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 "vvp_internal.h"
50 #include "range_lock.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 /* 4UL * 1024 * 1024 */
71 #define LL_MAX_BLKSIZE_BITS 22
73 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
74 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
76 struct ll_dentry_data {
77 struct lookup_intent *lld_it;
78 unsigned int lld_sa_generation;
79 unsigned int lld_invalid:1;
80 unsigned int lld_nfs_dentry:1;
81 struct rcu_head lld_rcu_head;
84 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
86 #define LLI_INODE_MAGIC 0x111d0de5
87 #define LLI_INODE_DEAD 0xdeadd00d
89 struct ll_getname_data {
90 #ifdef HAVE_DIR_CONTEXT
91 struct dir_context ctx;
93 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
94 struct lu_fid lgd_fid; /* target fid we are looking for */
95 int lgd_found; /* inode matched? */
99 struct lu_env *lg_env;
101 struct cl_lock *lg_lock;
102 unsigned long lg_gid;
105 struct ll_inode_info {
106 __u32 lli_inode_magic;
109 volatile unsigned long lli_flags;
110 struct posix_acl *lli_posix_acl;
112 /* identifying fields for both metadata and data stacks. */
113 struct lu_fid lli_fid;
114 /* master inode fid for stripe directory */
115 struct lu_fid lli_pfid;
117 /* We need all three because every inode may be opened in different
119 struct obd_client_handle *lli_mds_read_och;
120 struct obd_client_handle *lli_mds_write_och;
121 struct obd_client_handle *lli_mds_exec_och;
122 __u64 lli_open_fd_read_count;
123 __u64 lli_open_fd_write_count;
124 __u64 lli_open_fd_exec_count;
125 /* Protects access to och pointers and their usage counters */
126 struct mutex lli_och_mutex;
128 struct inode lli_vfs_inode;
130 /* the most recent timestamps obtained from mds */
134 spinlock_t lli_agl_lock;
136 /* update atime from MDS no matter if it's older than
137 * local inode atime. */
138 unsigned int lli_update_atime:1;
140 /* Try to make the d::member and f::member are aligned. Before using
141 * these members, make clear whether it is directory or not. */
145 /* serialize normal readdir and statahead-readdir. */
146 struct mutex lli_readdir_mutex;
148 /* metadata statahead */
149 /* since parent-child threads can share the same @file
150 * struct, "opendir_key" is the token when dir close for
151 * case of parent exit before child -- it is me should
152 * cleanup the dir readahead. */
153 void *lli_opendir_key;
154 struct ll_statahead_info *lli_sai;
155 /* protect statahead stuff. */
156 spinlock_t lli_sa_lock;
157 /* "opendir_pid" is the token when lookup/revalid
158 * -- I am the owner of dir statahead. */
159 pid_t lli_opendir_pid;
160 /* stat will try to access statahead entries or start
161 * statahead if this flag is set, and this flag will be
162 * set upon dir open, and cleared when dir is closed,
163 * statahead hit ratio is too low, or start statahead
165 unsigned int lli_sa_enabled:1;
166 /* generation for statahead */
167 unsigned int lli_sa_generation;
168 /* directory stripe information */
169 struct lmv_stripe_md *lli_lsm_md;
170 /* default directory stripe offset. This is extracted
171 * from the "dmv" xattr in order to decide which MDT to
172 * create a subdirectory on. The MDS itself fetches
173 * "dmv" and gets the rest of the default layout itself
174 * (count, hash, etc). */
175 __u32 lli_def_stripe_offset;
178 /* for non-directory */
180 struct mutex lli_size_mutex;
181 char *lli_symlink_name;
183 * struct rw_semaphore {
184 * signed long count; // align d.d_def_acl
185 * spinlock_t wait_lock; // align d.d_sa_lock
186 * struct list_head wait_list;
189 struct rw_semaphore lli_trunc_sem;
190 struct range_lock_tree lli_write_tree;
192 struct rw_semaphore lli_glimpse_sem;
193 cfs_time_t lli_glimpse_time;
194 struct list_head lli_agl_list;
197 /* for writepage() only to communicate to fsync */
201 * whenever a process try to read/write the file, the
202 * jobid of the process will be saved here, and it'll
203 * be packed into the write PRC when flush later.
205 * so the read/write statistics for jobid will not be
206 * accurate if the file is shared by different jobs.
