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, 2015, 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 "../include/lustre_debug.h"
36 #include "../include/lustre_ver.h"
37 #include "../include/lustre_disk.h" /* for s2sbi */
38 #include "../include/lustre_eacl.h"
39 #include "../include/lustre_linkea.h"
41 /* for struct cl_lock_descr and struct cl_io */
42 #include "../include/lustre_patchless_compat.h"
43 #include "../include/lustre_compat.h"
44 #include "../include/cl_object.h"
45 #include "../include/lustre_lmv.h"
46 #include "../include/lustre_mdc.h"
47 #include "../include/lustre_intent.h"
48 #include <linux/compat.h>
49 #include <linux/namei.h>
50 #include <linux/xattr.h>
51 #include <linux/posix_acl_xattr.h>
52 #include "vvp_internal.h"
53 #include "range_lock.h"
59 #ifndef VM_FAULT_RETRY
60 #define VM_FAULT_RETRY 0
63 /** Only used on client-side for indicating the tail of dir hash/offset. */
64 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
65 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
67 /* 4UL * 1024 * 1024 */
68 #define LL_MAX_BLKSIZE_BITS 22
70 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
71 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
73 struct ll_dentry_data
{
74 struct lookup_intent
*lld_it
;
75 unsigned int lld_sa_generation
;
76 unsigned int lld_invalid
:1;
77 unsigned int lld_nfs_dentry
:1;
78 struct rcu_head lld_rcu_head
;
81 #define ll_d2d(de) ((struct ll_dentry_data *)((de)->d_fsdata))
83 #define LLI_INODE_MAGIC 0x111d0de5
84 #define LLI_INODE_DEAD 0xdeadd00d
86 struct ll_getname_data
{
87 struct dir_context ctx
;
88 char *lgd_name
; /* points to a buffer with NAME_MAX+1 size */
89 struct lu_fid lgd_fid
; /* target fid we are looking for */
90 int lgd_found
; /* inode matched? */
94 struct lu_env
*lg_env
;
96 struct cl_lock
*lg_lock
;
101 /* MDS has an authority for the Size-on-MDS attributes. */
102 LLIF_MDS_SIZE_LOCK
= (1 << 0),
103 /* Epoch close is postponed. */
104 LLIF_EPOCH_PENDING
= (1 << 1),
105 /* DONE WRITING is allowed. */
106 LLIF_DONE_WRITING
= (1 << 2),
107 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
110 LLIF_SOM_DIRTY
= (1 << 3),
111 /* File data is modified. */
112 LLIF_DATA_MODIFIED
= (1 << 4),
113 /* File is being restored */
114 LLIF_FILE_RESTORING
= (1 << 5),
115 /* Xattr cache is attached to the file */
116 LLIF_XATTR_CACHE
= (1 << 6),
119 struct ll_inode_info
{
120 __u32 lli_inode_magic
;
125 struct posix_acl
*lli_posix_acl
;
127 /* identifying fields for both metadata and data stacks. */
128 struct lu_fid lli_fid
;
129 /* master inode fid for stripe directory */
130 struct lu_fid lli_pfid
;
132 struct list_head lli_close_list
;
134 /* handle is to be sent to MDS later on done_writing and setattr.
135 * Open handle data are needed for the recovery to reconstruct
136 * the inode state on the MDS. XXX: recovery is not ready yet.
138 struct obd_client_handle
*lli_pending_och
;
140 /* We need all three because every inode may be opened in different
143 struct obd_client_handle
*lli_mds_read_och
;
144 struct obd_client_handle
*lli_mds_write_och
;
145 struct obd_client_handle
*lli_mds_exec_och
;
146 __u64 lli_open_fd_read_count
;
147 __u64 lli_open_fd_write_count
;
148 __u64 lli_open_fd_exec_count
;
149 /* Protects access to och pointers and their usage counters */
150 struct mutex lli_och_mutex
;
152 struct inode lli_vfs_inode
;
154 /* the most recent timestamps obtained from mds */
158 spinlock_t lli_agl_lock
;
160 /* Try to make the d::member and f::member are aligned. Before using
161 * these members, make clear whether it is directory or not.
166 /* serialize normal readdir and statahead-readdir. */
167 struct mutex lli_readdir_mutex
;
169 /* metadata statahead */
170 /* since parent-child threads can share the same @file
171 * struct, "opendir_key" is the token when dir close for
172 * case of parent exit before child -- it is me should
173 * cleanup the dir readahead.
175 void *lli_opendir_key
;
176 struct ll_statahead_info
*lli_sai
;
177 /* protect statahead stuff. */
178 spinlock_t lli_sa_lock
;
179 /* "opendir_pid" is the token when lookup/revalidate
180 * -- I am the owner of dir statahead.
182 pid_t lli_opendir_pid
;
183 /* stat will try to access statahead entries or start
184 * statahead if this flag is set, and this flag will be
185 * set upon dir open, and cleared when dir is closed,
186 * statahead hit ratio is too low, or start statahead
189 unsigned int lli_sa_enabled
:1;
190 /* generation for statahead */
191 unsigned int lli_sa_generation
;
192 /* directory stripe information */
193 struct lmv_stripe_md
*lli_lsm_md
;
194 /* default directory stripe offset. This is extracted
195 * from the "dmv" xattr in order to decide which MDT to
196 * create a subdirectory on. The MDS itself fetches
197 * "dmv" and gets the rest of the default layout itself
198 * (count, hash, etc).
200 __u32 lli_def_stripe_offset
;
203 /* for non-directory */
205 struct mutex lli_size_mutex
;
206 char *lli_symlink_name
;
209 * struct rw_semaphore {
210 * signed long count; // align d.d_def_acl
211 * spinlock_t wait_lock; // align d.d_sa_lock
212 * struct list_head wait_list;
215 struct rw_semaphore lli_trunc_sem
;
216 struct range_lock_tree lli_write_tree
;
218 struct rw_semaphore lli_glimpse_sem
;
219 unsigned long lli_glimpse_time
;
220 struct list_head lli_agl_list
;
223 /* for writepage() only to communicate to fsync */
227 * whenever a process try to read/write the file, the
228 * jobid of the process will be saved here, and it'll
229 * be packed into the write PRC when flush later.
