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.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2016, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
33 #ifndef LLITE_INTERNAL_H
34 #define LLITE_INTERNAL_H
35 #include <lustre_debug.h>
36 #include <lustre_ver.h>
37 #include <lustre_disk.h> /* for s2sbi */
38 #include <lustre_eacl.h>
39 #include <lustre_linkea.h>
41 /* for struct cl_lock_descr and struct cl_io */
42 #include <cl_object.h>
43 #include <lustre_lmv.h>
44 #include <lustre_mdc.h>
45 #include <lustre_intent.h>
46 #include <linux/compat.h>
47 #include <linux/aio.h>
49 #include <lustre_compat.h>
50 #include "vvp_internal.h"
51 #include "range_lock.h"
57 #ifndef VM_FAULT_RETRY
58 #define VM_FAULT_RETRY 0
61 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
62 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
63 #ifndef LOOKUP_CONTINUE
64 #define LOOKUP_CONTINUE LOOKUP_PARENT
67 /** Only used on client-side for indicating the tail of dir hash/offset. */
68 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
69 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
71 /* 4UL * 1024 * 1024 */
72 #define LL_MAX_BLKSIZE_BITS 22
74 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
75 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
77 struct ll_dentry_data {
78 struct lookup_intent *lld_it;
79 unsigned int lld_sa_generation;
80 unsigned int lld_invalid:1;
81 unsigned int lld_nfs_dentry:1;
82 struct rcu_head lld_rcu_head;
85 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
87 #define LLI_INODE_MAGIC 0x111d0de5
88 #define LLI_INODE_DEAD 0xdeadd00d
90 struct ll_getname_data {
91 #ifdef HAVE_DIR_CONTEXT
92 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? */
100 struct lu_env *lg_env;
102 struct cl_lock *lg_lock;
103 unsigned long lg_gid;
106 struct ll_inode_info {
107 __u32 lli_inode_magic;
110 volatile unsigned long lli_flags;
111 struct posix_acl *lli_posix_acl;
113 /* identifying fields for both metadata and data stacks. */
114 struct lu_fid lli_fid;
115 /* master inode fid for stripe directory */
116 struct lu_fid lli_pfid;
118 /* We need all three because every inode may be opened in different
120 struct obd_client_handle *lli_mds_read_och;
121 struct obd_client_handle *lli_mds_write_och;
122 struct obd_client_handle *lli_mds_exec_och;
123 __u64 lli_open_fd_read_count;
124 __u64 lli_open_fd_write_count;
125 __u64 lli_open_fd_exec_count;
126 /* Protects access to och pointers and their usage counters */
127 struct mutex lli_och_mutex;
129 struct inode lli_vfs_inode;
131 /* the most recent timestamps obtained from mds */
135 spinlock_t lli_agl_lock;
137 /* update atime from MDS no matter if it's older than
138 * local inode atime. */
139 unsigned int lli_update_atime:1,
142 /* Try to make the d::member and f::member are aligned. Before using
143 * these members, make clear whether it is directory or not. */
147 /* serialize normal readdir and statahead-readdir. */
148 struct mutex lli_readdir_mutex;
150 /* metadata statahead */
151 /* since parent-child threads can share the same @file
152 * struct, "opendir_key" is the token when dir close for
153 * case of parent exit before child -- it is me should
154 * cleanup the dir readahead. */
155 void *lli_opendir_key;
156 struct ll_statahead_info *lli_sai;
157 /* protect statahead stuff. */
158 spinlock_t lli_sa_lock;
159 /* "opendir_pid" is the token when lookup/revalid
160 * -- I am the owner of dir statahead. */
161 pid_t lli_opendir_pid;
162 /* stat will try to access statahead entries or start
163 * statahead if this flag is set, and this flag will be
164 * set upon dir open, and cleared when dir is closed,
165 * statahead hit ratio is too low, or start statahead
167 unsigned int lli_sa_enabled:1;
168 /* generation for statahead */
169 unsigned int lli_sa_generation;
170 /* directory stripe information */
171 struct lmv_stripe_md *lli_lsm_md;
172 /* default directory stripe offset. This is extracted
173 * from the "dmv" xattr in order to decide which MDT to
174 * create a subdirectory on. The MDS itself fetches
175 * "dmv" and gets the rest of the default layout itself
176 * (count, hash, etc). */
177 __u32 lli_def_stripe_offset;
180 /* for non-directory */
182 struct mutex lli_size_mutex;
183 char *lli_symlink_name;
185 * struct rw_semaphore {
186 * signed long count; // align d.d_def_acl
187 * spinlock_t wait_lock; // align d.d_sa_lock
188 * struct list_head wait_list;
191 struct rw_semaphore lli_trunc_sem;
192 struct range_lock_tree lli_write_tree;
194 struct rw_semaphore lli_glimpse_sem;
195 cfs_time_t lli_glimpse_time;
196 struct list_head lli_agl_list;
199 /* for writepage() only to communicate to fsync */
203 * whenever a process try to read/write the file, the
204 * jobid of the process will be saved here, and it'll
205 * be packed into the write PRC when flush later.
