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, 2015, 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>
43 #include <lustre_linkea.h>
45 /* for struct cl_lock_descr and struct cl_io */
46 #include <cl_object.h>
47 #include <lustre_lmv.h>
48 #include <lustre_mdc.h>
49 #include <lustre_intent.h>
50 #include <linux/compat.h>
51 #include <linux/aio.h>
53 #include "vvp_internal.h"
54 #include "range_lock.h"
60 #ifndef VM_FAULT_RETRY
61 #define VM_FAULT_RETRY 0
64 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
65 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
66 #ifndef LOOKUP_CONTINUE
67 #define LOOKUP_CONTINUE LOOKUP_PARENT
70 /** Only used on client-side for indicating the tail of dir hash/offset. */
71 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
72 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
74 /* 4UL * 1024 * 1024 */
75 #define LL_MAX_BLKSIZE_BITS 22
77 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
78 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
80 struct ll_dentry_data {
81 struct lookup_intent *lld_it;
82 unsigned int lld_sa_generation;
83 unsigned int lld_invalid:1;
84 unsigned int lld_nfs_dentry:1;
85 struct rcu_head lld_rcu_head;
88 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
90 #define LLI_INODE_MAGIC 0x111d0de5
91 #define LLI_INODE_DEAD 0xdeadd00d
93 struct ll_getname_data {
94 #ifdef HAVE_DIR_CONTEXT
95 struct dir_context ctx;
97 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
98 struct lu_fid lgd_fid; /* target fid we are looking for */
99 int lgd_found; /* inode matched? */
102 struct ll_grouplock {
103 struct lu_env *lg_env;
105 struct cl_lock *lg_lock;
106 unsigned long lg_gid;
109 struct ll_inode_info {
110 __u32 lli_inode_magic;
113 volatile unsigned long lli_flags;
114 struct posix_acl *lli_posix_acl;
116 /* identifying fields for both metadata and data stacks. */
117 struct lu_fid lli_fid;
118 /* master inode fid for stripe directory */
119 struct lu_fid lli_pfid;
121 /* open count currently used by capability only, indicate whether
122 * capability needs renewal */
123 atomic_t lli_open_count;
125 /* We need all three because every inode may be opened in different
127 struct obd_client_handle *lli_mds_read_och;
128 struct obd_client_handle *lli_mds_write_och;
129 struct obd_client_handle *lli_mds_exec_och;
130 __u64 lli_open_fd_read_count;
131 __u64 lli_open_fd_write_count;
132 __u64 lli_open_fd_exec_count;
133 /* Protects access to och pointers and their usage counters */
134 struct mutex lli_och_mutex;
136 struct inode lli_vfs_inode;
138 /* the most recent timestamps obtained from mds */
142 spinlock_t lli_agl_lock;
144 /* Try to make the d::member and f::member are aligned. Before using
145 * these members, make clear whether it is directory or not. */
149 /* serialize normal readdir and statahead-readdir. */
150 struct mutex lli_readdir_mutex;
152 /* metadata statahead */
153 /* since parent-child threads can share the same @file
154 * struct, "opendir_key" is the token when dir close for
155 * case of parent exit before child -- it is me should
156 * cleanup the dir readahead. */
157 void *lli_opendir_key;
158 struct ll_statahead_info *lli_sai;
159 /* protect statahead stuff. */
160 spinlock_t lli_sa_lock;
161 /* "opendir_pid" is the token when lookup/revalid
162 * -- I am the owner of dir statahead. */
163 pid_t lli_opendir_pid;
164 /* stat will try to access statahead entries or start
165 * statahead if this flag is set, and this flag will be
166 * set upon dir open, and cleared when dir is closed,
167 * statahead hit ratio is too low, or start statahead
169 unsigned int lli_sa_enabled:1;
170 /* generation for statahead */
171 unsigned int lli_sa_generation;
172 /* directory stripe information */
173 struct lmv_stripe_md *lli_lsm_md;
174 /* default directory stripe offset. This is extracted
175 * from the "dmv" xattr in order to decide which MDT to
176 * create a subdirectory on. The MDS itself fetches
177 * "dmv" and gets the rest of the default layout itself
178 * (count, hash, etc). */
179 __u32 lli_def_stripe_offset;
182 /* for non-directory */
184 struct mutex lli_size_mutex;
185 char *lli_symlink_name;
187 * struct rw_semaphore {
188 * signed long count; // align d.d_def_acl
189 * spinlock_t wait_lock; // align d.d_sa_lock
190 * struct list_head wait_list;
193 struct rw_semaphore lli_trunc_sem;
194 struct range_lock_tree lli_write_tree;
196 struct rw_semaphore lli_glimpse_sem;
197 cfs_time_t lli_glimpse_time;
198 struct list_head lli_agl_list;
201 /* for writepage() only to communicate to fsync */
205 * whenever a process try to read/write the file, the
206 * jobid of the process will be saved here, and it'll
207 * be packed into the write PRC when flush later.
