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 struct dir_context ctx;
94 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
95 struct lu_fid lgd_fid; /* target fid we are looking for */
96 int lgd_found; /* inode matched? */
99 /* llite setxid/access permission for user on remote client */
100 struct ll_remote_perm {
101 struct hlist_node lrp_list;
106 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
107 is access permission with
108 lrp_fsuid/lrp_fsgid. */
112 /* MDS has an authority for the Size-on-MDS attributes. */
113 LLIF_MDS_SIZE_LOCK = (1 << 0),
114 /* Epoch close is postponed. */
115 LLIF_EPOCH_PENDING = (1 << 1),
116 /* DONE WRITING is allowed. */
117 LLIF_DONE_WRITING = (1 << 2),
118 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
120 LLIF_SOM_DIRTY = (1 << 3),
121 /* File is contented */
122 LLIF_CONTENDED = (1 << 4),
123 /* Truncate uses server lock for this file */
124 LLIF_SRVLOCK = (1 << 5),
125 /* File data is modified. */
126 LLIF_DATA_MODIFIED = (1 << 6),
129 struct ll_inode_info {
130 __u32 lli_inode_magic;
135 struct posix_acl *lli_posix_acl;
137 struct hlist_head *lli_remote_perms;
138 struct mutex lli_rmtperm_mutex;
140 /* identifying fields for both metadata and data stacks. */
141 struct lu_fid lli_fid;
142 /* Parent fid for accessing default stripe data on parent directory
143 * for allocating OST objects after a mknod() and later open-by-FID. */
144 struct lu_fid lli_pfid;
146 struct list_head lli_close_list;
147 struct list_head lli_oss_capas;
148 /* open count currently used by capability only, indicate whether
149 * capability needs renewal */
150 atomic_t lli_open_count;
151 struct obd_capa *lli_mds_capa;
152 cfs_time_t lli_rmtperm_time;
154 /* handle is to be sent to MDS later on done_writing and setattr.
155 * Open handle data are needed for the recovery to reconstruct
156 * the inode state on the MDS. XXX: recovery is not ready yet. */
157 struct obd_client_handle *lli_pending_och;
159 /* We need all three because every inode may be opened in different
161 struct obd_client_handle *lli_mds_read_och;
162 struct obd_client_handle *lli_mds_write_och;
163 struct obd_client_handle *lli_mds_exec_och;
164 __u64 lli_open_fd_read_count;
165 __u64 lli_open_fd_write_count;
166 __u64 lli_open_fd_exec_count;
167 /* Protects access to och pointers and their usage counters */
168 struct mutex lli_och_mutex;
170 struct inode lli_vfs_inode;
172 /* the most recent timestamps obtained from mds */
173 struct ost_lvb lli_lvb;
174 spinlock_t lli_agl_lock;
176 /* Try to make the d::member and f::member are aligned. Before using
177 * these members, make clear whether it is directory or not. */
181 /* serialize normal readdir and statahead-readdir. */
182 struct mutex d_readdir_mutex;
184 /* metadata statahead */
185 /* since parent-child threads can share the same @file
186 * struct, "opendir_key" is the token when dir close for
187 * case of parent exit before child -- it is me should
188 * cleanup the dir readahead. */
190 struct ll_statahead_info *d_sai;
191 struct posix_acl *d_def_acl;
192 /* protect statahead stuff. */
193 spinlock_t d_sa_lock;
194 /* "opendir_pid" is the token when lookup/revalid
195 * -- I am the owner of dir statahead. */
199 #define lli_readdir_mutex u.d.d_readdir_mutex
200 #define lli_opendir_key u.d.d_opendir_key
201 #define lli_sai u.d.d_sai
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 struct semaphore f_size_sem;
209 void *f_size_sem_owner;
210 char *f_symlink_name;
213 * struct rw_semaphore {
214 * signed long count; // align d.d_def_acl
215 * spinlock_t wait_lock; // align d.d_sa_lock
216 * struct list_head wait_list;
219 struct rw_semaphore f_trunc_sem;
220 struct mutex f_write_mutex;
222 struct rw_semaphore f_glimpse_sem;
223 cfs_time_t f_glimpse_time;
224 struct list_head f_agl_list;
227 /* for writepage() only to communicate to fsync */
230 /* volatile file criteria is based on file name, this
231 * flag is used to keep the test result, so the strcmp
236 * whenever a process try to read/write the file, the
237 * jobid of the process will be saved here, and it'll
238 * be packed into the write PRC when flush later.
240 * so the read/write statistics for jobid will not be
241 * accurate if the file is shared by different jobs.