207 * so the read/write statistics for jobid will not be
208 * accurate if the file is shared by different jobs.
210 char lli_jobid[LUSTRE_JOBID_SIZE];
214 /* XXX: For following frequent used members, although they maybe special
215 * used for non-directory object, it is some time-wasting to check
216 * whether the object is directory or not before using them. On the
217 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
218 * the "ll_inode_info" size even if moving those members into u.f.
219 * So keep them out side.
221 * In the future, if more members are added only for directory,
222 * some of the following members can be moved into u.f.
224 struct cl_object *lli_clob;
226 /* mutex to request for layout lock exclusively. */
227 struct mutex lli_layout_mutex;
228 /* Layout version, protected by lli_layout_lock */
229 __u32 lli_layout_gen;
230 spinlock_t lli_layout_lock;
232 __u32 lli_projid; /* project id */
234 struct rw_semaphore lli_xattrs_list_rwsem;
235 struct mutex lli_xattrs_enq_lock;
236 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
239 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
243 spin_lock(&lli->lli_layout_lock);
244 gen = lli->lli_layout_gen;
245 spin_unlock(&lli->lli_layout_lock);
250 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
252 spin_lock(&lli->lli_layout_lock);
253 lli->lli_layout_gen = gen;
254 spin_unlock(&lli->lli_layout_lock);
258 /* File data is modified. */
259 LLIF_DATA_MODIFIED = 0,
260 /* File is being restored */
261 LLIF_FILE_RESTORING = 1,
262 /* Xattr cache is attached to the file */
263 LLIF_XATTR_CACHE = 2,
266 static inline void ll_file_set_flag(struct ll_inode_info *lli,
267 enum ll_file_flags flag)
269 set_bit(flag, &lli->lli_flags);
272 static inline void ll_file_clear_flag(struct ll_inode_info *lli,
273 enum ll_file_flags flag)
275 clear_bit(flag, &lli->lli_flags);
278 static inline bool ll_file_test_flag(struct ll_inode_info *lli,
279 enum ll_file_flags flag)
281 return test_bit(flag, &lli->lli_flags);
284 static inline bool ll_file_test_and_clear_flag(struct ll_inode_info *lli,
285 enum ll_file_flags flag)
287 return test_and_clear_bit(flag, &lli->lli_flags);
290 int ll_xattr_cache_destroy(struct inode *inode);
292 int ll_xattr_cache_get(struct inode *inode,
298 int ll_dentry_init_security(struct dentry *dentry, int mode, struct qstr *name,
299 const char **secctx_name, void **secctx,
301 int ll_inode_init_security(struct dentry *dentry, struct inode *inode,
305 * Locking to guarantee consistency of non-atomic updates to long long i_size,
306 * consistency between file size and KMS.
308 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
311 void ll_inode_size_lock(struct inode *inode);
312 void ll_inode_size_unlock(struct inode *inode);
314 // FIXME: replace the name of this with LL_I to conform to kernel stuff
315 // static inline struct ll_inode_info *LL_I(struct inode *inode)
316 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
318 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
321 /* default to about 64M of readahead on a given system. */
322 #define SBI_DEFAULT_READAHEAD_MAX (64UL << (20 - PAGE_SHIFT))
324 /* default to read-ahead full files smaller than 2MB on the second read */
325 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_SHIFT))
330 RA_STAT_DISTANT_READPAGE,
331 RA_STAT_MISS_IN_WINDOW,
332 RA_STAT_FAILED_GRAB_PAGE,
333 RA_STAT_FAILED_MATCH,
338 RA_STAT_MAX_IN_FLIGHT,
339 RA_STAT_WRONG_GRAB_PAGE,
340 RA_STAT_FAILED_REACH_END,
345 atomic_t ra_cur_pages;
346 unsigned long ra_max_pages;
347 unsigned long ra_max_pages_per_file;
348 unsigned long ra_max_read_ahead_whole_pages;
351 /* ra_io_arg will be filled in the beginning of ll_readahead with
352 * ras_lock, then the following ll_read_ahead_pages will read RA
353 * pages according to this arg, all the items in this structure are
354 * counted by page index.
