4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2012, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
37 #ifndef LLITE_INTERNAL_H
38 #define LLITE_INTERNAL_H
39 #include <lustre_debug.h>
40 #include <lustre_ver.h>
41 #include <lustre_disk.h> /* for s2sbi */
42 #include <lustre_eacl.h>
44 /* for struct cl_lock_descr and struct cl_io */
45 #include <cl_object.h>
47 #include <lustre_mdc.h>
48 #include <linux/lustre_intent.h>
54 #ifndef VM_FAULT_RETRY
55 #define VM_FAULT_RETRY 0
58 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
59 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
60 #ifndef LOOKUP_CONTINUE
61 #define LOOKUP_CONTINUE LOOKUP_PARENT
64 /** Only used on client-side for indicating the tail of dir hash/offset. */
65 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
66 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
68 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
69 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
71 struct ll_dentry_data {
74 struct obd_client_handle lld_cwd_och;
75 struct obd_client_handle lld_mnt_och;
76 struct lookup_intent *lld_it;
77 unsigned int lld_sa_generation;
78 unsigned int lld_invalid:1;
79 struct rcu_head lld_rcu_head;
82 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
84 extern struct file_operations ll_pgcache_seq_fops;
86 #define LLI_INODE_MAGIC 0x111d0de5
87 #define LLI_INODE_DEAD 0xdeadd00d
89 /* remote client permission cache */
90 #define REMOTE_PERM_HASHSIZE 16
92 struct ll_getname_data {
93 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
94 struct lu_fid lgd_fid; /* target fid we are looking for */
95 int lgd_found; /* inode matched? */
98 /* llite setxid/access permission for user on remote client */
99 struct ll_remote_perm {
100 struct hlist_node lrp_list;
105 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
106 is access permission with
107 lrp_fsuid/lrp_fsgid. */
111 /* MDS has an authority for the Size-on-MDS attributes. */
112 LLIF_MDS_SIZE_LOCK = (1 << 0),
113 /* Epoch close is postponed. */
114 LLIF_EPOCH_PENDING = (1 << 1),
115 /* DONE WRITING is allowed. */
116 LLIF_DONE_WRITING = (1 << 2),
117 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
119 LLIF_SOM_DIRTY = (1 << 3),
120 /* File is contented */
121 LLIF_CONTENDED = (1 << 4),
122 /* Truncate uses server lock for this file */
123 LLIF_SRVLOCK = (1 << 5),
124 /* File data is modified. */
125 LLIF_DATA_MODIFIED = (1 << 6),
128 struct ll_inode_info {
129 __u32 lli_inode_magic;
134 struct posix_acl *lli_posix_acl;
136 struct hlist_head *lli_remote_perms;
137 struct mutex lli_rmtperm_mutex;
139 /* identifying fields for both metadata and data stacks. */
140 struct lu_fid lli_fid;
141 /* Parent fid for accessing default stripe data on parent directory
142 * for allocating OST objects after a mknod() and later open-by-FID. */
143 struct lu_fid lli_pfid;
145 struct list_head lli_close_list;
146 struct list_head lli_oss_capas;
147 /* open count currently used by capability only, indicate whether
148 * capability needs renewal */
149 atomic_t lli_open_count;
150 struct obd_capa *lli_mds_capa;
151 cfs_time_t lli_rmtperm_time;
153 /* handle is to be sent to MDS later on done_writing and setattr.
154 * Open handle data are needed for the recovery to reconstruct
155 * the inode state on the MDS. XXX: recovery is not ready yet. */
156 struct obd_client_handle *lli_pending_och;
158 /* We need all three because every inode may be opened in different
160 struct obd_client_handle *lli_mds_read_och;
161 struct obd_client_handle *lli_mds_write_och;
162 struct obd_client_handle *lli_mds_exec_och;
163 __u64 lli_open_fd_read_count;
164 __u64 lli_open_fd_write_count;
165 __u64 lli_open_fd_exec_count;
166 /* Protects access to och pointers and their usage counters */
167 struct mutex lli_och_mutex;
169 struct inode lli_vfs_inode;
171 /* the most recent timestamps obtained from mds */
172 struct ost_lvb lli_lvb;
173 spinlock_t lli_agl_lock;
175 /* Try to make the d::member and f::member are aligned. Before using
176 * these members, make clear whether it is directory or not. */
180 /* serialize normal readdir and statahead-readdir. */
181 struct mutex d_readdir_mutex;
183 /* metadata statahead */
184 /* since parent-child threads can share the same @file
185 * struct, "opendir_key" is the token when dir close for
186 * case of parent exit before child -- it is me should
187 * cleanup the dir readahead. */
189 struct ll_statahead_info *d_sai;
190 struct posix_acl *d_def_acl;
191 /* protect statahead stuff. */
192 spinlock_t d_sa_lock;
193 /* "opendir_pid" is the token when lookup/revalid
194 * -- I am the owner of dir statahead. */
198 #define lli_readdir_mutex u.d.d_readdir_mutex
199 #define lli_opendir_key u.d.d_opendir_key
200 #define lli_sai u.d.d_sai
201 #define lli_def_acl u.d.d_def_acl
202 #define lli_sa_lock u.d.d_sa_lock
203 #define lli_opendir_pid u.d.d_opendir_pid
205 /* for non-directory */
207 struct semaphore f_size_sem;
208 void *f_size_sem_owner;
209 char *f_symlink_name;
212 * struct rw_semaphore {
213 * signed long count; // align d.d_def_acl
214 * spinlock_t wait_lock; // align d.d_sa_lock
215 * struct list_head wait_list;
218 struct rw_semaphore f_trunc_sem;
219 struct mutex f_write_mutex;
221 struct rw_semaphore f_glimpse_sem;
222 cfs_time_t f_glimpse_time;
223 struct list_head f_agl_list;
226 /* for writepage() only to communicate to fsync */
229 /* volatile file criteria is based on file name, this
230 * flag is used to keep the test result, so the strcmp
235 * whenever a process try to read/write the file, the
236 * jobid of the process will be saved here, and it'll
237 * be packed into the write PRC when flush later.
