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, 2017, 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_disk.h> /* for s2sbi */
37 #include <lustre_eacl.h>
38 #include <lustre_linkea.h>
40 /* for struct cl_lock_descr and struct cl_io */
41 #include <cl_object.h>
42 #include <lustre_lmv.h>
43 #include <lustre_mdc.h>
44 #include <lustre_intent.h>
45 #include <linux/compat.h>
46 #include <linux/aio.h>
47 #include <lustre_compat.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 /* rw lock protects lli_lsm_md */
169 struct rw_semaphore lli_lsm_sem;
170 /* directory stripe information */
171 struct lmv_stripe_md *lli_lsm_md;
172 /* default directory stripe offset. This is extracted
173 * from the "dmv" xattr in order to decide which MDT to
174 * create a subdirectory on. The MDS itself fetches
175 * "dmv" and gets the rest of the default layout itself
176 * (count, hash, etc). */
177 __u32 lli_def_stripe_offset;
180 /* for non-directory */
182 struct mutex lli_size_mutex;
183 char *lli_symlink_name;
185 * struct rw_semaphore {
186 * signed long count; // align d.d_def_acl
187 * spinlock_t wait_lock; // align d.d_sa_lock
188 * struct list_head wait_list;
191 struct rw_semaphore lli_trunc_sem;
192 struct range_lock_tree lli_write_tree;
194 struct rw_semaphore lli_glimpse_sem;
195 ktime_t lli_glimpse_time;
196 struct list_head lli_agl_list;
199 /* for writepage() only to communicate to fsync */
203 * Whenever a process try to read/write the file, the
204 * jobid of the process will be saved here, and it'll
205 * be packed into the write PRC when flush later.
207 * So the read/write statistics for jobid will not be
208 * accurate if the file is shared by different jobs.
210 char lli_jobid[LUSTRE_JOBID_SIZE];
214 /* XXX: For following frequent used members, although they maybe special
215 * used for non-directory object, it is some time-wasting to check
216 * whether the object is directory or not before using them. On the
217 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
218 * the "ll_inode_info" size even if moving those members into u.f.
219 * So keep them out side.
221 * In the future, if more members are added only for directory,
222 * some of the following members can be moved into u.f.
224 struct cl_object *lli_clob;
226 /* mutex to request for layout lock exclusively. */
227 struct mutex lli_layout_mutex;
228 /* Layout version, protected by lli_layout_lock */
229 __u32 lli_layout_gen;
230 spinlock_t lli_layout_lock;
232 __u32 lli_projid; /* project id */
234 struct rw_semaphore lli_xattrs_list_rwsem;
235 struct mutex lli_xattrs_enq_lock;
236 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
239 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
243 spin_lock(&lli->lli_layout_lock);
244 gen = lli->lli_layout_gen;
245 spin_unlock(&lli->lli_layout_lock);
250 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
252 spin_lock(&lli->lli_layout_lock);
253 lli->lli_layout_gen = gen;
254 spin_unlock(&lli->lli_layout_lock);
258 /* File data is modified. */
259 LLIF_DATA_MODIFIED = 0,
260 /* File is being restored */
261 LLIF_FILE_RESTORING = 1,