243 char f_jobid[JOBSTATS_JOBID_SIZE];
246 #define lli_size_sem u.f.f_size_sem
247 #define lli_size_sem_owner u.f.f_size_sem_owner
248 #define lli_symlink_name u.f.f_symlink_name
249 #define lli_maxbytes u.f.f_maxbytes
250 #define lli_trunc_sem u.f.f_trunc_sem
251 #define lli_write_mutex u.f.f_write_mutex
252 #define lli_glimpse_sem u.f.f_glimpse_sem
253 #define lli_glimpse_time u.f.f_glimpse_time
254 #define lli_agl_list u.f.f_agl_list
255 #define lli_agl_index u.f.f_agl_index
256 #define lli_async_rc u.f.f_async_rc
257 #define lli_jobid u.f.f_jobid
258 #define lli_volatile u.f.f_volatile
262 /* XXX: For following frequent used members, although they maybe special
263 * used for non-directory object, it is some time-wasting to check
264 * whether the object is directory or not before using them. On the
265 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
266 * the "ll_inode_info" size even if moving those members into u.f.
267 * So keep them out side.
269 * In the future, if more members are added only for directory,
270 * some of the following members can be moved into u.f.
273 struct cl_object *lli_clob;
275 /* mutex to request for layout lock exclusively. */
276 struct mutex lli_layout_mutex;
277 /* valid only inside LAYOUT ibits lock, protected by lli_layout_mutex */
278 __u32 lli_layout_gen;
282 * Locking to guarantee consistency of non-atomic updates to long long i_size,
283 * consistency between file size and KMS.
285 * Implemented by ->lli_size_sem and ->lsm_lock, nested in that order.
288 void ll_inode_size_lock(struct inode *inode);
289 void ll_inode_size_unlock(struct inode *inode);
291 // FIXME: replace the name of this with LL_I to conform to kernel stuff
292 // static inline struct ll_inode_info *LL_I(struct inode *inode)
293 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
295 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
298 /* default to about 40meg of readahead on a given system. That much tied
299 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
300 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT))
302 /* default to read-ahead full files smaller than 2MB on the second read */
303 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
308 RA_STAT_DISTANT_READPAGE,
309 RA_STAT_MISS_IN_WINDOW,
310 RA_STAT_FAILED_GRAB_PAGE,
311 RA_STAT_FAILED_MATCH,
316 RA_STAT_MAX_IN_FLIGHT,
317 RA_STAT_WRONG_GRAB_PAGE,
322 atomic_t ra_cur_pages;
323 unsigned long ra_max_pages;
324 unsigned long ra_max_pages_per_file;
325 unsigned long ra_max_read_ahead_whole_pages;
328 /* ra_io_arg will be filled in the beginning of ll_readahead with
329 * ras_lock, then the following ll_read_ahead_pages will read RA
330 * pages according to this arg, all the items in this structure are
331 * counted by page index.
334 unsigned long ria_start; /* start offset of read-ahead*/
335 unsigned long ria_end; /* end offset of read-ahead*/
336 /* If stride read pattern is detected, ria_stoff means where
337 * stride read is started. Note: for normal read-ahead, the
338 * value here is meaningless, and also it will not be accessed*/
340 /* ria_length and ria_pages are the length and pages length in the
341 * stride I/O mode. And they will also be used to check whether
342 * it is stride I/O read-ahead in the read-ahead pages*/
343 unsigned long ria_length;
344 unsigned long ria_pages;
347 /* LL_HIST_MAX=32 causes an overflow */
348 #define LL_HIST_MAX 28
349 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
350 #define LL_PROCESS_HIST_MAX 10
351 struct per_process_info {
353 struct obd_histogram pp_r_hist;
354 struct obd_histogram pp_w_hist;
357 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
358 struct ll_rw_extents_info {
359 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
362 #define LL_OFFSET_HIST_MAX 100
363 struct ll_rw_process_info {
366 loff_t rw_range_start;
368 loff_t rw_last_file_pos;
370 size_t rw_smallest_extent;
371 size_t rw_largest_extent;
372 struct ll_file_data *rw_last_file;
375 enum stats_track_type {
376 STATS_TRACK_ALL = 0, /* track all processes */
377 STATS_TRACK_PID, /* track process with this pid */
378 STATS_TRACK_PPID, /* track processes with this ppid */
379 STATS_TRACK_GID, /* track processes with this gid */
383 /* flags for sbi->ll_flags */
384 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
385 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
386 #define LL_SBI_FLOCK 0x04
387 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
388 #define LL_SBI_ACL 0x10 /* support ACL */
389 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
390 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
391 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
392 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
393 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
394 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
395 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
396 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
397 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
398 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
399 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
