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, 2013, 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_lmv.h>
48 #include <lustre_mdc.h>
49 #include <linux/lustre_intent.h>
50 #include <linux/compat.h>
56 #ifndef VM_FAULT_RETRY
57 #define VM_FAULT_RETRY 0
60 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
61 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
62 #ifndef LOOKUP_CONTINUE
63 #define LOOKUP_CONTINUE LOOKUP_PARENT
66 /** Only used on client-side for indicating the tail of dir hash/offset. */
67 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
68 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
70 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
71 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
73 struct ll_dentry_data {
74 struct lookup_intent *lld_it;
75 unsigned int lld_sa_generation;
76 unsigned int lld_invalid:1;
77 struct rcu_head lld_rcu_head;
80 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
82 #define LLI_INODE_MAGIC 0x111d0de5
83 #define LLI_INODE_DEAD 0xdeadd00d
85 /* remote client permission cache */
86 #define REMOTE_PERM_HASHSIZE 16
88 struct ll_getname_data {
89 #ifdef HAVE_DIR_CONTEXT
90 struct dir_context ctx;
92 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
93 struct lu_fid lgd_fid; /* target fid we are looking for */
94 int lgd_found; /* inode matched? */
97 /* llite setxid/access permission for user on remote client */
98 struct ll_remote_perm {
99 struct hlist_node lrp_list;
104 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
105 is access permission with
106 lrp_fsuid/lrp_fsgid. */
110 /* MDS has an authority for the Size-on-MDS attributes. */
111 LLIF_MDS_SIZE_LOCK = (1 << 0),
112 /* Epoch close is postponed. */
113 LLIF_EPOCH_PENDING = (1 << 1),
114 /* DONE WRITING is allowed. */
115 LLIF_DONE_WRITING = (1 << 2),
116 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
118 LLIF_SOM_DIRTY = (1 << 3),
119 /* File data is modified. */
120 LLIF_DATA_MODIFIED = (1 << 4),
121 /* File is being restored */
122 LLIF_FILE_RESTORING = (1 << 5),
123 /* Xattr cache is attached to the file */
124 LLIF_XATTR_CACHE = (1 << 6),
127 struct ll_inode_info {
128 __u32 lli_inode_magic;
133 struct posix_acl *lli_posix_acl;
135 struct hlist_head *lli_remote_perms;
136 struct mutex lli_rmtperm_mutex;
138 /* identifying fields for both metadata and data stacks. */
139 struct lu_fid lli_fid;
140 /* master inode fid for stripe directory */
141 struct lu_fid lli_pfid;
143 struct list_head lli_close_list;
144 struct list_head lli_oss_capas;
145 /* open count currently used by capability only, indicate whether
146 * capability needs renewal */
147 atomic_t lli_open_count;
148 struct obd_capa *lli_mds_capa;
149 cfs_time_t lli_rmtperm_time;
151 /* handle is to be sent to MDS later on done_writing and setattr.
152 * Open handle data are needed for the recovery to reconstruct
153 * the inode state on the MDS. XXX: recovery is not ready yet. */
154 struct obd_client_handle *lli_pending_och;
156 /* We need all three because every inode may be opened in different
158 struct obd_client_handle *lli_mds_read_och;
159 struct obd_client_handle *lli_mds_write_och;
160 struct obd_client_handle *lli_mds_exec_och;
161 __u64 lli_open_fd_read_count;
162 __u64 lli_open_fd_write_count;
163 __u64 lli_open_fd_exec_count;
164 /* Protects access to och pointers and their usage counters */
165 struct mutex lli_och_mutex;
167 struct inode lli_vfs_inode;
169 /* the most recent timestamps obtained from mds */
170 struct ost_lvb lli_lvb;
171 spinlock_t lli_agl_lock;
173 /* Try to make the d::member and f::member are aligned. Before using
174 * these members, make clear whether it is directory or not. */
178 /* serialize normal readdir and statahead-readdir. */
179 struct mutex d_readdir_mutex;
181 /* metadata statahead */
182 /* since parent-child threads can share the same @file
183 * struct, "opendir_key" is the token when dir close for
184 * case of parent exit before child -- it is me should
185 * cleanup the dir readahead. */
187 struct ll_statahead_info *d_sai;
188 /* protect statahead stuff. */
189 spinlock_t d_sa_lock;
190 /* "opendir_pid" is the token when lookup/revalid
191 * -- I am the owner of dir statahead. */
193 /* directory stripe information */
194 struct lmv_stripe_md *d_lsm_md;
195 /* striped directory size */
196 loff_t d_stripe_size;
197 /* striped directory nlink */
198 __u64 d_stripe_nlink;
201 #define lli_readdir_mutex u.d.d_readdir_mutex
202 #define lli_opendir_key u.d.d_opendir_key
203 #define lli_sai u.d.d_sai
204 #define lli_sa_lock u.d.d_sa_lock
205 #define lli_opendir_pid u.d.d_opendir_pid
206 #define lli_lsm_md u.d.d_lsm_md
207 #define lli_stripe_dir_size u.d.d_stripe_size
208 #define lli_stripe_dir_nlink u.d.d_stripe_nlink
210 /* for non-directory */
212 struct mutex f_size_mutex;
213 char *f_symlink_name;
216 * struct rw_semaphore {
217 * signed long count; // align d.d_def_acl
218 * spinlock_t wait_lock; // align d.d_sa_lock
219 * struct list_head wait_list;
222 struct rw_semaphore f_trunc_sem;
223 struct mutex f_write_mutex;
225 struct rw_semaphore f_glimpse_sem;
226 cfs_time_t f_glimpse_time;
227 struct list_head f_agl_list;
230 /* for writepage() only to communicate to fsync */
234 * whenever a process try to read/write the file, the
235 * jobid of the process will be saved here, and it'll
236 * be packed into the write PRC when flush later.