208 char lli_jobid[LUSTRE_JOBID_SIZE];
212 /* XXX: For following frequent used members, although they maybe special
213 * used for non-directory object, it is some time-wasting to check
214 * whether the object is directory or not before using them. On the
215 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
216 * the "ll_inode_info" size even if moving those members into u.f.
217 * So keep them out side.
219 * In the future, if more members are added only for directory,
220 * some of the following members can be moved into u.f.
222 struct cl_object *lli_clob;
224 /* mutex to request for layout lock exclusively. */
225 struct mutex lli_layout_mutex;
226 /* Layout version, protected by lli_layout_lock */
227 __u32 lli_layout_gen;
228 spinlock_t lli_layout_lock;
230 __u32 lli_projid; /* project id */
232 struct rw_semaphore lli_xattrs_list_rwsem;
233 struct mutex lli_xattrs_enq_lock;
234 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
237 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
241 spin_lock(&lli->lli_layout_lock);
242 gen = lli->lli_layout_gen;
243 spin_unlock(&lli->lli_layout_lock);
248 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
250 spin_lock(&lli->lli_layout_lock);
251 lli->lli_layout_gen = gen;
252 spin_unlock(&lli->lli_layout_lock);
256 /* File data is modified. */
257 LLIF_DATA_MODIFIED = 0,
258 /* File is being restored */
259 LLIF_FILE_RESTORING = 1,
260 /* Xattr cache is attached to the file */
261 LLIF_XATTR_CACHE = 2,
264 static inline void ll_file_set_flag(struct ll_inode_info *lli,
265 enum ll_file_flags flag)
267 set_bit(flag, &lli->lli_flags);
270 static inline void ll_file_clear_flag(struct ll_inode_info *lli,
271 enum ll_file_flags flag)
273 clear_bit(flag, &lli->lli_flags);
276 static inline bool ll_file_test_flag(struct ll_inode_info *lli,
277 enum ll_file_flags flag)
279 return test_bit(flag, &lli->lli_flags);
282 static inline bool ll_file_test_and_clear_flag(struct ll_inode_info *lli,
283 enum ll_file_flags flag)
285 return test_and_clear_bit(flag, &lli->lli_flags);
288 int ll_xattr_cache_destroy(struct inode *inode);
290 int ll_xattr_cache_get(struct inode *inode,
296 int ll_dentry_init_security(struct dentry *dentry, int mode, struct qstr *name,
297 const char **secctx_name, void **secctx,
299 int ll_inode_init_security(struct dentry *dentry, struct inode *inode,
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_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_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 */
431 #define LL_SBI_FILE_SECCTX 0x800000 /* set file security context at create */
433 #define LL_SBI_FLAGS { \
460 /* This is embedded into llite super-blocks to keep track of connect
461 * flags (capabilities) supported by all imports given mount is
463 struct lustre_client_ocd {
464 /* This is conjunction of connect_flags across all imports
465 * (LOVs) this mount is connected to. This field is updated by
466 * cl_ocd_update() under ->lco_lock. */
468 struct mutex lco_lock;
469 struct obd_export *lco_md_exp;
470 struct obd_export *lco_dt_exp;
474 /* this protects pglist and ra_info. It isn't safe to
475 * grab from interrupt contexts */
477 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
478 spinlock_t ll_process_lock; /* ll_rw_process_info */
479 struct obd_uuid ll_sb_uuid;
480 struct obd_export *ll_md_exp;
481 struct obd_export *ll_dt_exp;
482 struct proc_dir_entry* ll_proc_root;
483 struct lu_fid ll_root_fid; /* root object fid */
486 unsigned int ll_umounting:1,
487 ll_xattr_cache_enabled:1,
488 ll_client_common_fill_super_succeeded:1;
490 struct lustre_client_ocd ll_lco;
492 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
494 /* Used to track "unstable" pages on a client, and maintain a
495 * LRU list of clean pages. An "unstable" page is defined as
496 * any page which is sent to a server as part of a bulk request,
497 * but is uncommitted to stable storage. */
498 struct cl_client_cache *ll_cache;
500 struct lprocfs_stats *ll_ra_stats;
502 struct ll_ra_info ll_ra_info;
503 unsigned int ll_namelen;
504 struct file_operations *ll_fop;
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;
537 * per file-descriptor read-ahead data.
539 struct ll_readahead_state {
542 * index of the last page that read(2) needed and that wasn't in the
543 * cache. Used by ras_update() to detect seeks.