262 /* Xattr cache is attached to the file */
263 LLIF_XATTR_CACHE = 2,
264 /* Project inherit */
265 LLIF_PROJECT_INHERIT = 3,
268 static inline void ll_file_set_flag(struct ll_inode_info *lli,
269 enum ll_file_flags flag)
271 set_bit(flag, &lli->lli_flags);
274 static inline void ll_file_clear_flag(struct ll_inode_info *lli,
275 enum ll_file_flags flag)
277 clear_bit(flag, &lli->lli_flags);
280 static inline bool ll_file_test_flag(struct ll_inode_info *lli,
281 enum ll_file_flags flag)
283 return test_bit(flag, &lli->lli_flags);
286 static inline bool ll_file_test_and_clear_flag(struct ll_inode_info *lli,
287 enum ll_file_flags flag)
289 return test_and_clear_bit(flag, &lli->lli_flags);
292 int ll_xattr_cache_destroy(struct inode *inode);
294 int ll_xattr_cache_get(struct inode *inode,
300 static inline bool obd_connect_has_secctx(struct obd_connect_data *data)
302 #if defined(HAVE_SECURITY_DENTRY_INIT_SECURITY) && defined(CONFIG_SECURITY)
303 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
304 data->ocd_connect_flags2 & OBD_CONNECT2_FILE_SECCTX;
307 #endif /* HAVE_SECURITY_DENTRY_INIT_SECURITY */
310 static inline void obd_connect_set_secctx(struct obd_connect_data *data)
312 #if defined(HAVE_SECURITY_DENTRY_INIT_SECURITY) && defined(CONFIG_SECURITY)
313 data->ocd_connect_flags2 |= OBD_CONNECT2_FILE_SECCTX;
314 #endif /* HAVE_SECURITY_DENTRY_INIT_SECURITY */
317 int ll_dentry_init_security(struct dentry *dentry, int mode, struct qstr *name,
318 const char **secctx_name, void **secctx,
320 int ll_inode_init_security(struct dentry *dentry, struct inode *inode,
323 int ll_listsecurity(struct inode *inode, char *secctx_name,
324 size_t secctx_name_size);
327 * Locking to guarantee consistency of non-atomic updates to long long i_size,
328 * consistency between file size and KMS.
330 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
333 void ll_inode_size_lock(struct inode *inode);
334 void ll_inode_size_unlock(struct inode *inode);
336 // FIXME: replace the name of this with LL_I to conform to kernel stuff
337 // static inline struct ll_inode_info *LL_I(struct inode *inode)
338 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
340 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
343 /* default to about 64M of readahead on a given system. */
344 #define SBI_DEFAULT_READAHEAD_MAX (64UL << (20 - PAGE_SHIFT))
346 /* default to read-ahead full files smaller than 2MB on the second read */
347 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_SHIFT))
352 RA_STAT_DISTANT_READPAGE,
353 RA_STAT_MISS_IN_WINDOW,
354 RA_STAT_FAILED_GRAB_PAGE,
355 RA_STAT_FAILED_MATCH,
360 RA_STAT_MAX_IN_FLIGHT,
361 RA_STAT_WRONG_GRAB_PAGE,
362 RA_STAT_FAILED_REACH_END,
367 atomic_t ra_cur_pages;
368 unsigned long ra_max_pages;
369 unsigned long ra_max_pages_per_file;
370 unsigned long ra_max_read_ahead_whole_pages;
373 /* ra_io_arg will be filled in the beginning of ll_readahead with
374 * ras_lock, then the following ll_read_ahead_pages will read RA
375 * pages according to this arg, all the items in this structure are
376 * counted by page index.