400 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
401 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
403 #define LL_SBI_FLAGS { \
423 /* default value for ll_sb_info->contention_time */
424 #define SBI_DEFAULT_CONTENTION_SECONDS 60
425 /* default value for lockless_truncate_enable */
426 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
427 #define RCE_HASHES 32
429 struct rmtacl_ctl_entry {
430 struct list_head rce_list;
431 pid_t rce_key; /* hash key */
432 int rce_ops; /* acl operation type */
435 struct rmtacl_ctl_table {
437 struct list_head rct_entries[RCE_HASHES];
444 struct list_head et_entries[EE_HASHES];
448 struct list_head ll_list;
449 /* this protects pglist and ra_info. It isn't safe to
450 * grab from interrupt contexts */
452 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
453 spinlock_t ll_process_lock; /* ll_rw_process_info */
454 struct obd_uuid ll_sb_uuid;
455 struct obd_export *ll_md_exp;
456 struct obd_export *ll_dt_exp;
457 struct proc_dir_entry* ll_proc_root;
458 struct lu_fid ll_root_fid; /* root object fid */
462 struct list_head ll_conn_chain; /* per-conn chain of SBs */
463 struct lustre_client_ocd ll_lco;
465 struct list_head ll_orphan_dentry_list; /*please don't ask -p*/
466 struct ll_close_queue *ll_lcq;
468 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
470 struct cl_client_cache ll_cache;
472 struct lprocfs_stats *ll_ra_stats;
474 struct ll_ra_info ll_ra_info;
475 unsigned int ll_namelen;
476 struct file_operations *ll_fop;
478 /* =0 - hold lock over whole read/write
479 * >0 - max. chunk to be read/written w/o lock re-acquiring */
480 unsigned long ll_max_rw_chunk;
481 unsigned int ll_md_brw_size; /* used by readdir */
483 struct lu_site *ll_site;
484 struct cl_device *ll_cl;
486 struct ll_rw_extents_info ll_rw_extents_info;
487 int ll_extent_process_count;
488 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
489 unsigned int ll_offset_process_count;
490 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
491 unsigned int ll_rw_offset_entry_count;
492 int ll_stats_track_id;
493 enum stats_track_type ll_stats_track_type;
496 /* metadata stat-ahead */
497 unsigned int ll_sa_max; /* max statahead RPCs */
498 atomic_t ll_sa_total; /* statahead thread started
500 atomic_t ll_sa_wrong; /* statahead thread stopped for
502 atomic_t ll_agl_total; /* AGL thread started count */
504 dev_t ll_sdev_orig; /* save s_dev before assign for
506 struct rmtacl_ctl_table ll_rct;
507 struct eacl_table ll_et;
508 __kernel_fsid_t ll_fsid;
511 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
516 struct task_struct *lrr_reader;
517 struct list_head lrr_linkage;
521 * per file-descriptor read-ahead data.
523 struct ll_readahead_state {
526 * index of the last page that read(2) needed and that wasn't in the
527 * cache. Used by ras_update() to detect seeks.
529 * XXX nikita: if access seeks into cached region, Lustre doesn't see
532 unsigned long ras_last_readpage;
534 * number of pages read after last read-ahead window reset. As window
535 * is reset on each seek, this is effectively a number of consecutive
536 * accesses. Maybe ->ras_accessed_in_window is better name.
538 * XXX nikita: window is also reset (by ras_update()) when Lustre
539 * believes that memory pressure evicts read-ahead pages. In that
540 * case, it probably doesn't make sense to expand window to
541 * PTLRPC_MAX_BRW_PAGES on the third access.
543 unsigned long ras_consecutive_pages;
545 * number of read requests after the last read-ahead window reset
546 * As window is reset on each seek, this is effectively the number
547 * on consecutive read request and is used to trigger read-ahead.
549 unsigned long ras_consecutive_requests;
551 * Parameters of current read-ahead window. Handled by
552 * ras_update(). On the initial access to the file or after a seek,
553 * window is reset to 0. After 3 consecutive accesses, window is
554 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
555 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
557 unsigned long ras_window_start, ras_window_len;
559 * Where next read-ahead should start at. This lies within read-ahead
560 * window. Read-ahead window is read in pieces rather than at once
561 * because: 1. lustre limits total number of pages under read-ahead by
562 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
563 * not covered by DLM lock.
565 unsigned long ras_next_readahead;
567 * Total number of ll_file_read requests issued, reads originating
568 * due to mmap are not counted in this total. This value is used to
569 * trigger full file read-ahead after multiple reads to a small file.
571 unsigned long ras_requests;
573 * Page index with respect to the current request, these value
574 * will not be accurate when dealing with reads issued via mmap.
576 unsigned long ras_request_index;
578 * list of struct ll_ra_read's one per read(2) call current in
579 * progress against this file descriptor. Used by read-ahead code,
580 * protected by ->ras_lock.
582 struct list_head ras_read_beads;
584 * The following 3 items are used for detecting the stride I/O
586 * In stride I/O mode,
587 * ...............|-----data-----|****gap*****|--------|******|....