545 * XXX nikita: if access seeks into cached region, Lustre doesn't see
548 unsigned long ras_last_readpage;
550 * number of pages read after last read-ahead window reset. As window
551 * is reset on each seek, this is effectively a number of consecutive
552 * accesses. Maybe ->ras_accessed_in_window is better name.
554 * XXX nikita: window is also reset (by ras_update()) when Lustre
555 * believes that memory pressure evicts read-ahead pages. In that
556 * case, it probably doesn't make sense to expand window to
557 * PTLRPC_MAX_BRW_PAGES on the third access.
559 unsigned long ras_consecutive_pages;
561 * number of read requests after the last read-ahead window reset
562 * As window is reset on each seek, this is effectively the number
563 * on consecutive read request and is used to trigger read-ahead.
565 unsigned long ras_consecutive_requests;
567 * Parameters of current read-ahead window. Handled by
568 * ras_update(). On the initial access to the file or after a seek,
569 * window is reset to 0. After 3 consecutive accesses, window is
570 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
571 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
573 unsigned long ras_window_start, ras_window_len;
575 * Optimal RPC size. It decides how many pages will be sent
576 * for each read-ahead.
578 unsigned long ras_rpc_size;
580 * Where next read-ahead should start at. This lies within read-ahead
581 * window. Read-ahead window is read in pieces rather than at once
582 * because: 1. lustre limits total number of pages under read-ahead by
583 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
584 * not covered by DLM lock.
586 unsigned long ras_next_readahead;
588 * Total number of ll_file_read requests issued, reads originating
589 * due to mmap are not counted in this total. This value is used to
590 * trigger full file read-ahead after multiple reads to a small file.
592 unsigned long ras_requests;
594 * Page index with respect to the current request, these value
595 * will not be accurate when dealing with reads issued via mmap.
597 unsigned long ras_request_index;
599 * The following 3 items are used for detecting the stride I/O
601 * In stride I/O mode,
602 * ...............|-----data-----|****gap*****|--------|******|....
603 * offset |-stride_pages-|-stride_gap-|
604 * ras_stride_offset = offset;
605 * ras_stride_length = stride_pages + stride_gap;
606 * ras_stride_pages = stride_pages;
607 * Note: all these three items are counted by pages.
609 unsigned long ras_stride_length;
610 unsigned long ras_stride_pages;
611 pgoff_t ras_stride_offset;
613 * number of consecutive stride request count, and it is similar as
614 * ras_consecutive_requests, but used for stride I/O mode.
615 * Note: only more than 2 consecutive stride request are detected,
616 * stride read-ahead will be enable
618 unsigned long ras_consecutive_stride_requests;
621 extern struct kmem_cache *ll_file_data_slab;
622 struct lustre_handle;
623 struct ll_file_data {
624 struct ll_readahead_state fd_ras;
625 struct ll_grouplock fd_grouplock;
629 /* openhandle if lease exists for this file.
630 * Borrow lli->lli_och_mutex to protect assignment */
631 struct obd_client_handle *fd_lease_och;
632 struct obd_client_handle *fd_och;
633 struct file *fd_file;
634 /* Indicate whether need to report failure when close.
635 * true: failure is known, not report again.
636 * false: unknown failure, should report. */
637 bool fd_write_failed;
638 rwlock_t fd_lock; /* protect lcc list */
639 struct list_head fd_lccs; /* list of ll_cl_context */
642 extern struct proc_dir_entry *proc_lustre_fs_root;
644 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
646 return &lli->lli_vfs_inode;
649 __u32 ll_i2suppgid(struct inode *i);
650 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
652 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
654 #if BITS_PER_LONG == 32
656 #elif defined(CONFIG_COMPAT)
657 return unlikely(in_compat_syscall() || (sbi->ll_flags & LL_SBI_32BIT_API));
659 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
663 static inline bool ll_sbi_has_fast_read(struct ll_sb_info *sbi)
665 return !!(sbi->ll_flags & LL_SBI_FAST_READ);
668 void ll_ras_enter(struct file *f);
670 /* llite/lcommon_misc.c */
671 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
672 enum obd_notify_event ev, void *owner);
673 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
674 struct ll_grouplock *lg);
675 void cl_put_grouplock(struct ll_grouplock *lg);
677 /* llite/lproc_llite.c */
678 #ifdef CONFIG_PROC_FS
679 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
680 struct super_block *sb, char *osc, char *mdc);
681 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
682 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
683 extern struct lprocfs_vars lprocfs_llite_obd_vars[];
685 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
686 struct super_block *sb, char *osc, char *mdc){return 0;}
687 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
688 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
