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 /** Only used on client-side for indicating the tail of dir hash/offset. */
62 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
63 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
65 #ifndef DCACHE_LUSTRE_INVALID
66 #define DCACHE_LUSTRE_INVALID 0x4000000
69 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
70 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
72 struct ll_dentry_data {
75 struct obd_client_handle lld_cwd_och;
76 struct obd_client_handle lld_mnt_och;
77 struct lookup_intent *lld_it;
78 unsigned int lld_sa_generation;
81 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
83 extern struct file_operations ll_pgcache_seq_fops;
85 #define LLI_INODE_MAGIC 0x111d0de5
86 #define LLI_INODE_DEAD 0xdeadd00d
88 /* remote client permission cache */
89 #define REMOTE_PERM_HASHSIZE 16
91 struct ll_getname_data {
92 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
93 struct lu_fid lgd_fid; /* target fid we are looking for */
94 int lgd_found; /* inode matched? */
97 /* llite setxid/access permission for user on remote client */
98 struct ll_remote_perm {
99 cfs_hlist_node_t lrp_list;
104 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
105 is access permission with
106 lrp_fsuid/lrp_fsgid. */
110 /* MDS has an authority for the Size-on-MDS attributes. */
111 LLIF_MDS_SIZE_LOCK = (1 << 0),
112 /* Epoch close is postponed. */
113 LLIF_EPOCH_PENDING = (1 << 1),
114 /* DONE WRITING is allowed. */
115 LLIF_DONE_WRITING = (1 << 2),
116 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
118 LLIF_SOM_DIRTY = (1 << 3),
119 /* File is contented */
120 LLIF_CONTENDED = (1 << 4),
121 /* Truncate uses server lock for this file */
122 LLIF_SRVLOCK = (1 << 5),
126 struct ll_inode_info {
127 __u32 lli_inode_magic;
131 cfs_spinlock_t lli_lock;
132 struct posix_acl *lli_posix_acl;
134 cfs_hlist_head_t *lli_remote_perms;
135 cfs_mutex_t lli_rmtperm_mutex;
137 /* identifying fields for both metadata and data stacks. */
138 struct lu_fid lli_fid;
139 /* Parent fid for accessing default stripe data on parent directory
140 * for allocating OST objects after a mknod() and later open-by-FID. */
141 struct lu_fid lli_pfid;
143 cfs_list_t lli_close_list;
144 cfs_list_t lli_oss_capas;
145 /* open count currently used by capability only, indicate whether
146 * capability needs renewal */
147 cfs_atomic_t lli_open_count;
148 struct obd_capa *lli_mds_capa;
149 cfs_time_t lli_rmtperm_time;
151 /* handle is to be sent to MDS later on done_writing and setattr.
152 * Open handle data are needed for the recovery to reconstruct
153 * the inode state on the MDS. XXX: recovery is not ready yet. */
154 struct obd_client_handle *lli_pending_och;
156 /* We need all three because every inode may be opened in different
158 struct obd_client_handle *lli_mds_read_och;
159 struct obd_client_handle *lli_mds_write_och;
160 struct obd_client_handle *lli_mds_exec_och;
161 __u64 lli_open_fd_read_count;
162 __u64 lli_open_fd_write_count;
163 __u64 lli_open_fd_exec_count;
164 /* Protects access to och pointers and their usage counters, also
165 * atomicity of check-update of lli_has_smd */
166 cfs_mutex_t lli_och_mutex;
168 struct inode lli_vfs_inode;
170 /* the most recent timestamps obtained from mds */
171 struct ost_lvb lli_lvb;
172 cfs_spinlock_t lli_agl_lock;
174 /* Try to make the d::member and f::member are aligned. Before using
175 * these members, make clear whether it is directory or not. */
179 /* serialize normal readdir and statahead-readdir. */
180 cfs_mutex_t d_readdir_mutex;
182 /* metadata statahead */
183 /* since parent-child threads can share the same @file
184 * struct, "opendir_key" is the token when dir close for
185 * case of parent exit before child -- it is me should
186 * cleanup the dir readahead. */
188 struct ll_statahead_info *d_sai;
190 struct posix_acl *d_def_acl;
191 /* protect statahead stuff. */
192 cfs_spinlock_t d_sa_lock;
193 /* "opendir_pid" is the token when lookup/revalid
194 * -- I am the owner of dir statahead. */
198 #define lli_readdir_mutex u.d.d_readdir_mutex
199 #define lli_opendir_key u.d.d_opendir_key
200 #define lli_sai u.d.d_sai
201 #define lli_sa_pos u.d.d_sa_pos
202 #define lli_def_acl u.d.d_def_acl
203 #define lli_sa_lock u.d.d_sa_lock
204 #define lli_opendir_pid u.d.d_opendir_pid
206 /* for non-directory */
208 cfs_semaphore_t f_size_sem;
209 void *f_size_sem_owner;
210 char *f_symlink_name;
213 * cfs_rw_semaphore_t {
214 * signed long count; // align u.d.d_def_acl
215 * cfs_spinlock_t wait_lock; // align u.d.d_sa_lock
216 * struct list_head wait_list;
219 cfs_rw_semaphore_t f_trunc_sem;
220 cfs_mutex_t f_write_mutex;
222 /* for writepage() only to communicate to fsync */
226 cfs_rw_semaphore_t f_glimpse_sem;
227 cfs_time_t f_glimpse_time;
228 cfs_list_t f_agl_list;
231 * whenever a process try to read/write the file, the
232 * jobid of the process will be saved here, and it'll
233 * be packed into the write PRC when flush later.