588 * offset |-stride_pages-|-stride_gap-|
589 * ras_stride_offset = offset;
590 * ras_stride_length = stride_pages + stride_gap;
591 * ras_stride_pages = stride_pages;
592 * Note: all these three items are counted by pages.
594 unsigned long ras_stride_length;
595 unsigned long ras_stride_pages;
596 pgoff_t ras_stride_offset;
598 * number of consecutive stride request count, and it is similar as
599 * ras_consecutive_requests, but used for stride I/O mode.
600 * Note: only more than 2 consecutive stride request are detected,
601 * stride read-ahead will be enable
603 unsigned long ras_consecutive_stride_requests;
606 extern struct kmem_cache *ll_file_data_slab;
607 struct lustre_handle;
608 struct ll_file_data {
609 struct ll_readahead_state fd_ras;
611 struct ccc_grouplock fd_grouplock;
614 struct file *fd_file;
615 /* Indicate whether need to report failure when close.
616 * true: failure is known, not report again.
617 * false: unknown failure, should report. */
618 bool fd_write_failed;
621 struct lov_stripe_md;
623 extern spinlock_t inode_lock;
625 extern struct proc_dir_entry *proc_lustre_fs_root;
627 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
629 return &lli->lli_vfs_inode;
633 struct inode *icbd_parent;
634 struct dentry **icbd_childp;
638 __u32 ll_i2suppgid(struct inode *i);
639 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
641 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
643 #if BITS_PER_LONG == 32
646 return unlikely(current_is_32bit() || (sbi->ll_flags & LL_SBI_32BIT_API));
650 #define LLAP_MAGIC 98764321
652 extern struct kmem_cache *ll_async_page_slab;
653 extern size_t ll_async_page_slab_size;
655 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
656 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
657 struct ll_ra_read *ll_ra_read_get(struct file *f);
659 /* llite/lproc_llite.c */
661 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
662 struct super_block *sb, char *osc, char *mdc);
663 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
664 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
665 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
667 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
668 struct super_block *sb, char *osc, char *mdc){return 0;}
669 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
670 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
671 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
673 memset(lvars, 0, sizeof(*lvars));
679 void ll_release_page(struct page *page, int remove);
680 extern struct file_operations ll_dir_operations;
681 extern struct inode_operations ll_dir_inode_operations;
682 struct page *ll_get_dir_page(struct inode *dir, __u64 hash,
683 struct ll_dir_chain *chain);
684 int ll_dir_read(struct inode *inode, struct dir_context *ctx);
686 int ll_get_mdt_idx(struct inode *inode);
688 int ll_objects_destroy(struct ptlrpc_request *request,
690 struct inode *ll_iget(struct super_block *sb, ino_t hash,
691 struct lustre_md *lic);
692 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
693 void *data, int flag);
694 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
695 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
698 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
699 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
700 int ll_writepage(struct page *page, struct writeback_control *wbc);
701 int ll_writepages(struct address_space *, struct writeback_control *wbc);
702 void ll_removepage(struct page *page);
703 int ll_readpage(struct file *file, struct page *page);
704 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
705 int ll_file_punch(struct inode *, loff_t, int);
706 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
707 void ll_clear_file_contended(struct inode*);
708 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
709 int ll_readahead(const struct lu_env *env, struct cl_io *io,
710 struct ll_readahead_state *ras, struct address_space *mapping,
711 struct cl_page_list *queue, int flags);
714 extern struct file_operations ll_file_operations;
715 extern struct file_operations ll_file_operations_flock;
716 extern struct file_operations ll_file_operations_noflock;
717 extern struct inode_operations ll_file_inode_operations;
718 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
720 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
721 ldlm_mode_t l_req_mode);
722 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
723 struct lustre_handle *lockh, __u64 flags);
724 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
726 int ll_revalidate_nd(struct dentry *dentry, unsigned int flags);
727 int ll_file_open(struct inode *inode, struct file *file);
728 int ll_file_release(struct inode *inode, struct file *file);
729 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
730 struct lov_stripe_md *lsm, lstat_t *st);
731 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
732 int ll_local_open(struct file *file,
733 struct lookup_intent *it, struct ll_file_data *fd,
734 struct obd_client_handle *och);
735 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
736 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
738 int ll_md_real_close(struct inode *inode, int flags);
739 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
740 struct obd_client_handle **och, unsigned long flags);
741 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
742 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
743 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
744 __u64 ioepoch, int sync);
745 int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data,
746 struct md_open_data **mod);
747 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
748 struct lustre_handle *fh);
749 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
750 struct ll_file_data *file, loff_t pos,