693 LPROC_LL_DIRTY_MISSES,
695 LPROC_LL_WRITE_BYTES,
720 LPROC_LL_ALLOC_INODE,
723 LPROC_LL_GETXATTR_HITS,
725 LPROC_LL_REMOVEXATTR,
727 LPROC_LL_FILE_OPCODES
731 struct ll_dir_chain {
734 static inline void ll_dir_chain_init(struct ll_dir_chain *chain)
738 static inline void ll_dir_chain_fini(struct ll_dir_chain *chain)
742 extern const struct file_operations ll_dir_operations;
743 extern const struct inode_operations ll_dir_inode_operations;
744 #ifdef HAVE_DIR_CONTEXT
745 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
746 struct dir_context *ctx);
748 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
749 void *cookie, filldir_t filldir);
751 int ll_get_mdt_idx(struct inode *inode);
752 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
753 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
754 __u64 offset, struct ll_dir_chain *chain);
755 void ll_release_page(struct inode *inode, struct page *page, bool remove);
758 extern const struct inode_operations ll_special_inode_operations;
760 struct inode *ll_iget(struct super_block *sb, ino_t hash,
761 struct lustre_md *lic);
762 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
763 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
764 void *data, int flag);
765 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
766 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
767 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
770 int ll_writepage(struct page *page, struct writeback_control *wbc);
771 int ll_writepages(struct address_space *, struct writeback_control *wbc);
772 int ll_readpage(struct file *file, struct page *page);
773 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
774 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
777 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io,
779 void ll_cl_remove(struct file *file, const struct lu_env *env);
780 struct ll_cl_context *ll_cl_find(struct file *file);
782 extern const struct address_space_operations ll_aops;
785 extern struct file_operations ll_file_operations;
786 extern struct file_operations ll_file_operations_flock;
787 extern struct file_operations ll_file_operations_noflock;
788 extern struct inode_operations ll_file_inode_operations;
789 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
790 enum ldlm_mode l_req_mode);
791 extern enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
792 struct lustre_handle *lockh, __u64 flags,
793 enum ldlm_mode mode);
795 int ll_file_open(struct inode *inode, struct file *file);
796 int ll_file_release(struct inode *inode, struct file *file);
797 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
798 int ll_md_real_close(struct inode *inode, fmode_t fmode);
799 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
800 struct ll_file_data *file, loff_t pos,
801 size_t count, int rw);
802 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
803 struct posix_acl *ll_get_acl(struct inode *inode, int type);
804 int ll_migrate(struct inode *parent, struct file *file, int mdtidx,
805 const char *name, int namelen);
806 int ll_get_fid_by_name(struct inode *parent, const char *name,
807 int namelen, struct lu_fid *fid, struct inode **inode);
808 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
809 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
811 # ifndef HAVE_INODE_PERMISION_2ARGS
812 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
814 int ll_inode_permission(struct inode *inode, int mask);
817 int ll_ioctl_fsgetxattr(struct inode *inode, unsigned int cmd,
819 int ll_ioctl_fssetxattr(struct inode *inode, unsigned int cmd,
822 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
823 __u64 flags, struct lov_user_md *lum,
825 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
826 struct lov_mds_md **lmm, int *lmm_size,
827 struct ptlrpc_request **request);
828 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
830 int ll_dir_getstripe(struct inode *inode, void **lmmp,
831 int *lmm_size, struct ptlrpc_request **request,
833 #ifdef HAVE_FILE_FSYNC_4ARGS
834 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
835 #elif defined(HAVE_FILE_FSYNC_2ARGS)
836 int ll_fsync(struct file *file, int data);
838 int ll_fsync(struct file *file, struct dentry *dentry, int data);
840 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
841 int ll_fid2path(struct inode *inode, void __user *arg);
842 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
843 int ll_hsm_release(struct inode *inode);
844 int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
848 int ll_d_init(struct dentry *de);
849 extern const struct dentry_operations ll_d_ops;
850 void ll_intent_drop_lock(struct lookup_intent *);
851 void ll_intent_release(struct lookup_intent *);
852 void ll_invalidate_aliases(struct inode *);
853 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
854 int ll_revalidate_it_finish(struct ptlrpc_request *request,
855 struct lookup_intent *it, struct dentry *de);
857 /* llite/llite_lib.c */
858 extern struct super_operations lustre_super_operations;