231 * so the read/write statistics for jobid will not be
232 * accurate if the file is shared by different jobs.
234 char lli_jobid
[LUSTRE_JOBID_SIZE
];
238 /* XXX: For following frequent used members, although they maybe special
239 * used for non-directory object, it is some time-wasting to check
240 * whether the object is directory or not before using them. On the
241 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
242 * the "ll_inode_info" size even if moving those members into u.f.
243 * So keep them out side.
245 * In the future, if more members are added only for directory,
246 * some of the following members can be moved into u.f.
249 struct cl_object
*lli_clob
;
251 /* mutex to request for layout lock exclusively. */
252 struct mutex lli_layout_mutex
;
253 /* Layout version, protected by lli_layout_lock */
254 __u32 lli_layout_gen
;
255 spinlock_t lli_layout_lock
;
257 struct rw_semaphore lli_xattrs_list_rwsem
;
258 struct mutex lli_xattrs_enq_lock
;
259 struct list_head lli_xattrs
;/* ll_xattr_entry->xe_list */
262 static inline __u32
ll_layout_version_get(struct ll_inode_info
*lli
)
266 spin_lock(&lli
->lli_layout_lock
);
267 gen
= lli
->lli_layout_gen
;
268 spin_unlock(&lli
->lli_layout_lock
);
273 static inline void ll_layout_version_set(struct ll_inode_info
*lli
, __u32 gen
)
275 spin_lock(&lli
->lli_layout_lock
);
276 lli
->lli_layout_gen
= gen
;
277 spin_unlock(&lli
->lli_layout_lock
);
280 int ll_xattr_cache_destroy(struct inode
*inode
);
282 int ll_xattr_cache_get(struct inode
*inode
, const char *name
,
283 char *buffer
, size_t size
, __u64 valid
);
286 * Locking to guarantee consistency of non-atomic updates to long long i_size,
287 * consistency between file size and KMS.
289 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
292 void ll_inode_size_lock(struct inode
*inode
);
293 void ll_inode_size_unlock(struct inode
*inode
);
295 /* FIXME: replace the name of this with LL_I to conform to kernel stuff */
296 /* static inline struct ll_inode_info *LL_I(struct inode *inode) */
297 static inline struct ll_inode_info
*ll_i2info(struct inode
*inode
)
299 return container_of(inode
, struct ll_inode_info
, lli_vfs_inode
);
302 /* default to about 40meg of readahead on a given system. That much tied
303 * up in 512k readahead requests serviced at 40ms each is about 1GB/s.
305 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_SHIFT))
307 /* default to read-ahead full files smaller than 2MB on the second read */
308 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_SHIFT))
313 RA_STAT_DISTANT_READPAGE
,
314 RA_STAT_MISS_IN_WINDOW
,
315 RA_STAT_FAILED_GRAB_PAGE
,
316 RA_STAT_FAILED_MATCH
,
321 RA_STAT_MAX_IN_FLIGHT
,
322 RA_STAT_WRONG_GRAB_PAGE
,
323 RA_STAT_FAILED_REACH_END
,
328 atomic_t ra_cur_pages
;
329 unsigned long ra_max_pages
;
330 unsigned long ra_max_pages_per_file
;
331 unsigned long ra_max_read_ahead_whole_pages
;
334 /* ra_io_arg will be filled in the beginning of ll_readahead with
335 * ras_lock, then the following ll_read_ahead_pages will read RA
336 * pages according to this arg, all the items in this structure are
337 * counted by page index.
340 unsigned long ria_start
; /* start offset of read-ahead*/
341 unsigned long ria_end
; /* end offset of read-ahead*/
342 /* If stride read pattern is detected, ria_stoff means where
343 * stride read is started. Note: for normal read-ahead, the
344 * value here is meaningless, and also it will not be accessed
347 /* ria_length and ria_pages are the length and pages length in the
348 * stride I/O mode. And they will also be used to check whether
349 * it is stride I/O read-ahead in the read-ahead pages
351 unsigned long ria_length
;
352 unsigned long ria_pages
;
355 /* LL_HIST_MAX=32 causes an overflow */
356 #define LL_HIST_MAX 28
357 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
358 #define LL_PROCESS_HIST_MAX 10
359 struct per_process_info
{
361 struct obd_histogram pp_r_hist
;
362 struct obd_histogram pp_w_hist
;
365 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
366 struct ll_rw_extents_info
{
367 struct per_process_info pp_extents
[LL_PROCESS_HIST_MAX
+ 1];
370 #define LL_OFFSET_HIST_MAX 100
371 struct ll_rw_process_info
{
374 loff_t rw_range_start
;
376 loff_t rw_last_file_pos
;
378 size_t rw_smallest_extent
;
379 size_t rw_largest_extent
;
380 struct ll_file_data
*rw_last_file
;
383 enum stats_track_type
{
384 STATS_TRACK_ALL
= 0, /* track all processes */
385 STATS_TRACK_PID
, /* track process with this pid */
386 STATS_TRACK_PPID
, /* track processes with this ppid */
387 STATS_TRACK_GID
, /* track processes with this gid */
391 /* flags for sbi->ll_flags */
392 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
393 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
394 #define LL_SBI_FLOCK 0x04
395 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
396 #define LL_SBI_ACL 0x10 /* support ACL */
397 /* LL_SBI_RMT_CLIENT 0x40 remote client */
398 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa, obsolete */
399 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa, obsolete */
400 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
401 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
402 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
403 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
404 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
405 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
406 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
407 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
408 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
409 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
410 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
411 #define LL_SBI_NOROOTSQUASH 0x100000 /* do not apply root squash */
413 #define LL_SBI_FLAGS { \
438 * This is embedded into llite super-blocks to keep track of connect
439 * flags (capabilities) supported by all imports given mount is
442 struct lustre_client_ocd
{
444 * This is conjunction of connect_flags across all imports
445 * (LOVs) this mount is connected to. This field is updated by
446 * cl_ocd_update() under ->lco_lock.