751 size_t count, int rw);
752 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
753 struct lookup_intent *it, struct kstat *stat);
754 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
755 struct ll_file_data *ll_file_data_get(void);
756 struct posix_acl * ll_get_acl(struct inode *inode, int type);
758 int ll_inode_permission(struct inode *inode, int mask);
760 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
761 int flags, struct lov_user_md *lum,
763 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
764 struct lov_mds_md **lmm, int *lmm_size,
765 struct ptlrpc_request **request);
766 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
768 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmmp,
769 int *lmm_size, struct ptlrpc_request **request);
770 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
771 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
773 int ll_merge_lvb(const struct lu_env *env, struct inode *inode);
774 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
775 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
776 int ll_fid2path(struct inode *inode, void *arg);
777 int ll_data_version(struct inode *inode, __u64 *data_version, int extent_lock);
781 int ll_dops_init(struct dentry *de, int block, int init_sa);
782 extern struct dentry_operations ll_d_ops;
783 void ll_intent_drop_lock(struct lookup_intent *);
784 void ll_intent_release(struct lookup_intent *);
785 void ll_invalidate_aliases(struct inode *);
786 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
787 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
788 int ll_dcompare(const struct dentry *parent, const struct dentry *dentry,
789 unsigned int len, const char *str, const struct qstr *d_name);
790 int ll_revalidate_it_finish(struct ptlrpc_request *request,
791 struct lookup_intent *it, struct dentry *de);
793 /* llite/llite_lib.c */
794 extern struct super_operations lustre_super_operations;
796 char *ll_read_opt(const char *opt, char *data);
797 void ll_lli_init(struct ll_inode_info *lli);
798 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
799 void ll_put_super(struct super_block *sb);
800 void ll_kill_super(struct super_block *sb);
801 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
802 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
803 void ll_clear_inode(struct inode *inode);
804 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr);
805 int ll_setattr(struct dentry *de, struct iattr *attr);
806 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
807 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
808 __u64 max_age, __u32 flags);
809 void ll_update_inode(struct inode *inode, struct lustre_md *md);
810 void ll_read_inode2(struct inode *inode, void *opaque);
811 void ll_delete_inode(struct inode *inode);
812 int ll_iocontrol(struct inode *inode, struct file *file,
813 unsigned int cmd, unsigned long arg);
814 int ll_flush_ctx(struct inode *inode);
815 void ll_umount_begin(struct super_block *sb);
816 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
817 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
818 void ll_dirty_page_discard_warn(struct page *page, int ioret);
819 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
820 struct super_block *, struct lookup_intent *);
821 void lustre_dump_dentry(struct dentry *, int recur);
822 void lustre_dump_inode(struct inode *);
823 int ll_obd_statfs(struct inode *inode, void *arg);
824 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
825 int ll_process_config(struct lustre_cfg *lcfg);
826 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
827 struct inode *i1, struct inode *i2,
828 const char *name, int namelen,
829 int mode, __u32 opc, void *data);
830 void ll_finish_md_op_data(struct md_op_data *op_data);
831 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
832 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
834 /* llite/llite_nfs.c */
835 extern struct export_operations lustre_export_operations;
836 __u32 get_uuid2int(const char *name, int len);
837 void get_uuid2fsid(const char *name, int len, __kernel_fsid_t *fsid);
838 struct inode *search_inode_for_lustre(struct super_block *sb,
839 const struct lu_fid *fid);
841 /* llite/special.c */
842 extern struct inode_operations ll_special_inode_operations;
843 extern struct file_operations ll_special_chr_inode_fops;
844 extern struct file_operations ll_special_chr_file_fops;
845 extern struct file_operations ll_special_blk_inode_fops;
846 extern struct file_operations ll_special_fifo_inode_fops;
847 extern struct file_operations ll_special_fifo_file_fops;
848 extern struct file_operations ll_special_sock_inode_fops;
850 /* llite/symlink.c */
851 extern struct inode_operations ll_fast_symlink_inode_operations;
853 /* llite/llite_close.c */
854 struct ll_close_queue {
856 struct list_head lcq_head;
857 wait_queue_head_t lcq_waitq;
858 struct completion lcq_comp;
862 struct ccc_object *cl_inode2ccc(struct inode *inode);
865 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
866 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
868 /* specific achitecture can implement only part of this list */
869 enum vvp_io_subtype {
872 /** io called from .sendfile */
874 /** io started from splice_{read|write} */
881 enum vvp_io_subtype cui_io_subtype;
885 read_actor_t cui_actor;
889 struct pipe_inode_info *cui_pipe;
890 unsigned int cui_flags;
892 struct vvp_fault_io {
894 * Inode modification time that is checked across DLM
898 struct vm_area_struct *ft_vma;
900 * locked page returned from vvp_io
902 struct page *ft_vmpage;
903 struct vm_fault_api {
907 struct vm_fault *ft_vmf;
909 * fault API used bitflags for return code.
911 unsigned int ft_flags;
916 * Read-ahead state used by read and page-fault IO contexts.
918 struct ll_ra_read cui_bead;
920 * Set when cui_bead has been initialized.