239 * so the read/write statistics for jobid will not be
240 * accurate if the file is shared by different jobs.
242 char f_jobid[JOBSTATS_JOBID_SIZE];
245 #define lli_size_sem u.f.f_size_sem
246 #define lli_size_sem_owner u.f.f_size_sem_owner
247 #define lli_symlink_name u.f.f_symlink_name
248 #define lli_maxbytes u.f.f_maxbytes
249 #define lli_trunc_sem u.f.f_trunc_sem
250 #define lli_write_mutex u.f.f_write_mutex
251 #define lli_glimpse_sem u.f.f_glimpse_sem
252 #define lli_glimpse_time u.f.f_glimpse_time
253 #define lli_agl_list u.f.f_agl_list
254 #define lli_agl_index u.f.f_agl_index
255 #define lli_async_rc u.f.f_async_rc
256 #define lli_jobid u.f.f_jobid
257 #define lli_volatile u.f.f_volatile
261 /* XXX: For following frequent used members, although they maybe special
262 * used for non-directory object, it is some time-wasting to check
263 * whether the object is directory or not before using them. On the
264 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
265 * the "ll_inode_info" size even if moving those members into u.f.
266 * So keep them out side.
268 * In the future, if more members are added only for directory,
269 * some of the following members can be moved into u.f.
272 struct cl_object *lli_clob;
274 /* mutex to request for layout lock exclusively. */
275 struct mutex lli_layout_mutex;
276 /* valid only inside LAYOUT ibits lock, protected by lli_layout_mutex */
277 __u32 lli_layout_gen;
281 * Locking to guarantee consistency of non-atomic updates to long long i_size,
282 * consistency between file size and KMS.
284 * Implemented by ->lli_size_sem and ->lsm_lock, nested in that order.
287 void ll_inode_size_lock(struct inode *inode);
288 void ll_inode_size_unlock(struct inode *inode);
290 // FIXME: replace the name of this with LL_I to conform to kernel stuff
291 // static inline struct ll_inode_info *LL_I(struct inode *inode)
292 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
294 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
297 /* default to about 40meg of readahead on a given system. That much tied
298 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
299 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT))
301 /* default to read-ahead full files smaller than 2MB on the second read */
302 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
307 RA_STAT_DISTANT_READPAGE,
308 RA_STAT_MISS_IN_WINDOW,
309 RA_STAT_FAILED_GRAB_PAGE,
310 RA_STAT_FAILED_MATCH,
315 RA_STAT_MAX_IN_FLIGHT,
316 RA_STAT_WRONG_GRAB_PAGE,
321 atomic_t ra_cur_pages;
322 unsigned long ra_max_pages;
323 unsigned long ra_max_pages_per_file;
324 unsigned long ra_max_read_ahead_whole_pages;
327 /* ra_io_arg will be filled in the beginning of ll_readahead with
328 * ras_lock, then the following ll_read_ahead_pages will read RA
329 * pages according to this arg, all the items in this structure are
330 * counted by page index.