357 unsigned long ria_start; /* start offset of read-ahead*/
358 unsigned long ria_end; /* end offset of read-ahead*/
359 unsigned long ria_reserved; /* reserved pages for read-ahead */
360 unsigned long ria_end_min; /* minimum end to cover current read */
361 bool ria_eof; /* reach end of file */
362 /* If stride read pattern is detected, ria_stoff means where
363 * stride read is started. Note: for normal read-ahead, the
364 * value here is meaningless, and also it will not be accessed*/
366 /* ria_length and ria_pages are the length and pages length in the
367 * stride I/O mode. And they will also be used to check whether
368 * it is stride I/O read-ahead in the read-ahead pages*/
369 unsigned long ria_length;
370 unsigned long ria_pages;
373 /* LL_HIST_MAX=32 causes an overflow */
374 #define LL_HIST_MAX 28
375 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
376 #define LL_PROCESS_HIST_MAX 10
377 struct per_process_info {
379 struct obd_histogram pp_r_hist;
380 struct obd_histogram pp_w_hist;
383 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
384 struct ll_rw_extents_info {
385 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
388 #define LL_OFFSET_HIST_MAX 100
389 struct ll_rw_process_info {
392 loff_t rw_range_start;
394 loff_t rw_last_file_pos;
396 size_t rw_smallest_extent;
397 size_t rw_largest_extent;
398 struct ll_file_data *rw_last_file;
401 enum stats_track_type {
402 STATS_TRACK_ALL = 0, /* track all processes */
403 STATS_TRACK_PID, /* track process with this pid */
404 STATS_TRACK_PPID, /* track processes with this ppid */
405 STATS_TRACK_GID, /* track processes with this gid */
409 /* flags for sbi->ll_flags */
410 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
411 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
412 #define LL_SBI_FLOCK 0x04
413 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
414 #define LL_SBI_ACL 0x10 /* support ACL */
415 /* LL_SBI_RMT_CLIENT 0x40 remote client */
416 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa, obsolete */
417 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa, obsolete */
418 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
419 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
420 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
421 /* LL_SBI_SOM_PREVIEW 0x1000 SOM preview mount option, obsolete */
422 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
423 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
424 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
425 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
426 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
427 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
428 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
429 #define LL_SBI_NOROOTSQUASH 0x100000 /* do not apply root squash */
430 #define LL_SBI_ALWAYS_PING 0x200000 /* always ping even if server
432 #define LL_SBI_FAST_READ 0x400000 /* fast read support */
433 #define LL_SBI_FILE_SECCTX 0x800000 /* set file security context at create */
434 #define LL_SBI_PIO 0x1000000 /* parallel IO support */
436 #define LL_SBI_FLAGS { \
464 /* This is embedded into llite super-blocks to keep track of connect
465 * flags (capabilities) supported by all imports given mount is
467 struct lustre_client_ocd {
468 /* This is conjunction of connect_flags across all imports
469 * (LOVs) this mount is connected to. This field is updated by
470 * cl_ocd_update() under ->lco_lock. */
472 struct mutex lco_lock;
473 struct obd_export *lco_md_exp;
474 struct obd_export *lco_dt_exp;
478 /* this protects pglist and ra_info. It isn't safe to
479 * grab from interrupt contexts */
481 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
482 spinlock_t ll_process_lock; /* ll_rw_process_info */
483 struct obd_uuid ll_sb_uuid;
484 struct obd_export *ll_md_exp;
485 struct obd_export *ll_dt_exp;
486 struct proc_dir_entry* ll_proc_root;
487 struct lu_fid ll_root_fid; /* root object fid */
490 unsigned int ll_umounting:1,
491 ll_xattr_cache_enabled:1,
492 ll_xattr_cache_set:1, /* already set to 0/1 */
493 ll_client_common_fill_super_succeeded:1;
495 struct lustre_client_ocd ll_lco;
497 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
499 /* Used to track "unstable" pages on a client, and maintain a
500 * LRU list of clean pages. An "unstable" page is defined as
501 * any page which is sent to a server as part of a bulk request,
502 * but is uncommitted to stable storage. */
503 struct cl_client_cache *ll_cache;
505 struct lprocfs_stats *ll_ra_stats;
507 struct ll_ra_info ll_ra_info;
508 unsigned int ll_namelen;
509 struct file_operations *ll_fop;
511 struct lu_site *ll_site;
512 struct cl_device *ll_cl;
514 struct ll_rw_extents_info ll_rw_extents_info;
515 int ll_extent_process_count;
516 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
517 unsigned int ll_offset_process_count;
518 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
519 unsigned int ll_rw_offset_entry_count;
520 int ll_stats_track_id;
521 enum stats_track_type ll_stats_track_type;
524 /* metadata stat-ahead */
525 unsigned int ll_sa_max; /* max statahead RPCs */
526 atomic_t ll_sa_total; /* statahead thread started
528 atomic_t ll_sa_wrong; /* statahead thread stopped for
530 atomic_t ll_sa_running; /* running statahead thread
532 atomic_t ll_agl_total; /* AGL thread started count */
534 dev_t ll_sdev_orig; /* save s_dev before assign for
537 struct root_squash_info ll_squash;
540 /* st_blksize returned by stat(2), when non-zero */
541 unsigned int ll_stat_blksize;
545 * per file-descriptor read-ahead data.
547 struct ll_readahead_state {
550 * index of the last page that read(2) needed and that wasn't in the
551 * cache. Used by ras_update() to detect seeks.
553 * XXX nikita: if access seeks into cached region, Lustre doesn't see
556 unsigned long ras_last_readpage;
558 * number of pages read after last read-ahead window reset. As window
559 * is reset on each seek, this is effectively a number of consecutive
560 * accesses. Maybe ->ras_accessed_in_window is better name.
562 * XXX nikita: window is also reset (by ras_update()) when Lustre
563 * believes that memory pressure evicts read-ahead pages. In that
564 * case, it probably doesn't make sense to expand window to
565 * PTLRPC_MAX_BRW_PAGES on the third access.
567 unsigned long ras_consecutive_pages;
569 * number of read requests after the last read-ahead window reset
570 * As window is reset on each seek, this is effectively the number
571 * on consecutive read request and is used to trigger read-ahead.