238 * so the read/write statistics for jobid will not be
239 * accurate if the file is shared by different jobs.
241 char f_jobid[JOBSTATS_JOBID_SIZE];
244 #define lli_size_mutex u.f.f_size_mutex
245 #define lli_symlink_name u.f.f_symlink_name
246 #define lli_maxbytes u.f.f_maxbytes
247 #define lli_trunc_sem u.f.f_trunc_sem
248 #define lli_write_mutex u.f.f_write_mutex
249 #define lli_glimpse_sem u.f.f_glimpse_sem
250 #define lli_glimpse_time u.f.f_glimpse_time
251 #define lli_agl_list u.f.f_agl_list
252 #define lli_agl_index u.f.f_agl_index
253 #define lli_async_rc u.f.f_async_rc
254 #define lli_jobid u.f.f_jobid
258 /* XXX: For following frequent used members, although they maybe special
259 * used for non-directory object, it is some time-wasting to check
260 * whether the object is directory or not before using them. On the
261 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
262 * the "ll_inode_info" size even if moving those members into u.f.
263 * So keep them out side.
265 * In the future, if more members are added only for directory,
266 * some of the following members can be moved into u.f.
269 struct cl_object *lli_clob;
271 /* mutex to request for layout lock exclusively. */
272 struct mutex lli_layout_mutex;
273 /* Layout version, protected by lli_layout_lock */
274 __u32 lli_layout_gen;
275 spinlock_t lli_layout_lock;
277 struct rw_semaphore lli_xattrs_list_rwsem;
278 struct mutex lli_xattrs_enq_lock;
279 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
282 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
286 spin_lock(&lli->lli_layout_lock);
287 gen = lli->lli_layout_gen;
288 spin_unlock(&lli->lli_layout_lock);
293 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
295 spin_lock(&lli->lli_layout_lock);
296 lli->lli_layout_gen = gen;
297 spin_unlock(&lli->lli_layout_lock);
300 int ll_xattr_cache_destroy(struct inode *inode);
302 int ll_xattr_cache_get(struct inode *inode,
309 * Locking to guarantee consistency of non-atomic updates to long long i_size,
310 * consistency between file size and KMS.
312 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
315 void ll_inode_size_lock(struct inode *inode);
316 void ll_inode_size_unlock(struct inode *inode);
318 // FIXME: replace the name of this with LL_I to conform to kernel stuff
319 // static inline struct ll_inode_info *LL_I(struct inode *inode)
320 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
322 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
325 /* default to about 40meg of readahead on a given system. That much tied
326 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
327 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT))
329 /* default to read-ahead full files smaller than 2MB on the second read */
330 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
335 RA_STAT_DISTANT_READPAGE,
336 RA_STAT_MISS_IN_WINDOW,
337 RA_STAT_FAILED_GRAB_PAGE,
338 RA_STAT_FAILED_MATCH,
343 RA_STAT_MAX_IN_FLIGHT,
344 RA_STAT_WRONG_GRAB_PAGE,
345 RA_STAT_FAILED_REACH_END,
350 atomic_t ra_cur_pages;
351 unsigned long ra_max_pages;
352 unsigned long ra_max_pages_per_file;
353 unsigned long ra_max_read_ahead_whole_pages;
356 /* ra_io_arg will be filled in the beginning of ll_readahead with
357 * ras_lock, then the following ll_read_ahead_pages will read RA
358 * pages according to this arg, all the items in this structure are
359 * counted by page index.