333 unsigned long ria_start; /* start offset of read-ahead*/
334 unsigned long ria_end; /* end offset of read-ahead*/
335 /* If stride read pattern is detected, ria_stoff means where
336 * stride read is started. Note: for normal read-ahead, the
337 * value here is meaningless, and also it will not be accessed*/
339 /* ria_length and ria_pages are the length and pages length in the
340 * stride I/O mode. And they will also be used to check whether
341 * it is stride I/O read-ahead in the read-ahead pages*/
342 unsigned long ria_length;
343 unsigned long ria_pages;
346 /* LL_HIST_MAX=32 causes an overflow */
347 #define LL_HIST_MAX 28
348 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
349 #define LL_PROCESS_HIST_MAX 10
350 struct per_process_info {
352 struct obd_histogram pp_r_hist;
353 struct obd_histogram pp_w_hist;
356 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
357 struct ll_rw_extents_info {
358 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
361 #define LL_OFFSET_HIST_MAX 100
362 struct ll_rw_process_info {
365 loff_t rw_range_start;
367 loff_t rw_last_file_pos;
369 size_t rw_smallest_extent;
370 size_t rw_largest_extent;
371 struct ll_file_data *rw_last_file;
374 enum stats_track_type {
375 STATS_TRACK_ALL = 0, /* track all processes */
376 STATS_TRACK_PID, /* track process with this pid */
377 STATS_TRACK_PPID, /* track processes with this ppid */
378 STATS_TRACK_GID, /* track processes with this gid */
382 /* flags for sbi->ll_flags */
383 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
384 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
385 #define LL_SBI_FLOCK 0x04
386 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
387 #define LL_SBI_ACL 0x10 /* support ACL */
388 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
389 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
390 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
391 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
392 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
393 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
394 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
395 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
396 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
397 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
398 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
399 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
400 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
402 #define LL_SBI_FLAGS { \
422 /* default value for ll_sb_info->contention_time */
423 #define SBI_DEFAULT_CONTENTION_SECONDS 60
424 /* default value for lockless_truncate_enable */
425 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
426 #define RCE_HASHES 32
428 struct rmtacl_ctl_entry {
429 struct list_head rce_list;
430 pid_t rce_key; /* hash key */
431 int rce_ops; /* acl operation type */
434 struct rmtacl_ctl_table {
436 struct list_head rct_entries[RCE_HASHES];
442 struct list_head ee_list;
443 pid_t ee_key; /* hash key */
444 struct lu_fid ee_fid;
445 int ee_type; /* ACL type for ACCESS or DEFAULT */
446 ext_acl_xattr_header *ee_acl;
451 struct list_head et_entries[EE_HASHES];
455 struct list_head ll_list;
456 /* this protects pglist and ra_info. It isn't safe to
457 * grab from interrupt contexts */
459 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
460 spinlock_t ll_process_lock; /* ll_rw_process_info */
461 struct obd_uuid ll_sb_uuid;
462 struct obd_export *ll_md_exp;
463 struct obd_export *ll_dt_exp;
464 struct proc_dir_entry* ll_proc_root;
465 struct lu_fid ll_root_fid; /* root object fid */
469 struct list_head ll_conn_chain; /* per-conn chain of SBs */
470 struct lustre_client_ocd ll_lco;
472 struct list_head ll_orphan_dentry_list; /*please don't ask -p*/
473 struct ll_close_queue *ll_lcq;
475 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
477 struct cl_client_cache ll_cache;
479 struct lprocfs_stats *ll_ra_stats;
481 struct ll_ra_info ll_ra_info;
482 unsigned int ll_namelen;
483 struct file_operations *ll_fop;
485 /* =0 - hold lock over whole read/write
486 * >0 - max. chunk to be read/written w/o lock re-acquiring */
487 unsigned long ll_max_rw_chunk;
488 unsigned int ll_md_brw_size; /* used by readdir */
490 struct lu_site *ll_site;
491 struct cl_device *ll_cl;
493 struct ll_rw_extents_info ll_rw_extents_info;
494 int ll_extent_process_count;
495 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
496 unsigned int ll_offset_process_count;
497 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
498 unsigned int ll_rw_offset_entry_count;
499 int ll_stats_track_id;
500 enum stats_track_type ll_stats_track_type;
503 /* metadata stat-ahead */
504 unsigned int ll_sa_max; /* max statahead RPCs */
505 atomic_t ll_sa_total; /* statahead thread started
507 atomic_t ll_sa_wrong; /* statahead thread stopped for
509 atomic_t ll_agl_total; /* AGL thread started count */
511 dev_t ll_sdev_orig; /* save s_dev before assign for
513 struct rmtacl_ctl_table ll_rct;
514 struct eacl_table ll_et;
517 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
522 struct task_struct *lrr_reader;
523 struct list_head lrr_linkage;
527 * per file-descriptor read-ahead data.
529 struct ll_readahead_state {
532 * index of the last page that read(2) needed and that wasn't in the
533 * cache. Used by ras_update() to detect seeks.
535 * XXX nikita: if access seeks into cached region, Lustre doesn't see
538 unsigned long ras_last_readpage;
540 * number of pages read after last read-ahead window reset. As window
541 * is reset on each seek, this is effectively a number of consecutive
542 * accesses. Maybe ->ras_accessed_in_window is better name.
544 * XXX nikita: window is also reset (by ras_update()) when Lustre
545 * believes that memory pressure evicts read-ahead pages. In that
546 * case, it probably doesn't make sense to expand window to
547 * PTLRPC_MAX_BRW_PAGES on the third access.
549 unsigned long ras_consecutive_pages;
551 * number of read requests after the last read-ahead window reset
552 * As window is reset on each seek, this is effectively the number
553 * on consecutive read request and is used to trigger read-ahead.
555 unsigned long ras_consecutive_requests;
557 * Parameters of current read-ahead window. Handled by
558 * ras_update(). On the initial access to the file or after a seek,
559 * window is reset to 0. After 3 consecutive accesses, window is
560 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
561 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
563 unsigned long ras_window_start, ras_window_len;
565 * Where next read-ahead should start at. This lies within read-ahead
566 * window. Read-ahead window is read in pieces rather than at once
567 * because: 1. lustre limits total number of pages under read-ahead by
568 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
569 * not covered by DLM lock.
571 unsigned long ras_next_readahead;
573 * Total number of ll_file_read requests issued, reads originating
574 * due to mmap are not counted in this total. This value is used to
575 * trigger full file read-ahead after multiple reads to a small file.
577 unsigned long ras_requests;
579 * Page index with respect to the current request, these value
580 * will not be accurate when dealing with reads issued via mmap.