449 struct mutex lco_lock
;
450 struct obd_export
*lco_md_exp
;
451 struct obd_export
*lco_dt_exp
;
455 /* this protects pglist and ra_info. It isn't safe to
456 * grab from interrupt contexts
459 spinlock_t ll_pp_extent_lock
; /* pp_extent entry*/
460 spinlock_t ll_process_lock
; /* ll_rw_process_info */
461 struct obd_uuid ll_sb_uuid
;
462 struct obd_export
*ll_md_exp
;
463 struct obd_export
*ll_dt_exp
;
464 struct dentry
*ll_debugfs_entry
;
465 struct lu_fid ll_root_fid
; /* root object fid */
468 unsigned int ll_umounting
:1,
469 ll_xattr_cache_enabled
:1;
470 struct lustre_client_ocd ll_lco
;
472 struct ll_close_queue
*ll_lcq
;
474 struct lprocfs_stats
*ll_stats
; /* lprocfs stats counter */
477 * Used to track "unstable" pages on a client, and maintain a
478 * LRU list of clean pages. An "unstable" page is defined as
479 * any page which is sent to a server as part of a bulk request,
480 * but is uncommitted to stable storage.
482 struct cl_client_cache
*ll_cache
;
484 struct lprocfs_stats
*ll_ra_stats
;
486 struct ll_ra_info ll_ra_info
;
487 unsigned int ll_namelen
;
488 struct file_operations
*ll_fop
;
490 unsigned int ll_md_brw_pages
; /* readdir pages per RPC */
492 struct lu_site
*ll_site
;
493 struct cl_device
*ll_cl
;
495 struct ll_rw_extents_info ll_rw_extents_info
;
496 int ll_extent_process_count
;
497 struct ll_rw_process_info ll_rw_process_info
[LL_PROCESS_HIST_MAX
];
498 unsigned int ll_offset_process_count
;
499 struct ll_rw_process_info ll_rw_offset_info
[LL_OFFSET_HIST_MAX
];
500 unsigned int ll_rw_offset_entry_count
;
501 int ll_stats_track_id
;
502 enum stats_track_type ll_stats_track_type
;
505 /* metadata stat-ahead */
506 unsigned int ll_sa_max
; /* max statahead RPCs */
507 atomic_t ll_sa_total
; /* statahead thread started
510 atomic_t ll_sa_wrong
; /* statahead thread stopped for
513 atomic_t ll_sa_running
; /* running statahead thread
516 atomic_t ll_agl_total
; /* AGL thread started count */
518 dev_t ll_sdev_orig
; /* save s_dev before assign for
522 struct root_squash_info ll_squash
;
524 __kernel_fsid_t ll_fsid
;
525 struct kobject ll_kobj
; /* sysfs object */
526 struct super_block
*ll_sb
; /* struct super_block (for sysfs code)*/
527 struct completion ll_kobj_unregister
;
531 * per file-descriptor read-ahead data.
533 struct ll_readahead_state
{
536 * index of the last page that read(2) needed and that wasn't in the
537 * cache. Used by ras_update() to detect seeks.
539 * XXX nikita: if access seeks into cached region, Lustre doesn't see
542 unsigned long ras_last_readpage
;
544 * number of pages read after last read-ahead window reset. As window
545 * is reset on each seek, this is effectively a number of consecutive
546 * accesses. Maybe ->ras_accessed_in_window is better name.
548 * XXX nikita: window is also reset (by ras_update()) when Lustre
549 * believes that memory pressure evicts read-ahead pages. In that
550 * case, it probably doesn't make sense to expand window to
551 * PTLRPC_MAX_BRW_PAGES on the third access.
553 unsigned long ras_consecutive_pages
;
555 * number of read requests after the last read-ahead window reset
556 * As window is reset on each seek, this is effectively the number
557 * on consecutive read request and is used to trigger read-ahead.
559 unsigned long ras_consecutive_requests
;
561 * Parameters of current read-ahead window. Handled by
562 * ras_update(). On the initial access to the file or after a seek,
563 * window is reset to 0. After 3 consecutive accesses, window is
564 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
565 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
567 unsigned long ras_window_start
, ras_window_len
;
569 * Where next read-ahead should start at. This lies within read-ahead
570 * window. Read-ahead window is read in pieces rather than at once
571 * because: 1. lustre limits total number of pages under read-ahead by
572 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
573 * not covered by DLM lock.
575 unsigned long ras_next_readahead
;
577 * Total number of ll_file_read requests issued, reads originating
578 * due to mmap are not counted in this total. This value is used to
579 * trigger full file read-ahead after multiple reads to a small file.
581 unsigned long ras_requests
;
583 * Page index with respect to the current request, these value
584 * will not be accurate when dealing with reads issued via mmap.
586 unsigned long ras_request_index
;
588 * The following 3 items are used for detecting the stride I/O
590 * In stride I/O mode,
591 * ...............|-----data-----|****gap*****|--------|******|....
592 * offset |-stride_pages-|-stride_gap-|
593 * ras_stride_offset = offset;
594 * ras_stride_length = stride_pages + stride_gap;
595 * ras_stride_pages = stride_pages;
596 * Note: all these three items are counted by pages.
598 unsigned long ras_stride_length
;
599 unsigned long ras_stride_pages
;
600 pgoff_t ras_stride_offset
;
602 * number of consecutive stride request count, and it is similar as
603 * ras_consecutive_requests, but used for stride I/O mode.
604 * Note: only more than 2 consecutive stride request are detected,
605 * stride read-ahead will be enable
607 unsigned long ras_consecutive_stride_requests
;
610 extern struct kmem_cache
*ll_file_data_slab
;
611 struct lustre_handle
;
612 struct ll_file_data
{
613 struct ll_readahead_state fd_ras
;
614 struct ll_grouplock fd_grouplock
;
618 /* openhandle if lease exists for this file.
619 * Borrow lli->lli_och_mutex to protect assignment
621 struct obd_client_handle
*fd_lease_och
;
622 struct obd_client_handle
*fd_och
;
623 struct file
*fd_file
;
624 /* Indicate whether need to report failure when close.
625 * true: failure is known, not report again.
626 * false: unknown failure, should report.