209 * so the read/write statistics for jobid will not be
210 * accurate if the file is shared by different jobs.
212 char lli_jobid[LUSTRE_JOBID_SIZE];
216 /* XXX: For following frequent used members, although they maybe special
217 * used for non-directory object, it is some time-wasting to check
218 * whether the object is directory or not before using them. On the
219 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
220 * the "ll_inode_info" size even if moving those members into u.f.
221 * So keep them out side.
223 * In the future, if more members are added only for directory,
224 * some of the following members can be moved into u.f.
226 struct cl_object *lli_clob;
228 /* mutex to request for layout lock exclusively. */
229 struct mutex lli_layout_mutex;
230 /* Layout version, protected by lli_layout_lock */
231 __u32 lli_layout_gen;
232 spinlock_t lli_layout_lock;
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_init_security(struct dentry *dentry,
303 * Locking to guarantee consistency of non-atomic updates to long long i_size,
304 * consistency between file size and KMS.
306 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
309 void ll_inode_size_lock(struct inode *inode);
310 void ll_inode_size_unlock(struct inode *inode);
312 // FIXME: replace the name of this with LL_I to conform to kernel stuff
313 // static inline struct ll_inode_info *LL_I(struct inode *inode)
314 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
316 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
319 /* default to about 64M of readahead on a given system. */
320 #define SBI_DEFAULT_READAHEAD_MAX (64UL << (20 - PAGE_CACHE_SHIFT))
322 /* default to read-ahead full files smaller than 2MB on the second read */
323 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
328 RA_STAT_DISTANT_READPAGE,
329 RA_STAT_MISS_IN_WINDOW,
330 RA_STAT_FAILED_GRAB_PAGE,
331 RA_STAT_FAILED_MATCH,
336 RA_STAT_MAX_IN_FLIGHT,
337 RA_STAT_WRONG_GRAB_PAGE,
338 RA_STAT_FAILED_REACH_END,
343 atomic_t ra_cur_pages;
344 unsigned long ra_max_pages;
345 unsigned long ra_max_pages_per_file;
346 unsigned long ra_max_read_ahead_whole_pages;
349 /* ra_io_arg will be filled in the beginning of ll_readahead with
350 * ras_lock, then the following ll_read_ahead_pages will read RA
351 * pages according to this arg, all the items in this structure are
352 * counted by page index.
355 unsigned long ria_start; /* start offset of read-ahead*/
356 unsigned long ria_end; /* end offset of read-ahead*/
357 unsigned long ria_reserved; /* reserved pages for read-ahead */
358 unsigned long ria_end_min; /* minimum end to cover current read */
359 bool ria_eof; /* reach end of file */
360 /* If stride read pattern is detected, ria_stoff means where
361 * stride read is started. Note: for normal read-ahead, the
362 * value here is meaningless, and also it will not be accessed*/
364 /* ria_length and ria_pages are the length and pages length in the
365 * stride I/O mode. And they will also be used to check whether
366 * it is stride I/O read-ahead in the read-ahead pages*/
367 unsigned long ria_length;
368 unsigned long ria_pages;
371 /* LL_HIST_MAX=32 causes an overflow */
372 #define LL_HIST_MAX 28
373 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
374 #define LL_PROCESS_HIST_MAX 10
375 struct per_process_info {
377 struct obd_histogram pp_r_hist;
378 struct obd_histogram pp_w_hist;
381 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
382 struct ll_rw_extents_info {
383 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
386 #define LL_OFFSET_HIST_MAX 100
387 struct ll_rw_process_info {
390 loff_t rw_range_start;
392 loff_t rw_last_file_pos;
394 size_t rw_smallest_extent;
395 size_t rw_largest_extent;
396 struct ll_file_data *rw_last_file;
399 enum stats_track_type {
400 STATS_TRACK_ALL = 0, /* track all processes */
401 STATS_TRACK_PID, /* track process with this pid */
402 STATS_TRACK_PPID, /* track processes with this ppid */
403 STATS_TRACK_GID, /* track processes with this gid */
407 /* flags for sbi->ll_flags */
408 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
409 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
410 #define LL_SBI_FLOCK 0x04
411 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
412 #define LL_SBI_ACL 0x10 /* support ACL */
413 /* LL_SBI_RMT_CLIENT 0x40 remote client */
414 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa, obsolete */
415 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa, obsolete */
416 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
417 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
418 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
419 /* LL_SBI_SOM_PREVIEW 0x1000 SOM preview mount option, obsolete */
420 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
421 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
422 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
423 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
424 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
425 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