922 int cui_ra_window_set;
924 * Partially truncated page, that vvp_io_trunc_start() keeps locked
927 struct cl_page *cui_partpage;
931 * IO arguments for various VFS I/O interfaces.
934 /** normal/sendfile/splice */
935 enum vvp_io_subtype via_io_subtype;
939 struct kiocb *via_iocb;
940 struct iovec *via_iov;
941 unsigned long via_nrsegs;
944 read_actor_t via_actor;
948 struct pipe_inode_info *via_pipe;
949 unsigned int via_flags;
954 struct ll_cl_context {
956 struct cl_io *lcc_io;
957 struct cl_page *lcc_page;
958 struct lu_env *lcc_env;
963 struct vvp_thread_info {
964 struct ost_lvb vti_lvb;
965 struct cl_2queue vti_queue;
966 struct iovec vti_local_iov;
967 struct vvp_io_args vti_args;
968 struct ra_io_arg vti_ria;
969 struct kiocb vti_kiocb;
970 struct ll_cl_context vti_io_ctx;
973 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
975 extern struct lu_context_key vvp_key;
976 struct vvp_thread_info *info;
978 info = lu_context_key_get(&env->le_ctx, &vvp_key);
979 LASSERT(info != NULL);
983 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
984 enum vvp_io_subtype type)
986 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
988 ret->via_io_subtype = type;
994 struct vvp_io vs_ios;
997 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
999 extern struct lu_context_key vvp_session_key;
1000 struct vvp_session *ses;
1002 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1003 LASSERT(ses != NULL);
1007 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1009 return &vvp_env_session(env)->vs_ios;
1012 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1013 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1014 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1016 /* llite/llite_mmap.c */
1017 typedef struct rb_root rb_root_t;
1018 typedef struct rb_node rb_node_t;
1020 struct ll_lock_tree_node;
1021 struct ll_lock_tree {
1023 struct list_head lt_locked_list;
1024 struct ll_file_data *lt_fd;
1027 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1028 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1029 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
1030 __u64 end, ldlm_mode_t mode);
1031 void policy_from_vma(ldlm_policy_data_t *policy,
1032 struct vm_area_struct *vma, unsigned long addr, size_t count);
1033 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1036 static inline void ll_invalidate_page(struct page *vmpage)
1038 struct address_space *mapping = vmpage->mapping;
1039 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1041 LASSERT(PageLocked(vmpage));
1042 if (mapping == NULL)
1045 ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE);
1046 truncate_complete_page(mapping, vmpage);
1049 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1051 /* don't need an addref as the sb_info should be holding one */
1052 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1054 return ll_s2sbi(sb)->ll_dt_exp;
1057 /* don't need an addref as the sb_info should be holding one */
1058 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1060 return ll_s2sbi(sb)->ll_md_exp;
1063 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1065 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1071 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1072 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1074 return ll_s2sbi(inode->i_sb);
1077 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1079 return ll_s2dtexp(inode->i_sb);
1082 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1084 return ll_s2mdexp(inode->i_sb);
1087 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1091 LASSERT(inode != NULL);
1092 fid = &ll_i2info(inode)->lli_fid;
1097 static inline int ll_mds_max_easize(struct super_block *sb)
1099 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1102 static inline __u64 ll_file_maxbytes(struct inode *inode)
1104 return ll_i2info(inode)->lli_maxbytes;
1108 int ll_setxattr(struct dentry *dentry, const char *name,
1109 const void *value, size_t size, int flags);
1110 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1111 void *buffer, size_t size);
1112 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1113 int ll_removexattr(struct dentry *dentry, const char *name);
1115 /* llite/remote_perm.c */
1116 extern struct kmem_cache *ll_remote_perm_cachep;
1117 extern struct kmem_cache *ll_rmtperm_hash_cachep;
1119 struct hlist_head *alloc_rmtperm_hash(void);
1120 void free_rmtperm_hash(struct hlist_head *hash);
1121 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1122 int lustre_check_remote_perm(struct inode *inode, int mask);
1124 /* llite/llite_capa.c */
1125 extern struct timer_list ll_capa_timer;
1127 int ll_capa_thread_start(void);
1128 void ll_capa_thread_stop(void);
1129 void ll_capa_timer_callback(unsigned long unused);
1131 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1132 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1134 void ll_capa_open(struct inode *inode);
1135 void ll_capa_close(struct inode *inode);
1137 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1138 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1140 void ll_truncate_free_capa(struct obd_capa *ocapa);
1141 void ll_clear_inode_capas(struct inode *inode);
1142 void ll_print_capa_stat(struct ll_sb_info *sbi);
1144 /* llite/llite_cl.c */
1145 extern struct lu_device_type vvp_device_type;
1148 * Common IO arguments for various VFS I/O interfaces.