573 unsigned long ras_consecutive_requests;
575 * Parameters of current read-ahead window. Handled by
576 * ras_update(). On the initial access to the file or after a seek,
577 * window is reset to 0. After 3 consecutive accesses, window is
578 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
579 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
581 unsigned long ras_window_start, ras_window_len;
583 * Optimal RPC size. It decides how many pages will be sent
584 * for each read-ahead.
586 unsigned long ras_rpc_size;
588 * Where next read-ahead should start at. This lies within read-ahead
589 * window. Read-ahead window is read in pieces rather than at once
590 * because: 1. lustre limits total number of pages under read-ahead by
591 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
592 * not covered by DLM lock.
594 unsigned long ras_next_readahead;
596 * Total number of ll_file_read requests issued, reads originating
597 * due to mmap are not counted in this total. This value is used to
598 * trigger full file read-ahead after multiple reads to a small file.
600 unsigned long ras_requests;
602 * Page index with respect to the current request, these value
603 * will not be accurate when dealing with reads issued via mmap.
605 unsigned long ras_request_index;
607 * The following 3 items are used for detecting the stride I/O
609 * In stride I/O mode,
610 * ...............|-----data-----|****gap*****|--------|******|....
611 * offset |-stride_pages-|-stride_gap-|
612 * ras_stride_offset = offset;
613 * ras_stride_length = stride_pages + stride_gap;
614 * ras_stride_pages = stride_pages;
615 * Note: all these three items are counted by pages.
617 unsigned long ras_stride_length;
618 unsigned long ras_stride_pages;
619 pgoff_t ras_stride_offset;
621 * number of consecutive stride request count, and it is similar as
622 * ras_consecutive_requests, but used for stride I/O mode.
623 * Note: only more than 2 consecutive stride request are detected,
624 * stride read-ahead will be enable
626 unsigned long ras_consecutive_stride_requests;
629 extern struct kmem_cache *ll_file_data_slab;
630 struct lustre_handle;
631 struct ll_file_data {
632 struct ll_readahead_state fd_ras;
633 struct ll_grouplock fd_grouplock;
637 /* openhandle if lease exists for this file.
638 * Borrow lli->lli_och_mutex to protect assignment */
639 struct obd_client_handle *fd_lease_och;
640 struct obd_client_handle *fd_och;
641 struct file *fd_file;
642 /* Indicate whether need to report failure when close.
643 * true: failure is known, not report again.
644 * false: unknown failure, should report. */
645 bool fd_write_failed;
646 rwlock_t fd_lock; /* protect lcc list */
647 struct list_head fd_lccs; /* list of ll_cl_context */
650 extern struct proc_dir_entry *proc_lustre_fs_root;
652 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
654 return &lli->lli_vfs_inode;
657 __u32 ll_i2suppgid(struct inode *i);
658 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
660 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
662 #if BITS_PER_LONG == 32
664 #elif defined(CONFIG_COMPAT)
665 if (unlikely(sbi->ll_flags & LL_SBI_32BIT_API))
668 # ifdef CONFIG_X86_X32
669 /* in_compat_syscall() returns true when called from a kthread
670 * and CONFIG_X86_X32 is enabled, which is wrong. So check
671 * whether the caller comes from a syscall (ie. not a kthread)
672 * before calling in_compat_syscall(). */
673 if (current->flags & PF_KTHREAD)
677 return unlikely(in_compat_syscall());
679 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
683 static inline bool ll_sbi_has_fast_read(struct ll_sb_info *sbi)
685 return !!(sbi->ll_flags & LL_SBI_FAST_READ);
688 void ll_ras_enter(struct file *f);
690 /* llite/lcommon_misc.c */
691 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
692 enum obd_notify_event ev, void *owner);
693 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
694 struct ll_grouplock *lg);
695 void cl_put_grouplock(struct ll_grouplock *lg);
697 /* llite/lproc_llite.c */
698 #ifdef CONFIG_PROC_FS
699 int lprocfs_ll_register_mountpoint(struct proc_dir_entry *parent,
700 struct super_block *sb);
701 int lprocfs_ll_register_obd(struct super_block *sb, const char *obdname);
702 void lprocfs_ll_unregister_mountpoint(struct ll_sb_info *sbi);
703 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
704 extern struct lprocfs_vars lprocfs_llite_obd_vars[];
706 static inline int lprocfs_ll_register_mountpoint(struct proc_dir_entry *parent,
707 struct super_block *sb) {return 0; }
708 static inline int lprocfs_ll_register_obd(struct super_block *sb,
709 const char *obdname) {return 0; }
710 static inline void lprocfs_ll_unregister_mountpoint(struct ll_sb_info *sbi) {}
711 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
716 LPROC_LL_DIRTY_MISSES,
718 LPROC_LL_WRITE_BYTES,
743 LPROC_LL_ALLOC_INODE,
746 LPROC_LL_GETXATTR_HITS,
748 LPROC_LL_REMOVEXATTR,
750 LPROC_LL_FILE_OPCODES
754 struct ll_dir_chain {
757 static inline void ll_dir_chain_init(struct ll_dir_chain *chain)
761 static inline void ll_dir_chain_fini(struct ll_dir_chain *chain)
765 extern const struct file_operations ll_dir_operations;
766 extern const struct inode_operations ll_dir_inode_operations;
767 #ifdef HAVE_DIR_CONTEXT
768 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
769 struct dir_context *ctx);
771 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
772 void *cookie, filldir_t filldir);
774 int ll_get_mdt_idx(struct inode *inode);
775 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
776 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
777 __u64 offset, struct ll_dir_chain *chain);