426 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
427 #define LL_SBI_NOROOTSQUASH 0x100000 /* do not apply root squash */
428 #define LL_SBI_ALWAYS_PING 0x200000 /* always ping even if server
430 #define LL_SBI_FAST_READ 0x400000 /* fast read support */
432 #define LL_SBI_FLAGS { \
458 /* This is embedded into llite super-blocks to keep track of connect
459 * flags (capabilities) supported by all imports given mount is
461 struct lustre_client_ocd {
462 /* This is conjunction of connect_flags across all imports
463 * (LOVs) this mount is connected to. This field is updated by
464 * cl_ocd_update() under ->lco_lock. */
466 struct mutex lco_lock;
467 struct obd_export *lco_md_exp;
468 struct obd_export *lco_dt_exp;
472 /* this protects pglist and ra_info. It isn't safe to
473 * grab from interrupt contexts */
475 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
476 spinlock_t ll_process_lock; /* ll_rw_process_info */
477 struct obd_uuid ll_sb_uuid;
478 struct obd_export *ll_md_exp;
479 struct obd_export *ll_dt_exp;
480 struct proc_dir_entry* ll_proc_root;
481 struct lu_fid ll_root_fid; /* root object fid */
484 unsigned int ll_umounting:1,
485 ll_xattr_cache_enabled:1,
486 ll_client_common_fill_super_succeeded:1;
488 struct lustre_client_ocd ll_lco;
490 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
492 /* Used to track "unstable" pages on a client, and maintain a
493 * LRU list of clean pages. An "unstable" page is defined as
494 * any page which is sent to a server as part of a bulk request,
495 * but is uncommitted to stable storage. */
496 struct cl_client_cache *ll_cache;
498 struct lprocfs_stats *ll_ra_stats;
500 struct ll_ra_info ll_ra_info;
501 unsigned int ll_namelen;
502 struct file_operations *ll_fop;
504 unsigned int ll_md_brw_pages; /* readdir pages per RPC */
506 struct lu_site *ll_site;
507 struct cl_device *ll_cl;
509 struct ll_rw_extents_info ll_rw_extents_info;
510 int ll_extent_process_count;
511 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
512 unsigned int ll_offset_process_count;
513 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
514 unsigned int ll_rw_offset_entry_count;
515 int ll_stats_track_id;
516 enum stats_track_type ll_stats_track_type;
519 /* metadata stat-ahead */
520 unsigned int ll_sa_max; /* max statahead RPCs */
521 atomic_t ll_sa_total; /* statahead thread started
523 atomic_t ll_sa_wrong; /* statahead thread stopped for
525 atomic_t ll_sa_running; /* running statahead thread
527 atomic_t ll_agl_total; /* AGL thread started count */
529 dev_t ll_sdev_orig; /* save s_dev before assign for
532 struct root_squash_info ll_squash;
536 * per file-descriptor read-ahead data.
538 struct ll_readahead_state {
541 * index of the last page that read(2) needed and that wasn't in the
542 * cache. Used by ras_update() to detect seeks.
544 * XXX nikita: if access seeks into cached region, Lustre doesn't see
547 unsigned long ras_last_readpage;
549 * number of pages read after last read-ahead window reset. As window
550 * is reset on each seek, this is effectively a number of consecutive
551 * accesses. Maybe ->ras_accessed_in_window is better name.
553 * XXX nikita: window is also reset (by ras_update()) when Lustre
554 * believes that memory pressure evicts read-ahead pages. In that
555 * case, it probably doesn't make sense to expand window to
556 * PTLRPC_MAX_BRW_PAGES on the third access.
558 unsigned long ras_consecutive_pages;
560 * number of read requests after the last read-ahead window reset
561 * As window is reset on each seek, this is effectively the number
562 * on consecutive read request and is used to trigger read-ahead.
564 unsigned long ras_consecutive_requests;
566 * Parameters of current read-ahead window. Handled by
567 * ras_update(). On the initial access to the file or after a seek,
568 * window is reset to 0. After 3 consecutive accesses, window is
569 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
570 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
572 unsigned long ras_window_start, ras_window_len;
574 * Optimal RPC size. It decides how many pages will be sent
575 * for each read-ahead.
577 unsigned long ras_rpc_size;
579 * Where next read-ahead should start at. This lies within read-ahead
580 * window. Read-ahead window is read in pieces rather than at once
581 * because: 1. lustre limits total number of pages under read-ahead by
582 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
583 * not covered by DLM lock.
585 unsigned long ras_next_readahead;
587 * Total number of ll_file_read requests issued, reads originating
588 * due to mmap are not counted in this total. This value is used to
589 * trigger full file read-ahead after multiple reads to a small file.
591 unsigned long ras_requests;
593 * Page index with respect to the current request, these value
594 * will not be accurate when dealing with reads issued via mmap.
596 unsigned long ras_request_index;
598 * The following 3 items are used for detecting the stride I/O
600 * In stride I/O mode,
601 * ...............|-----data-----|****gap*****|--------|******|....