235 * so the read/write statistics for jobid will not be
236 * accurate if the file is shared by different jobs.
238 char f_jobid[JOBSTATS_JOBID_SIZE];
241 #define lli_size_sem u.f.f_size_sem
242 #define lli_size_sem_owner u.f.f_size_sem_owner
243 #define lli_symlink_name u.f.f_symlink_name
244 #define lli_maxbytes u.f.f_maxbytes
245 #define lli_trunc_sem u.f.f_trunc_sem
246 #define lli_write_mutex u.f.f_write_mutex
247 #define lli_async_rc u.f.f_async_rc
248 #define lli_write_rc u.f.f_write_rc
249 #define lli_glimpse_sem u.f.f_glimpse_sem
250 #define lli_glimpse_time u.f.f_glimpse_time
251 #define lli_agl_list u.f.f_agl_list
252 #define lli_agl_index u.f.f_agl_index
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;
272 * Locking to guarantee consistency of non-atomic updates to long long i_size,
273 * consistency between file size and KMS.
275 * Implemented by ->lli_size_sem and ->lsm_lock, nested in that order.
278 void ll_inode_size_lock(struct inode *inode);
279 void ll_inode_size_unlock(struct inode *inode);
281 // FIXME: replace the name of this with LL_I to conform to kernel stuff
282 // static inline struct ll_inode_info *LL_I(struct inode *inode)
283 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
285 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
288 /* default to about 40meg of readahead on a given system. That much tied
289 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
290 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - CFS_PAGE_SHIFT))
292 /* default to read-ahead full files smaller than 2MB on the second read */
293 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - CFS_PAGE_SHIFT))
298 RA_STAT_DISTANT_READPAGE,
299 RA_STAT_MISS_IN_WINDOW,
300 RA_STAT_FAILED_GRAB_PAGE,
301 RA_STAT_FAILED_MATCH,
306 RA_STAT_MAX_IN_FLIGHT,
307 RA_STAT_WRONG_GRAB_PAGE,
312 cfs_atomic_t ra_cur_pages;
313 unsigned long ra_max_pages;
314 unsigned long ra_max_pages_per_file;
315 unsigned long ra_max_read_ahead_whole_pages;
318 /* ra_io_arg will be filled in the beginning of ll_readahead with
319 * ras_lock, then the following ll_read_ahead_pages will read RA
320 * pages according to this arg, all the items in this structure are
321 * counted by page index.
324 unsigned long ria_start; /* start offset of read-ahead*/
325 unsigned long ria_end; /* end offset of read-ahead*/
326 /* If stride read pattern is detected, ria_stoff means where
327 * stride read is started. Note: for normal read-ahead, the
328 * value here is meaningless, and also it will not be accessed*/
330 /* ria_length and ria_pages are the length and pages length in the
331 * stride I/O mode. And they will also be used to check whether
332 * it is stride I/O read-ahead in the read-ahead pages*/
333 unsigned long ria_length;
334 unsigned long ria_pages;
337 /* LL_HIST_MAX=32 causes an overflow */
338 #define LL_HIST_MAX 28
339 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
340 #define LL_PROCESS_HIST_MAX 10
341 struct per_process_info {
343 struct obd_histogram pp_r_hist;
344 struct obd_histogram pp_w_hist;
347 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
348 struct ll_rw_extents_info {
349 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
352 #define LL_OFFSET_HIST_MAX 100
353 struct ll_rw_process_info {
356 loff_t rw_range_start;
358 loff_t rw_last_file_pos;
360 size_t rw_smallest_extent;
361 size_t rw_largest_extent;
362 struct ll_file_data *rw_last_file;
365 enum stats_track_type {
366 STATS_TRACK_ALL = 0, /* track all processes */
367 STATS_TRACK_PID, /* track process with this pid */
368 STATS_TRACK_PPID, /* track processes with this ppid */
369 STATS_TRACK_GID, /* track processes with this gid */
373 /* flags for sbi->ll_flags */
374 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
375 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
376 #define LL_SBI_FLOCK 0x04
377 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
378 #define LL_SBI_ACL 0x10 /* support ACL */
379 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
380 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
381 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
382 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
383 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
384 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
385 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
386 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
387 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
388 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
389 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
391 /* default value for ll_sb_info->contention_time */
392 #define SBI_DEFAULT_CONTENTION_SECONDS 60
393 /* default value for lockless_truncate_enable */
394 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
395 #define RCE_HASHES 32
397 struct rmtacl_ctl_entry {
399 pid_t rce_key; /* hash key */
400 int rce_ops; /* acl operation type */