582 unsigned long ras_request_index;
584 * list of struct ll_ra_read's one per read(2) call current in
585 * progress against this file descriptor. Used by read-ahead code,
586 * protected by ->ras_lock.
588 struct list_head ras_read_beads;
590 * The following 3 items are used for detecting the stride I/O
592 * In stride I/O mode,
593 * ...............|-----data-----|****gap*****|--------|******|....
594 * offset |-stride_pages-|-stride_gap-|
595 * ras_stride_offset = offset;
596 * ras_stride_length = stride_pages + stride_gap;
597 * ras_stride_pages = stride_pages;
598 * Note: all these three items are counted by pages.
600 unsigned long ras_stride_length;
601 unsigned long ras_stride_pages;
602 pgoff_t ras_stride_offset;
604 * number of consecutive stride request count, and it is similar as
605 * ras_consecutive_requests, but used for stride I/O mode.
606 * Note: only more than 2 consecutive stride request are detected,
607 * stride read-ahead will be enable
609 unsigned long ras_consecutive_stride_requests;
612 extern struct kmem_cache *ll_file_data_slab;
613 struct lustre_handle;
614 struct ll_file_data {
615 struct ll_readahead_state fd_ras;
617 struct ccc_grouplock fd_grouplock;
620 struct file *fd_file;
621 /* Indicate whether need to report failure when close.
622 * true: failure is known, not report again.
623 * false: unknown failure, should report. */
624 bool fd_write_failed;
627 struct lov_stripe_md;
629 extern spinlock_t inode_lock;
631 extern struct proc_dir_entry *proc_lustre_fs_root;
633 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
635 return &lli->lli_vfs_inode;
639 struct inode *icbd_parent;
640 struct dentry **icbd_childp;
644 __u32 ll_i2suppgid(struct inode *i);
645 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
647 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
649 #if BITS_PER_LONG == 32
652 return unlikely(current_is_32bit() || (sbi->ll_flags & LL_SBI_32BIT_API));
656 #define LLAP_MAGIC 98764321
658 extern struct kmem_cache *ll_async_page_slab;
659 extern size_t ll_async_page_slab_size;
661 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
662 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
663 struct ll_ra_read *ll_ra_read_get(struct file *f);
665 /* llite/lproc_llite.c */
667 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
668 struct super_block *sb, char *osc, char *mdc);
669 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
670 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
671 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
673 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
674 struct super_block *sb, char *osc, char *mdc){return 0;}
675 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
676 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
677 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
679 memset(lvars, 0, sizeof(*lvars));
685 void ll_release_page(struct page *page, int remove);
686 extern struct file_operations ll_dir_operations;
687 extern struct inode_operations ll_dir_inode_operations;
688 struct page *ll_get_dir_page(struct inode *dir, __u64 hash,
689 struct ll_dir_chain *chain);
690 int ll_dir_read(struct inode *inode, __u64 *_pos, void *cookie,
693 int ll_get_mdt_idx(struct inode *inode);
695 int ll_objects_destroy(struct ptlrpc_request *request,
697 struct inode *ll_iget(struct super_block *sb, ino_t hash,
698 struct lustre_md *lic);
699 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
700 void *data, int flag);
701 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
702 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
705 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
706 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
707 int ll_writepage(struct page *page, struct writeback_control *wbc);
708 int ll_writepages(struct address_space *, struct writeback_control *wbc);
709 void ll_removepage(struct page *page);
710 int ll_readpage(struct file *file, struct page *page);
711 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
712 int ll_file_punch(struct inode *, loff_t, int);
713 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
714 void ll_clear_file_contended(struct inode*);
715 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
716 int ll_readahead(const struct lu_env *env, struct cl_io *io,
717 struct ll_readahead_state *ras, struct address_space *mapping,
718 struct cl_page_list *queue, int flags);
721 extern struct file_operations ll_file_operations;
722 extern struct file_operations ll_file_operations_flock;
723 extern struct file_operations ll_file_operations_noflock;
724 extern struct inode_operations ll_file_inode_operations;
725 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
727 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
728 ldlm_mode_t l_req_mode);
729 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
730 struct lustre_handle *lockh, __u64 flags);
731 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
733 int ll_revalidate_nd(struct dentry *dentry, unsigned int flags);
734 int ll_file_open(struct inode *inode, struct file *file);
735 int ll_file_release(struct inode *inode, struct file *file);
736 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
737 struct lov_stripe_md *lsm, lstat_t *st);
738 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
739 int ll_local_open(struct file *file,
740 struct lookup_intent *it, struct ll_file_data *fd,
741 struct obd_client_handle *och);
742 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
743 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
745 int ll_md_real_close(struct inode *inode, int flags);
746 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
747 struct obd_client_handle **och, unsigned long flags);
748 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
749 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
750 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
751 __u64 ioepoch, int sync);
752 int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data,