379 unsigned long ria_start; /* start offset of read-ahead*/
380 unsigned long ria_end; /* end offset of read-ahead*/
381 unsigned long ria_reserved; /* reserved pages for read-ahead */
382 unsigned long ria_end_min; /* minimum end to cover current read */
383 bool ria_eof; /* reach end of file */
384 /* If stride read pattern is detected, ria_stoff means where
385 * stride read is started. Note: for normal read-ahead, the
386 * value here is meaningless, and also it will not be accessed*/
388 /* ria_length and ria_pages are the length and pages length in the
389 * stride I/O mode. And they will also be used to check whether
390 * it is stride I/O read-ahead in the read-ahead pages*/
391 unsigned long ria_length;
392 unsigned long ria_pages;
395 /* LL_HIST_MAX=32 causes an overflow */
396 #define LL_HIST_MAX 28
397 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
398 #define LL_PROCESS_HIST_MAX 10
399 struct per_process_info {
401 struct obd_histogram pp_r_hist;
402 struct obd_histogram pp_w_hist;
405 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
406 struct ll_rw_extents_info {
407 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
410 #define LL_OFFSET_HIST_MAX 100
411 struct ll_rw_process_info {
414 loff_t rw_range_start;
416 loff_t rw_last_file_pos;
418 size_t rw_smallest_extent;
419 size_t rw_largest_extent;
420 struct ll_file_data *rw_last_file;
423 enum stats_track_type {
424 STATS_TRACK_ALL = 0, /* track all processes */
425 STATS_TRACK_PID, /* track process with this pid */
426 STATS_TRACK_PPID, /* track processes with this ppid */
427 STATS_TRACK_GID, /* track processes with this gid */
431 /* flags for sbi->ll_flags */
432 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
433 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
434 #define LL_SBI_FLOCK 0x04
435 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
436 #define LL_SBI_ACL 0x10 /* support ACL */
437 /* LL_SBI_RMT_CLIENT 0x40 remote client */
438 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa, obsolete */
439 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa, obsolete */
440 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
441 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
442 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
443 /* LL_SBI_SOM_PREVIEW 0x1000 SOM preview mount option, obsolete */
444 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
445 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
446 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
447 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
448 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
449 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
450 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
451 #define LL_SBI_NOROOTSQUASH 0x100000 /* do not apply root squash */
452 #define LL_SBI_ALWAYS_PING 0x200000 /* always ping even if server
454 #define LL_SBI_FAST_READ 0x400000 /* fast read support */
455 #define LL_SBI_FILE_SECCTX 0x800000 /* set file security context at create */
456 /* LL_SBI_PIO 0x1000000 parallel IO support, introduced in
458 #define LL_SBI_TINY_WRITE 0x2000000 /* tiny write support */
460 #define LL_SBI_FLAGS { \
489 /* This is embedded into llite super-blocks to keep track of connect
490 * flags (capabilities) supported by all imports given mount is
492 struct lustre_client_ocd {
493 /* This is conjunction of connect_flags across all imports
494 * (LOVs) this mount is connected to. This field is updated by
495 * cl_ocd_update() under ->lco_lock. */
497 struct mutex lco_lock;
498 struct obd_export *lco_md_exp;
499 struct obd_export *lco_dt_exp;
503 /* this protects pglist and ra_info. It isn't safe to
504 * grab from interrupt contexts */
506 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
507 spinlock_t ll_process_lock; /* ll_rw_process_info */
508 struct obd_uuid ll_sb_uuid;
509 struct obd_export *ll_md_exp;
510 struct obd_export *ll_dt_exp;
511 struct obd_device *ll_md_obd;
512 struct obd_device *ll_dt_obd;
513 struct dentry *ll_debugfs_entry;
514 struct lu_fid ll_root_fid; /* root object fid */
517 unsigned int ll_umounting:1,
518 ll_xattr_cache_enabled:1,
519 ll_xattr_cache_set:1, /* already set to 0/1 */
520 ll_client_common_fill_super_succeeded:1,
523 struct lustre_client_ocd ll_lco;
525 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
527 /* Used to track "unstable" pages on a client, and maintain a
528 * LRU list of clean pages. An "unstable" page is defined as
529 * any page which is sent to a server as part of a bulk request,
530 * but is uncommitted to stable storage. */
531 struct cl_client_cache *ll_cache;
533 struct lprocfs_stats *ll_ra_stats;
535 struct ll_ra_info ll_ra_info;
536 unsigned int ll_namelen;
537 struct file_operations *ll_fop;
539 struct lu_site *ll_site;
540 struct cl_device *ll_cl;
542 struct ll_rw_extents_info ll_rw_extents_info;
543 int ll_extent_process_count;
544 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
545 unsigned int ll_offset_process_count;
546 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
547 unsigned int ll_rw_offset_entry_count;
548 int ll_stats_track_id;
549 enum stats_track_type ll_stats_track_type;
552 /* metadata stat-ahead */
553 unsigned int ll_sa_running_max;/* max concurrent
554 * statahead instances */
555 unsigned int ll_sa_max; /* max statahead RPCs */
556 atomic_t ll_sa_total; /* statahead thread started
558 atomic_t ll_sa_wrong; /* statahead thread stopped for
560 atomic_t ll_sa_running; /* running statahead thread
562 atomic_t ll_agl_total; /* AGL thread started count */
564 dev_t ll_sdev_orig; /* save s_dev before assign for
567 struct root_squash_info ll_squash;
570 /* st_blksize returned by stat(2), when non-zero */
571 unsigned int ll_stat_blksize;
573 struct kset ll_kset; /* sysfs object */
574 struct completion ll_kobj_unregister;
578 * per file-descriptor read-ahead data.