403 struct rmtacl_ctl_table {
404 cfs_spinlock_t rct_lock;
405 cfs_list_t rct_entries[RCE_HASHES];
412 pid_t ee_key; /* hash key */
413 struct lu_fid ee_fid;
414 int ee_type; /* ACL type for ACCESS or DEFAULT */
415 ext_acl_xattr_header *ee_acl;
419 cfs_spinlock_t et_lock;
420 cfs_list_t et_entries[EE_HASHES];
425 /* this protects pglist and ra_info. It isn't safe to
426 * grab from interrupt contexts */
427 cfs_spinlock_t ll_lock;
428 cfs_spinlock_t ll_pp_extent_lock; /* Lock for pp_extent entries */
429 cfs_spinlock_t ll_process_lock; /* Lock for ll_rw_process_info */
430 struct obd_uuid ll_sb_uuid;
431 struct obd_export *ll_md_exp;
432 struct obd_export *ll_dt_exp;
433 struct proc_dir_entry* ll_proc_root;
434 struct lu_fid ll_root_fid; /* root object fid */
437 cfs_list_t ll_conn_chain; /* per-conn chain of SBs */
438 struct lustre_client_ocd ll_lco;
440 cfs_list_t ll_orphan_dentry_list; /*please don't ask -p*/
441 struct ll_close_queue *ll_lcq;
443 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
445 unsigned long ll_async_page_max;
446 unsigned long ll_async_page_count;
448 struct lprocfs_stats *ll_ra_stats;
450 struct ll_ra_info ll_ra_info;
451 unsigned int ll_namelen;
452 struct file_operations *ll_fop;
454 /* =0 - hold lock over whole read/write
455 * >0 - max. chunk to be read/written w/o lock re-acquiring */
456 unsigned long ll_max_rw_chunk;
457 unsigned int ll_md_brw_size; /* used by readdir */
459 struct lu_site *ll_site;
460 struct cl_device *ll_cl;
462 struct ll_rw_extents_info ll_rw_extents_info;
463 int ll_extent_process_count;
464 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
465 unsigned int ll_offset_process_count;
466 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
467 unsigned int ll_rw_offset_entry_count;
468 int ll_stats_track_id;
469 enum stats_track_type ll_stats_track_type;
472 /* metadata stat-ahead */
473 unsigned int ll_sa_max; /* max statahead RPCs */
474 atomic_t ll_sa_total; /* statahead thread started
476 atomic_t ll_sa_wrong; /* statahead thread stopped for
478 atomic_t ll_agl_total; /* AGL thread started count */
480 dev_t ll_sdev_orig; /* save s_dev before assign for
482 struct rmtacl_ctl_table ll_rct;
483 struct eacl_table ll_et;
484 struct vfsmount *ll_mnt;
487 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
492 struct task_struct *lrr_reader;
493 cfs_list_t lrr_linkage;
497 * per file-descriptor read-ahead data.
499 struct ll_readahead_state {
500 cfs_spinlock_t ras_lock;
502 * index of the last page that read(2) needed and that wasn't in the
503 * cache. Used by ras_update() to detect seeks.
505 * XXX nikita: if access seeks into cached region, Lustre doesn't see
508 unsigned long ras_last_readpage;
510 * number of pages read after last read-ahead window reset. As window
511 * is reset on each seek, this is effectively a number of consecutive
512 * accesses. Maybe ->ras_accessed_in_window is better name.
514 * XXX nikita: window is also reset (by ras_update()) when Lustre
515 * believes that memory pressure evicts read-ahead pages. In that
516 * case, it probably doesn't make sense to expand window to
517 * PTLRPC_MAX_BRW_PAGES on the third access.
519 unsigned long ras_consecutive_pages;
521 * number of read requests after the last read-ahead window reset
522 * As window is reset on each seek, this is effectively the number
523 * on consecutive read request and is used to trigger read-ahead.
525 unsigned long ras_consecutive_requests;
527 * Parameters of current read-ahead window. Handled by
528 * ras_update(). On the initial access to the file or after a seek,
529 * window is reset to 0. After 3 consecutive accesses, window is
530 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
531 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
533 unsigned long ras_window_start, ras_window_len;
535 * Where next read-ahead should start at. This lies within read-ahead
536 * window. Read-ahead window is read in pieces rather than at once
537 * because: 1. lustre limits total number of pages under read-ahead by
538 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
539 * not covered by DLM lock.
541 unsigned long ras_next_readahead;
543 * Total number of ll_file_read requests issued, reads originating
544 * due to mmap are not counted in this total. This value is used to
545 * trigger full file read-ahead after multiple reads to a small file.
547 unsigned long ras_requests;
549 * Page index with respect to the current request, these value
550 * will not be accurate when dealing with reads issued via mmap.
552 unsigned long ras_request_index;
554 * list of struct ll_ra_read's one per read(2) call current in
555 * progress against this file descriptor. Used by read-ahead code,
556 * protected by ->ras_lock.
558 cfs_list_t ras_read_beads;
560 * The following 3 items are used for detecting the stride I/O
562 * In stride I/O mode,
563 * ...............|-----data-----|****gap*****|--------|******|....