1150 int cl_sb_init(struct super_block *sb);
1151 int cl_sb_fini(struct super_block *sb);
1152 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1153 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1155 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1156 struct ll_readahead_state *ras, unsigned long index,
1158 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1159 int ll_is_file_contended(struct file *file);
1160 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1162 /* llite/llite_rmtacl.c */
1163 #ifdef CONFIG_FS_POSIX_ACL
1165 struct list_head ee_list;
1166 pid_t ee_key; /* hash key */
1167 struct lu_fid ee_fid;
1168 int ee_type; /* ACL type for ACCESS or DEFAULT */
1169 ext_acl_xattr_header *ee_acl;
1172 obd_valid rce_ops2valid(int ops);
1173 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1174 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1175 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1176 void rct_init(struct rmtacl_ctl_table *rct);
1177 void rct_fini(struct rmtacl_ctl_table *rct);
1179 void ee_free(struct eacl_entry *ee);
1180 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1181 ext_acl_xattr_header *header);
1182 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1183 struct lu_fid *fid, int type);
1184 void et_search_free(struct eacl_table *et, pid_t key);
1185 void et_init(struct eacl_table *et);
1186 void et_fini(struct eacl_table *et);
1188 static inline obd_valid rce_ops2valid(int ops)
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 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 wait_queue_head_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 struct list_head sai_entries; /* entry list */
1238 struct list_head sai_entries_received; /* entries returned */
1239 struct list_head sai_entries_stated; /* entries stated */
1240 struct list_head sai_entries_agl; /* AGL entries to be sent */
1241 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1242 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1243 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 down_read(&lli->lli_glimpse_sem);
1256 rc = cl_glimpse_size(inode);
1257 lli->lli_glimpse_time = cfs_time_current();
1258 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 != current_pid())
1273 if (sai != NULL && ldd != NULL)
1274 ldd->lld_sa_generation = sai->sai_generation;
1278 ll_need_statahead(struct inode *dir, struct dentry *dentryp)
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 != current_pid())
1291 /* statahead has been stopped */
1292 if (lli->lli_opendir_key == NULL)
1295 ldd = ll_d2d(dentryp);
1297 * When stats a dentry, the system trigger more than once "revalidate"
1298 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1299 * Under patchless client mode, the operation intent is not accurate,
1300 * which maybe misguide the statahead thread. For example:
1301 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1302 * have the same operation intent -- "IT_GETATTR".
1303 * In fact, one dentry should has only one chance to interact with the
1304 * statahead thread, otherwise the statahead windows will be confused.
1305 * The solution is as following:
1306 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1307 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1308 * will bypass interacting with statahead thread for checking:
1309 * "lld_sa_generation == lli_sai->sai_generation"
1311 if (ldd && lli->lli_sai &&
1312 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1319 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1323 ret = ll_need_statahead(dir, *dentryp);
1327 return do_statahead_enter(dir, dentryp, only_unplug);
1330 /* llite ioctl register support rountine */
1336 #define LLIOC_MAX_CMD 256
1339 * Rules to write a callback function:
1342 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1343 * returned to ll_iocontrol_register. Callback functions should use this
1344 * data to check the potential collasion of ioctl cmd. If collasion is
1345 * found, callback function should return LLIOC_CONT.
1346 * @rcp: The result of ioctl command.
1349 * If @magic matches the pointer returned by ll_iocontrol_data, the
1350 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1352 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1353 struct file *file, unsigned int cmd, unsigned long arg,
1354 void *magic, int *rcp);
1356 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1357 unsigned int cmd, unsigned long arg, int *rcp);
1359 /* export functions */
1360 /* Register ioctl block dynamatically for a regular file.
1362 * @cmd: the array of ioctl command set
1363 * @count: number of commands in the @cmd
1364 * @cb: callback function, it will be called if an ioctl command is found to
1365 * belong to the command list @cmd.
1368 * A magic pointer will be returned if success;
1369 * otherwise, NULL will be returned.