602 * offset |-stride_pages-|-stride_gap-|
603 * ras_stride_offset = offset;
604 * ras_stride_length = stride_pages + stride_gap;
605 * ras_stride_pages = stride_pages;
606 * Note: all these three items are counted by pages.
608 unsigned long ras_stride_length;
609 unsigned long ras_stride_pages;
610 pgoff_t ras_stride_offset;
612 * number of consecutive stride request count, and it is similar as
613 * ras_consecutive_requests, but used for stride I/O mode.
614 * Note: only more than 2 consecutive stride request are detected,
615 * stride read-ahead will be enable
617 unsigned long ras_consecutive_stride_requests;
620 extern struct kmem_cache *ll_file_data_slab;
621 struct lustre_handle;
622 struct ll_file_data {
623 struct ll_readahead_state fd_ras;
624 struct ll_grouplock fd_grouplock;
628 /* openhandle if lease exists for this file.
629 * Borrow lli->lli_och_mutex to protect assignment */
630 struct obd_client_handle *fd_lease_och;
631 struct obd_client_handle *fd_och;
632 struct file *fd_file;
633 /* Indicate whether need to report failure when close.
634 * true: failure is known, not report again.
635 * false: unknown failure, should report. */
636 bool fd_write_failed;
637 rwlock_t fd_lock; /* protect lcc list */
638 struct list_head fd_lccs; /* list of ll_cl_context */
641 extern spinlock_t inode_lock;
643 extern struct proc_dir_entry *proc_lustre_fs_root;
645 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
647 return &lli->lli_vfs_inode;
650 __u32 ll_i2suppgid(struct inode *i);
651 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
653 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
655 #if BITS_PER_LONG == 32
657 #elif defined(CONFIG_COMPAT)
658 return unlikely(is_compat_task() || (sbi->ll_flags & LL_SBI_32BIT_API));
660 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
664 static inline bool ll_sbi_has_fast_read(struct ll_sb_info *sbi)
666 return !!(sbi->ll_flags & LL_SBI_FAST_READ);
669 void ll_ras_enter(struct file *f);
671 /* llite/lcommon_misc.c */
672 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
673 enum obd_notify_event ev, void *owner, void *data);
674 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
675 struct ll_grouplock *lg);
676 void cl_put_grouplock(struct ll_grouplock *lg);
678 /* llite/lproc_llite.c */
679 #ifdef CONFIG_PROC_FS
680 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
681 struct super_block *sb, char *osc, char *mdc);
682 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
683 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
684 extern struct lprocfs_vars lprocfs_llite_obd_vars[];
686 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
687 struct super_block *sb, char *osc, char *mdc){return 0;}
688 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
689 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
694 LPROC_LL_DIRTY_MISSES,
696 LPROC_LL_WRITE_BYTES,
721 LPROC_LL_ALLOC_INODE,
724 LPROC_LL_GETXATTR_HITS,
726 LPROC_LL_REMOVEXATTR,
728 LPROC_LL_FILE_OPCODES
732 struct ll_dir_chain {
735 static inline void ll_dir_chain_init(struct ll_dir_chain *chain)
739 static inline void ll_dir_chain_fini(struct ll_dir_chain *chain)
743 extern const struct file_operations ll_dir_operations;
744 extern const struct inode_operations ll_dir_inode_operations;
745 #ifdef HAVE_DIR_CONTEXT
746 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
747 struct dir_context *ctx);
749 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
750 void *cookie, filldir_t filldir);
752 int ll_get_mdt_idx(struct inode *inode);
753 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
754 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
755 __u64 offset, struct ll_dir_chain *chain);
756 void ll_release_page(struct inode *inode, struct page *page, bool remove);
759 extern const struct inode_operations ll_special_inode_operations;
761 struct inode *ll_iget(struct super_block *sb, ino_t hash,
762 struct lustre_md *lic);
763 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
764 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
765 void *data, int flag);
766 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
767 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
768 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
771 int ll_writepage(struct page *page, struct writeback_control *wbc);
772 int ll_writepages(struct address_space *, struct writeback_control *wbc);
773 int ll_readpage(struct file *file, struct page *page);
774 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
775 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
778 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io,
780 void ll_cl_remove(struct file *file, const struct lu_env *env);
781 struct ll_cl_context *ll_cl_find(struct file *file);
783 extern const struct address_space_operations ll_aops;
786 extern struct file_operations ll_file_operations;
787 extern struct file_operations ll_file_operations_flock;
788 extern struct file_operations ll_file_operations_noflock;
789 extern struct inode_operations ll_file_inode_operations;
790 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
791 enum ldlm_mode l_req_mode);
792 extern enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
793 struct lustre_handle *lockh, __u64 flags,
794 enum ldlm_mode mode);
796 int ll_file_open(struct inode *inode, struct file *file);
797 int ll_file_release(struct inode *inode, struct file *file);
798 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
799 int ll_md_real_close(struct inode *inode, fmode_t fmode);