753 struct md_open_data **mod);
754 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
755 struct lustre_handle *fh);
756 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
757 struct ll_file_data *file, loff_t pos,
758 size_t count, int rw);
759 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
760 struct lookup_intent *it, struct kstat *stat);
761 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
762 struct ll_file_data *ll_file_data_get(void);
763 struct posix_acl * ll_get_acl(struct inode *inode, int type);
765 int ll_inode_permission(struct inode *inode, int mask);
767 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
768 int flags, struct lov_user_md *lum,
770 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
771 struct lov_mds_md **lmm, int *lmm_size,
772 struct ptlrpc_request **request);
773 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
775 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmmp,
776 int *lmm_size, struct ptlrpc_request **request);
777 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
778 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
780 int ll_merge_lvb(const struct lu_env *env, 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);
784 int ll_data_version(struct inode *inode, __u64 *data_version, int extent_lock);
788 int ll_dops_init(struct dentry *de, int block, int init_sa);
789 extern struct dentry_operations ll_d_ops;
790 void ll_intent_drop_lock(struct lookup_intent *);
791 void ll_intent_release(struct lookup_intent *);
792 void ll_invalidate_aliases(struct inode *);
793 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
794 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
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);
798 int ll_revalidate_it_finish(struct ptlrpc_request *request,
799 struct lookup_intent *it, struct dentry *de);
801 /* llite/llite_lib.c */
802 extern struct super_operations lustre_super_operations;
804 char *ll_read_opt(const char *opt, char *data);
805 void ll_lli_init(struct ll_inode_info *lli);
806 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
807 void ll_put_super(struct super_block *sb);
808 void ll_kill_super(struct super_block *sb);
809 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
810 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
811 void ll_clear_inode(struct inode *inode);
812 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr);
813 int ll_setattr(struct dentry *de, struct iattr *attr);
814 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
815 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
816 __u64 max_age, __u32 flags);
817 void ll_update_inode(struct inode *inode, struct lustre_md *md);
818 void ll_read_inode2(struct inode *inode, void *opaque);
819 void ll_delete_inode(struct inode *inode);
820 int ll_iocontrol(struct inode *inode, struct file *file,
821 unsigned int cmd, unsigned long arg);
822 int ll_flush_ctx(struct inode *inode);
823 void ll_umount_begin(struct super_block *sb);
824 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
825 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
826 void ll_dirty_page_discard_warn(struct page *page, int ioret);
827 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
828 struct super_block *, struct lookup_intent *);
829 void lustre_dump_dentry(struct dentry *, int recur);
830 void lustre_dump_inode(struct inode *);
831 int ll_obd_statfs(struct inode *inode, void *arg);
832 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
833 int ll_process_config(struct lustre_cfg *lcfg);
834 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
835 struct inode *i1, struct inode *i2,
836 const char *name, int namelen,
837 int mode, __u32 opc, void *data);
838 void ll_finish_md_op_data(struct md_op_data *op_data);
839 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
840 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
842 /* llite/llite_nfs.c */
843 extern struct export_operations lustre_export_operations;
844 __u32 get_uuid2int(const char *name, int len);
845 struct inode *search_inode_for_lustre(struct super_block *sb,
846 const struct lu_fid *fid);
848 /* llite/special.c */
849 extern struct inode_operations ll_special_inode_operations;
850 extern struct file_operations ll_special_chr_inode_fops;
851 extern struct file_operations ll_special_chr_file_fops;
852 extern struct file_operations ll_special_blk_inode_fops;
853 extern struct file_operations ll_special_fifo_inode_fops;
854 extern struct file_operations ll_special_fifo_file_fops;
855 extern struct file_operations ll_special_sock_inode_fops;
857 /* llite/symlink.c */
858 extern struct inode_operations ll_fast_symlink_inode_operations;
860 /* llite/llite_close.c */
861 struct ll_close_queue {
863 struct list_head lcq_head;
864 wait_queue_head_t lcq_waitq;
865 struct completion lcq_comp;
869 struct ccc_object *cl_inode2ccc(struct inode *inode);
872 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
873 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
875 /* specific achitecture can implement only part of this list */
876 enum vvp_io_subtype {
879 /** io called from .sendfile */
881 /** io started from splice_{read|write} */
888 enum vvp_io_subtype cui_io_subtype;
892 read_actor_t cui_actor;
896 struct pipe_inode_info *cui_pipe;
897 unsigned int cui_flags;
899 struct vvp_fault_io {
901 * Inode modification time that is checked across DLM
905 struct vm_area_struct *ft_vma;
907 * locked page returned from vvp_io
909 struct page *ft_vmpage;
910 struct vm_fault_api {
914 struct vm_fault *ft_vmf;
916 * fault API used bitflags for return code.
918 unsigned int ft_flags;
923 * Read-ahead state used by read and page-fault IO contexts.
925 struct ll_ra_read cui_bead;
927 * Set when cui_bead has been initialized.
929 int cui_ra_window_set;
931 * Partially truncated page, that vvp_io_trunc_start() keeps locked
934 struct cl_page *cui_partpage;
938 * IO arguments for various VFS I/O interfaces.
