1112 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1115 LASSERT(inode != NULL);
1116 fid = &ll_i2info(inode)->lli_fid;
1117 LASSERT(fid_is_igif(fid) || fid_ver(fid) == 0);
1121 static inline int ll_mds_max_easize(struct super_block *sb)
1123 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1126 static inline __u64 ll_file_maxbytes(struct inode *inode)
1128 return ll_i2info(inode)->lli_maxbytes;
1132 int ll_setxattr(struct dentry *dentry, const char *name,
1133 const void *value, size_t size, int flags);
1134 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1135 void *buffer, size_t size);
1136 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1137 int ll_removexattr(struct dentry *dentry, const char *name);
1139 /* llite/remote_perm.c */
1140 extern cfs_mem_cache_t *ll_remote_perm_cachep;
1141 extern cfs_mem_cache_t *ll_rmtperm_hash_cachep;
1143 cfs_hlist_head_t *alloc_rmtperm_hash(void);
1144 void free_rmtperm_hash(cfs_hlist_head_t *hash);
1145 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1146 int lustre_check_remote_perm(struct inode *inode, int mask);
1148 /* llite/llite_capa.c */
1149 extern cfs_timer_t ll_capa_timer;
1151 int ll_capa_thread_start(void);
1152 void ll_capa_thread_stop(void);
1153 void ll_capa_timer_callback(unsigned long unused);
1155 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1156 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1158 void ll_capa_open(struct inode *inode);
1159 void ll_capa_close(struct inode *inode);
1161 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1162 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1164 void ll_truncate_free_capa(struct obd_capa *ocapa);
1165 void ll_clear_inode_capas(struct inode *inode);
1166 void ll_print_capa_stat(struct ll_sb_info *sbi);
1168 /* llite/llite_cl.c */
1169 extern struct lu_device_type vvp_device_type;
1172 * Common IO arguments for various VFS I/O interfaces.
1174 int cl_sb_init(struct super_block *sb);
1175 int cl_sb_fini(struct super_block *sb);
1176 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1177 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1179 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1180 struct ll_readahead_state *ras, unsigned long index,
1182 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1183 int ll_is_file_contended(struct file *file);
1184 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1186 /* llite/llite_rmtacl.c */
1187 #ifdef CONFIG_FS_POSIX_ACL
1188 obd_valid rce_ops2valid(int ops);
1189 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1190 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1191 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1192 void rct_init(struct rmtacl_ctl_table *rct);
1193 void rct_fini(struct rmtacl_ctl_table *rct);
1195 void ee_free(struct eacl_entry *ee);
1196 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1197 ext_acl_xattr_header *header);
1198 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1199 struct lu_fid *fid, int type);
1200 void et_search_free(struct eacl_table *et, pid_t key);
1201 void et_init(struct eacl_table *et);
1202 void et_fini(struct eacl_table *et);
1207 #define LL_SA_RPC_MIN 2
1208 #define LL_SA_RPC_DEF 32
1209 #define LL_SA_RPC_MAX 8192
1211 #define LL_SA_CACHE_BIT 5
1212 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1213 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1215 /* per inode struct, for dir only */
1216 struct ll_statahead_info {
1217 struct inode *sai_inode;
1218 cfs_atomic_t sai_refcount; /* when access this struct, hold
1220 unsigned int sai_generation; /* generation for statahead */
1221 unsigned int sai_max; /* max ahead of lookup */
1222 __u64 sai_sent; /* stat requests sent count */
1223 __u64 sai_replied; /* stat requests which received
1225 __u64 sai_index; /* index of statahead entry */
1226 __u64 sai_index_wait; /* index of entry which is the
1227 * caller is waiting for */
1228 __u64 sai_hit; /* hit count */
1229 __u64 sai_miss; /* miss count:
1230 * for "ls -al" case, it includes
1231 * hidden dentry miss;
1232 * for "ls -l" case, it does not
1233 * include hidden dentry miss.