628 bool fd_write_failed
;
629 rwlock_t fd_lock
; /* protect lcc list */
630 struct list_head fd_lccs
; /* list of ll_cl_context */
633 struct lov_stripe_md
;
635 extern struct dentry
*llite_root
;
636 extern struct kset
*llite_kset
;
638 static inline struct inode
*ll_info2i(struct ll_inode_info
*lli
)
640 return &lli
->lli_vfs_inode
;
643 __u32
ll_i2suppgid(struct inode
*i
);
644 void ll_i2gids(__u32
*suppgids
, struct inode
*i1
, struct inode
*i2
);
646 static inline int ll_need_32bit_api(struct ll_sb_info
*sbi
)
648 #if BITS_PER_LONG == 32
650 #elif defined(CONFIG_COMPAT)
651 return unlikely(in_compat_syscall() || (sbi
->ll_flags
& LL_SBI_32BIT_API
));
653 return unlikely(sbi
->ll_flags
& LL_SBI_32BIT_API
);
657 void ll_ras_enter(struct file
*f
);
659 /* llite/lcommon_misc.c */
660 int cl_init_ea_size(struct obd_export
*md_exp
, struct obd_export
*dt_exp
);
661 int cl_ocd_update(struct obd_device
*host
,
662 struct obd_device
*watched
,
663 enum obd_notify_event ev
, void *owner
, void *data
);
664 int cl_get_grouplock(struct cl_object
*obj
, unsigned long gid
, int nonblock
,
665 struct ll_grouplock
*cg
);
666 void cl_put_grouplock(struct ll_grouplock
*cg
);
668 /* llite/lproc_llite.c */
669 int ldebugfs_register_mountpoint(struct dentry
*parent
,
670 struct super_block
*sb
, char *osc
, char *mdc
);
671 void ldebugfs_unregister_mountpoint(struct ll_sb_info
*sbi
);
672 void ll_stats_ops_tally(struct ll_sb_info
*sbi
, int op
, int count
);
673 void lprocfs_llite_init_vars(struct lprocfs_static_vars
*lvars
);
674 void ll_rw_stats_tally(struct ll_sb_info
*sbi
, pid_t pid
,
675 struct ll_file_data
*file
, loff_t pos
,
676 size_t count
, int rw
);
680 LPROC_LL_DIRTY_MISSES
,
682 LPROC_LL_WRITE_BYTES
,
707 LPROC_LL_ALLOC_INODE
,
710 LPROC_LL_GETXATTR_HITS
,
712 LPROC_LL_REMOVEXATTR
,
714 LPROC_LL_FILE_OPCODES
718 extern const struct file_operations ll_dir_operations
;
719 extern const struct inode_operations ll_dir_inode_operations
;
720 int ll_dir_read(struct inode
*inode
, __u64
*ppos
, struct md_op_data
*op_data
,
721 struct dir_context
*ctx
);
722 int ll_get_mdt_idx(struct inode
*inode
);
723 int ll_get_mdt_idx_by_fid(struct ll_sb_info
*sbi
, const struct lu_fid
*fid
);
724 struct page
*ll_get_dir_page(struct inode
*dir
, struct md_op_data
*op_data
,
726 void ll_release_page(struct inode
*inode
, struct page
*page
, bool remove
);
729 extern const struct inode_operations ll_special_inode_operations
;
731 struct inode
*ll_iget(struct super_block
*sb
, ino_t hash
,
732 struct lustre_md
*lic
);
733 int ll_test_inode_by_fid(struct inode
*inode
, void *opaque
);
734 int ll_md_blocking_ast(struct ldlm_lock
*, struct ldlm_lock_desc
*,
735 void *data
, int flag
);
736 struct dentry
*ll_splice_alias(struct inode
*inode
, struct dentry
*de
);
737 void ll_update_times(struct ptlrpc_request
*request
, struct inode
*inode
);
740 int ll_writepage(struct page
*page
, struct writeback_control
*wbc
);
741 int ll_writepages(struct address_space
*, struct writeback_control
*wbc
);
742 int ll_readpage(struct file
*file
, struct page
*page
);
743 void ll_readahead_init(struct inode
*inode
, struct ll_readahead_state
*ras
);
744 int ll_readahead(const struct lu_env
*env
, struct cl_io
*io
,
745 struct cl_page_list
*queue
, struct ll_readahead_state
*ras
,
747 struct ll_cl_context
*ll_cl_find(struct file
*file
);
748 void ll_cl_add(struct file
*file
, const struct lu_env
*env
, struct cl_io
*io
);
749 void ll_cl_remove(struct file
*file
, const struct lu_env
*env
);
751 extern const struct address_space_operations ll_aops
;
754 extern struct file_operations ll_file_operations
;
755 extern struct file_operations ll_file_operations_flock
;
756 extern struct file_operations ll_file_operations_noflock
;
757 extern const struct inode_operations ll_file_inode_operations
;
758 int ll_have_md_lock(struct inode
*inode
, __u64
*bits
,
759 enum ldlm_mode l_req_mode
);
760 enum ldlm_mode
ll_take_md_lock(struct inode
*inode
, __u64 bits
,
761 struct lustre_handle
*lockh
, __u64 flags
,
762 enum ldlm_mode mode
);
763 int ll_file_open(struct inode
*inode
, struct file
*file
);
764 int ll_file_release(struct inode
*inode
, struct file
*file
);
765 int ll_glimpse_ioctl(struct ll_sb_info
*sbi
,
766 struct lov_stripe_md
*lsm
, lstat_t
*st
);
767 void ll_ioepoch_open(struct ll_inode_info
*lli
, __u64 ioepoch
);
768 int ll_release_openhandle(struct inode
*, struct lookup_intent
*);
769 int ll_md_real_close(struct inode
*inode
, fmode_t fmode
);
770 void ll_ioepoch_close(struct inode
*inode
, struct md_op_data
*op_data
,
771 struct obd_client_handle
**och
, unsigned long flags
);
772 void ll_done_writing_attr(struct inode
*inode
, struct md_op_data
*op_data
);
773 int ll_som_update(struct inode
*inode
, struct md_op_data
*op_data
);
774 int ll_inode_getattr(struct inode
*inode
, struct obdo
*obdo
,
775 __u64 ioepoch
, int sync
);
776 void ll_pack_inode2opdata(struct inode
*inode
, struct md_op_data
*op_data
,
777 struct lustre_handle
*fh
);
778 int ll_getattr(struct vfsmount
*mnt
, struct dentry
*de
, struct kstat
*stat
);
779 struct posix_acl
*ll_get_acl(struct inode
*inode
, int type
);
780 int ll_migrate(struct inode
*parent
, struct file
*file
, int mdtidx
,
781 const char *name
, int namelen
);
782 int ll_get_fid_by_name(struct inode
*parent
, const char *name
,