564 * offset |-stride_pages-|-stride_gap-|
565 * ras_stride_offset = offset;
566 * ras_stride_length = stride_pages + stride_gap;
567 * ras_stride_pages = stride_pages;
568 * Note: all these three items are counted by pages.
570 unsigned long ras_stride_length;
571 unsigned long ras_stride_pages;
572 pgoff_t ras_stride_offset;
574 * number of consecutive stride request count, and it is similar as
575 * ras_consecutive_requests, but used for stride I/O mode.
576 * Note: only more than 2 consecutive stride request are detected,
577 * stride read-ahead will be enable
579 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;
593 struct ll_file_dir fd_dir;
595 struct file *fd_file;
598 struct lov_stripe_md;
600 extern cfs_spinlock_t inode_lock;
602 extern struct proc_dir_entry *proc_lustre_fs_root;
604 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
606 return &lli->lli_vfs_inode;
610 struct inode *icbd_parent;
611 struct dentry **icbd_childp;
615 __u32 ll_i2suppgid(struct inode *i);
616 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
618 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
620 #if BITS_PER_LONG == 32
623 return unlikely(cfs_curproc_is_32bit() || (sbi->ll_flags & LL_SBI_32BIT_API));
627 #define LLAP_MAGIC 98764321
629 extern cfs_mem_cache_t *ll_async_page_slab;
630 extern size_t ll_async_page_slab_size;
632 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
633 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
634 struct ll_ra_read *ll_ra_read_get(struct file *f);
636 /* llite/lproc_llite.c */
638 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
639 struct super_block *sb, char *osc, char *mdc);
640 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
641 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
642 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
644 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
645 struct super_block *sb, char *osc, char *mdc){return 0;}
646 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
647 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
648 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
650 memset(lvars, 0, sizeof(*lvars));
656 void ll_release_page(struct page *page, int remove);
657 extern struct file_operations ll_dir_operations;
658 extern struct inode_operations ll_dir_inode_operations;
659 struct page *ll_get_dir_page(struct file *filp, struct inode *dir, __u64 hash,
660 struct ll_dir_chain *chain);
661 int ll_readdir(struct file *filp, void *cookie, filldir_t filldir);
663 int ll_get_mdt_idx(struct inode *inode);
665 int ll_objects_destroy(struct ptlrpc_request *request,
667 struct inode *ll_iget(struct super_block *sb, ino_t hash,
668 struct lustre_md *lic);
669 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
670 void *data, int flag);
671 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
673 int ll_lookup_it_finish(struct ptlrpc_request *request,
674 struct lookup_intent *it, void *data);
675 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
678 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
679 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
680 int ll_writepage(struct page *page, struct writeback_control *wbc);
681 int ll_writepages(struct address_space *, struct writeback_control *wbc);
682 void ll_removepage(struct page *page);
683 int ll_readpage(struct file *file, struct page *page);
684 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
685 void ll_truncate(struct inode *inode);
686 int ll_file_punch(struct inode *, loff_t, int);
687 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
688 void ll_clear_file_contended(struct inode*);
689 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
690 int ll_readahead(const struct lu_env *env, struct cl_io *io,
691 struct ll_readahead_state *ras, struct address_space *mapping,
692 struct cl_page_list *queue, int flags);
695 extern struct file_operations ll_file_operations;
696 extern struct file_operations ll_file_operations_flock;
697 extern struct file_operations ll_file_operations_noflock;
698 extern struct inode_operations ll_file_inode_operations;
699 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
701 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
702 ldlm_mode_t l_req_mode);
703 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
704 struct lustre_handle *lockh);
705 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
707 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
708 int ll_file_open(struct inode *inode, struct file *file);
709 int ll_file_release(struct inode *inode, struct file *file);
710 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
711 struct lov_stripe_md *lsm, lstat_t *st);
712 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
713 int ll_local_open(struct file *file,
714 struct lookup_intent *it, struct ll_file_data *fd,
715 struct obd_client_handle *och);
716 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
717 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
719 int ll_md_real_close(struct inode *inode, int flags);
720 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
721 struct obd_client_handle **och, unsigned long flags);
722 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
723 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
724 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
725 __u64 ioepoch, int sync);
726 int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data,
727 struct md_open_data **mod);
728 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
729 struct lustre_handle *fh);
730 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
731 struct ll_file_data *file, loff_t pos,
732 size_t count, int rw);
733 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
734 struct lookup_intent *it, struct kstat *stat);
735 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
736 struct ll_file_data *ll_file_data_get(void);
737 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
738 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
740 # ifndef HAVE_INODE_PERMISION_2ARGS