362 unsigned long ria_start; /* start offset of read-ahead*/
363 unsigned long ria_end; /* end offset of read-ahead*/
364 /* If stride read pattern is detected, ria_stoff means where
365 * stride read is started. Note: for normal read-ahead, the
366 * value here is meaningless, and also it will not be accessed*/
368 /* ria_length and ria_pages are the length and pages length in the
369 * stride I/O mode. And they will also be used to check whether
370 * it is stride I/O read-ahead in the read-ahead pages*/
371 unsigned long ria_length;
372 unsigned long ria_pages;
375 /* LL_HIST_MAX=32 causes an overflow */
376 #define LL_HIST_MAX 28
377 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
378 #define LL_PROCESS_HIST_MAX 10
379 struct per_process_info {
381 struct obd_histogram pp_r_hist;
382 struct obd_histogram pp_w_hist;
385 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
386 struct ll_rw_extents_info {
387 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
390 #define LL_OFFSET_HIST_MAX 100
391 struct ll_rw_process_info {
394 loff_t rw_range_start;
396 loff_t rw_last_file_pos;
398 size_t rw_smallest_extent;
399 size_t rw_largest_extent;
400 struct ll_file_data *rw_last_file;
403 enum stats_track_type {
404 STATS_TRACK_ALL = 0, /* track all processes */
405 STATS_TRACK_PID, /* track process with this pid */
406 STATS_TRACK_PPID, /* track processes with this ppid */
407 STATS_TRACK_GID, /* track processes with this gid */
411 /* flags for sbi->ll_flags */
412 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
413 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
414 #define LL_SBI_FLOCK 0x04
415 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
416 #define LL_SBI_ACL 0x10 /* support ACL */
417 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
418 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
419 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
420 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
421 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
422 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
423 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
424 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
425 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
426 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
427 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
428 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
429 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
430 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
431 #define LL_SBI_NOROOTSQUASH 0x100000 /* do not apply root squash */
433 #define LL_SBI_FLAGS { \
457 #define RCE_HASHES 32
459 struct rmtacl_ctl_entry {
460 struct list_head rce_list;
461 pid_t rce_key; /* hash key */
462 int rce_ops; /* acl operation type */
465 struct rmtacl_ctl_table {
467 struct list_head rct_entries[RCE_HASHES];
473 struct list_head ee_list;
474 pid_t ee_key; /* hash key */
475 struct lu_fid ee_fid;
476 int ee_type; /* ACL type for ACCESS or DEFAULT */
477 ext_acl_xattr_header *ee_acl;
482 struct list_head et_entries[EE_HASHES];
486 struct list_head ll_list;
487 /* this protects pglist and ra_info. It isn't safe to
488 * grab from interrupt contexts */
490 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
491 spinlock_t ll_process_lock; /* ll_rw_process_info */
492 struct obd_uuid ll_sb_uuid;
493 struct obd_export *ll_md_exp;
494 struct obd_export *ll_dt_exp;
495 struct proc_dir_entry* ll_proc_root;
496 struct lu_fid ll_root_fid; /* root object fid */
499 unsigned int ll_umounting:1,
500 ll_xattr_cache_enabled:1;
501 /* per-conn chain of SBs */
502 struct list_head ll_conn_chain;
503 struct lustre_client_ocd ll_lco;
505 /*please don't ask -p*/
506 struct list_head ll_orphan_dentry_list;
507 struct ll_close_queue *ll_lcq;
509 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
511 /* Used to track "unstable" pages on a client, and maintain a
512 * LRU list of clean pages. An "unstable" page is defined as
513 * any page which is sent to a server as part of a bulk request,
514 * but is uncommitted to stable storage. */
515 struct cl_client_cache ll_cache;
517 struct lprocfs_stats *ll_ra_stats;
519 struct ll_ra_info ll_ra_info;
520 unsigned int ll_namelen;
521 struct file_operations *ll_fop;
523 /* =0 - hold lock over whole read/write
524 * >0 - max. chunk to be read/written w/o lock re-acquiring */
525 unsigned long ll_max_rw_chunk;
526 unsigned int ll_md_brw_pages; /* readdir pages per RPC */
528 struct lu_site *ll_site;
529 struct cl_device *ll_cl;
531 struct ll_rw_extents_info ll_rw_extents_info;
532 int ll_extent_process_count;
533 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
534 unsigned int ll_offset_process_count;
535 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
536 unsigned int ll_rw_offset_entry_count;
537 int ll_stats_track_id;
538 enum stats_track_type ll_stats_track_type;
541 /* metadata stat-ahead */
542 unsigned int ll_sa_max; /* max statahead RPCs */
543 atomic_t ll_sa_total; /* statahead thread started
545 atomic_t ll_sa_wrong; /* statahead thread stopped for
547 atomic_t ll_agl_total; /* AGL thread started count */
549 dev_t ll_sdev_orig; /* save s_dev before assign for
551 struct rmtacl_ctl_table ll_rct;
552 struct eacl_table ll_et;
555 struct root_squash_info ll_squash;
558 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
563 struct task_struct *lrr_reader;
564 struct list_head lrr_linkage;
568 * per file-descriptor read-ahead data.
570 struct ll_readahead_state {
573 * index of the last page that read(2) needed and that wasn't in the
574 * cache. Used by ras_update() to detect seeks.
576 * XXX nikita: if access seeks into cached region, Lustre doesn't see
579 unsigned long ras_last_readpage;
581 * number of pages read after last read-ahead window reset. As window
582 * is reset on each seek, this is effectively a number of consecutive
583 * accesses. Maybe ->ras_accessed_in_window is better name.
585 * XXX nikita: window is also reset (by ras_update()) when Lustre
586 * believes that memory pressure evicts read-ahead pages. In that
587 * case, it probably doesn't make sense to expand window to
588 * PTLRPC_MAX_BRW_PAGES on the third access.
590 unsigned long ras_consecutive_pages;
592 * number of read requests after the last read-ahead window reset
593 * As window is reset on each seek, this is effectively the number
594 * on consecutive read request and is used to trigger read-ahead.
596 unsigned long ras_consecutive_requests;
598 * Parameters of current read-ahead window. Handled by
599 * ras_update(). On the initial access to the file or after a seek,
600 * window is reset to 0. After 3 consecutive accesses, window is
601 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
602 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
604 unsigned long ras_window_start, ras_window_len;
606 * Where next read-ahead should start at. This lies within read-ahead
607 * window. Read-ahead window is read in pieces rather than at once
608 * because: 1. lustre limits total number of pages under read-ahead by
609 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
610 * not covered by DLM lock.