800 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
801 struct ll_file_data *file, loff_t pos,
802 size_t count, int rw);
803 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
804 struct posix_acl *ll_get_acl(struct inode *inode, int type);
805 int ll_migrate(struct inode *parent, struct file *file, int mdtidx,
806 const char *name, int namelen);
807 int ll_get_fid_by_name(struct inode *parent, const char *name,
808 int namelen, struct lu_fid *fid, struct inode **inode);
809 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
810 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
812 # ifndef HAVE_INODE_PERMISION_2ARGS
813 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
815 int ll_inode_permission(struct inode *inode, int mask);
819 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
820 __u64 flags, struct lov_user_md *lum,
822 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
823 struct lov_mds_md **lmm, int *lmm_size,
824 struct ptlrpc_request **request);
825 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
827 int ll_dir_getstripe(struct inode *inode, void **lmmp,
828 int *lmm_size, struct ptlrpc_request **request,
830 #ifdef HAVE_FILE_FSYNC_4ARGS
831 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
832 #elif defined(HAVE_FILE_FSYNC_2ARGS)
833 int ll_fsync(struct file *file, int data);
835 int ll_fsync(struct file *file, struct dentry *dentry, int data);
837 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
838 int ll_fid2path(struct inode *inode, void __user *arg);
839 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
840 int ll_hsm_release(struct inode *inode);
841 int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
845 int ll_d_init(struct dentry *de);
846 extern const struct dentry_operations ll_d_ops;
847 void ll_intent_drop_lock(struct lookup_intent *);
848 void ll_intent_release(struct lookup_intent *);
849 void ll_invalidate_aliases(struct inode *);
850 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
851 int ll_revalidate_it_finish(struct ptlrpc_request *request,
852 struct lookup_intent *it, struct dentry *de);
854 /* llite/llite_lib.c */
855 extern struct super_operations lustre_super_operations;
857 void ll_lli_init(struct ll_inode_info *lli);
858 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
859 void ll_put_super(struct super_block *sb);
860 void ll_kill_super(struct super_block *sb);
861 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
862 void ll_dir_clear_lsm_md(struct inode *inode);
863 void ll_clear_inode(struct inode *inode);
864 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import);
865 int ll_setattr(struct dentry *de, struct iattr *attr);
866 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
867 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
868 __u64 max_age, __u32 flags);
869 int ll_update_inode(struct inode *inode, struct lustre_md *md);
870 int ll_read_inode2(struct inode *inode, void *opaque);
871 void ll_delete_inode(struct inode *inode);
872 int ll_iocontrol(struct inode *inode, struct file *file,
873 unsigned int cmd, unsigned long arg);
874 int ll_flush_ctx(struct inode *inode);
875 void ll_umount_begin(struct super_block *sb);
876 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
877 #ifdef HAVE_SUPEROPS_USE_DENTRY
878 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
880 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
882 void ll_dirty_page_discard_warn(struct page *page, int ioret);
883 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
884 struct super_block *, struct lookup_intent *);
885 void lustre_dump_dentry(struct dentry *, int recur);
886 int ll_obd_statfs(struct inode *inode, void __user *arg);
887 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
888 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
889 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
890 int ll_process_config(struct lustre_cfg *lcfg);
893 LUSTRE_OPC_MKDIR = 0,
894 LUSTRE_OPC_SYMLINK = 1,
895 LUSTRE_OPC_MKNOD = 2,
896 LUSTRE_OPC_CREATE = 3,
900 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
901 struct inode *i1, struct inode *i2,
902 const char *name, size_t namelen,
903 __u32 mode, __u32 opc, void *data);
904 void ll_finish_md_op_data(struct md_op_data *op_data);
905 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
906 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
907 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
908 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
909 struct lov_user_md **kbuf);
910 void ll_open_cleanup(struct super_block *sb, struct ptlrpc_request *open_req);
912 /* Compute expected user md size when passing in a md from user space */
913 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
915 switch (lum->lmm_magic) {
916 case LOV_USER_MAGIC_V1:
917 return sizeof(struct lov_user_md_v1);
918 case LOV_USER_MAGIC_V3:
919 return sizeof(struct lov_user_md_v3);
920 case LOV_USER_MAGIC_SPECIFIC:
921 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
924 return lov_user_md_size(lum->lmm_stripe_count,
925 LOV_USER_MAGIC_SPECIFIC);
931 /* llite/llite_nfs.c */
932 extern struct export_operations lustre_export_operations;
933 __u32 get_uuid2int(const char *name, int len);
934 struct inode *search_inode_for_lustre(struct super_block *sb,
935 const struct lu_fid *fid);
936 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
938 /* llite/symlink.c */
939 extern struct inode_operations ll_fast_symlink_inode_operations;
942 * IO arguments for various VFS I/O interfaces.