778 void ll_release_page(struct inode *inode, struct page *page, bool remove);
781 extern const struct inode_operations ll_special_inode_operations;
783 struct inode *ll_iget(struct super_block *sb, ino_t hash,
784 struct lustre_md *lic);
785 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
786 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
787 void *data, int flag);
788 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
789 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
790 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
793 int ll_writepage(struct page *page, struct writeback_control *wbc);
794 int ll_writepages(struct address_space *, struct writeback_control *wbc);
795 int ll_readpage(struct file *file, struct page *page);
796 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
797 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
800 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io,
802 void ll_cl_remove(struct file *file, const struct lu_env *env);
803 struct ll_cl_context *ll_cl_find(struct file *file);
805 extern const struct address_space_operations ll_aops;
808 extern struct file_operations ll_file_operations;
809 extern struct file_operations ll_file_operations_flock;
810 extern struct file_operations ll_file_operations_noflock;
811 extern struct inode_operations ll_file_inode_operations;
812 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
813 enum ldlm_mode l_req_mode);
814 extern enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
815 struct lustre_handle *lockh, __u64 flags,
816 enum ldlm_mode mode);
818 int ll_file_open(struct inode *inode, struct file *file);
819 int ll_file_release(struct inode *inode, struct file *file);
820 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
821 int ll_md_real_close(struct inode *inode, fmode_t fmode);
822 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
823 struct ll_file_data *file, loff_t pos,
824 size_t count, int rw);
825 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
826 struct posix_acl *ll_get_acl(struct inode *inode, int type);
827 #ifdef HAVE_IOP_SET_ACL
828 #ifdef CONFIG_FS_POSIX_ACL
829 int ll_set_acl(struct inode *inode, struct posix_acl *acl, int type);
830 #else /* !CONFIG_FS_POSIX_ACL */
831 #define ll_set_acl NULL
832 #endif /* CONFIG_FS_POSIX_ACL */
835 int ll_migrate(struct inode *parent, struct file *file, int mdtidx,
836 const char *name, int namelen);
837 int ll_get_fid_by_name(struct inode *parent, const char *name,
838 int namelen, struct lu_fid *fid, struct inode **inode);
839 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
840 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
842 # ifndef HAVE_INODE_PERMISION_2ARGS
843 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
845 int ll_inode_permission(struct inode *inode, int mask);
848 int ll_ioctl_fsgetxattr(struct inode *inode, unsigned int cmd,
850 int ll_ioctl_fssetxattr(struct inode *inode, unsigned int cmd,
853 int ll_lov_setstripe_ea_info(struct inode *inode, struct dentry *dentry,
854 __u64 flags, struct lov_user_md *lum,
856 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
857 struct lov_mds_md **lmm, int *lmm_size,
858 struct ptlrpc_request **request);
859 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
861 int ll_dir_getstripe(struct inode *inode, void **lmmp,
862 int *lmm_size, struct ptlrpc_request **request,
864 #ifdef HAVE_FILE_FSYNC_4ARGS
865 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
866 #elif defined(HAVE_FILE_FSYNC_2ARGS)
867 int ll_fsync(struct file *file, int data);
869 int ll_fsync(struct file *file, struct dentry *dentry, int data);
871 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
872 int ll_fid2path(struct inode *inode, void __user *arg);
873 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
874 int ll_hsm_release(struct inode *inode);
875 int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
879 int ll_d_init(struct dentry *de);
880 extern const struct dentry_operations ll_d_ops;
881 void ll_intent_drop_lock(struct lookup_intent *);
882 void ll_intent_release(struct lookup_intent *);
883 void ll_invalidate_aliases(struct inode *);
884 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
885 int ll_revalidate_it_finish(struct ptlrpc_request *request,
886 struct lookup_intent *it, struct dentry *de);
888 /* llite/llite_lib.c */
889 extern struct super_operations lustre_super_operations;
891 void ll_lli_init(struct ll_inode_info *lli);
892 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
893 void ll_put_super(struct super_block *sb);
894 void ll_kill_super(struct super_block *sb);
895 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
896 void ll_dir_clear_lsm_md(struct inode *inode);
897 void ll_clear_inode(struct inode *inode);
898 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import);
899 int ll_setattr(struct dentry *de, struct iattr *attr);
900 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
901 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
902 __u64 max_age, __u32 flags);
903 int ll_update_inode(struct inode *inode, struct lustre_md *md);
904 int ll_read_inode2(struct inode *inode, void *opaque);
905 void ll_delete_inode(struct inode *inode);
906 int ll_iocontrol(struct inode *inode, struct file *file,
907 unsigned int cmd, unsigned long arg);
908 int ll_flush_ctx(struct inode *inode);
909 void ll_umount_begin(struct super_block *sb);
910 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
911 #ifdef HAVE_SUPEROPS_USE_DENTRY
912 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
914 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