612 unsigned long ras_next_readahead;
614 * Total number of ll_file_read requests issued, reads originating
615 * due to mmap are not counted in this total. This value is used to
616 * trigger full file read-ahead after multiple reads to a small file.
618 unsigned long ras_requests;
620 * Page index with respect to the current request, these value
621 * will not be accurate when dealing with reads issued via mmap.
623 unsigned long ras_request_index;
625 * list of struct ll_ra_read's one per read(2) call current in
626 * progress against this file descriptor. Used by read-ahead code,
627 * protected by ->ras_lock.
629 struct list_head ras_read_beads;
631 * The following 3 items are used for detecting the stride I/O
633 * In stride I/O mode,
634 * ...............|-----data-----|****gap*****|--------|******|....
635 * offset |-stride_pages-|-stride_gap-|
636 * ras_stride_offset = offset;
637 * ras_stride_length = stride_pages + stride_gap;
638 * ras_stride_pages = stride_pages;
639 * Note: all these three items are counted by pages.
641 unsigned long ras_stride_length;
642 unsigned long ras_stride_pages;
643 pgoff_t ras_stride_offset;
645 * number of consecutive stride request count, and it is similar as
646 * ras_consecutive_requests, but used for stride I/O mode.
647 * Note: only more than 2 consecutive stride request are detected,
648 * stride read-ahead will be enable
650 unsigned long ras_consecutive_stride_requests;
653 extern struct kmem_cache *ll_file_data_slab;
654 struct lustre_handle;
655 struct ll_file_data {
656 struct ll_readahead_state fd_ras;
657 struct ccc_grouplock fd_grouplock;
661 /* openhandle if lease exists for this file.
662 * Borrow lli->lli_och_mutex to protect assignment */
663 struct obd_client_handle *fd_lease_och;
664 struct obd_client_handle *fd_och;
665 struct file *fd_file;
666 /* Indicate whether need to report failure when close.
667 * true: failure is known, not report again.
668 * false: unknown failure, should report. */
669 bool fd_write_failed;
670 rwlock_t fd_lock; /* protect lcc list */
671 struct list_head fd_lccs; /* list of ll_cl_context */
674 struct lov_stripe_md;
676 extern spinlock_t inode_lock;
678 extern struct proc_dir_entry *proc_lustre_fs_root;
680 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
682 return &lli->lli_vfs_inode;
685 __u32 ll_i2suppgid(struct inode *i);
686 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
688 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
690 #if BITS_PER_LONG == 32
692 #elif defined(CONFIG_COMPAT)
693 return unlikely(is_compat_task() || (sbi->ll_flags & LL_SBI_32BIT_API));
695 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
699 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
700 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
701 struct ll_ra_read *ll_ra_read_get(struct file *f);
703 /* llite/lproc_llite.c */
705 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
706 struct super_block *sb, char *osc, char *mdc);
707 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
708 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
709 extern struct lprocfs_seq_vars lprocfs_llite_obd_vars[];
711 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
712 struct super_block *sb, char *osc, char *mdc){return 0;}
713 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
714 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
719 extern const struct file_operations ll_dir_operations;
720 extern const struct inode_operations ll_dir_inode_operations;
721 #ifdef HAVE_DIR_CONTEXT
722 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
723 struct dir_context *ctx);
725 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
726 void *cookie, filldir_t filldir);
728 int ll_get_mdt_idx(struct inode *inode);
729 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
730 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
731 __u64 offset, struct ll_dir_chain *chain);
732 void ll_release_page(struct inode *inode, struct page *page, bool remove);
735 extern const struct inode_operations ll_special_inode_operations;
737 int ll_objects_destroy(struct ptlrpc_request *request,
739 struct inode *ll_iget(struct super_block *sb, ino_t hash,
740 struct lustre_md *lic);
741 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
742 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
743 void *data, int flag);
744 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
745 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
746 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
749 int ll_writepage(struct page *page, struct writeback_control *wbc);
750 int ll_writepages(struct address_space *, struct writeback_control *wbc);
751 int ll_readpage(struct file *file, struct page *page);
752 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
753 int ll_readahead(const struct lu_env *env, struct cl_io *io,
754 struct cl_page_list *queue, struct ll_readahead_state *ras,
756 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
757 struct ll_cl_context *ll_cl_find(struct file *file);
758 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io);
759 void ll_cl_remove(struct file *file, const struct lu_env *env);
761 #ifndef MS_HAS_NEW_AOPS
762 extern const struct address_space_operations ll_aops;
764 extern const struct address_space_operations_ext ll_aops;
768 extern struct file_operations ll_file_operations;
769 extern struct file_operations ll_file_operations_flock;
770 extern struct file_operations ll_file_operations_noflock;
771 extern struct inode_operations ll_file_inode_operations;
772 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
773 ldlm_mode_t l_req_mode);
774 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
775 struct lustre_handle *lockh, __u64 flags,
778 int ll_file_open(struct inode *inode, struct file *file);
779 int ll_file_release(struct inode *inode, struct file *file);
780 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
781 struct lov_stripe_md *lsm, lstat_t *st);
782 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
783 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
784 int ll_md_real_close(struct inode *inode, fmode_t fmode);