783 int namelen
, struct lu_fid
*fid
);
784 int ll_inode_permission(struct inode
*inode
, int mask
);
786 int ll_lov_setstripe_ea_info(struct inode
*inode
, struct dentry
*dentry
,
787 __u64 flags
, struct lov_user_md
*lum
,
789 int ll_lov_getstripe_ea_info(struct inode
*inode
, const char *filename
,
790 struct lov_mds_md
**lmm
, int *lmm_size
,
791 struct ptlrpc_request
**request
);
792 int ll_dir_setstripe(struct inode
*inode
, struct lov_user_md
*lump
,
794 int ll_dir_getstripe(struct inode
*inode
, void **lmmp
, int *lmm_size
,
795 struct ptlrpc_request
**request
, u64 valid
);
796 int ll_fsync(struct file
*file
, loff_t start
, loff_t end
, int data
);
797 int ll_merge_attr(const struct lu_env
*env
, struct inode
*inode
);
798 int ll_fid2path(struct inode
*inode
, void __user
*arg
);
799 int ll_data_version(struct inode
*inode
, __u64
*data_version
, int flags
);
800 int ll_hsm_release(struct inode
*inode
);
804 int ll_d_init(struct dentry
*de
);
805 extern const struct dentry_operations ll_d_ops
;
806 void ll_intent_drop_lock(struct lookup_intent
*);
807 void ll_intent_release(struct lookup_intent
*);
808 void ll_invalidate_aliases(struct inode
*);
809 void ll_lookup_finish_locks(struct lookup_intent
*it
, struct inode
*inode
);
810 int ll_revalidate_it_finish(struct ptlrpc_request
*request
,
811 struct lookup_intent
*it
, struct inode
*inode
);
813 /* llite/llite_lib.c */
814 extern struct super_operations lustre_super_operations
;
816 void ll_lli_init(struct ll_inode_info
*lli
);
817 int ll_fill_super(struct super_block
*sb
, struct vfsmount
*mnt
);
818 void ll_put_super(struct super_block
*sb
);
819 void ll_kill_super(struct super_block
*sb
);
820 struct inode
*ll_inode_from_resource_lock(struct ldlm_lock
*lock
);
821 void ll_clear_inode(struct inode
*inode
);
822 int ll_setattr_raw(struct dentry
*dentry
, struct iattr
*attr
, bool hsm_import
);
823 int ll_setattr(struct dentry
*de
, struct iattr
*attr
);
824 int ll_statfs(struct dentry
*de
, struct kstatfs
*sfs
);
825 int ll_statfs_internal(struct super_block
*sb
, struct obd_statfs
*osfs
,
826 __u64 max_age
, __u32 flags
);
827 int ll_update_inode(struct inode
*inode
, struct lustre_md
*md
);
828 int ll_read_inode2(struct inode
*inode
, void *opaque
);
829 void ll_delete_inode(struct inode
*inode
);
830 int ll_iocontrol(struct inode
*inode
, struct file
*file
,
831 unsigned int cmd
, unsigned long arg
);
832 int ll_flush_ctx(struct inode
*inode
);
833 void ll_umount_begin(struct super_block
*sb
);
834 int ll_remount_fs(struct super_block
*sb
, int *flags
, char *data
);
835 int ll_show_options(struct seq_file
*seq
, struct dentry
*dentry
);
836 void ll_dirty_page_discard_warn(struct page
*page
, int ioret
);
837 int ll_prep_inode(struct inode
**inode
, struct ptlrpc_request
*req
,
838 struct super_block
*, struct lookup_intent
*);
839 int ll_obd_statfs(struct inode
*inode
, void __user
*arg
);
840 int ll_get_max_mdsize(struct ll_sb_info
*sbi
, int *max_mdsize
);
841 int ll_get_default_mdsize(struct ll_sb_info
*sbi
, int *default_mdsize
);
842 int ll_set_default_mdsize(struct ll_sb_info
*sbi
, int default_mdsize
);
843 int ll_process_config(struct lustre_cfg
*lcfg
);
846 LUSTRE_OPC_MKDIR
= 0,
847 LUSTRE_OPC_SYMLINK
= 1,
848 LUSTRE_OPC_MKNOD
= 2,
849 LUSTRE_OPC_CREATE
= 3,
853 struct md_op_data
*ll_prep_md_op_data(struct md_op_data
*op_data
,
854 struct inode
*i1
, struct inode
*i2
,
855 const char *name
, size_t namelen
,
856 u32 mode
, __u32 opc
, void *data
);
857 void ll_finish_md_op_data(struct md_op_data
*op_data
);
858 int ll_get_obd_name(struct inode
*inode
, unsigned int cmd
, unsigned long arg
);
859 char *ll_get_fsname(struct super_block
*sb
, char *buf
, int buflen
);
860 void ll_compute_rootsquash_state(struct ll_sb_info
*sbi
);
861 void ll_open_cleanup(struct super_block
*sb
, struct ptlrpc_request
*open_req
);
862 ssize_t
ll_copy_user_md(const struct lov_user_md __user
*md
,
863 struct lov_user_md
**kbuf
);
865 /* Compute expected user md size when passing in a md from user space */
866 static inline ssize_t
ll_lov_user_md_size(const struct lov_user_md
*lum
)
868 switch (lum
->lmm_magic
) {
869 case LOV_USER_MAGIC_V1
:
870 return sizeof(struct lov_user_md_v1
);
871 case LOV_USER_MAGIC_V3
:
872 return sizeof(struct lov_user_md_v3
);
873 case LOV_USER_MAGIC_SPECIFIC
:
874 if (lum
->lmm_stripe_count
> LOV_MAX_STRIPE_COUNT
)
877 return lov_user_md_size(lum
->lmm_stripe_count
,
878 LOV_USER_MAGIC_SPECIFIC
);
883 /* llite/llite_nfs.c */
884 extern const struct export_operations lustre_export_operations
;
885 __u32
get_uuid2int(const char *name
, int len
);
886 void get_uuid2fsid(const char *name
, int len
, __kernel_fsid_t
*fsid
);
887 struct inode
*search_inode_for_lustre(struct super_block
*sb
,
888 const struct lu_fid
*fid
);
889 int ll_dir_get_parent_fid(struct inode
*dir
, struct lu_fid
*parent_fid
);
891 /* llite/symlink.c */
892 extern const struct inode_operations ll_fast_symlink_inode_operations
;
894 /* llite/llite_close.c */
895 struct ll_close_queue
{
897 struct list_head lcq_head
;
898 wait_queue_head_t lcq_waitq
;
899 struct completion lcq_comp
;
903 void vvp_write_pending(struct vvp_object
*club
, struct vvp_page
*page
);
904 void vvp_write_complete(struct vvp_object
*club
, struct vvp_page
*page
);
907 * IO arguments for various VFS I/O interfaces.