741 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
743 int ll_inode_permission(struct inode *inode, int mask);
746 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
747 int flags, struct lov_user_md *lum,
749 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
750 struct lov_mds_md **lmm, int *lmm_size,
751 struct ptlrpc_request **request);
752 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
754 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmm,
755 int *lmm_size, struct ptlrpc_request **request);
756 #ifdef HAVE_FILE_FSYNC_4ARGS
757 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
758 #elif defined(HAVE_FILE_FSYNC_2ARGS)
759 int ll_fsync(struct file *file, int data);
761 int ll_fsync(struct file *file, struct dentry *dentry, int data);
763 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
765 int ll_merge_lvb(struct inode *inode);
766 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
767 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
768 int ll_fid2path(struct obd_export *exp, void *arg);
772 int ll_dops_init(struct dentry *de, int block, int init_sa);
773 extern struct dentry_operations ll_d_ops;
774 void ll_intent_drop_lock(struct lookup_intent *);
775 void ll_intent_release(struct lookup_intent *);
776 void ll_invalidate_aliases(struct inode *);
777 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
778 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
779 #ifdef HAVE_D_COMPARE_7ARGS
780 int ll_dcompare(const struct dentry *parent, const struct inode *pinode,
781 const struct dentry *dentry, const struct inode *inode,
782 unsigned int len, const char *str, const struct qstr *d_name);
784 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
786 int ll_revalidate_it_finish(struct ptlrpc_request *request,
787 struct lookup_intent *it, struct dentry *de);
789 /* llite/llite_lib.c */
790 extern struct super_operations lustre_super_operations;
792 char *ll_read_opt(const char *opt, char *data);
793 void ll_lli_init(struct ll_inode_info *lli);
794 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
795 void ll_put_super(struct super_block *sb);
796 void ll_kill_super(struct super_block *sb);
797 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
798 void ll_clear_inode(struct inode *inode);
799 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr);
800 int ll_setattr(struct dentry *de, struct iattr *attr);
801 #ifndef HAVE_STATFS_DENTRY_PARAM
802 int ll_statfs(struct super_block *sb, struct kstatfs *sfs);
804 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
806 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
807 __u64 max_age, __u32 flags);
808 void ll_update_inode(struct inode *inode, struct lustre_md *md);
809 void ll_read_inode2(struct inode *inode, void *opaque);
810 void ll_delete_inode(struct inode *inode);
811 int ll_iocontrol(struct inode *inode, struct file *file,
812 unsigned int cmd, unsigned long arg);
813 int ll_flush_ctx(struct inode *inode);
814 #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT
815 void ll_umount_begin(struct vfsmount *vfsmnt, int flags);
817 void ll_umount_begin(struct super_block *sb);
819 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
820 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
821 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
822 struct super_block *);
823 void lustre_dump_dentry(struct dentry *, int recur);
824 void lustre_dump_inode(struct inode *);
825 int ll_obd_statfs(struct inode *inode, void *arg);
826 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
827 int ll_process_config(struct lustre_cfg *lcfg);
828 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
829 struct inode *i1, struct inode *i2,
830 const char *name, int namelen,
831 int mode, __u32 opc, void *data);
832 void ll_finish_md_op_data(struct md_op_data *op_data);
833 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
835 /* llite/llite_nfs.c */
836 extern struct export_operations lustre_export_operations;
837 __u32 get_uuid2int(const char *name, int len);
839 /* llite/special.c */
840 extern struct inode_operations ll_special_inode_operations;
841 extern struct file_operations ll_special_chr_inode_fops;
842 extern struct file_operations ll_special_chr_file_fops;
843 extern struct file_operations ll_special_blk_inode_fops;
844 extern struct file_operations ll_special_fifo_inode_fops;
845 extern struct file_operations ll_special_fifo_file_fops;
846 extern struct file_operations ll_special_sock_inode_fops;
848 /* llite/symlink.c */
849 extern struct inode_operations ll_fast_symlink_inode_operations;
851 /* llite/llite_close.c */
852 struct ll_close_queue {
853 cfs_spinlock_t lcq_lock;
855 cfs_waitq_t lcq_waitq;
856 cfs_completion_t lcq_comp;
857 cfs_atomic_t lcq_stop;
860 struct ccc_object *cl_inode2ccc(struct inode *inode);
863 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
864 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
866 /* specific achitecture can implement only part of this list */
867 enum vvp_io_subtype {
870 /** io called from .sendfile */
872 /** io started from splice_{read|write} */
879 enum vvp_io_subtype cui_io_subtype;
883 read_actor_t cui_actor;
887 struct pipe_inode_info *cui_pipe;
888 unsigned int cui_flags;
890 struct vvp_fault_io {
892 * Inode modification time that is checked across DLM
896 struct vm_area_struct *ft_vma;
898 * locked page returned from vvp_io
900 cfs_page_t *ft_vmpage;
901 #ifndef HAVE_VM_OP_FAULT
902 struct vm_nopage_api {
904 * Virtual address at which fault occurred.
906 unsigned long ft_address;
908 * Fault type, as to be supplied to
914 struct vm_fault_api {
918 struct vm_fault *ft_vmf;
920 * fault API used bitflags for return code.
922 unsigned int ft_flags;
928 * Read-ahead state used by read and page-fault IO contexts.
930 struct ll_ra_read cui_bead;
932 * Set when cui_bead has been initialized.
934 int cui_ra_window_set;
936 * Partially truncated page, that vvp_io_trunc_start() keeps locked
939 struct cl_page *cui_partpage;
943 * IO arguments for various VFS I/O interfaces.