580 struct ll_readahead_state {
583 * index of the last page that read(2) needed and that wasn't in the
584 * cache. Used by ras_update() to detect seeks.
586 * XXX nikita: if access seeks into cached region, Lustre doesn't see
589 unsigned long ras_last_readpage;
591 * number of pages read after last read-ahead window reset. As window
592 * is reset on each seek, this is effectively a number of consecutive
593 * accesses. Maybe ->ras_accessed_in_window is better name.
595 * XXX nikita: window is also reset (by ras_update()) when Lustre
596 * believes that memory pressure evicts read-ahead pages. In that
597 * case, it probably doesn't make sense to expand window to
598 * PTLRPC_MAX_BRW_PAGES on the third access.
600 unsigned long ras_consecutive_pages;
602 * number of read requests after the last read-ahead window reset
603 * As window is reset on each seek, this is effectively the number
604 * on consecutive read request and is used to trigger read-ahead.
606 unsigned long ras_consecutive_requests;
608 * Parameters of current read-ahead window. Handled by
609 * ras_update(). On the initial access to the file or after a seek,
610 * window is reset to 0. After 3 consecutive accesses, window is
611 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
612 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
614 unsigned long ras_window_start, ras_window_len;
616 * Optimal RPC size. It decides how many pages will be sent
617 * for each read-ahead.
619 unsigned long ras_rpc_size;
621 * Where next read-ahead should start at. This lies within read-ahead
622 * window. Read-ahead window is read in pieces rather than at once
623 * because: 1. lustre limits total number of pages under read-ahead by
624 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
625 * not covered by DLM lock.
627 unsigned long ras_next_readahead;
629 * Total number of ll_file_read requests issued, reads originating
630 * due to mmap are not counted in this total. This value is used to
631 * trigger full file read-ahead after multiple reads to a small file.
633 unsigned long ras_requests;
635 * Page index with respect to the current request, these value
636 * will not be accurate when dealing with reads issued via mmap.
638 unsigned long ras_request_index;
640 * The following 3 items are used for detecting the stride I/O
642 * In stride I/O mode,
643 * ...............|-----data-----|****gap*****|--------|******|....
644 * offset |-stride_pages-|-stride_gap-|
645 * ras_stride_offset = offset;
646 * ras_stride_length = stride_pages + stride_gap;
647 * ras_stride_pages = stride_pages;
648 * Note: all these three items are counted by pages.
650 unsigned long ras_stride_length;
651 unsigned long ras_stride_pages;
652 pgoff_t ras_stride_offset;
654 * number of consecutive stride request count, and it is similar as
655 * ras_consecutive_requests, but used for stride I/O mode.
656 * Note: only more than 2 consecutive stride request are detected,
657 * stride read-ahead will be enable
659 unsigned long ras_consecutive_stride_requests;
662 extern struct kmem_cache *ll_file_data_slab;
663 struct lustre_handle;
664 struct ll_file_data {
665 struct ll_readahead_state fd_ras;
666 struct ll_grouplock fd_grouplock;
670 /* openhandle if lease exists for this file.