913 struct kiocb
*via_iocb
;
914 struct iov_iter
*via_iter
;
919 struct ll_cl_context
{
920 struct list_head lcc_list
;
922 const struct lu_env
*lcc_env
;
923 struct cl_io
*lcc_io
;
924 struct cl_page
*lcc_page
;
927 struct ll_thread_info
{
928 struct vvp_io_args lti_args
;
929 struct ra_io_arg lti_ria
;
930 struct ll_cl_context lti_io_ctx
;
933 extern struct lu_context_key ll_thread_key
;
934 static inline struct ll_thread_info
*ll_env_info(const struct lu_env
*env
)
936 struct ll_thread_info
*lti
;
938 lti
= lu_context_key_get(&env
->le_ctx
, &ll_thread_key
);
943 static inline struct vvp_io_args
*ll_env_args(const struct lu_env
*env
)
945 return &ll_env_info(env
)->lti_args
;
948 void ll_queue_done_writing(struct inode
*inode
, unsigned long flags
);
949 void ll_close_thread_shutdown(struct ll_close_queue
*lcq
);
950 int ll_close_thread_start(struct ll_close_queue
**lcq_ret
);
952 /* llite/llite_mmap.c */
954 int ll_teardown_mmaps(struct address_space
*mapping
, __u64 first
, __u64 last
);
955 int ll_file_mmap(struct file
*file
, struct vm_area_struct
*vma
);
956 void policy_from_vma(ldlm_policy_data_t
*policy
, struct vm_area_struct
*vma
,
957 unsigned long addr
, size_t count
);
958 struct vm_area_struct
*our_vma(struct mm_struct
*mm
, unsigned long addr
,
961 static inline void ll_invalidate_page(struct page
*vmpage
)
963 struct address_space
*mapping
= vmpage
->mapping
;
964 loff_t offset
= vmpage
->index
<< PAGE_SHIFT
;
966 LASSERT(PageLocked(vmpage
));
971 * truncate_complete_page() calls
972 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
974 ll_teardown_mmaps(mapping
, offset
, offset
+ PAGE_SIZE
);
975 truncate_complete_page(mapping
, vmpage
);
978 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
980 /* don't need an addref as the sb_info should be holding one */
981 static inline struct obd_export
*ll_s2dtexp(struct super_block
*sb
)
983 return ll_s2sbi(sb
)->ll_dt_exp
;
986 /* don't need an addref as the sb_info should be holding one */
987 static inline struct obd_export
*ll_s2mdexp(struct super_block
*sb
)
989 return ll_s2sbi(sb
)->ll_md_exp
;
992 static inline struct client_obd
*sbi2mdc(struct ll_sb_info
*sbi
)
994 struct obd_device
*obd
= sbi
->ll_md_exp
->exp_obd
;
1001 /* FIXME: replace the name of this with LL_SB to conform to kernel stuff */
1002 static inline struct ll_sb_info
*ll_i2sbi(struct inode
*inode
)
1004 return ll_s2sbi(inode
->i_sb
);
1007 static inline struct obd_export
*ll_i2dtexp(struct inode
*inode
)
1009 return ll_s2dtexp(inode
->i_sb
);
1012 static inline struct obd_export
*ll_i2mdexp(struct inode
*inode
)
1014 return ll_s2mdexp(inode
->i_sb
);
1017 static inline struct lu_fid
*ll_inode2fid(struct inode
*inode
)
1022 fid
= &ll_i2info(inode
)->lli_fid
;
1027 static inline __u64
ll_file_maxbytes(struct inode
*inode
)
1029 return ll_i2info(inode
)->lli_maxbytes
;
1033 extern const struct xattr_handler
*ll_xattr_handlers
[];
1035 #define XATTR_USER_T 1
1036 #define XATTR_TRUSTED_T 2
1037 #define XATTR_SECURITY_T 3
1038 #define XATTR_ACL_ACCESS_T 4
1039 #define XATTR_ACL_DEFAULT_T 5
1040 #define XATTR_LUSTRE_T 6
1041 #define XATTR_OTHER_T 7
1043 ssize_t
ll_listxattr(struct dentry
*dentry
, char *buffer
, size_t size
);
1044 int ll_xattr_list(struct inode
*inode
, const char *name
, int type
,
1045 void *buffer
, size_t size
, __u64 valid
);
1048 * Common IO arguments for various VFS I/O interfaces.
1050 int cl_sb_init(struct super_block
*sb
);
1051 int cl_sb_fini(struct super_block
*sb
);
1052 void ll_io_init(struct cl_io
*io
, const struct file
*file
, int write
);
1054 void ras_update(struct ll_sb_info
*sbi
, struct inode
*inode
,
1055 struct ll_readahead_state
*ras
, unsigned long index
,
1057 void ll_ra_count_put(struct ll_sb_info
*sbi
, unsigned long len
);
1058 void ll_ra_stats_inc(struct inode
*inode
, enum ra_stat which
);
1061 #define LL_SA_RPC_MIN 2
1062 #define LL_SA_RPC_DEF 32
1063 #define LL_SA_RPC_MAX 8192
1065 #define LL_SA_CACHE_BIT 5
1066 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1067 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1069 /* per inode struct, for dir only */
1070 struct ll_statahead_info
{
1071 struct dentry
*sai_dentry
;
1072 atomic_t sai_refcount
; /* when access this struct, hold
1075 unsigned int sai_max
; /* max ahead of lookup */
1076 __u64 sai_sent
; /* stat requests sent count */
1077 __u64 sai_replied
; /* stat requests which received
1080 __u64 sai_index
; /* index of statahead entry */
1081 __u64 sai_index_wait
; /* index of entry which is the
1082 * caller is waiting for
1084 __u64 sai_hit
; /* hit count */
1085 __u64 sai_miss
; /* miss count:
1086 * for "ls -al" case, it includes
1087 * hidden dentry miss;
1088 * for "ls -l" case, it does not
1089 * include hidden dentry miss.