1371 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1372 void ll_iocontrol_unregister(void *magic);
1375 /* lclient compat stuff */
1376 #define cl_inode_info ll_inode_info
1377 #define cl_i2info(info) ll_i2info(info)
1378 #define cl_inode_mode(inode) ((inode)->i_mode)
1379 #define cl_i2sbi ll_i2sbi
1381 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1382 const struct iattr *attr)
1384 LASSERT(attr->ia_valid & ATTR_FILE);
1385 return LUSTRE_FPRIVATE(attr->ia_file);
1388 static inline void cl_isize_lock(struct inode *inode)
1390 ll_inode_size_lock(inode);
1393 static inline void cl_isize_unlock(struct inode *inode)
1395 ll_inode_size_unlock(inode);
1398 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1400 LASSERT(down_trylock(&ll_i2info(inode)->lli_size_sem) != 0);
1401 i_size_write(inode, kms);
1404 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1406 ll_inode_size_lock(inode);
1407 i_size_write(inode, kms);
1408 ll_inode_size_unlock(inode);
1411 #define cl_isize_read(inode) i_size_read(inode)
1413 static inline int cl_merge_lvb(const struct lu_env *env, struct inode *inode)
1415 return ll_merge_lvb(env, inode);
1418 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1419 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1420 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1422 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1424 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1425 enum cl_fsync_mode mode, int ignore_layout);
1427 /** direct write pages */
1428 struct ll_dio_pages {
1429 /** page array to be written. we don't support
1430 * partial pages except the last one. */
1431 struct page **ldp_pages;
1432 /* offset of each page */
1433 loff_t *ldp_offsets;
1434 /** if ldp_offsets is NULL, it means a sequential
1435 * pages to be written, then this is the file offset
1436 * of the * first page. */
1437 loff_t ldp_start_offset;
1438 /** how many bytes are to be written. */
1440 /** # of pages in the array. */
1444 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1447 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1450 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1453 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1454 int rw, struct inode *inode,
1455 struct ll_dio_pages *pv);
1457 static inline int ll_file_nolock(const struct file *file)
1459 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1460 struct inode *inode = file->f_dentry->d_inode;
1462 LASSERT(fd != NULL);
1463 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1464 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1467 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1468 struct lookup_intent *it, __u64 *bits)
1470 if (!it->d.lustre.it_lock_set) {
1471 struct lustre_handle handle;
1473 /* If this inode is a remote object, it will get two
1474 * separate locks in different namespaces, Master MDT,
1475 * where the name entry is, will grant LOOKUP lock,
1476 * remote MDT, where the object is, will grant
1477 * UPDATE|PERM lock. The inode will be attched to both
1478 * LOOKUP and PERM locks, so revoking either locks will
1479 * case the dcache being cleared */
1480 if (it->d.lustre.it_remote_lock_mode) {
1481 handle.cookie = it->d.lustre.it_remote_lock_handle;
1482 CDEBUG(D_DLMTRACE, "setting l_data to inode %p"
1483 "(%lu/%u) for remote lock "LPX64"\n", inode,
1484 inode->i_ino, inode->i_generation,
1486 md_set_lock_data(exp, &handle.cookie, inode, NULL);
1489 handle.cookie = it->d.lustre.it_lock_handle;
1491 CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)"
1492 " for lock "LPX64"\n", inode, inode->i_ino,
1493 inode->i_generation, handle.cookie);
1495 md_set_lock_data(exp, &handle.cookie, inode,
1496 &it->d.lustre.it_lock_bits);
1497 it->d.lustre.it_lock_set = 1;
1501 *bits = it->d.lustre.it_lock_bits;
1504 static inline void ll_lock_dcache(struct inode *inode)
1506 spin_lock(&inode->i_lock);
1509 static inline void ll_unlock_dcache(struct inode *inode)
1511 spin_unlock(&inode->i_lock);
1514 static inline int d_lustre_invalid(const struct dentry *dentry)
1516 struct ll_dentry_data *lld = ll_d2d(dentry);
1518 return (lld == NULL) || lld->lld_invalid;
1521 static inline void __d_lustre_invalidate(struct dentry *dentry)
1523 struct ll_dentry_data *lld = ll_d2d(dentry);
1526 lld->lld_invalid = 1;
1530 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1531 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1532 * else dput() of the last refcount will unhash this dentry and kill it.
1534 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1536 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1537 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1538 dentry->d_parent, dentry->d_inode, d_count(dentry));
1540 spin_lock_nested(&dentry->d_lock,
1541 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1542 __d_lustre_invalidate(dentry);
1543 if (d_count(dentry) == 0)
1545 spin_unlock(&dentry->d_lock);
1548 static inline void d_lustre_revalidate(struct dentry *dentry)
1550 spin_lock(&dentry->d_lock);
1551 LASSERT(ll_d2d(dentry) != NULL);
1552 ll_d2d(dentry)->lld_invalid = 0;
1553 spin_unlock(&dentry->d_lock);
1556 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0)
1557 /* Compatibility for old (1.8) compiled userspace quota code */
1558 struct if_quotactl_18 {
1563 struct obd_dqinfo qc_dqinfo;
1564 struct obd_dqblk qc_dqblk;
1566 struct obd_uuid obd_uuid;
1568 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1569 /* End compatibility for old (1.8) compiled userspace quota code */
1571 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1572 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */
1575 LL_LAYOUT_GEN_NONE = ((__u32)-2), /* layout lock was cancelled */
1576 LL_LAYOUT_GEN_EMPTY = ((__u32)-1) /* for empty layout */
1579 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1580 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1582 #endif /* LLITE_INTERNAL_H */