860 void ll_lli_init(struct ll_inode_info *lli);
861 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
862 void ll_put_super(struct super_block *sb);
863 void ll_kill_super(struct super_block *sb);
864 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
865 void ll_dir_clear_lsm_md(struct inode *inode);
866 void ll_clear_inode(struct inode *inode);
867 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import);
868 int ll_setattr(struct dentry *de, struct iattr *attr);
869 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
870 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
871 __u64 max_age, __u32 flags);
872 int ll_update_inode(struct inode *inode, struct lustre_md *md);
873 int ll_read_inode2(struct inode *inode, void *opaque);
874 void ll_delete_inode(struct inode *inode);
875 int ll_iocontrol(struct inode *inode, struct file *file,
876 unsigned int cmd, unsigned long arg);
877 int ll_flush_ctx(struct inode *inode);
878 void ll_umount_begin(struct super_block *sb);
879 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
880 #ifdef HAVE_SUPEROPS_USE_DENTRY
881 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
883 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
885 void ll_dirty_page_discard_warn(struct page *page, int ioret);
886 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
887 struct super_block *, struct lookup_intent *);
888 int ll_obd_statfs(struct inode *inode, void __user *arg);
889 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
890 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
891 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
892 int ll_process_config(struct lustre_cfg *lcfg);
895 LUSTRE_OPC_MKDIR = 0,
896 LUSTRE_OPC_SYMLINK = 1,
897 LUSTRE_OPC_MKNOD = 2,
898 LUSTRE_OPC_CREATE = 3,
902 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
903 struct inode *i1, struct inode *i2,
904 const char *name, size_t namelen,
905 __u32 mode, __u32 opc, void *data);
906 void ll_finish_md_op_data(struct md_op_data *op_data);
907 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
908 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
909 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
910 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
911 struct lov_user_md **kbuf);
912 void ll_open_cleanup(struct super_block *sb, struct ptlrpc_request *open_req);
914 /* Compute expected user md size when passing in a md from user space */
915 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
917 switch (lum->lmm_magic) {
918 case LOV_USER_MAGIC_V1:
919 return sizeof(struct lov_user_md_v1);
920 case LOV_USER_MAGIC_V3:
921 return sizeof(struct lov_user_md_v3);
922 case LOV_USER_MAGIC_SPECIFIC:
923 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
926 return lov_user_md_size(lum->lmm_stripe_count,
927 LOV_USER_MAGIC_SPECIFIC);
928 case LOV_USER_MAGIC_COMP_V1:
929 return ((struct lov_comp_md_v1 *)lum)->lcm_size;
935 /* llite/llite_nfs.c */
936 extern struct export_operations lustre_export_operations;
937 __u32 get_uuid2int(const char *name, int len);
938 struct inode *search_inode_for_lustre(struct super_block *sb,
939 const struct lu_fid *fid);
940 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
942 /* llite/symlink.c */
943 extern struct inode_operations ll_fast_symlink_inode_operations;
946 * IO arguments for various VFS I/O interfaces.
949 /** normal/sendfile/splice */
950 enum vvp_io_subtype via_io_subtype;
954 struct kiocb *via_iocb;
955 struct iov_iter *via_iter;
958 struct pipe_inode_info *via_pipe;
959 unsigned int via_flags;
969 struct ll_cl_context {
970 struct list_head lcc_list;
972 const struct lu_env *lcc_env;
973 struct cl_io *lcc_io;
974 struct cl_page *lcc_page;
975 enum lcc_type lcc_type;
978 struct ll_thread_info {
979 struct iov_iter lti_iter;
980 struct vvp_io_args lti_args;
981 struct ra_io_arg lti_ria;
982 struct kiocb lti_kiocb;
983 struct ll_cl_context lti_io_ctx;
986 extern struct lu_context_key ll_thread_key;
988 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
990 struct ll_thread_info *lti;
992 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
993 LASSERT(lti != NULL);
998 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env,
999 enum vvp_io_subtype type)
1001 struct vvp_io_args *via = &ll_env_info(env)->lti_args;
1003 via->via_io_subtype = type;
1008 /* llite/llite_mmap.c */
1010 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1011 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1012 void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
1013 unsigned long addr, size_t count);
1014 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1017 static inline void ll_invalidate_page(struct page *vmpage)
1019 struct address_space *mapping = vmpage->mapping;
1020 loff_t offset = vmpage->index << PAGE_SHIFT;
1022 LASSERT(PageLocked(vmpage));
1023 if (mapping == NULL)
1027 * truncate_complete_page() calls
1028 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
1030 ll_teardown_mmaps(mapping, offset, offset + PAGE_SIZE);
1031 truncate_complete_page(mapping, vmpage);