946 /** normal/sendfile/splice */
947 enum vvp_io_subtype via_io_subtype;
951 #ifndef HAVE_FILE_WRITEV
952 struct kiocb *via_iocb;
954 struct iovec *via_iov;
955 unsigned long via_nrsegs;
958 read_actor_t via_actor;
962 struct pipe_inode_info *via_pipe;
963 unsigned int via_flags;
968 struct ll_cl_context {
970 struct cl_io *lcc_io;
971 struct cl_page *lcc_page;
972 struct lu_env *lcc_env;
977 struct vvp_thread_info {
978 struct ost_lvb vti_lvb;
979 struct cl_2queue vti_queue;
980 struct iovec vti_local_iov;
981 struct vvp_io_args vti_args;
982 struct ra_io_arg vti_ria;
983 struct kiocb vti_kiocb;
984 struct ll_cl_context vti_io_ctx;
987 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
989 extern struct lu_context_key vvp_key;
990 struct vvp_thread_info *info;
992 info = lu_context_key_get(&env->le_ctx, &vvp_key);
993 LASSERT(info != NULL);
997 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
998 enum vvp_io_subtype type)
1000 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
1002 ret->via_io_subtype = type;
1007 struct vvp_session {
1008 struct vvp_io vs_ios;
1011 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1013 extern struct lu_context_key vvp_session_key;
1014 struct vvp_session *ses;
1016 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1017 LASSERT(ses != NULL);
1021 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1023 return &vvp_env_session(env)->vs_ios;
1026 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1027 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1028 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1030 /* llite/llite_mmap.c */
1031 typedef struct rb_root rb_root_t;
1032 typedef struct rb_node rb_node_t;
1034 struct ll_lock_tree_node;
1035 struct ll_lock_tree {
1037 cfs_list_t lt_locked_list;
1038 struct ll_file_data *lt_fd;
1041 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1042 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1043 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
1044 __u64 end, ldlm_mode_t mode);
1045 void policy_from_vma(ldlm_policy_data_t *policy,
1046 struct vm_area_struct *vma, unsigned long addr, size_t count);
1047 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1050 static inline void ll_invalidate_page(struct page *vmpage)
1052 struct address_space *mapping = vmpage->mapping;
1053 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1055 LASSERT(PageLocked(vmpage));
1056 if (mapping == NULL)
1059 ll_teardown_mmaps(mapping, offset, offset + CFS_PAGE_SIZE);
1060 truncate_complete_page(mapping, vmpage);
1063 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1065 /* don't need an addref as the sb_info should be holding one */
1066 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1068 return ll_s2sbi(sb)->ll_dt_exp;
1071 /* don't need an addref as the sb_info should be holding one */
1072 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1074 return ll_s2sbi(sb)->ll_md_exp;
1077 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1079 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1085 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1086 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1088 return ll_s2sbi(inode->i_sb);
1091 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1093 return ll_s2dtexp(inode->i_sb);
1096 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1098 return ll_s2mdexp(inode->i_sb);
1101 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1104 LASSERT(inode != NULL);
1105 fid = &ll_i2info(inode)->lli_fid;
1106 LASSERT(fid_is_igif(fid) || fid_ver(fid) == 0);
1110 static inline int ll_mds_max_easize(struct super_block *sb)
1112 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1115 static inline __u64 ll_file_maxbytes(struct inode *inode)
1117 return ll_i2info(inode)->lli_maxbytes;
1121 int ll_setxattr(struct dentry *dentry, const char *name,
1122 const void *value, size_t size, int flags);
1123 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1124 void *buffer, size_t size);
1125 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1126 int ll_removexattr(struct dentry *dentry, const char *name);
1128 /* llite/remote_perm.c */
1129 extern cfs_mem_cache_t *ll_remote_perm_cachep;
1130 extern cfs_mem_cache_t *ll_rmtperm_hash_cachep;
1132 cfs_hlist_head_t *alloc_rmtperm_hash(void);
1133 void free_rmtperm_hash(cfs_hlist_head_t *hash);
1134 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1135 int lustre_check_remote_perm(struct inode *inode, int mask);
1137 /* llite/llite_capa.c */
1138 extern cfs_timer_t ll_capa_timer;
1140 int ll_capa_thread_start(void);
1141 void ll_capa_thread_stop(void);
1142 void ll_capa_timer_callback(unsigned long unused);
1144 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1145 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1147 void ll_capa_open(struct inode *inode);
1148 void ll_capa_close(struct inode *inode);
1150 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1151 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1153 void ll_truncate_free_capa(struct obd_capa *ocapa);
1154 void ll_clear_inode_capas(struct inode *inode);
1155 void ll_print_capa_stat(struct ll_sb_info *sbi);
1157 /* llite/llite_cl.c */
1158 extern struct lu_device_type vvp_device_type;
1161 * Common IO arguments for various VFS I/O interfaces.