1234 * "sai_miss_hidden" is used for
1237 unsigned int sai_consecutive_miss; /* consecutive miss */
1238 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1239 * is not a hidden one */
1240 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1241 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1243 sai_in_readpage:1,/* statahead is in readdir()*/
1244 sai_agl_valid:1;/* AGL is valid for the dir */
1245 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
1246 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1247 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1248 cfs_list_t sai_entries_sent; /* entries sent out */
1249 cfs_list_t sai_entries_received; /* entries returned */
1250 cfs_list_t sai_entries_stated; /* entries stated */
1251 cfs_list_t sai_entries_agl; /* AGL entries to be sent */
1252 cfs_list_t sai_cache[LL_SA_CACHE_SIZE];
1253 cfs_spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1254 cfs_atomic_t sai_cache_count; /* entry count in cache */
1257 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1259 void ll_stop_statahead(struct inode *dir, void *key);
1261 static inline int ll_glimpse_size(struct inode *inode)
1263 struct ll_inode_info *lli = ll_i2info(inode);
1266 cfs_down_read(&lli->lli_glimpse_sem);
1267 rc = cl_glimpse_size(inode);
1268 lli->lli_glimpse_time = cfs_time_current();
1269 cfs_up_read(&lli->lli_glimpse_sem);
1274 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1276 struct ll_inode_info *lli = ll_i2info(dir);
1277 struct ll_statahead_info *sai = lli->lli_sai;
1278 struct ll_dentry_data *ldd = ll_d2d(dentry);
1280 /* not the same process, don't mark */
1281 if (lli->lli_opendir_pid != cfs_curproc_pid())
1284 if (sai != NULL && ldd != NULL)
1285 ldd->lld_sa_generation = sai->sai_generation;
1289 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1291 struct ll_inode_info *lli;
1292 struct ll_dentry_data *ldd;
1294 if (ll_i2sbi(dir)->ll_sa_max == 0)
1297 lli = ll_i2info(dir);
1298 /* not the same process, don't statahead */
1299 if (lli->lli_opendir_pid != cfs_curproc_pid())
1302 ldd = ll_d2d(*dentryp);
1304 * When stats a dentry, the system trigger more than once "revalidate"
1305 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1306 * Under patchless client mode, the operation intent is not accurate,
1307 * which maybe misguide the statahead thread. For example:
1308 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1309 * have the same operation intent -- "IT_GETATTR".
1310 * In fact, one dentry should has only one chance to interact with the
1311 * statahead thread, otherwise the statahead windows will be confused.
1312 * The solution is as following:
1313 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1314 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1315 * will bypass interacting with statahead thread for checking:
1316 * "lld_sa_generation == lli_sai->sai_generation"
1318 if (ldd && lli->lli_sai &&
1319 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1322 return do_statahead_enter(dir, dentryp, only_unplug);
1325 /* llite ioctl register support rountine */
1332 #define LLIOC_MAX_CMD 256
1335 * Rules to write a callback function:
1338 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1339 * returned to ll_iocontrol_register. Callback functions should use this
1340 * data to check the potential collasion of ioctl cmd. If collasion is
1341 * found, callback function should return LLIOC_CONT.
1342 * @rcp: The result of ioctl command.
1345 * If @magic matches the pointer returned by ll_iocontrol_data, the
1346 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1348 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1349 struct file *file, unsigned int cmd, unsigned long arg,
1350 void *magic, int *rcp);
1352 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1353 unsigned int cmd, unsigned long arg, int *rcp);
1355 /* export functions */
1356 /* Register ioctl block dynamatically for a regular file.
1358 * @cmd: the array of ioctl command set
1359 * @count: number of commands in the @cmd
1360 * @cb: callback function, it will be called if an ioctl command is found to
1361 * belong to the command list @cmd.
1364 * A magic pointer will be returned if success;
1365 * otherwise, NULL will be returned.