785 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
786 struct obd_client_handle **och, unsigned long flags);
787 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
788 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
789 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
790 __u64 ioepoch, int sync);
791 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
792 struct lustre_handle *fh);
793 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
794 struct ll_file_data *file, loff_t pos,
795 size_t count, int rw);
796 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
797 struct posix_acl *ll_get_acl(struct inode *inode, int type);
798 int ll_migrate(struct inode *parent, struct file *file, int mdtidx,
799 const char *name, int namelen);
800 int ll_get_fid_by_name(struct inode *parent, const char *name,
801 int namelen, struct lu_fid *fid);
802 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
803 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
805 # ifndef HAVE_INODE_PERMISION_2ARGS
806 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
808 int ll_inode_permission(struct inode *inode, int mask);
812 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
813 __u64 flags, struct lov_user_md *lum,
815 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
816 struct lov_mds_md **lmm, int *lmm_size,
817 struct ptlrpc_request **request);
818 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
820 int ll_dir_getstripe(struct inode *inode, void **lmmp,
821 int *lmm_size, struct ptlrpc_request **request,
823 #ifdef HAVE_FILE_FSYNC_4ARGS
824 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
825 #elif defined(HAVE_FILE_FSYNC_2ARGS)
826 int ll_fsync(struct file *file, int data);
828 int ll_fsync(struct file *file, struct dentry *dentry, int data);
830 int ll_merge_lvb(const struct lu_env *env, struct inode *inode);
831 int ll_fid2path(struct inode *inode, void __user *arg);
832 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
833 int ll_hsm_release(struct inode *inode);
837 int ll_d_init(struct dentry *de);
838 extern const struct dentry_operations ll_d_ops;
839 void ll_intent_drop_lock(struct lookup_intent *);
840 void ll_intent_release(struct lookup_intent *);
841 void ll_invalidate_aliases(struct inode *);
842 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
843 int ll_revalidate_it_finish(struct ptlrpc_request *request,
844 struct lookup_intent *it, struct dentry *de);
846 /* llite/llite_lib.c */
847 extern struct super_operations lustre_super_operations;
849 void ll_lli_init(struct ll_inode_info *lli);
850 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
851 void ll_put_super(struct super_block *sb);
852 void ll_kill_super(struct super_block *sb);
853 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
854 void ll_clear_inode(struct inode *inode);
855 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import);
856 int ll_setattr(struct dentry *de, struct iattr *attr);
857 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
858 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
859 __u64 max_age, __u32 flags);
860 int ll_update_inode(struct inode *inode, struct lustre_md *md);
861 int ll_read_inode2(struct inode *inode, void *opaque);
862 void ll_delete_inode(struct inode *inode);
863 int ll_iocontrol(struct inode *inode, struct file *file,
864 unsigned int cmd, unsigned long arg);
865 int ll_flush_ctx(struct inode *inode);
866 void ll_umount_begin(struct super_block *sb);
867 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
868 #ifdef HAVE_SUPEROPS_USE_DENTRY
869 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
871 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
873 void ll_dirty_page_discard_warn(struct page *page, int ioret);
874 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
875 struct super_block *, struct lookup_intent *);
876 void lustre_dump_dentry(struct dentry *, int recur);
877 int ll_obd_statfs(struct inode *inode, void *arg);
878 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
879 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
880 int ll_get_max_cookiesize(struct ll_sb_info *sbi, int *max_cookiesize);
881 int ll_get_default_cookiesize(struct ll_sb_info *sbi, int *default_cookiesize);
882 int ll_process_config(struct lustre_cfg *lcfg);
883 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
884 struct inode *i1, struct inode *i2,
885 const char *name, int namelen,
886 int mode, __u32 opc, void *data);
887 void ll_finish_md_op_data(struct md_op_data *op_data);
888 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
889 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
890 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
892 /* llite/llite_nfs.c */
893 extern struct export_operations lustre_export_operations;
894 __u32 get_uuid2int(const char *name, int len);
895 struct inode *search_inode_for_lustre(struct super_block *sb,
896 const struct lu_fid *fid);
897 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
899 /* llite/symlink.c */
900 extern struct inode_operations ll_fast_symlink_inode_operations;
902 /* llite/llite_close.c */
903 struct ll_close_queue {
905 struct list_head lcq_head;
906 wait_queue_head_t lcq_waitq;
907 struct completion lcq_comp;
911 struct ccc_object *cl_inode2ccc(struct inode *inode);
914 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
915 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
917 /* specific achitecture can implement only part of this list */
918 enum vvp_io_subtype {
921 /** io started from splice_{read|write} */
928 enum vvp_io_subtype cui_io_subtype;
932 struct pipe_inode_info *cui_pipe;
933 unsigned int cui_flags;
935 struct vvp_fault_io {
937 * Inode modification time that is checked across DLM
941 struct vm_area_struct *ft_vma;
943 * locked page returned from vvp_io
945 struct page *ft_vmpage;
946 struct vm_fault_api {
950 struct vm_fault *ft_vmf;
952 * fault API used bitflags for return code.
954 unsigned int ft_flags;
956 * check that flags are from filemap_fault
963 * Read-ahead state used by read and page-fault IO contexts.
965 struct ll_ra_read cui_bead;
967 * Set when cui_bead has been initialized.