941 /** normal/sendfile/splice */
942 enum vvp_io_subtype via_io_subtype;
946 struct kiocb *via_iocb;
947 struct iovec *via_iov;
948 unsigned long via_nrsegs;
951 read_actor_t via_actor;
955 struct pipe_inode_info *via_pipe;
956 unsigned int via_flags;
961 struct ll_cl_context {
963 struct cl_io *lcc_io;
964 struct cl_page *lcc_page;
965 struct lu_env *lcc_env;
970 struct vvp_thread_info {
971 struct ost_lvb vti_lvb;
972 struct cl_2queue vti_queue;
973 struct iovec vti_local_iov;
974 struct vvp_io_args vti_args;
975 struct ra_io_arg vti_ria;
976 struct kiocb vti_kiocb;
977 struct ll_cl_context vti_io_ctx;
980 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
982 extern struct lu_context_key vvp_key;
983 struct vvp_thread_info *info;
985 info = lu_context_key_get(&env->le_ctx, &vvp_key);
986 LASSERT(info != NULL);
990 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
991 enum vvp_io_subtype type)
993 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
995 ret->via_io_subtype = type;
1000 struct vvp_session {
1001 struct vvp_io vs_ios;
1004 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1006 extern struct lu_context_key vvp_session_key;
1007 struct vvp_session *ses;
1009 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1010 LASSERT(ses != NULL);
1014 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1016 return &vvp_env_session(env)->vs_ios;
1019 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1020 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1021 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1023 /* llite/llite_mmap.c */
1024 typedef struct rb_root rb_root_t;
1025 typedef struct rb_node rb_node_t;
1027 struct ll_lock_tree_node;
1028 struct ll_lock_tree {
1030 struct list_head lt_locked_list;
1031 struct ll_file_data *lt_fd;
1034 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1035 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1036 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
1037 __u64 end, ldlm_mode_t mode);
1038 void policy_from_vma(ldlm_policy_data_t *policy,
1039 struct vm_area_struct *vma, unsigned long addr, size_t count);
1040 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1043 static inline void ll_invalidate_page(struct page *vmpage)
1045 struct address_space *mapping = vmpage->mapping;
1046 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1048 LASSERT(PageLocked(vmpage));
1049 if (mapping == NULL)
1052 ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE);
1053 truncate_complete_page(mapping, vmpage);
1056 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1058 /* don't need an addref as the sb_info should be holding one */
1059 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1061 return ll_s2sbi(sb)->ll_dt_exp;
1064 /* don't need an addref as the sb_info should be holding one */
1065 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1067 return ll_s2sbi(sb)->ll_md_exp;
1070 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1072 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1078 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1079 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1081 return ll_s2sbi(inode->i_sb);
1084 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1086 return ll_s2dtexp(inode->i_sb);
1089 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1091 return ll_s2mdexp(inode->i_sb);
1094 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1098 LASSERT(inode != NULL);
1099 fid = &ll_i2info(inode)->lli_fid;
1104 static inline int ll_mds_max_easize(struct super_block *sb)
1106 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1109 static inline __u64 ll_file_maxbytes(struct inode *inode)
1111 return ll_i2info(inode)->lli_maxbytes;
1115 int ll_setxattr(struct dentry *dentry, const char *name,
1116 const void *value, size_t size, int flags);
1117 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1118 void *buffer, size_t size);
1119 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1120 int ll_removexattr(struct dentry *dentry, const char *name);
1122 /* llite/remote_perm.c */
1123 extern struct kmem_cache *ll_remote_perm_cachep;
1124 extern struct kmem_cache *ll_rmtperm_hash_cachep;
1126 struct hlist_head *alloc_rmtperm_hash(void);
1127 void free_rmtperm_hash(struct hlist_head *hash);
1128 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1129 int lustre_check_remote_perm(struct inode *inode, int mask);
1131 /* llite/llite_capa.c */
1132 extern timer_list_t ll_capa_timer;
1134 int ll_capa_thread_start(void);
1135 void ll_capa_thread_stop(void);
1136 void ll_capa_timer_callback(unsigned long unused);
1138 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1139 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1141 void ll_capa_open(struct inode *inode);
1142 void ll_capa_close(struct inode *inode);
1144 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1145 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1147 void ll_truncate_free_capa(struct obd_capa *ocapa);
1148 void ll_clear_inode_capas(struct inode *inode);
1149 void ll_print_capa_stat(struct ll_sb_info *sbi);
1151 /* llite/llite_cl.c */
1152 extern struct lu_device_type vvp_device_type;
1155 * Common IO arguments for various VFS I/O interfaces.
1157 int cl_sb_init(struct super_block *sb);
1158 int cl_sb_fini(struct super_block *sb);
1159 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1160 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1162 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1163 struct ll_readahead_state *ras, unsigned long index,
1165 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1166 int ll_is_file_contended(struct file *file);
1167 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1169 /* llite/llite_rmtacl.c */
1170 #ifdef CONFIG_FS_POSIX_ACL
1171 obd_valid rce_ops2valid(int ops);
1172 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1173 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1174 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1175 void rct_init(struct rmtacl_ctl_table *rct);
1176 void rct_fini(struct rmtacl_ctl_table *rct);
1178 void ee_free(struct eacl_entry *ee);
1179 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1180 ext_acl_xattr_header *header);
1181 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1182 struct lu_fid *fid, int type);
1183 void et_search_free(struct eacl_table *et, pid_t key);
1184 void et_init(struct eacl_table *et);
1185 void et_fini(struct eacl_table *et);
1190 #define LL_SA_RPC_MIN 2
1191 #define LL_SA_RPC_DEF 32
1192 #define LL_SA_RPC_MAX 8192
1194 #define LL_SA_CACHE_BIT 5
1195 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1196 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1198 /* per inode struct, for dir only */
1199 struct ll_statahead_info {
1200 struct inode *sai_inode;
1201 atomic_t sai_refcount; /* when access this struct, hold
1203 unsigned int sai_generation; /* generation for statahead */
1204 unsigned int sai_max; /* max ahead of lookup */
1205 __u64 sai_sent; /* stat requests sent count */
1206 __u64 sai_replied; /* stat requests which received
1208 __u64 sai_index; /* index of statahead entry */
1209 __u64 sai_index_wait; /* index of entry which is the
1210 * caller is waiting for */
1211 __u64 sai_hit; /* hit count */
1212 __u64 sai_miss; /* miss count:
1213 * for "ls -al" case, it includes
1214 * hidden dentry miss;
1215 * for "ls -l" case, it does not
1216 * include hidden dentry miss.