916 void ll_dirty_page_discard_warn(struct page *page, int ioret);
917 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
918 struct super_block *, struct lookup_intent *);
919 int ll_obd_statfs(struct inode *inode, void __user *arg);
920 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
921 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
922 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
923 int ll_process_config(struct lustre_cfg *lcfg);
926 LUSTRE_OPC_MKDIR = 0,
927 LUSTRE_OPC_SYMLINK = 1,
928 LUSTRE_OPC_MKNOD = 2,
929 LUSTRE_OPC_CREATE = 3,
933 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
934 struct inode *i1, struct inode *i2,
935 const char *name, size_t namelen,
936 __u32 mode, __u32 opc, void *data);
937 void ll_finish_md_op_data(struct md_op_data *op_data);
938 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
939 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
940 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
941 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
942 struct lov_user_md **kbuf);
943 void ll_open_cleanup(struct super_block *sb, struct ptlrpc_request *open_req);
945 /* Compute expected user md size when passing in a md from user space */
946 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
948 switch (lum->lmm_magic) {
949 case LOV_USER_MAGIC_V1:
950 return sizeof(struct lov_user_md_v1);
951 case LOV_USER_MAGIC_V3:
952 return sizeof(struct lov_user_md_v3);
953 case LOV_USER_MAGIC_SPECIFIC:
954 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
957 return lov_user_md_size(lum->lmm_stripe_count,
958 LOV_USER_MAGIC_SPECIFIC);
959 case LOV_USER_MAGIC_COMP_V1:
960 return ((struct lov_comp_md_v1 *)lum)->lcm_size;
966 /* llite/llite_nfs.c */
967 extern struct export_operations lustre_export_operations;
968 __u32 get_uuid2int(const char *name, int len);
969 struct inode *search_inode_for_lustre(struct super_block *sb,
970 const struct lu_fid *fid);
971 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
973 /* llite/symlink.c */
974 extern struct inode_operations ll_fast_symlink_inode_operations;
977 * IO arguments for various VFS I/O interfaces.
980 /** normal/sendfile/splice */
981 enum vvp_io_subtype via_io_subtype;
985 struct kiocb *via_iocb;
986 struct iov_iter *via_iter;
989 struct pipe_inode_info *via_pipe;
990 unsigned int via_flags;
1000 struct ll_cl_context {
1001 struct list_head lcc_list;
1003 const struct lu_env *lcc_env;
1004 struct cl_io *lcc_io;
1005 struct cl_page *lcc_page;
1006 enum lcc_type lcc_type;
1009 struct ll_thread_info {
1010 struct iov_iter lti_iter;
1011 struct vvp_io_args lti_args;
1012 struct ra_io_arg lti_ria;
1013 struct kiocb lti_kiocb;
1014 struct ll_cl_context lti_io_ctx;
1017 extern struct lu_context_key ll_thread_key;
1019 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
1021 struct ll_thread_info *lti;
1023 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
1024 LASSERT(lti != NULL);
1029 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env,
1030 enum vvp_io_subtype type)
1032 struct vvp_io_args *via = &ll_env_info(env)->lti_args;
1034 via->via_io_subtype = type;
1039 /* llite/llite_mmap.c */
1041 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1042 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1043 void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
1044 unsigned long addr, size_t count);
1045 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1048 static inline void ll_invalidate_page(struct page *vmpage)
1050 struct address_space *mapping = vmpage->mapping;
1051 loff_t offset = vmpage->index << PAGE_SHIFT;
1053 LASSERT(PageLocked(vmpage));
1054 if (mapping == NULL)
1058 * truncate_complete_page() calls
1059 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
1061 ll_teardown_mmaps(mapping, offset, offset + PAGE_SIZE);
1062 truncate_complete_page(mapping, vmpage);
1065 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1067 /* don't need an addref as the sb_info should be holding one */
1068 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1070 return ll_s2sbi(sb)->ll_dt_exp;
1073 /* don't need an addref as the sb_info should be holding one */
1074 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1076 return ll_s2sbi(sb)->ll_md_exp;
1079 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1081 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1087 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1088 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1090 return ll_s2sbi(inode->i_sb);
1093 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1095 return ll_s2dtexp(inode->i_sb);
1098 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1100 return ll_s2mdexp(inode->i_sb);
1103 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1107 LASSERT(inode != NULL);
1108 fid = &ll_i2info(inode)->lli_fid;
1113 static inline loff_t ll_file_maxbytes(struct inode *inode)
1115 struct cl_object *obj = ll_i2info(inode)->lli_clob;
1118 return MAX_LFS_FILESIZE;
1120 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
1124 extern const struct xattr_handler *ll_xattr_handlers[];
1126 #define XATTR_USER_T 1
1127 #define XATTR_TRUSTED_T 2
1128 #define XATTR_SECURITY_T 3
1129 #define XATTR_ACL_ACCESS_T 4
1130 #define XATTR_ACL_DEFAULT_T 5
1131 #define XATTR_LUSTRE_T 6
1132 #define XATTR_OTHER_T 7
1134 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1135 int ll_xattr_list(struct inode *inode, const char *name, int type,
1136 void *buffer, size_t size, u64 valid);
1137 const struct xattr_handler *get_xattr_type(const char *name);
1140 * Common IO arguments for various VFS I/O interfaces.