1163 int cl_sb_init(struct super_block *sb);
1164 int cl_sb_fini(struct super_block *sb);
1165 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1166 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1168 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1169 struct ll_readahead_state *ras, unsigned long index,
1171 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1172 int ll_is_file_contended(struct file *file);
1173 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1175 /* llite/llite_rmtacl.c */
1176 #ifdef CONFIG_FS_POSIX_ACL
1177 obd_valid rce_ops2valid(int ops);
1178 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1179 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1180 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1181 void rct_init(struct rmtacl_ctl_table *rct);
1182 void rct_fini(struct rmtacl_ctl_table *rct);
1184 void ee_free(struct eacl_entry *ee);
1185 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1186 ext_acl_xattr_header *header);
1187 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1188 struct lu_fid *fid, int type);
1189 void et_search_free(struct eacl_table *et, pid_t key);
1190 void et_init(struct eacl_table *et);
1191 void et_fini(struct eacl_table *et);
1196 #define LL_SA_RPC_MIN 2
1197 #define LL_SA_RPC_DEF 32
1198 #define LL_SA_RPC_MAX 8192
1200 #define LL_SA_CACHE_BIT 5
1201 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1202 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1204 /* per inode struct, for dir only */
1205 struct ll_statahead_info {
1206 struct inode *sai_inode;
1207 cfs_atomic_t sai_refcount; /* when access this struct, hold
1209 unsigned int sai_generation; /* generation for statahead */
1210 unsigned int sai_max; /* max ahead of lookup */
1211 __u64 sai_sent; /* stat requests sent count */
1212 __u64 sai_replied; /* stat requests which received
1214 __u64 sai_index; /* index of statahead entry */
1215 __u64 sai_index_wait; /* index of entry which is the
1216 * caller is waiting for */
1217 __u64 sai_hit; /* hit count */
1218 __u64 sai_miss; /* miss count:
1219 * for "ls -al" case, it includes
1220 * hidden dentry miss;
1221 * for "ls -l" case, it does not
1222 * include hidden dentry miss.
1223 * "sai_miss_hidden" is used for
1226 unsigned int sai_consecutive_miss; /* consecutive miss */
1227 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1228 * is not a hidden one */
1229 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1230 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1232 sai_in_readpage:1,/* statahead is in readdir()*/
1233 sai_agl_valid:1;/* AGL is valid for the dir */
1234 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
1235 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1236 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1237 cfs_list_t sai_entries_sent; /* entries sent out */
1238 cfs_list_t sai_entries_received; /* entries returned */
1239 cfs_list_t sai_entries_stated; /* entries stated */
1240 cfs_list_t sai_entries_agl; /* AGL entries to be sent */
1241 cfs_list_t sai_cache[LL_SA_CACHE_SIZE];
1242 cfs_spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1243 cfs_atomic_t sai_cache_count; /* entry count in cache */
1246 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1248 void ll_stop_statahead(struct inode *dir, void *key);
1250 static inline int ll_glimpse_size(struct inode *inode)
1252 struct ll_inode_info *lli = ll_i2info(inode);
1255 cfs_down_read(&lli->lli_glimpse_sem);
1256 rc = cl_glimpse_size(inode);
1257 lli->lli_glimpse_time = cfs_time_current();
1258 cfs_up_read(&lli->lli_glimpse_sem);
1263 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1265 struct ll_inode_info *lli = ll_i2info(dir);
1266 struct ll_statahead_info *sai = lli->lli_sai;
1267 struct ll_dentry_data *ldd = ll_d2d(dentry);
1269 /* not the same process, don't mark */
1270 if (lli->lli_opendir_pid != cfs_curproc_pid())
1273 if (sai != NULL && ldd != NULL)
1274 ldd->lld_sa_generation = sai->sai_generation;
1278 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1280 struct ll_inode_info *lli;
1281 struct ll_dentry_data *ldd;
1283 if (ll_i2sbi(dir)->ll_sa_max == 0)
1286 lli = ll_i2info(dir);
1287 /* not the same process, don't statahead */
1288 if (lli->lli_opendir_pid != cfs_curproc_pid())
1291 ldd = ll_d2d(*dentryp);
1293 * When stats a dentry, the system trigger more than once "revalidate"
1294 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1295 * Under patchless client mode, the operation intent is not accurate,
1296 * which maybe misguide the statahead thread. For example:
1297 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1298 * have the same operation intent -- "IT_GETATTR".
1299 * In fact, one dentry should has only one chance to interact with the
1300 * statahead thread, otherwise the statahead windows will be confused.
1301 * The solution is as following:
1302 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1303 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1304 * will bypass interacting with statahead thread for checking:
1305 * "lld_sa_generation == lli_sai->sai_generation"
1307 if (ldd && lli->lli_sai &&
1308 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1311 return do_statahead_enter(dir, dentryp, only_unplug);
1314 /* llite ioctl register support rountine */
1321 #define LLIOC_MAX_CMD 256
1324 * Rules to write a callback function:
1327 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1328 * returned to ll_iocontrol_register. Callback functions should use this
1329 * data to check the potential collasion of ioctl cmd. If collasion is
1330 * found, callback function should return LLIOC_CONT.
1331 * @rcp: The result of ioctl command.