1217 * "sai_miss_hidden" is used for
1220 unsigned int sai_consecutive_miss; /* consecutive miss */
1221 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1222 * is not a hidden one */
1223 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1224 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1226 sai_in_readpage:1,/* statahead is in readdir()*/
1227 sai_agl_valid:1;/* AGL is valid for the dir */
1228 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1229 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1230 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1231 struct list_head sai_entries; /* entry list */
1232 struct list_head sai_entries_received; /* entries returned */
1233 struct list_head sai_entries_stated; /* entries stated */
1234 struct list_head sai_entries_agl; /* AGL entries to be sent */
1235 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1236 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1237 atomic_t sai_cache_count; /* entry count in cache */
1240 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1242 void ll_stop_statahead(struct inode *dir, void *key);
1244 static inline int ll_glimpse_size(struct inode *inode)
1246 struct ll_inode_info *lli = ll_i2info(inode);
1249 down_read(&lli->lli_glimpse_sem);
1250 rc = cl_glimpse_size(inode);
1251 lli->lli_glimpse_time = cfs_time_current();
1252 up_read(&lli->lli_glimpse_sem);
1257 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1259 struct ll_inode_info *lli = ll_i2info(dir);
1260 struct ll_statahead_info *sai = lli->lli_sai;
1261 struct ll_dentry_data *ldd = ll_d2d(dentry);
1263 /* not the same process, don't mark */
1264 if (lli->lli_opendir_pid != current_pid())
1267 if (sai != NULL && ldd != NULL)
1268 ldd->lld_sa_generation = sai->sai_generation;
1272 ll_need_statahead(struct inode *dir, struct dentry *dentryp)
1274 struct ll_inode_info *lli;
1275 struct ll_dentry_data *ldd;
1277 if (ll_i2sbi(dir)->ll_sa_max == 0)
1280 lli = ll_i2info(dir);
1281 /* not the same process, don't statahead */
1282 if (lli->lli_opendir_pid != current_pid())
1285 /* statahead has been stopped */
1286 if (lli->lli_opendir_key == NULL)
1289 ldd = ll_d2d(dentryp);
1291 * When stats a dentry, the system trigger more than once "revalidate"
1292 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1293 * Under patchless client mode, the operation intent is not accurate,
1294 * which maybe misguide the statahead thread. For example:
1295 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1296 * have the same operation intent -- "IT_GETATTR".
1297 * In fact, one dentry should has only one chance to interact with the
1298 * statahead thread, otherwise the statahead windows will be confused.
1299 * The solution is as following:
1300 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1301 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1302 * will bypass interacting with statahead thread for checking:
1303 * "lld_sa_generation == lli_sai->sai_generation"
1305 if (ldd && lli->lli_sai &&
1306 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1313 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1317 ret = ll_need_statahead(dir, *dentryp);
1321 return do_statahead_enter(dir, dentryp, only_unplug);
1324 /* llite ioctl register support rountine */
1330 #define LLIOC_MAX_CMD 256
1333 * Rules to write a callback function:
1336 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1337 * returned to ll_iocontrol_register. Callback functions should use this
1338 * data to check the potential collasion of ioctl cmd. If collasion is
1339 * found, callback function should return LLIOC_CONT.
1340 * @rcp: The result of ioctl command.
1343 * If @magic matches the pointer returned by ll_iocontrol_data, the
1344 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1346 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1347 struct file *file, unsigned int cmd, unsigned long arg,
1348 void *magic, int *rcp);
1350 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1351 unsigned int cmd, unsigned long arg, int *rcp);
1353 /* export functions */
1354 /* Register ioctl block dynamatically for a regular file.
1356 * @cmd: the array of ioctl command set
1357 * @count: number of commands in the @cmd
1358 * @cb: callback function, it will be called if an ioctl command is found to
1359 * belong to the command list @cmd.
1362 * A magic pointer will be returned if success;
1363 * otherwise, NULL will be returned.