1090 * "sai_miss_hidden" is used for
1093 unsigned int sai_consecutive_miss
; /* consecutive miss */
1094 unsigned int sai_miss_hidden
;/* "ls -al", but first dentry
1095 * is not a hidden one
1097 unsigned int sai_skip_hidden
;/* skipped hidden dentry count */
1098 unsigned int sai_ls_all
:1, /* "ls -al", do stat-ahead for
1101 sai_agl_valid
:1,/* AGL is valid for the dir */
1102 sai_in_readpage
:1;/* statahead is in readdir() */
1103 wait_queue_head_t sai_waitq
; /* stat-ahead wait queue */
1104 struct ptlrpc_thread sai_thread
; /* stat-ahead thread */
1105 struct ptlrpc_thread sai_agl_thread
; /* AGL thread */
1106 struct list_head sai_interim_entries
; /* entries which got async
1107 * stat reply, but not
1110 struct list_head sai_entries
; /* completed entries */
1111 struct list_head sai_agls
; /* AGLs to be sent */
1112 struct list_head sai_cache
[LL_SA_CACHE_SIZE
];
1113 spinlock_t sai_cache_lock
[LL_SA_CACHE_SIZE
];
1114 atomic_t sai_cache_count
; /* entry count in cache */
1117 int ll_statahead(struct inode
*dir
, struct dentry
**dentry
, bool unplug
);
1118 void ll_authorize_statahead(struct inode
*dir
, void *key
);
1119 void ll_deauthorize_statahead(struct inode
*dir
, void *key
);
1121 blkcnt_t
dirty_cnt(struct inode
*inode
);
1123 int cl_glimpse_size0(struct inode
*inode
, int agl
);
1124 int cl_glimpse_lock(const struct lu_env
*env
, struct cl_io
*io
,
1125 struct inode
*inode
, struct cl_object
*clob
, int agl
);
1127 static inline int cl_glimpse_size(struct inode
*inode
)
1129 return cl_glimpse_size0(inode
, 0);
1132 static inline int cl_agl(struct inode
*inode
)
1134 return cl_glimpse_size0(inode
, 1);
1137 static inline int ll_glimpse_size(struct inode
*inode
)
1139 struct ll_inode_info
*lli
= ll_i2info(inode
);
1142 down_read(&lli
->lli_glimpse_sem
);
1143 rc
= cl_glimpse_size(inode
);
1144 lli
->lli_glimpse_time
= cfs_time_current();
1145 up_read(&lli
->lli_glimpse_sem
);
1150 * dentry may statahead when statahead is enabled and current process has opened
1151 * parent directory, and this dentry hasn't accessed statahead cache before
1154 dentry_may_statahead(struct inode
*dir
, struct dentry
*dentry
)
1156 struct ll_inode_info
*lli
;
1157 struct ll_dentry_data
*ldd
;
1159 if (ll_i2sbi(dir
)->ll_sa_max
== 0)
1162 lli
= ll_i2info(dir
);
1165 * statahead is not allowed for this dir, there may be three causes:
1166 * 1. dir is not opened.
1167 * 2. statahead hit ratio is too low.
1168 * 3. previous stat started statahead thread failed.
1170 if (!lli
->lli_sa_enabled
)
1173 /* not the same process, don't statahead */
1174 if (lli
->lli_opendir_pid
!= current_pid())
1178 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1179 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1180 * For patchless client, lookup intent is not accurate, which may
1181 * misguide statahead. For example:
1182 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1183 * have the same intent -- IT_GETATTR, while one dentry should access
1184 * statahead cache once, otherwise statahead windows is messed up.
1185 * The solution is as following:
1186 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1187 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1188 * bypass interacting with statahead cache by checking
1189 * 'lld_sa_generation == lli->lli_sa_generation'.
1191 ldd
= ll_d2d(dentry
);
1192 if (ldd
&& ldd
->lld_sa_generation
== lli
->lli_sa_generation
)
1198 /* llite ioctl register support routine */
1204 #define LLIOC_MAX_CMD 256
1207 * Rules to write a callback function:
1210 * @magic: Dynamic ioctl call routine will feed this value with the pointer
1211 * returned to ll_iocontrol_register. Callback functions should use this
1212 * data to check the potential collasion of ioctl cmd. If collasion is
1213 * found, callback function should return LLIOC_CONT.
1214 * @rcp: The result of ioctl command.
1217 * If @magic matches the pointer returned by ll_iocontrol_data, the
1218 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1220 typedef enum llioc_iter (*llioc_callback_t
)(struct inode
*inode
,
1221 struct file
*file
, unsigned int cmd
, unsigned long arg
,
1222 void *magic
, int *rcp
);
1224 /* export functions */
1225 /* Register ioctl block dynamatically for a regular file.
1227 * @cmd: the array of ioctl command set
1228 * @count: number of commands in the @cmd
1229 * @cb: callback function, it will be called if an ioctl command is found to
1230 * belong to the command list @cmd.
1233 * A magic pointer will be returned if success;
1234 * otherwise, NULL will be returned.
1236 void *ll_iocontrol_register(llioc_callback_t cb
, int count
, unsigned int *cmd
);
1237 void ll_iocontrol_unregister(void *magic
);
1239 int cl_sync_file_range(struct inode
*inode
, loff_t start
, loff_t end
,
1240 enum cl_fsync_mode mode
, int ignore_layout
);
1242 /** direct write pages */
1243 struct ll_dio_pages
{
1244 /** page array to be written. we don't support
1245 * partial pages except the last one.