1142 int cl_sb_init(struct super_block *sb);
1143 int cl_sb_fini(struct super_block *sb);
1145 enum ras_update_flags {
1149 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1150 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1154 #define LL_SA_RPC_MIN 2
1155 #define LL_SA_RPC_DEF 32
1156 #define LL_SA_RPC_MAX 8192
1158 #define LL_SA_CACHE_BIT 5
1159 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1160 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1162 /* per inode struct, for dir only */
1163 struct ll_statahead_info {
1164 struct dentry *sai_dentry;
1165 atomic_t sai_refcount; /* when access this struct, hold
1167 unsigned int sai_max; /* max ahead of lookup */
1168 __u64 sai_sent; /* stat requests sent count */
1169 __u64 sai_replied; /* stat requests which received
1171 __u64 sai_index; /* index of statahead entry */
1172 __u64 sai_index_wait; /* index of entry which is the
1173 * caller is waiting for */
1174 __u64 sai_hit; /* hit count */
1175 __u64 sai_miss; /* miss count:
1176 * for "ls -al" case, includes
1177 * hidden dentry miss;
1178 * for "ls -l" case, it does not
1179 * include hidden dentry miss.
1180 * "sai_miss_hidden" is used for
1183 unsigned int sai_consecutive_miss; /* consecutive miss */
1184 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1185 * is not a hidden one */
1186 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1188 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1190 sai_agl_valid:1,/* AGL is valid for the dir */
1191 sai_in_readpage:1;/* statahead is in readdir()*/
1192 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1193 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1194 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1195 struct list_head sai_interim_entries; /* entries which got async
1196 * stat reply, but not
1198 struct list_head sai_entries; /* completed entries */
1199 struct list_head sai_agls; /* AGLs to be sent */
1200 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1201 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1202 atomic_t sai_cache_count; /* entry count in cache */
1205 int ll_statahead(struct inode *dir, struct dentry **dentry, bool unplug);
1206 void ll_authorize_statahead(struct inode *dir, void *key);
1207 void ll_deauthorize_statahead(struct inode *dir, void *key);
1210 blkcnt_t dirty_cnt(struct inode *inode);
1212 int cl_glimpse_size0(struct inode *inode, int agl);
1213 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1214 struct inode *inode, struct cl_object *clob, int agl);
1216 static inline int cl_glimpse_size(struct inode *inode)
1218 return cl_glimpse_size0(inode, 0);
1221 static inline int cl_agl(struct inode *inode)
1223 return cl_glimpse_size0(inode, 1);
1226 static inline int ll_glimpse_size(struct inode *inode)
1228 struct ll_inode_info *lli = ll_i2info(inode);
1231 down_read(&lli->lli_glimpse_sem);
1232 rc = cl_glimpse_size(inode);
1233 lli->lli_glimpse_time = cfs_time_current();
1234 up_read(&lli->lli_glimpse_sem);
1238 /* dentry may statahead when statahead is enabled and current process has opened
1239 * parent directory, and this dentry hasn't accessed statahead cache before */
1241 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1243 struct ll_inode_info *lli;
1244 struct ll_dentry_data *ldd;
1246 if (ll_i2sbi(dir)->ll_sa_max == 0)
1249 lli = ll_i2info(dir);
1251 /* statahead is not allowed for this dir, there may be three causes:
1252 * 1. dir is not opened.
1253 * 2. statahead hit ratio is too low.
1254 * 3. previous stat started statahead thread failed. */
1255 if (!lli->lli_sa_enabled)
1258 /* not the same process, don't statahead */
1259 if (lli->lli_opendir_pid != current_pid())
1263 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1264 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1265 * For patchless client, lookup intent is not accurate, which may
1266 * misguide statahead. For example:
1267 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1268 * have the same intent -- IT_GETATTR, while one dentry should access
1269 * statahead cache once, otherwise statahead windows is messed up.
1270 * The solution is as following:
1271 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1272 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1273 * bypass interacting with statahead cache by checking
1274 * 'lld_sa_generation == lli->lli_sa_generation'.
1276 ldd = ll_d2d(dentry);
1277 if (ldd != NULL && ldd->lld_sa_generation == lli->lli_sa_generation)
1283 /* llite ioctl register support rountine */
1289 #define LLIOC_MAX_CMD 256
1292 * Rules to write a callback function:
1295 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1296 * returned to ll_iocontrol_register. Callback functions should use this
1297 * data to check the potential collasion of ioctl cmd. If collasion is
1298 * found, callback function should return LLIOC_CONT.
1299 * @rcp: The result of ioctl command.