671 * Borrow lli->lli_och_mutex to protect assignment */
672 struct obd_client_handle *fd_lease_och;
673 struct obd_client_handle *fd_och;
674 struct file *fd_file;
675 /* Indicate whether need to report failure when close.
676 * true: failure is known, not report again.
677 * false: unknown failure, should report. */
678 bool fd_write_failed;
679 bool ll_lock_no_expand;
680 rwlock_t fd_lock; /* protect lcc list */
681 struct list_head fd_lccs; /* list of ll_cl_context */
682 /* Used by mirrored file to lead IOs to a specific mirror, usually
683 * for mirror resync. 0 means default. */
684 __u32 fd_designated_mirror;
685 /* The layout version when resync starts. Resync I/O should carry this
686 * layout version for verification to OST objects */
687 __u32 fd_layout_version;
690 void llite_tunables_unregister(void);
691 int llite_tunables_register(void);
693 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
695 return &lli->lli_vfs_inode;
698 __u32 ll_i2suppgid(struct inode *i);
699 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
701 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
703 #if BITS_PER_LONG == 32
705 #elif defined(CONFIG_COMPAT)
706 if (unlikely(sbi->ll_flags & LL_SBI_32BIT_API))
709 # ifdef CONFIG_X86_X32
710 /* in_compat_syscall() returns true when called from a kthread
711 * and CONFIG_X86_X32 is enabled, which is wrong. So check
712 * whether the caller comes from a syscall (ie. not a kthread)
713 * before calling in_compat_syscall(). */
714 if (current->flags & PF_KTHREAD)
718 return unlikely(in_compat_syscall());
720 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
724 static inline bool ll_sbi_has_fast_read(struct ll_sb_info *sbi)
726 return !!(sbi->ll_flags & LL_SBI_FAST_READ);
729 static inline bool ll_sbi_has_tiny_write(struct ll_sb_info *sbi)
731 return !!(sbi->ll_flags & LL_SBI_TINY_WRITE);
734 void ll_ras_enter(struct file *f);
736 /* llite/lcommon_misc.c */
737 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
738 enum obd_notify_event ev, void *owner);
739 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
740 struct ll_grouplock *lg);
741 void cl_put_grouplock(struct ll_grouplock *lg);
743 /* llite/lproc_llite.c */
744 int ll_debugfs_register_super(struct super_block *sb, const char *name);
745 void ll_debugfs_unregister_super(struct super_block *sb);
746 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
750 LPROC_LL_DIRTY_MISSES,
752 LPROC_LL_WRITE_BYTES,
777 LPROC_LL_ALLOC_INODE,
780 LPROC_LL_GETXATTR_HITS,
782 LPROC_LL_REMOVEXATTR,
784 LPROC_LL_FILE_OPCODES
788 struct ll_dir_chain {
791 static inline void ll_dir_chain_init(struct ll_dir_chain *chain)
795 static inline void ll_dir_chain_fini(struct ll_dir_chain *chain)
799 extern const struct file_operations ll_dir_operations;
800 extern const struct inode_operations ll_dir_inode_operations;
801 #ifdef HAVE_DIR_CONTEXT
802 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
803 struct dir_context *ctx);
805 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
806 void *cookie, filldir_t filldir);
808 int ll_get_mdt_idx(struct inode *inode);
809 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
810 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
811 __u64 offset, struct ll_dir_chain *chain);
812 void ll_release_page(struct inode *inode, struct page *page, bool remove);
815 extern const struct inode_operations ll_special_inode_operations;
817 struct inode *ll_iget(struct super_block *sb, ino_t hash,
818 struct lustre_md *lic);
819 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
820 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
821 void *data, int flag);
822 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
823 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
824 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
827 int ll_writepage(struct page *page, struct writeback_control *wbc);