945 /** normal/sendfile/splice */
946 enum vvp_io_subtype via_io_subtype;
950 struct kiocb *via_iocb;
951 struct iov_iter *via_iter;
954 struct pipe_inode_info *via_pipe;
955 unsigned int via_flags;
965 struct ll_cl_context {
966 struct list_head lcc_list;
968 const struct lu_env *lcc_env;
969 struct cl_io *lcc_io;
970 struct cl_page *lcc_page;
971 enum lcc_type lcc_type;
974 struct ll_thread_info {
975 struct iov_iter lti_iter;
976 struct vvp_io_args lti_args;
977 struct ra_io_arg lti_ria;
978 struct kiocb lti_kiocb;
979 struct ll_cl_context lti_io_ctx;
982 extern struct lu_context_key ll_thread_key;
984 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
986 struct ll_thread_info *lti;
988 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
989 LASSERT(lti != NULL);
994 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env,
995 enum vvp_io_subtype type)
997 struct vvp_io_args *via = &ll_env_info(env)->lti_args;
999 via->via_io_subtype = type;
1004 /* llite/llite_mmap.c */
1006 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1007 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1008 void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
1009 unsigned long addr, size_t count);
1010 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1013 static inline void ll_invalidate_page(struct page *vmpage)
1015 struct address_space *mapping = vmpage->mapping;
1016 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1018 LASSERT(PageLocked(vmpage));
1019 if (mapping == NULL)
1023 * truncate_complete_page() calls
1024 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
1026 ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE);
1027 truncate_complete_page(mapping, vmpage);
1030 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1032 /* don't need an addref as the sb_info should be holding one */
1033 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1035 return ll_s2sbi(sb)->ll_dt_exp;
1038 /* don't need an addref as the sb_info should be holding one */
1039 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1041 return ll_s2sbi(sb)->ll_md_exp;
1044 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1046 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1052 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1053 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1055 return ll_s2sbi(inode->i_sb);
1058 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1060 return ll_s2dtexp(inode->i_sb);
1063 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1065 return ll_s2mdexp(inode->i_sb);
1068 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1072 LASSERT(inode != NULL);
1073 fid = &ll_i2info(inode)->lli_fid;
1078 static inline loff_t ll_file_maxbytes(struct inode *inode)
1080 struct cl_object *obj = ll_i2info(inode)->lli_clob;
1083 return MAX_LFS_FILESIZE;
1085 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
1089 int ll_setxattr(struct dentry *dentry, const char *name,
1090 const void *value, size_t size, int flags);
1091 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1092 void *buffer, size_t size);
1093 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1094 int ll_removexattr(struct dentry *dentry, const char *name);
1097 * Common IO arguments for various VFS I/O interfaces.
1099 int cl_sb_init(struct super_block *sb);
1100 int cl_sb_fini(struct super_block *sb);
1102 enum ras_update_flags {
1106 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1107 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1111 #define LL_SA_RPC_MIN 2
1112 #define LL_SA_RPC_DEF 32
1113 #define LL_SA_RPC_MAX 8192
1115 #define LL_SA_CACHE_BIT 5
1116 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1117 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1119 /* per inode struct, for dir only */
1120 struct ll_statahead_info {
1121 struct dentry *sai_dentry;
1122 atomic_t sai_refcount; /* when access this struct, hold
1124 unsigned int sai_max; /* max ahead of lookup */
1125 __u64 sai_sent; /* stat requests sent count */
1126 __u64 sai_replied; /* stat requests which received
1128 __u64 sai_index; /* index of statahead entry */
1129 __u64 sai_index_wait; /* index of entry which is the
1130 * caller is waiting for */
1131 __u64 sai_hit; /* hit count */
1132 __u64 sai_miss; /* miss count:
1133 * for "ls -al" case, includes
1134 * hidden dentry miss;
1135 * for "ls -l" case, it does not
1136 * include hidden dentry miss.