1367 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1368 void ll_iocontrol_unregister(void *magic);
1372 /* lclient compat stuff */
1373 #define cl_inode_info ll_inode_info
1374 #define cl_i2info(info) ll_i2info(info)
1375 #define cl_inode_mode(inode) ((inode)->i_mode)
1376 #define cl_i2sbi ll_i2sbi
1378 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1379 const struct iattr *attr)
1381 LASSERT(attr->ia_valid & ATTR_FILE);
1382 return LUSTRE_FPRIVATE(attr->ia_file);
1385 static inline void cl_isize_lock(struct inode *inode)
1387 ll_inode_size_lock(inode);
1390 static inline void cl_isize_unlock(struct inode *inode)
1392 ll_inode_size_unlock(inode);
1395 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1397 LASSERT_SEM_LOCKED(&ll_i2info(inode)->lli_size_sem);
1398 i_size_write(inode, kms);
1401 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1403 ll_inode_size_lock(inode);
1404 i_size_write(inode, kms);
1405 ll_inode_size_unlock(inode);
1408 #define cl_isize_read(inode) i_size_read(inode)
1410 static inline int cl_merge_lvb(struct inode *inode)
1412 return ll_merge_lvb(inode);
1415 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1416 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1417 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1419 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1421 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1422 enum cl_fsync_mode mode);
1424 /** direct write pages */
1425 struct ll_dio_pages {
1426 /** page array to be written. we don't support
1427 * partial pages except the last one. */
1428 struct page **ldp_pages;
1429 /* offset of each page */
1430 loff_t *ldp_offsets;
1431 /** if ldp_offsets is NULL, it means a sequential
1432 * pages to be written, then this is the file offset
1433 * of the * first page. */
1434 loff_t ldp_start_offset;
1435 /** how many bytes are to be written. */
1437 /** # of pages in the array. */
1441 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1444 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1447 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1450 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1451 int rw, struct inode *inode,
1452 struct ll_dio_pages *pv);
1454 static inline int ll_file_nolock(const struct file *file)
1456 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1457 struct inode *inode = file->f_dentry->d_inode;
1459 LASSERT(fd != NULL);
1460 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1461 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1464 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1465 struct lookup_intent *it, __u64 *bits)
1467 if (!it->d.lustre.it_lock_set) {
1468 CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)\n",
1469 inode, inode->i_ino, inode->i_generation);
1470 md_set_lock_data(exp, &it->d.lustre.it_lock_handle,
1471 inode, &it->d.lustre.it_lock_bits);
1472 it->d.lustre.it_lock_set = 1;
1476 *bits = it->d.lustre.it_lock_bits;
1479 static inline void ll_lock_dcache(struct inode *inode)
1481 #ifdef HAVE_DCACHE_LOCK
1482 spin_lock(&dcache_lock);
1484 spin_lock(&inode->i_lock);
1488 static inline void ll_unlock_dcache(struct inode *inode)
1490 #ifdef HAVE_DCACHE_LOCK
1491 spin_unlock(&dcache_lock);
1493 spin_unlock(&inode->i_lock);
1497 static inline int d_lustre_invalid(const struct dentry *dentry)
1499 return dentry->d_flags & DCACHE_LUSTRE_INVALID;
1502 static inline void __d_lustre_invalidate(struct dentry *dentry)
1504 dentry->d_flags |= DCACHE_LUSTRE_INVALID;
1508 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1509 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1510 * else dput() of the last refcount will unhash this dentry and kill it.
1512 static inline void d_lustre_invalidate(struct dentry *dentry)
1514 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1515 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1516 dentry->d_parent, dentry->d_inode, d_refcount(dentry));
1518 spin_lock(&dentry->d_lock);
1519 __d_lustre_invalidate(dentry);
1520 if (d_refcount(dentry) == 0)
1522 spin_unlock(&dentry->d_lock);
1525 static inline void d_lustre_revalidate(struct dentry *dentry)
1527 spin_lock(&dentry->d_lock);
1528 dentry->d_flags &= ~DCACHE_LUSTRE_INVALID;
1529 spin_unlock(&dentry->d_lock);
1532 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0)
1533 /* Compatibility for old (1.8) compiled userspace quota code */
1534 struct if_quotactl_18 {
1539 struct obd_dqinfo qc_dqinfo;
1540 struct obd_dqblk qc_dqblk;
1542 struct obd_uuid obd_uuid;
1544 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1545 /* End compatibility for old (1.8) compiled userspace quota code */
1547 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1548 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */
1550 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1551 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1553 #endif /* LLITE_INTERNAL_H */