828 int ll_writepages(struct address_space *, struct writeback_control *wbc);
829 int ll_readpage(struct file *file, struct page *page);
830 int ll_io_read_page(const struct lu_env *env, struct cl_io *io,
831 struct cl_page *page, struct file *file);
832 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
833 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
836 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io,
838 void ll_cl_remove(struct file *file, const struct lu_env *env);
839 struct ll_cl_context *ll_cl_find(struct file *file);
841 extern const struct address_space_operations ll_aops;
844 extern struct file_operations ll_file_operations;
845 extern struct file_operations ll_file_operations_flock;
846 extern struct file_operations ll_file_operations_noflock;
847 extern struct inode_operations ll_file_inode_operations;
848 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
849 enum ldlm_mode l_req_mode);
850 extern enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
851 struct lustre_handle *lockh, __u64 flags,
852 enum ldlm_mode mode);
854 int ll_file_open(struct inode *inode, struct file *file);
855 int ll_file_release(struct inode *inode, struct file *file);
856 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
857 int ll_md_real_close(struct inode *inode, fmode_t fmode);
858 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
859 struct ll_file_data *file, loff_t pos,
860 size_t count, int rw);
861 #ifdef HAVE_INODEOPS_ENHANCED_GETATTR
862 int ll_getattr(const struct path *path, struct kstat *stat,
863 u32 request_mask, unsigned int flags);
865 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
867 struct posix_acl *ll_get_acl(struct inode *inode, int type);
868 #ifdef HAVE_IOP_SET_ACL
869 #ifdef CONFIG_FS_POSIX_ACL
870 int ll_set_acl(struct inode *inode, struct posix_acl *acl, int type);
871 #else /* !CONFIG_FS_POSIX_ACL */
872 #define ll_set_acl NULL
873 #endif /* CONFIG_FS_POSIX_ACL */
876 int ll_migrate(struct inode *parent, struct file *file,
877 struct lmv_user_md *lum, const char *name);
878 int ll_get_fid_by_name(struct inode *parent, const char *name,
879 int namelen, struct lu_fid *fid, struct inode **inode);
880 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
881 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
883 # ifndef HAVE_INODE_PERMISION_2ARGS
884 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
886 int ll_inode_permission(struct inode *inode, int mask);
889 int ll_ioctl_check_project(struct inode *inode, struct fsxattr *fa);
890 int ll_ioctl_fsgetxattr(struct inode *inode, unsigned int cmd,
892 int ll_ioctl_fssetxattr(struct inode *inode, unsigned int cmd,
895 int ll_lov_setstripe_ea_info(struct inode *inode, struct dentry *dentry,
896 __u64 flags, struct lov_user_md *lum,
898 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
899 struct lov_mds_md **lmm, int *lmm_size,
900 struct ptlrpc_request **request);
901 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
903 int ll_dir_getstripe(struct inode *inode, void **lmmp,
904 int *lmm_size, struct ptlrpc_request **request,
906 #ifdef HAVE_FILE_FSYNC_4ARGS
907 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
908 #elif defined(HAVE_FILE_FSYNC_2ARGS)
909 int ll_fsync(struct file *file, int data);
911 int ll_fsync(struct file *file, struct dentry *dentry, int data);
913 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
914 int ll_fid2path(struct inode *inode, void __user *arg);
915 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
916 int ll_hsm_release(struct inode *inode);
917 int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
918 void ll_io_set_mirror(struct cl_io *io, const struct file *file);
922 int ll_d_init(struct dentry *de);
923 extern const struct dentry_operations ll_d_ops;
924 void ll_intent_drop_lock(struct lookup_intent *);
925 void ll_intent_release(struct lookup_intent *);