1034 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1036 /* don't need an addref as the sb_info should be holding one */
1037 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1039 return ll_s2sbi(sb)->ll_dt_exp;
1042 /* don't need an addref as the sb_info should be holding one */
1043 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1045 return ll_s2sbi(sb)->ll_md_exp;
1048 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1050 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1056 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1057 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1059 return ll_s2sbi(inode->i_sb);
1062 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1064 return ll_s2dtexp(inode->i_sb);
1067 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1069 return ll_s2mdexp(inode->i_sb);
1072 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1076 LASSERT(inode != NULL);
1077 fid = &ll_i2info(inode)->lli_fid;
1082 static inline loff_t ll_file_maxbytes(struct inode *inode)
1084 struct cl_object *obj = ll_i2info(inode)->lli_clob;
1087 return MAX_LFS_FILESIZE;
1089 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
1093 int ll_setxattr(struct dentry *dentry, const char *name,
1094 const void *value, size_t size, int flags);
1095 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1096 void *buffer, size_t size);
1097 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1098 int ll_removexattr(struct dentry *dentry, const char *name);
1101 * Common IO arguments for various VFS I/O interfaces.
1103 int cl_sb_init(struct super_block *sb);
1104 int cl_sb_fini(struct super_block *sb);
1106 enum ras_update_flags {
1110 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1111 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1115 #define LL_SA_RPC_MIN 2
1116 #define LL_SA_RPC_DEF 32
1117 #define LL_SA_RPC_MAX 8192
1119 #define LL_SA_CACHE_BIT 5
1120 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1121 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1123 /* per inode struct, for dir only */
1124 struct ll_statahead_info {
1125 struct dentry *sai_dentry;
1126 atomic_t sai_refcount; /* when access this struct, hold
1128 unsigned int sai_max; /* max ahead of lookup */
1129 __u64 sai_sent; /* stat requests sent count */
1130 __u64 sai_replied; /* stat requests which received
1132 __u64 sai_index; /* index of statahead entry */
1133 __u64 sai_index_wait; /* index of entry which is the
1134 * caller is waiting for */
1135 __u64 sai_hit; /* hit count */
1136 __u64 sai_miss; /* miss count:
1137 * for "ls -al" case, includes
1138 * hidden dentry miss;
1139 * for "ls -l" case, it does not
1140 * include hidden dentry miss.
1141 * "sai_miss_hidden" is used for
1144 unsigned int sai_consecutive_miss; /* consecutive miss */
1145 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1146 * is not a hidden one */
1147 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1149 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1151 sai_agl_valid:1,/* AGL is valid for the dir */
1152 sai_in_readpage:1;/* statahead is in readdir()*/
1153 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1154 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1155 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1156 struct list_head sai_interim_entries; /* entries which got async
1157 * stat reply, but not
1159 struct list_head sai_entries; /* completed entries */
1160 struct list_head sai_agls; /* AGLs to be sent */
1161 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1162 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1163 atomic_t sai_cache_count; /* entry count in cache */
1166 int ll_statahead(struct inode *dir, struct dentry **dentry, bool unplug);
1167 void ll_authorize_statahead(struct inode *dir, void *key);
1168 void ll_deauthorize_statahead(struct inode *dir, void *key);
1171 blkcnt_t dirty_cnt(struct inode *inode);
1173 int cl_glimpse_size0(struct inode *inode, int agl);
1174 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1175 struct inode *inode, struct cl_object *clob, int agl);
1177 static inline int cl_glimpse_size(struct inode *inode)
1179 return cl_glimpse_size0(inode, 0);
1182 static inline int cl_agl(struct inode *inode)
1184 return cl_glimpse_size0(inode, 1);
1187 static inline int ll_glimpse_size(struct inode *inode)
1189 struct ll_inode_info *lli = ll_i2info(inode);
1192 down_read(&lli->lli_glimpse_sem);
1193 rc = cl_glimpse_size(inode);
1194 lli->lli_glimpse_time = cfs_time_current();
1195 up_read(&lli->lli_glimpse_sem);
1199 /* dentry may statahead when statahead is enabled and current process has opened
1200 * parent directory, and this dentry hasn't accessed statahead cache before */
1202 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1204 struct ll_inode_info *lli;
1205 struct ll_dentry_data *ldd;
1207 if (ll_i2sbi(dir)->ll_sa_max == 0)
1210 lli = ll_i2info(dir);
1212 /* statahead is not allowed for this dir, there may be three causes:
1213 * 1. dir is not opened.