969 int cui_ra_window_set;
973 * IO arguments for various VFS I/O interfaces.
976 /** normal/sendfile/splice */
977 enum vvp_io_subtype via_io_subtype;
981 struct kiocb *via_iocb;
982 struct iovec *via_iov;
983 unsigned long via_nrsegs;
986 struct pipe_inode_info *via_pipe;
987 unsigned int via_flags;
992 struct ll_cl_context {
993 struct list_head lcc_list;
995 const struct lu_env *lcc_env;
996 struct cl_io *lcc_io;
997 struct cl_page *lcc_page;
1000 struct vvp_thread_info {
1001 struct iovec vti_local_iov;
1002 struct vvp_io_args vti_args;
1003 struct ra_io_arg vti_ria;
1004 struct kiocb vti_kiocb;
1005 struct ll_cl_context vti_io_ctx;
1008 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
1010 extern struct lu_context_key vvp_key;
1011 struct vvp_thread_info *info;
1013 info = lu_context_key_get(&env->le_ctx, &vvp_key);
1014 LASSERT(info != NULL);
1018 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
1019 enum vvp_io_subtype type)
1021 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
1023 ret->via_io_subtype = type;
1028 struct vvp_session {
1029 struct vvp_io vs_ios;
1032 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1034 extern struct lu_context_key vvp_session_key;
1035 struct vvp_session *ses;
1037 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1038 LASSERT(ses != NULL);
1042 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1044 return &vvp_env_session(env)->vs_ios;
1047 int vvp_global_init(void);
1048 void vvp_global_fini(void);
1050 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1051 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1052 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1054 /* llite/llite_mmap.c */
1056 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1057 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1058 void policy_from_vma(ldlm_policy_data_t *policy,
1059 struct vm_area_struct *vma, unsigned long addr, size_t count);
1060 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1063 static inline void ll_invalidate_page(struct page *vmpage)
1065 struct address_space *mapping = vmpage->mapping;
1066 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1068 LASSERT(PageLocked(vmpage));
1069 if (mapping == NULL)
1073 * truncate_complete_page() calls
1074 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
1076 ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE);
1077 truncate_complete_page(mapping, vmpage);
1080 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1082 /* don't need an addref as the sb_info should be holding one */
1083 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1085 return ll_s2sbi(sb)->ll_dt_exp;
1088 /* don't need an addref as the sb_info should be holding one */
1089 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1091 return ll_s2sbi(sb)->ll_md_exp;
1094 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1096 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1102 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1103 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1105 return ll_s2sbi(inode->i_sb);
1108 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1110 return ll_s2dtexp(inode->i_sb);
1113 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1115 return ll_s2mdexp(inode->i_sb);
1118 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1122 LASSERT(inode != NULL);
1123 fid = &ll_i2info(inode)->lli_fid;
1128 static inline __u64 ll_file_maxbytes(struct inode *inode)
1130 return ll_i2info(inode)->lli_maxbytes;
1134 int ll_setxattr(struct dentry *dentry, const char *name,
1135 const void *value, size_t size, int flags);
1136 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1137 void *buffer, size_t size);
1138 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1139 int ll_removexattr(struct dentry *dentry, const char *name);
1141 /* llite/remote_perm.c */
1142 extern struct kmem_cache *ll_remote_perm_cachep;
1143 extern struct kmem_cache *ll_rmtperm_hash_cachep;
1145 void free_rmtperm_hash(struct hlist_head *hash);
1146 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1147 int lustre_check_remote_perm(struct inode *inode, int mask);
1149 /* llite/llite_capa.c */
1150 extern struct timer_list ll_capa_timer;
1152 int ll_capa_thread_start(void);
1153 void ll_capa_thread_stop(void);
1154 void ll_capa_timer_callback(unsigned long unused);
1156 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1158 void ll_capa_open(struct inode *inode);
1159 void ll_capa_close(struct inode *inode);
1161 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1162 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1164 void ll_truncate_free_capa(struct obd_capa *ocapa);
1165 void ll_clear_inode_capas(struct inode *inode);
1166 void ll_print_capa_stat(struct ll_sb_info *sbi);
1168 /* llite/llite_cl.c */
1169 extern struct lu_device_type vvp_device_type;
1172 * Common IO arguments for various VFS I/O interfaces.
1174 int cl_sb_init(struct super_block *sb);
1175 int cl_sb_fini(struct super_block *sb);
1177 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1178 struct ll_readahead_state *ras, unsigned long index,
1180 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1181 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1183 /* llite/llite_rmtacl.c */
1184 #ifdef CONFIG_FS_POSIX_ACL
1185 obd_valid rce_ops2valid(int ops);
1186 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1187 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1188 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1189 void rct_init(struct rmtacl_ctl_table *rct);
1190 void rct_fini(struct rmtacl_ctl_table *rct);
1192 void ee_free(struct eacl_entry *ee);
1193 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1194 ext_acl_xattr_header *header);
1195 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1196 struct lu_fid *fid, int type);
1197 void et_search_free(struct eacl_table *et, pid_t key);
1198 void et_init(struct eacl_table *et);
1199 void et_fini(struct eacl_table *et);
1204 #define LL_SA_RPC_MIN 2
1205 #define LL_SA_RPC_DEF 32
1206 #define LL_SA_RPC_MAX 8192
1208 #define LL_SA_CACHE_BIT 5
1209 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1210 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1212 /* per inode struct, for dir only */
1213 struct ll_statahead_info {
1214 struct inode *sai_inode;
1215 atomic_t sai_refcount; /* when access this struct, hold
1217 unsigned int sai_generation; /* generation for statahead */
1218 unsigned int sai_max; /* max ahead of lookup */
1219 __u64 sai_sent; /* stat requests sent count */
1220 __u64 sai_replied; /* stat requests which received
1222 __u64 sai_index; /* index of statahead entry */
1223 __u64 sai_index_wait; /* index of entry which is the
1224 * caller is waiting for */
1225 __u64 sai_hit; /* hit count */
1226 __u64 sai_miss; /* miss count:
1227 * for "ls -al" case, it
1228 * includes hidden dentry miss;
1229 * for "ls -l" case, it does not
1230 * include hidden dentry miss.