926 void ll_invalidate_aliases(struct inode *);
927 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
928 int ll_revalidate_it_finish(struct ptlrpc_request *request,
929 struct lookup_intent *it, struct dentry *de);
931 /* llite/llite_lib.c */
932 extern struct super_operations lustre_super_operations;
934 void ll_lli_init(struct ll_inode_info *lli);
935 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
936 void ll_put_super(struct super_block *sb);
937 void ll_kill_super(struct super_block *sb);
938 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
939 void ll_dir_clear_lsm_md(struct inode *inode);
940 void ll_clear_inode(struct inode *inode);
941 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr,
942 enum op_xvalid xvalid, bool hsm_import);
943 int ll_setattr(struct dentry *de, struct iattr *attr);
944 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
945 int ll_statfs_internal(struct ll_sb_info *sbi, struct obd_statfs *osfs,
947 int ll_update_inode(struct inode *inode, struct lustre_md *md);
948 void ll_update_inode_flags(struct inode *inode, int ext_flags);
949 int ll_read_inode2(struct inode *inode, void *opaque);
950 void ll_delete_inode(struct inode *inode);
951 int ll_iocontrol(struct inode *inode, struct file *file,
952 unsigned int cmd, unsigned long arg);
953 int ll_flush_ctx(struct inode *inode);
954 void ll_umount_begin(struct super_block *sb);
955 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
956 #ifdef HAVE_SUPEROPS_USE_DENTRY
957 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
959 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
961 void ll_dirty_page_discard_warn(struct page *page, int ioret);
962 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
963 struct super_block *, struct lookup_intent *);
964 int ll_obd_statfs(struct inode *inode, void __user *arg);
965 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
966 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
967 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
970 LUSTRE_OPC_MKDIR = 0,
971 LUSTRE_OPC_SYMLINK = 1,
972 LUSTRE_OPC_MKNOD = 2,
973 LUSTRE_OPC_CREATE = 3,
977 void ll_unlock_md_op_lsm(struct md_op_data *op_data);
978 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
979 struct inode *i1, struct inode *i2,
980 const char *name, size_t namelen,
981 __u32 mode, __u32 opc, void *data);
982 void ll_finish_md_op_data(struct md_op_data *op_data);
983 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
984 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
985 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
986 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
987 struct lov_user_md **kbuf);
988 void ll_open_cleanup(struct super_block *sb, struct ptlrpc_request *open_req);
990 void ll_dom_finish_open(struct inode *inode, struct ptlrpc_request *req,
991 struct lookup_intent *it);
993 /* Compute expected user md size when passing in a md from user space */
994 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
996 switch (lum->lmm_magic) {
997 case LOV_USER_MAGIC_V1:
998 return sizeof(struct lov_user_md_v1);
999 case LOV_USER_MAGIC_V3:
1000 return sizeof(struct lov_user_md_v3);
1001 case LOV_USER_MAGIC_SPECIFIC:
1002 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
1005 return lov_user_md_size(lum->lmm_stripe_count,
1006 LOV_USER_MAGIC_SPECIFIC);
1007 case LOV_USER_MAGIC_COMP_V1:
1008 return ((struct lov_comp_md_v1 *)lum)->lcm_size;
1014 /* llite/llite_nfs.c */
1015 extern struct export_operations lustre_export_operations;
1016 __u32 get_uuid2int(const char *name, int len);
1017 struct inode *search_inode_for_lustre(struct super_block *sb,
1018 const struct lu_fid *fid);
1019 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
1021 /* llite/symlink.c */
1022 extern struct inode_operations ll_fast_symlink_inode_operations;
1025 * IO arguments for various VFS I/O interfaces.