1214 * 2. statahead hit ratio is too low.
1215 * 3. previous stat started statahead thread failed. */
1216 if (!lli->lli_sa_enabled)
1219 /* not the same process, don't statahead */
1220 if (lli->lli_opendir_pid != current_pid())
1224 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1225 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1226 * For patchless client, lookup intent is not accurate, which may
1227 * misguide statahead. For example:
1228 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1229 * have the same intent -- IT_GETATTR, while one dentry should access
1230 * statahead cache once, otherwise statahead windows is messed up.
1231 * The solution is as following:
1232 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1233 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1234 * bypass interacting with statahead cache by checking
1235 * 'lld_sa_generation == lli->lli_sa_generation'.
1237 ldd = ll_d2d(dentry);
1238 if (ldd != NULL && ldd->lld_sa_generation == lli->lli_sa_generation)
1244 /* llite ioctl register support rountine */
1250 #define LLIOC_MAX_CMD 256
1253 * Rules to write a callback function:
1256 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1257 * returned to ll_iocontrol_register. Callback functions should use this
1258 * data to check the potential collasion of ioctl cmd. If collasion is
1259 * found, callback function should return LLIOC_CONT.
1260 * @rcp: The result of ioctl command.
1263 * If @magic matches the pointer returned by ll_iocontrol_data, the
1264 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1266 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1267 struct file *file, unsigned int cmd, unsigned long arg,
1268 void *magic, int *rcp);
1270 /* export functions */
1271 /* Register ioctl block dynamatically for a regular file.
1273 * @cmd: the array of ioctl command set
1274 * @count: number of commands in the @cmd
1275 * @cb: callback function, it will be called if an ioctl command is found to
1276 * belong to the command list @cmd.
1279 * A magic pointer will be returned if success;
1280 * otherwise, NULL will be returned.
1282 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1283 void ll_iocontrol_unregister(void *magic);
1285 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1286 enum cl_fsync_mode mode, int ignore_layout);
1288 /** direct write pages */
1289 struct ll_dio_pages {
1290 /** page array to be written. we don't support
1291 * partial pages except the last one. */
1292 struct page **ldp_pages;
1293 /* offset of each page */
1294 loff_t *ldp_offsets;
1295 /** if ldp_offsets is NULL, it means a sequential
1296 * pages to be written, then this is the file offset
1297 * of the * first page. */
1298 loff_t ldp_start_offset;
1299 /** how many bytes are to be written. */
1301 /** # of pages in the array. */
1305 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1306 int rw, struct inode *inode,
1307 struct ll_dio_pages *pv);
1309 static inline int ll_file_nolock(const struct file *file)
1311 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1312 struct inode *inode = file_inode((struct file *)file);
1314 LASSERT(fd != NULL);
1315 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1316 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1319 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1320 struct lookup_intent *it, __u64 *bits)
1322 if (!it->it_lock_set) {
1323 struct lustre_handle handle;
1325 /* If this inode is a remote object, it will get two
1326 * separate locks in different namespaces, Master MDT,
1327 * where the name entry is, will grant LOOKUP lock,
1328 * remote MDT, where the object is, will grant
1329 * UPDATE|PERM lock. The inode will be attched to both
1330 * LOOKUP and PERM locks, so revoking either locks will
1331 * case the dcache being cleared */
1332 if (it->it_remote_lock_mode) {
1333 handle.cookie = it->it_remote_lock_handle;
1334 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1335 "(%p) for remote lock %#llx\n",
1336 PFID(ll_inode2fid(inode)), inode,
1338 md_set_lock_data(exp, &handle, inode, NULL);
1341 handle.cookie = it->it_lock_handle;
1343 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1344 " for lock %#llx\n",
1345 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1347 md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1348 it->it_lock_set = 1;
1352 *bits = it->it_lock_bits;
1355 static inline void ll_lock_dcache(struct inode *inode)
1357 #ifdef HAVE_DCACHE_LOCK
1358 spin_lock(&dcache_lock);
1360 spin_lock(&inode->i_lock);
1364 static inline void ll_unlock_dcache(struct inode *inode)
1366 #ifdef HAVE_DCACHE_LOCK
1367 spin_unlock(&dcache_lock);
1369 spin_unlock(&inode->i_lock);
1373 static inline int d_lustre_invalid(const struct dentry *dentry)
1375 struct ll_dentry_data *lld = ll_d2d(dentry);
1377 return (lld == NULL) || lld->lld_invalid;
1380 static inline void __d_lustre_invalidate(struct dentry *dentry)
1382 struct ll_dentry_data *lld = ll_d2d(dentry);
1385 lld->lld_invalid = 1;
1389 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1390 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1391 * else dput() of the last refcount will unhash this dentry and kill it.