1137 * "sai_miss_hidden" is used for
1140 unsigned int sai_consecutive_miss; /* consecutive miss */
1141 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1142 * is not a hidden one */
1143 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1145 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1147 sai_agl_valid:1,/* AGL is valid for the dir */
1148 sai_in_readpage:1;/* statahead is in readdir()*/
1149 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1150 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1151 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1152 struct list_head sai_interim_entries; /* entries which got async
1153 * stat reply, but not
1155 struct list_head sai_entries; /* completed entries */
1156 struct list_head sai_agls; /* AGLs to be sent */
1157 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1158 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1159 atomic_t sai_cache_count; /* entry count in cache */
1162 int ll_statahead(struct inode *dir, struct dentry **dentry, bool unplug);
1163 void ll_authorize_statahead(struct inode *dir, void *key);
1164 void ll_deauthorize_statahead(struct inode *dir, void *key);
1167 blkcnt_t dirty_cnt(struct inode *inode);
1169 int cl_glimpse_size0(struct inode *inode, int agl);
1170 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1171 struct inode *inode, struct cl_object *clob, int agl);
1173 static inline int cl_glimpse_size(struct inode *inode)
1175 return cl_glimpse_size0(inode, 0);
1178 static inline int cl_agl(struct inode *inode)
1180 return cl_glimpse_size0(inode, 1);
1183 static inline int ll_glimpse_size(struct inode *inode)
1185 struct ll_inode_info *lli = ll_i2info(inode);
1188 down_read(&lli->lli_glimpse_sem);
1189 rc = cl_glimpse_size(inode);
1190 lli->lli_glimpse_time = cfs_time_current();
1191 up_read(&lli->lli_glimpse_sem);
1195 /* dentry may statahead when statahead is enabled and current process has opened
1196 * parent directory, and this dentry hasn't accessed statahead cache before */
1198 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1200 struct ll_inode_info *lli;
1201 struct ll_dentry_data *ldd;
1203 if (ll_i2sbi(dir)->ll_sa_max == 0)
1206 lli = ll_i2info(dir);
1208 /* statahead is not allowed for this dir, there may be three causes:
1209 * 1. dir is not opened.
1210 * 2. statahead hit ratio is too low.
1211 * 3. previous stat started statahead thread failed. */
1212 if (!lli->lli_sa_enabled)
1215 /* not the same process, don't statahead */
1216 if (lli->lli_opendir_pid != current_pid())
1220 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1221 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1222 * For patchless client, lookup intent is not accurate, which may
1223 * misguide statahead. For example:
1224 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1225 * have the same intent -- IT_GETATTR, while one dentry should access
1226 * statahead cache once, otherwise statahead windows is messed up.
1227 * The solution is as following:
1228 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1229 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1230 * bypass interacting with statahead cache by checking
1231 * 'lld_sa_generation == lli->lli_sa_generation'.
1233 ldd = ll_d2d(dentry);
1234 if (ldd != NULL && ldd->lld_sa_generation == lli->lli_sa_generation)
1240 /* llite ioctl register support rountine */
1246 #define LLIOC_MAX_CMD 256
1249 * Rules to write a callback function:
1252 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1253 * returned to ll_iocontrol_register. Callback functions should use this
1254 * data to check the potential collasion of ioctl cmd. If collasion is
1255 * found, callback function should return LLIOC_CONT.
1256 * @rcp: The result of ioctl command.
1259 * If @magic matches the pointer returned by ll_iocontrol_data, the
1260 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1262 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1263 struct file *file, unsigned int cmd, unsigned long arg,
1264 void *magic, int *rcp);
1266 /* export functions */
1267 /* Register ioctl block dynamatically for a regular file.
1269 * @cmd: the array of ioctl command set
1270 * @count: number of commands in the @cmd
1271 * @cb: callback function, it will be called if an ioctl command is found to
1272 * belong to the command list @cmd.
1275 * A magic pointer will be returned if success;
1276 * otherwise, NULL will be returned.
1278 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1279 void ll_iocontrol_unregister(void *magic);
1281 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1282 enum cl_fsync_mode mode, int ignore_layout);
1284 /** direct write pages */
1285 struct ll_dio_pages {
1286 /** page array to be written. we don't support
1287 * partial pages except the last one. */
1288 struct page **ldp_pages;
1289 /* offset of each page */
1290 loff_t *ldp_offsets;
1291 /** if ldp_offsets is NULL, it means a sequential
1292 * pages to be written, then this is the file offset
1293 * of the * first page. */
1294 loff_t ldp_start_offset;
1295 /** how many bytes are to be written. */
1297 /** # of pages in the array. */
1301 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1302 int rw, struct inode *inode,
1303 struct ll_dio_pages *pv);
1305 static inline int ll_file_nolock(const struct file *file)
1307 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1308 struct inode *inode = file->f_path.dentry->d_inode;
1310 LASSERT(fd != NULL);
1311 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1312 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1315 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1316 struct lookup_intent *it, __u64 *bits)
1318 if (!it->it_lock_set) {
1319 struct lustre_handle handle;
1321 /* If this inode is a remote object, it will get two
1322 * separate locks in different namespaces, Master MDT,
1323 * where the name entry is, will grant LOOKUP lock,
1324 * remote MDT, where the object is, will grant
1325 * UPDATE|PERM lock. The inode will be attched to both
1326 * LOOKUP and PERM locks, so revoking either locks will
1327 * case the dcache being cleared */
1328 if (it->it_remote_lock_mode) {
1329 handle.cookie = it->it_remote_lock_handle;
1330 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1331 "(%p) for remote lock "LPX64"\n",
1332 PFID(ll_inode2fid(inode)), inode,
1334 md_set_lock_data(exp, &handle, inode, NULL);
1337 handle.cookie = it->it_lock_handle;
1339 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1340 " for lock "LPX64"\n",
1341 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1343 md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1344 it->it_lock_set = 1;
1348 *bits = it->it_lock_bits;
1351 static inline void ll_lock_dcache(struct inode *inode)
1353 #ifdef HAVE_DCACHE_LOCK
1354 spin_lock(&dcache_lock);
1356 spin_lock(&inode->i_lock);
1360 static inline void ll_unlock_dcache(struct inode *inode)
1362 #ifdef HAVE_DCACHE_LOCK
1363 spin_unlock(&dcache_lock);
1365 spin_unlock(&inode->i_lock);
1369 static inline int d_lustre_invalid(const struct dentry *dentry)
1371 struct ll_dentry_data *lld = ll_d2d(dentry);
1373 return (lld == NULL) || lld->lld_invalid;
1376 static inline void __d_lustre_invalidate(struct dentry *dentry)
1378 struct ll_dentry_data *lld = ll_d2d(dentry);
1381 lld->lld_invalid = 1;
1385 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1386 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1387 * else dput() of the last refcount will unhash this dentry and kill it.