1247 struct page
**ldp_pages
;
1248 /* offset of each page */
1249 loff_t
*ldp_offsets
;
1250 /** if ldp_offsets is NULL, it means a sequential
1251 * pages to be written, then this is the file offset
1252 * of the first page.
1254 loff_t ldp_start_offset
;
1255 /** how many bytes are to be written. */
1257 /** # of pages in the array. */
1261 static inline void cl_stats_tally(struct cl_device
*dev
, enum cl_req_type crt
,
1264 int opc
= (crt
== CRT_READ
) ? LPROC_LL_OSC_READ
:
1267 ll_stats_ops_tally(ll_s2sbi(cl2vvp_dev(dev
)->vdv_sb
), opc
, rc
);
1270 ssize_t
ll_direct_rw_pages(const struct lu_env
*env
, struct cl_io
*io
,
1271 int rw
, struct inode
*inode
,
1272 struct ll_dio_pages
*pv
);
1274 static inline int ll_file_nolock(const struct file
*file
)
1276 struct ll_file_data
*fd
= LUSTRE_FPRIVATE(file
);
1277 struct inode
*inode
= file_inode(file
);
1279 return ((fd
->fd_flags
& LL_FILE_IGNORE_LOCK
) ||
1280 (ll_i2sbi(inode
)->ll_flags
& LL_SBI_NOLCK
));
1283 static inline void ll_set_lock_data(struct obd_export
*exp
, struct inode
*inode
,
1284 struct lookup_intent
*it
, __u64
*bits
)
1286 if (!it
->it_lock_set
) {
1287 struct lustre_handle handle
;
1289 /* If this inode is a remote object, it will get two
1290 * separate locks in different namespaces, Master MDT,
1291 * where the name entry is, will grant LOOKUP lock,
1292 * remote MDT, where the object is, will grant
1293 * UPDATE|PERM lock. The inode will be attached to both
1294 * LOOKUP and PERM locks, so revoking either locks will
1295 * case the dcache being cleared
1297 if (it
->it_remote_lock_mode
) {
1298 handle
.cookie
= it
->it_remote_lock_handle
;
1299 CDEBUG(D_DLMTRACE
, "setting l_data to inode "DFID
"%p for remote lock %#llx\n",
1300 PFID(ll_inode2fid(inode
)), inode
,
1302 md_set_lock_data(exp
, &handle
, inode
, NULL
);
1305 handle
.cookie
= it
->it_lock_handle
;
1307 CDEBUG(D_DLMTRACE
, "setting l_data to inode "DFID
"%p for lock %#llx\n",
1308 PFID(ll_inode2fid(inode
)), inode
, handle
.cookie
);
1310 md_set_lock_data(exp
, &handle
, inode
, &it
->it_lock_bits
);
1311 it
->it_lock_set
= 1;
1315 *bits
= it
->it_lock_bits
;
1318 static inline int d_lustre_invalid(const struct dentry
*dentry
)
1320 struct ll_dentry_data
*lld
= ll_d2d(dentry
);
1322 return !lld
|| lld
->lld_invalid
;
1325 static inline void __d_lustre_invalidate(struct dentry
*dentry
)
1327 struct ll_dentry_data
*lld
= ll_d2d(dentry
);
1330 lld
->lld_invalid
= 1;
1334 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1335 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1336 * else dput() of the last refcount will unhash this dentry and kill it.
1338 static inline void d_lustre_invalidate(struct dentry
*dentry
, int nested
)
1340 CDEBUG(D_DENTRY
, "invalidate dentry %pd (%p) parent %p inode %p refc %d\n",
1342 dentry
->d_parent
, d_inode(dentry
), d_count(dentry
));
1344 spin_lock_nested(&dentry
->d_lock
,
1345 nested
? DENTRY_D_LOCK_NESTED
: DENTRY_D_LOCK_NORMAL
);
1346 __d_lustre_invalidate(dentry
);
1348 * We should be careful about dentries created by d_obtain_alias().
1349 * These dentries are not put in the dentry tree, instead they are
1350 * linked to sb->s_anon through dentry->d_hash.
1351 * shrink_dcache_for_umount() shrinks the tree and sb->s_anon list.
1352 * If we unhashed such a dentry, unmount would not be able to find
1353 * it and busy inodes would be reported.
1355 if (d_count(dentry
) == 0 && !(dentry
->d_flags
& DCACHE_DISCONNECTED
))
1357 spin_unlock(&dentry
->d_lock
);
1360 static inline void d_lustre_revalidate(struct dentry
*dentry
)
1362 spin_lock(&dentry
->d_lock
);
1363 LASSERT(ll_d2d(dentry
));
1364 ll_d2d(dentry
)->lld_invalid
= 0;
1365 spin_unlock(&dentry
->d_lock
);
1369 LL_LAYOUT_GEN_NONE
= ((__u32
)-2), /* layout lock was cancelled */
1370 LL_LAYOUT_GEN_EMPTY
= ((__u32
)-1) /* for empty layout */
1373 int ll_layout_conf(struct inode
*inode
, const struct cl_object_conf
*conf
);
1374 int ll_layout_refresh(struct inode
*inode
, __u32
*gen
);
1375 int ll_layout_restore(struct inode
*inode
, loff_t start
, __u64 length
);
1377 int ll_xattr_init(void);
1378 void ll_xattr_fini(void);
1380 int ll_page_sync_io(const struct lu_env
*env
, struct cl_io
*io
,
1381 struct cl_page
*page
, enum cl_req_type crt
);
1383 int ll_getparent(struct file
*file
, struct getparent __user
*arg
);
1386 int cl_setattr_ost(struct inode
*inode
, const struct iattr
*attr
);
1388 extern struct lu_env
*cl_inode_fini_env
;
1389 extern int cl_inode_fini_refcheck
;
1391 int cl_file_inode_init(struct inode
*inode
, struct lustre_md
*md
);
1392 void cl_inode_fini(struct inode
*inode
);
1393 int cl_local_size(struct inode
*inode
);
1395 __u64
cl_fid_build_ino(const struct lu_fid
*fid
, int api32
);
1396 __u32
cl_fid_build_gen(const struct lu_fid
*fid
);
1398 #endif /* LLITE_INTERNAL_H */