1334 * If @magic matches the pointer returned by ll_iocontrol_data, the
1335 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1337 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1338 struct file *file, unsigned int cmd, unsigned long arg,
1339 void *magic, int *rcp);
1341 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1342 unsigned int cmd, unsigned long arg, int *rcp);
1344 /* export functions */
1345 /* Register ioctl block dynamatically for a regular file.
1347 * @cmd: the array of ioctl command set
1348 * @count: number of commands in the @cmd
1349 * @cb: callback function, it will be called if an ioctl command is found to
1350 * belong to the command list @cmd.
1353 * A magic pointer will be returned if success;
1354 * otherwise, NULL will be returned.
1356 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1357 void ll_iocontrol_unregister(void *magic);
1361 /* lclient compat stuff */
1362 #define cl_inode_info ll_inode_info
1363 #define cl_i2info(info) ll_i2info(info)
1364 #define cl_inode_mode(inode) ((inode)->i_mode)
1365 #define cl_i2sbi ll_i2sbi
1367 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1368 const struct iattr *attr)
1370 LASSERT(attr->ia_valid & ATTR_FILE);
1371 return LUSTRE_FPRIVATE(attr->ia_file);
1374 static inline void cl_isize_lock(struct inode *inode)
1376 ll_inode_size_lock(inode);
1379 static inline void cl_isize_unlock(struct inode *inode)
1381 ll_inode_size_unlock(inode);
1384 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1386 LASSERT_SEM_LOCKED(&ll_i2info(inode)->lli_size_sem);
1387 i_size_write(inode, kms);
1390 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1392 ll_inode_size_lock(inode);
1393 i_size_write(inode, kms);
1394 ll_inode_size_unlock(inode);
1397 #define cl_isize_read(inode) i_size_read(inode)
1399 static inline int cl_merge_lvb(struct inode *inode)
1401 return ll_merge_lvb(inode);
1404 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1405 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1406 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1408 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1410 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1411 enum cl_fsync_mode mode);
1413 /** direct write pages */
1414 struct ll_dio_pages {
1415 /** page array to be written. we don't support
1416 * partial pages except the last one. */
1417 struct page **ldp_pages;
1418 /* offset of each page */
1419 loff_t *ldp_offsets;
1420 /** if ldp_offsets is NULL, it means a sequential
1421 * pages to be written, then this is the file offset
1422 * of the * first page. */
1423 loff_t ldp_start_offset;
1424 /** how many bytes are to be written. */
1426 /** # of pages in the array. */
1430 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1433 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1436 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1439 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1440 int rw, struct inode *inode,
1441 struct ll_dio_pages *pv);
1443 static inline int ll_file_nolock(const struct file *file)
1445 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1446 struct inode *inode = file->f_dentry->d_inode;
1448 LASSERT(fd != NULL);
1449 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1450 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1453 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1454 struct lookup_intent *it, __u64 *bits)
1456 if (!it->d.lustre.it_lock_set) {
1457 CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)\n",
1458 inode, inode->i_ino, inode->i_generation);
1459 md_set_lock_data(exp, &it->d.lustre.it_lock_handle,
1460 inode, &it->d.lustre.it_lock_bits);
1461 it->d.lustre.it_lock_set = 1;
1465 *bits = it->d.lustre.it_lock_bits;
1468 static inline void ll_lock_dcache(struct inode *inode)
1470 #ifdef HAVE_DCACHE_LOCK
1471 spin_lock(&dcache_lock);
1473 spin_lock(&inode->i_lock);
1477 static inline void ll_unlock_dcache(struct inode *inode)
1479 #ifdef HAVE_DCACHE_LOCK
1480 spin_unlock(&dcache_lock);
1482 spin_unlock(&inode->i_lock);
1486 static inline int d_lustre_invalid(const struct dentry *dentry)
1488 return dentry->d_flags & DCACHE_LUSTRE_INVALID;
1491 static inline void __d_lustre_invalidate(struct dentry *dentry)
1493 dentry->d_flags |= DCACHE_LUSTRE_INVALID;
1497 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1498 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1499 * else dput() of the last refcount will unhash this dentry and kill it.
1501 static inline void d_lustre_invalidate(struct dentry *dentry)
1503 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1504 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1505 dentry->d_parent, dentry->d_inode, d_refcount(dentry));
1507 spin_lock(&dentry->d_lock);
1508 __d_lustre_invalidate(dentry);
1509 if (d_refcount(dentry) == 0)
1511 spin_unlock(&dentry->d_lock);
1514 static inline void d_lustre_revalidate(struct dentry *dentry)
1516 spin_lock(&dentry->d_lock);
1517 dentry->d_flags &= ~DCACHE_LUSTRE_INVALID;
1518 spin_unlock(&dentry->d_lock);
1521 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2,7,50,0)
1522 /* Compatibility for old (1.8) compiled userspace quota code */
1523 struct if_quotactl_18 {
1528 struct obd_dqinfo qc_dqinfo;
1529 struct obd_dqblk qc_dqblk;
1531 struct obd_uuid obd_uuid;
1533 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1534 /* End compatibility for old (1.8) compiled userspace quota code */
1536 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1537 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2,7,50,0) */
1539 #endif /* LLITE_INTERNAL_H */