1389 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1391 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1392 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1393 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1395 spin_lock_nested(&dentry->d_lock,
1396 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1397 __d_lustre_invalidate(dentry);
1399 * We should be careful about dentries created by d_obtain_alias().
1400 * These dentries are not put in the dentry tree, instead they are
1401 * linked to sb->s_anon through dentry->d_hash.
1402 * shrink_dcache_for_umount() shrinks the tree and sb->s_anon list.
1403 * If we unhashed such a dentry, unmount would not be able to find
1404 * it and busy inodes would be reported.
1406 if (ll_d_count(dentry) == 0 && !(dentry->d_flags & DCACHE_DISCONNECTED))
1408 spin_unlock(&dentry->d_lock);
1411 static inline void d_lustre_revalidate(struct dentry *dentry)
1413 spin_lock(&dentry->d_lock);
1414 LASSERT(ll_d2d(dentry) != NULL);
1415 ll_d2d(dentry)->lld_invalid = 0;
1416 spin_unlock(&dentry->d_lock);
1419 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1420 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1421 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1423 int ll_xattr_init(void);
1424 void ll_xattr_fini(void);
1426 int ll_getxattr_common(struct inode *inode, const char *name,
1427 void *buffer, size_t size, __u64 valid);
1428 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1429 struct cl_page *page, enum cl_req_type crt);
1431 int ll_getparent(struct file *file, struct getparent __user *arg);
1434 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1435 unsigned int attr_flags);
1437 extern struct lu_env *cl_inode_fini_env;
1438 extern __u16 cl_inode_fini_refcheck;
1440 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1441 void cl_inode_fini(struct inode *inode);
1443 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1444 u32 cl_fid_build_gen(const struct lu_fid *fid);
1446 #ifndef HAVE_FILE_INODE
1447 static inline struct inode *file_inode(struct file *file)
1449 return file->f_path.dentry->d_inode;
1453 #ifndef HAVE_IOV_ITER_TRUNCATE
1454 static inline void iov_iter_truncate(struct iov_iter *i, u64 count)
1456 if (i->count > count)
1461 #ifndef HAVE_FILE_OPERATIONS_READ_WRITE_ITER
1462 static inline void iov_iter_reexpand(struct iov_iter *i, size_t count)
1467 static inline struct iovec iov_iter_iovec(const struct iov_iter *iter)
1469 return (struct iovec) {
1470 .iov_base = iter->iov->iov_base + iter->iov_offset,
1471 .iov_len = min(iter->count,
1472 iter->iov->iov_len - iter->iov_offset),
1476 #define iov_for_each(iov, iter, start) \
1477 for (iter = (start); \
1478 (iter).count && ((iov = iov_iter_iovec(&(iter))), 1); \
1479 iov_iter_advance(&(iter), (iov).iov_len))
1481 static inline ssize_t
1482 generic_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
1488 iov_for_each(iov, i, *iter) {
1491 res = generic_file_aio_read(iocb, &iov, 1, iocb->ki_pos);
1499 if (res < iov.iov_len)
1504 iov_iter_advance(iter, bytes);
1508 static inline ssize_t
1509 generic_file_write_iter(struct kiocb *iocb, struct iov_iter *iter)
1515 iov_for_each(iov, i, *iter) {
1518 res = __generic_file_aio_write(iocb, &iov, 1, &iocb->ki_pos);
1526 if (res < iov.iov_len)
1531 iov_iter_advance(iter, bytes);
1534 #endif /* HAVE_FILE_OPERATIONS_READ_WRITE_ITER */
1535 #endif /* LLITE_INTERNAL_H */