1393 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1395 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1396 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1397 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1399 spin_lock_nested(&dentry->d_lock,
1400 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1401 __d_lustre_invalidate(dentry);
1403 * We should be careful about dentries created by d_obtain_alias().
1404 * These dentries are not put in the dentry tree, instead they are
1405 * linked to sb->s_anon through dentry->d_hash.
1406 * shrink_dcache_for_umount() shrinks the tree and sb->s_anon list.
1407 * If we unhashed such a dentry, unmount would not be able to find
1408 * it and busy inodes would be reported.
1410 if (ll_d_count(dentry) == 0 && !(dentry->d_flags & DCACHE_DISCONNECTED))
1412 spin_unlock(&dentry->d_lock);
1415 static inline void d_lustre_revalidate(struct dentry *dentry)
1417 spin_lock(&dentry->d_lock);
1418 LASSERT(ll_d2d(dentry) != NULL);
1419 ll_d2d(dentry)->lld_invalid = 0;
1420 spin_unlock(&dentry->d_lock);
1423 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1424 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1425 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1426 int ll_layout_write_intent(struct inode *inode, __u64 start, __u64 end);
1428 int ll_xattr_init(void);
1429 void ll_xattr_fini(void);
1431 int ll_getxattr_common(struct inode *inode, const char *name,
1432 void *buffer, size_t size, __u64 valid);
1433 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1434 struct cl_page *page, enum cl_req_type crt);
1436 int ll_getparent(struct file *file, struct getparent __user *arg);
1439 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1440 unsigned int attr_flags);
1442 extern struct lu_env *cl_inode_fini_env;
1443 extern __u16 cl_inode_fini_refcheck;
1445 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1446 void cl_inode_fini(struct inode *inode);
1448 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1449 u32 cl_fid_build_gen(const struct lu_fid *fid);
1451 #ifndef HAVE_IOV_ITER_TRUNCATE
1452 static inline void iov_iter_truncate(struct iov_iter *i, u64 count)
1454 if (i->count > count)
1459 #ifndef HAVE_IS_SXID
1460 static inline bool is_sxid(umode_t mode)
1462 return (mode & S_ISUID) || ((mode & S_ISGID) && (mode & S_IXGRP));
1467 #define IS_NOSEC(inode) (!is_sxid(inode->i_mode))
1470 #ifndef HAVE_FILE_OPERATIONS_READ_WRITE_ITER
1471 static inline void iov_iter_reexpand(struct iov_iter *i, size_t count)
1476 static inline struct iovec iov_iter_iovec(const struct iov_iter *iter)
1478 return (struct iovec) {
1479 .iov_base = iter->iov->iov_base + iter->iov_offset,
1480 .iov_len = min(iter->count,
1481 iter->iov->iov_len - iter->iov_offset),
1485 #define iov_for_each(iov, iter, start) \
1486 for (iter = (start); \
1487 (iter).count && ((iov = iov_iter_iovec(&(iter))), 1); \
1488 iov_iter_advance(&(iter), (iov).iov_len))
1490 static inline ssize_t
1491 generic_file_read_iter(struct kiocb *iocb, struct iov_iter *iter)
1497 iov_for_each(iov, i, *iter) {
1500 res = generic_file_aio_read(iocb, &iov, 1, iocb->ki_pos);
1508 if (res < iov.iov_len)
1513 iov_iter_advance(iter, bytes);
1517 static inline ssize_t
1518 __generic_file_write_iter(struct kiocb *iocb, struct iov_iter *iter)
1524 iov_for_each(iov, i, *iter) {
1527 res = __generic_file_aio_write(iocb, &iov, 1, &iocb->ki_pos);
1535 if (res < iov.iov_len)
1540 iov_iter_advance(iter, bytes);
1543 #endif /* HAVE_FILE_OPERATIONS_READ_WRITE_ITER */
1544 #endif /* LLITE_INTERNAL_H */