1231 * "sai_miss_hidden" is used for
1234 unsigned int sai_consecutive_miss; /* consecutive miss */
1235 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1236 * is not a hidden one */
1237 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1238 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1240 sai_agl_valid:1,/* AGL is valid for the dir */
1241 sai_in_readpage:1;/* statahead is in readdir()*/
1242 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1243 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1244 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1245 struct list_head sai_entries; /* entry list */
1246 struct list_head sai_entries_received; /* entries returned */
1247 struct list_head sai_entries_stated; /* entries stated */
1248 struct list_head sai_entries_agl; /* AGL entries to be sent */
1249 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1250 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1251 atomic_t sai_cache_count; /* entry count in cache */
1254 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1256 void ll_stop_statahead(struct inode *dir, void *key);
1258 static inline int ll_glimpse_size(struct inode *inode)
1260 struct ll_inode_info *lli = ll_i2info(inode);
1263 down_read(&lli->lli_glimpse_sem);
1264 rc = cl_glimpse_size(inode);
1265 lli->lli_glimpse_time = cfs_time_current();
1266 up_read(&lli->lli_glimpse_sem);
1271 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1273 struct ll_inode_info *lli = ll_i2info(dir);
1274 struct ll_statahead_info *sai = lli->lli_sai;
1275 struct ll_dentry_data *ldd = ll_d2d(dentry);
1277 /* not the same process, don't mark */
1278 if (lli->lli_opendir_pid != current_pid())
1281 LASSERT(ldd != NULL);
1283 ldd->lld_sa_generation = sai->sai_generation;
1287 d_need_statahead(struct inode *dir, struct dentry *dentryp)
1289 struct ll_inode_info *lli;
1290 struct ll_dentry_data *ldd;
1292 if (ll_i2sbi(dir)->ll_sa_max == 0)
1295 lli = ll_i2info(dir);
1296 /* not the same process, don't statahead */
1297 if (lli->lli_opendir_pid != current_pid())
1300 /* statahead has been stopped */
1301 if (lli->lli_opendir_key == NULL)
1304 ldd = ll_d2d(dentryp);
1306 * When stats a dentry, the system trigger more than once "revalidate"
1307 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1308 * Under patchless client mode, the operation intent is not accurate,
1309 * which maybe misguide the statahead thread. For example:
1310 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1311 * have the same operation intent -- "IT_GETATTR".
1312 * In fact, one dentry should has only one chance to interact with the
1313 * statahead thread, otherwise the statahead windows will be confused.
1314 * The solution is as following:
1315 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1316 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1317 * will bypass interacting with statahead thread for checking:
1318 * "lld_sa_generation == lli_sai->sai_generation"
1320 if (ldd && lli->lli_sai &&
1321 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1328 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1332 ret = d_need_statahead(dir, *dentryp);
1336 return do_statahead_enter(dir, dentryp, only_unplug);
1339 /* llite ioctl register support rountine */
1345 #define LLIOC_MAX_CMD 256
1348 * Rules to write a callback function:
1351 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1352 * returned to ll_iocontrol_register. Callback functions should use this
1353 * data to check the potential collasion of ioctl cmd. If collasion is
1354 * found, callback function should return LLIOC_CONT.
1355 * @rcp: The result of ioctl command.
1358 * If @magic matches the pointer returned by ll_iocontrol_data, the
1359 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1361 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1362 struct file *file, unsigned int cmd, unsigned long arg,
1363 void *magic, int *rcp);
1365 /* export functions */
1366 /* Register ioctl block dynamatically for a regular file.
1368 * @cmd: the array of ioctl command set
1369 * @count: number of commands in the @cmd
1370 * @cb: callback function, it will be called if an ioctl command is found to
1371 * belong to the command list @cmd.
1374 * A magic pointer will be returned if success;
1375 * otherwise, NULL will be returned.