1365 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1366 void ll_iocontrol_unregister(void *magic);
1369 /* lclient compat stuff */
1370 #define cl_inode_info ll_inode_info
1371 #define cl_i2info(info) ll_i2info(info)
1372 #define cl_inode_mode(inode) ((inode)->i_mode)
1373 #define cl_i2sbi ll_i2sbi
1375 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1376 const struct iattr *attr)
1378 LASSERT(attr->ia_valid & ATTR_FILE);
1379 return LUSTRE_FPRIVATE(attr->ia_file);
1382 static inline void cl_isize_lock(struct inode *inode)
1384 ll_inode_size_lock(inode);
1387 static inline void cl_isize_unlock(struct inode *inode)
1389 ll_inode_size_unlock(inode);
1392 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1394 LASSERT(down_trylock(&ll_i2info(inode)->lli_size_sem) != 0);
1395 i_size_write(inode, kms);
1398 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1400 ll_inode_size_lock(inode);
1401 i_size_write(inode, kms);
1402 ll_inode_size_unlock(inode);
1405 #define cl_isize_read(inode) i_size_read(inode)
1407 static inline int cl_merge_lvb(const struct lu_env *env, struct inode *inode)
1409 return ll_merge_lvb(env, inode);
1412 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1413 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1414 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1416 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1418 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1419 enum cl_fsync_mode mode, int ignore_layout);
1421 /** direct write pages */
1422 struct ll_dio_pages {
1423 /** page array to be written. we don't support
1424 * partial pages except the last one. */
1425 struct page **ldp_pages;
1426 /* offset of each page */
1427 loff_t *ldp_offsets;
1428 /** if ldp_offsets is NULL, it means a sequential
1429 * pages to be written, then this is the file offset
1430 * of the * first page. */
1431 loff_t ldp_start_offset;
1432 /** how many bytes are to be written. */
1434 /** # of pages in the array. */
1438 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1441 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1444 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1447 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1448 int rw, struct inode *inode,
1449 struct ll_dio_pages *pv);
1451 static inline int ll_file_nolock(const struct file *file)
1453 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1454 struct inode *inode = file->f_dentry->d_inode;
1456 LASSERT(fd != NULL);
1457 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1458 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1461 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1462 struct lookup_intent *it, __u64 *bits)
1464 if (!it->d.lustre.it_lock_set) {
1465 struct lustre_handle handle;
1467 /* If this inode is a remote object, it will get two
1468 * separate locks in different namespaces, Master MDT,
1469 * where the name entry is, will grant LOOKUP lock,
1470 * remote MDT, where the object is, will grant
1471 * UPDATE|PERM lock. The inode will be attched to both
1472 * LOOKUP and PERM locks, so revoking either locks will
1473 * case the dcache being cleared */
1474 if (it->d.lustre.it_remote_lock_mode) {
1475 handle.cookie = it->d.lustre.it_remote_lock_handle;
1476 CDEBUG(D_DLMTRACE, "setting l_data to inode %p"
1477 "(%lu/%u) for remote lock "LPX64"\n", inode,
1478 inode->i_ino, inode->i_generation,
1480 md_set_lock_data(exp, &handle.cookie, inode, NULL);
1483 handle.cookie = it->d.lustre.it_lock_handle;
1485 CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)"
1486 " for lock "LPX64"\n", inode, inode->i_ino,
1487 inode->i_generation, handle.cookie);
1489 md_set_lock_data(exp, &handle.cookie, inode,
1490 &it->d.lustre.it_lock_bits);
1491 it->d.lustre.it_lock_set = 1;
1495 *bits = it->d.lustre.it_lock_bits;
1498 static inline void ll_lock_dcache(struct inode *inode)
1500 spin_lock(&inode->i_lock);
1503 static inline void ll_unlock_dcache(struct inode *inode)
1505 spin_unlock(&inode->i_lock);
1508 static inline int d_lustre_invalid(const struct dentry *dentry)
1510 struct ll_dentry_data *lld = ll_d2d(dentry);
1512 return (lld == NULL) || lld->lld_invalid;
1515 static inline void __d_lustre_invalidate(struct dentry *dentry)
1517 struct ll_dentry_data *lld = ll_d2d(dentry);
1520 lld->lld_invalid = 1;
1524 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1525 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1526 * else dput() of the last refcount will unhash this dentry and kill it.
1528 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1530 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1531 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1532 dentry->d_parent, dentry->d_inode, d_count(dentry));
1534 spin_lock_nested(&dentry->d_lock,
1535 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1536 __d_lustre_invalidate(dentry);
1537 if (d_count(dentry) == 0)
1539 spin_unlock(&dentry->d_lock);
1542 static inline void d_lustre_revalidate(struct dentry *dentry)
1544 spin_lock(&dentry->d_lock);
1545 LASSERT(ll_d2d(dentry) != NULL);
1546 ll_d2d(dentry)->lld_invalid = 0;
1547 spin_unlock(&dentry->d_lock);
1550 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0)
1551 /* Compatibility for old (1.8) compiled userspace quota code */
1552 struct if_quotactl_18 {
1557 struct obd_dqinfo qc_dqinfo;
1558 struct obd_dqblk qc_dqblk;
1560 struct obd_uuid obd_uuid;
1562 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1563 /* End compatibility for old (1.8) compiled userspace quota code */
1565 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1566 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */
1569 LL_LAYOUT_GEN_NONE = ((__u32)-2), /* layout lock was cancelled */
1570 LL_LAYOUT_GEN_EMPTY = ((__u32)-1) /* for empty layout */
1573 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1574 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1576 #endif /* LLITE_INTERNAL_H */