1302 * If @magic matches the pointer returned by ll_iocontrol_data, the
1303 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1305 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1306 struct file *file, unsigned int cmd, unsigned long arg,
1307 void *magic, int *rcp);
1309 /* export functions */
1310 /* Register ioctl block dynamatically for a regular file.
1312 * @cmd: the array of ioctl command set
1313 * @count: number of commands in the @cmd
1314 * @cb: callback function, it will be called if an ioctl command is found to
1315 * belong to the command list @cmd.
1318 * A magic pointer will be returned if success;
1319 * otherwise, NULL will be returned.
1321 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1322 void ll_iocontrol_unregister(void *magic);
1324 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1325 enum cl_fsync_mode mode, int ignore_layout);
1327 /** direct write pages */
1328 struct ll_dio_pages {
1329 /** page array to be written. we don't support
1330 * partial pages except the last one. */
1331 struct page **ldp_pages;
1332 /* offset of each page */
1333 loff_t *ldp_offsets;
1334 /** if ldp_offsets is NULL, it means a sequential
1335 * pages to be written, then this is the file offset
1336 * of the * first page. */
1337 loff_t ldp_start_offset;
1338 /** how many bytes are to be written. */
1340 /** # of pages in the array. */
1344 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1345 int rw, struct inode *inode,
1346 struct ll_dio_pages *pv);
1348 static inline int ll_file_nolock(const struct file *file)
1350 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1351 struct inode *inode = file_inode((struct file *)file);
1353 LASSERT(fd != NULL);
1354 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1355 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1358 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1359 struct lookup_intent *it, __u64 *bits)
1361 if (!it->it_lock_set) {
1362 struct lustre_handle handle;
1364 /* If this inode is a remote object, it will get two
1365 * separate locks in different namespaces, Master MDT,
1366 * where the name entry is, will grant LOOKUP lock,
1367 * remote MDT, where the object is, will grant
1368 * UPDATE|PERM lock. The inode will be attched to both
1369 * LOOKUP and PERM locks, so revoking either locks will
1370 * case the dcache being cleared */
1371 if (it->it_remote_lock_mode) {
1372 handle.cookie = it->it_remote_lock_handle;
1373 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1374 "(%p) for remote lock %#llx\n",
1375 PFID(ll_inode2fid(inode)), inode,
1377 md_set_lock_data(exp, &handle, inode, NULL);
1380 handle.cookie = it->it_lock_handle;
1382 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1383 " for lock %#llx\n",
1384 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1386 md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1387 it->it_lock_set = 1;
1391 *bits = it->it_lock_bits;
1394 static inline void ll_lock_dcache(struct inode *inode)
1396 #ifdef HAVE_DCACHE_LOCK
1397 spin_lock(&dcache_lock);
1399 spin_lock(&inode->i_lock);
1403 static inline void ll_unlock_dcache(struct inode *inode)
1405 #ifdef HAVE_DCACHE_LOCK
1406 spin_unlock(&dcache_lock);
1408 spin_unlock(&inode->i_lock);
1412 static inline int d_lustre_invalid(const struct dentry *dentry)
1414 struct ll_dentry_data *lld = ll_d2d(dentry);
1416 return (lld == NULL) || lld->lld_invalid;
1419 static inline void __d_lustre_invalidate(struct dentry *dentry)
1421 struct ll_dentry_data *lld = ll_d2d(dentry);
1424 lld->lld_invalid = 1;
1428 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1429 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1430 * else dput() of the last refcount will unhash this dentry and kill it.
1432 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1434 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1435 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1436 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1438 spin_lock_nested(&dentry->d_lock,
1439 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1440 __d_lustre_invalidate(dentry);
1442 * We should be careful about dentries created by d_obtain_alias().
1443 * These dentries are not put in the dentry tree, instead they are
1444 * linked to sb->s_anon through dentry->d_hash.
1445 * shrink_dcache_for_umount() shrinks the tree and sb->s_anon list.
1446 * If we unhashed such a dentry, unmount would not be able to find
1447 * it and busy inodes would be reported.
1449 if (ll_d_count(dentry) == 0 && !(dentry->d_flags & DCACHE_DISCONNECTED))
1451 spin_unlock(&dentry->d_lock);
1454 static inline void d_lustre_revalidate(struct dentry *dentry)
1456 spin_lock(&dentry->d_lock);
1457 LASSERT(ll_d2d(dentry) != NULL);
1458 ll_d2d(dentry)->lld_invalid = 0;
1459 spin_unlock(&dentry->d_lock);
1462 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1463 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1464 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1465 int ll_layout_write_intent(struct inode *inode, __u64 start, __u64 end);
1467 int ll_xattr_init(void);
1468 void ll_xattr_fini(void);
1470 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1471 struct cl_page *page, enum cl_req_type crt);
1473 int ll_getparent(struct file *file, struct getparent __user *arg);
1476 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1477 unsigned int attr_flags);
1479 extern struct lu_env *cl_inode_fini_env;
1480 extern __u16 cl_inode_fini_refcheck;
1482 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1483 void cl_inode_fini(struct inode *inode);
1485 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1486 u32 cl_fid_build_gen(const struct lu_fid *fid);
1488 #endif /* LLITE_INTERNAL_H */