1377 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1378 void ll_iocontrol_unregister(void *magic);
1381 /* lclient compat stuff */
1382 #define cl_inode_info ll_inode_info
1383 #define cl_i2info(info) ll_i2info(info)
1384 #define cl_inode_mode(inode) ((inode)->i_mode)
1385 #define cl_i2sbi ll_i2sbi
1387 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1388 const struct iattr *attr)
1390 LASSERT(attr->ia_valid & ATTR_FILE);
1391 return LUSTRE_FPRIVATE(attr->ia_file);
1394 static inline void cl_isize_lock(struct inode *inode)
1396 ll_inode_size_lock(inode);
1399 static inline void cl_isize_unlock(struct inode *inode)
1401 ll_inode_size_unlock(inode);
1404 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1406 LASSERT(mutex_is_locked(&ll_i2info(inode)->lli_size_mutex));
1407 i_size_write(inode, kms);
1410 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1412 ll_inode_size_lock(inode);
1413 i_size_write(inode, kms);
1414 ll_inode_size_unlock(inode);
1417 #define cl_isize_read(inode) i_size_read(inode)
1419 static inline int cl_merge_lvb(const struct lu_env *env, struct inode *inode)
1421 return ll_merge_lvb(env, inode);
1424 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1425 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1426 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1428 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1430 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1431 enum cl_fsync_mode mode, int ignore_layout);
1433 /** direct write pages */
1434 struct ll_dio_pages {
1435 /** page array to be written. we don't support
1436 * partial pages except the last one. */
1437 struct page **ldp_pages;
1438 /* offset of each page */
1439 loff_t *ldp_offsets;
1440 /** if ldp_offsets is NULL, it means a sequential
1441 * pages to be written, then this is the file offset
1442 * of the * first page. */
1443 loff_t ldp_start_offset;
1444 /** how many bytes are to be written. */
1446 /** # of pages in the array. */
1450 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1453 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1456 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1459 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1460 int rw, struct inode *inode,
1461 struct ll_dio_pages *pv);
1463 static inline int ll_file_nolock(const struct file *file)
1465 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1466 struct inode *inode = file->f_dentry->d_inode;
1468 LASSERT(fd != NULL);
1469 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1470 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1473 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1474 struct lookup_intent *it, __u64 *bits)
1476 if (!it->d.lustre.it_lock_set) {
1477 struct lustre_handle handle;
1479 /* If this inode is a remote object, it will get two
1480 * separate locks in different namespaces, Master MDT,
1481 * where the name entry is, will grant LOOKUP lock,
1482 * remote MDT, where the object is, will grant
1483 * UPDATE|PERM lock. The inode will be attched to both
1484 * LOOKUP and PERM locks, so revoking either locks will
1485 * case the dcache being cleared */
1486 if (it->d.lustre.it_remote_lock_mode) {
1487 handle.cookie = it->d.lustre.it_remote_lock_handle;
1488 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1489 "(%p) for remote lock "LPX64"\n",
1490 PFID(ll_inode2fid(inode)), inode,
1492 md_set_lock_data(exp, &handle.cookie, inode, NULL);
1495 handle.cookie = it->d.lustre.it_lock_handle;
1497 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1498 " for lock "LPX64"\n",
1499 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1501 md_set_lock_data(exp, &handle.cookie, inode,
1502 &it->d.lustre.it_lock_bits);
1503 it->d.lustre.it_lock_set = 1;
1507 *bits = it->d.lustre.it_lock_bits;
1510 static inline void ll_lock_dcache(struct inode *inode)
1512 #ifdef HAVE_DCACHE_LOCK
1513 spin_lock(&dcache_lock);
1515 spin_lock(&inode->i_lock);
1519 static inline void ll_unlock_dcache(struct inode *inode)
1521 #ifdef HAVE_DCACHE_LOCK
1522 spin_unlock(&dcache_lock);
1524 spin_unlock(&inode->i_lock);
1528 static inline int d_lustre_invalid(const struct dentry *dentry)
1530 struct ll_dentry_data *lld = ll_d2d(dentry);
1532 return (lld == NULL) || lld->lld_invalid;
1535 static inline void __d_lustre_invalidate(struct dentry *dentry)
1537 struct ll_dentry_data *lld = ll_d2d(dentry);
1540 lld->lld_invalid = 1;
1544 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1545 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1546 * else dput() of the last refcount will unhash this dentry and kill it.
1548 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1550 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1551 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1552 dentry->d_parent, dentry->d_inode, d_count(dentry));
1554 spin_lock_nested(&dentry->d_lock,
1555 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1556 __d_lustre_invalidate(dentry);
1557 if (d_count(dentry) == 0)
1559 spin_unlock(&dentry->d_lock);
1562 static inline void d_lustre_revalidate(struct dentry *dentry)
1564 spin_lock(&dentry->d_lock);
1565 LASSERT(ll_d2d(dentry) != NULL);
1566 ll_d2d(dentry)->lld_invalid = 0;
1567 spin_unlock(&dentry->d_lock);
1570 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0)
1571 /* Compatibility for old (1.8) compiled userspace quota code */
1572 struct if_quotactl_18 {
1577 struct obd_dqinfo qc_dqinfo;
1578 struct obd_dqblk qc_dqblk;
1580 struct obd_uuid obd_uuid;
1582 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1583 /* End compatibility for old (1.8) compiled userspace quota code */
1585 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1586 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */
1589 LL_LAYOUT_GEN_NONE = ((__u32)-2), /* layout lock was cancelled */
1590 LL_LAYOUT_GEN_EMPTY = ((__u32)-1) /* for empty layout */
1593 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1594 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1595 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1597 int ll_xattr_init(void);
1598 void ll_xattr_fini(void);
1600 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1601 struct cl_page *page, enum cl_req_type crt);
1603 #endif /* LLITE_INTERNAL_H */