1027 struct vvp_io_args {
1028 /** normal/sendfile/splice */
1029 enum vvp_io_subtype via_io_subtype;
1033 struct kiocb *via_iocb;
1034 struct iov_iter *via_iter;
1037 struct pipe_inode_info *via_pipe;
1038 unsigned int via_flags;
1048 struct ll_cl_context {
1049 struct list_head lcc_list;
1051 const struct lu_env *lcc_env;
1052 struct cl_io *lcc_io;
1053 struct cl_page *lcc_page;
1054 enum lcc_type lcc_type;
1057 struct ll_thread_info {
1058 struct iov_iter lti_iter;
1059 struct vvp_io_args lti_args;
1060 struct ra_io_arg lti_ria;
1061 struct ll_cl_context lti_io_ctx;
1064 extern struct lu_context_key ll_thread_key;
1066 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
1068 struct ll_thread_info *lti;
1070 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
1071 LASSERT(lti != NULL);
1076 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env,
1077 enum vvp_io_subtype type)
1079 struct vvp_io_args *via = &ll_env_info(env)->lti_args;
1081 via->via_io_subtype = type;
1086 /* llite/llite_mmap.c */
1088 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1089 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1090 void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
1091 unsigned long addr, size_t count);
1092 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1095 static inline void ll_invalidate_page(struct page *vmpage)
1097 struct address_space *mapping = vmpage->mapping;
1098 loff_t offset = vmpage->index << PAGE_SHIFT;
1100 LASSERT(PageLocked(vmpage));
1101 if (mapping == NULL)
1105 * truncate_complete_page() calls
1106 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
1108 ll_teardown_mmaps(mapping, offset, offset + PAGE_SIZE);
1109 truncate_complete_page(mapping, vmpage);
1112 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1114 /* don't need an addref as the sb_info should be holding one */
1115 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1117 return ll_s2sbi(sb)->ll_dt_exp;
1120 /* don't need an addref as the sb_info should be holding one */
1121 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1123 return ll_s2sbi(sb)->ll_md_exp;
1126 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1128 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1134 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1135 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1137 return ll_s2sbi(inode->i_sb);
1140 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1142 return ll_s2dtexp(inode->i_sb);
1145 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1147 return ll_s2mdexp(inode->i_sb);
1150 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1154 LASSERT(inode != NULL);
1155 fid = &ll_i2info(inode)->lli_fid;
1160 static inline loff_t ll_file_maxbytes(struct inode *inode)
1162 struct cl_object *obj = ll_i2info(inode)->lli_clob;
1165 return MAX_LFS_FILESIZE;
1167 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
1171 extern const struct xattr_handler *ll_xattr_handlers[];
1173 #define XATTR_USER_T 1
1174 #define XATTR_TRUSTED_T 2
1175 #define XATTR_SECURITY_T 3
1176 #define XATTR_ACL_ACCESS_T 4
1177 #define XATTR_ACL_DEFAULT_T 5
1178 #define XATTR_LUSTRE_T 6
1179 #define XATTR_OTHER_T 7
1181 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1182 int ll_xattr_list(struct inode *inode, const char *name, int type,
1183 void *buffer, size_t size, u64 valid);
1184 const struct xattr_handler *get_xattr_type(const char *name);
1187 * Common IO arguments for various VFS I/O interfaces.
1189 int cl_sb_init(struct super_block *sb);
1190 int cl_sb_fini(struct super_block *sb);
1192 enum ras_update_flags {
1196 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1197 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1201 #define LL_SA_RPC_MIN 2
1202 #define LL_SA_RPC_DEF 32
1203 #define LL_SA_RPC_MAX 512
1205 /* XXX: If want to support more concurrent statahead instances,
1206 * please consider to decentralize the RPC lists attached
1207 * on related import, such as imp_{sending,delayed}_list.
1209 #define LL_SA_RUNNING_MAX 256
1210 #define LL_SA_RUNNING_DEF 16
1212 #define LL_SA_CACHE_BIT 5
1213 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1214 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1216 /* per inode struct, for dir only */
1217 struct ll_statahead_info {
1218 struct dentry *sai_dentry;
1219 atomic_t sai_refcount; /* when access this struct, hold
1221 unsigned int sai_max; /* max ahead of lookup */
1222 __u64 sai_sent; /* stat requests sent count */
1223 __u64 sai_replied; /* stat requests which received
1225 __u64 sai_index; /* index of statahead entry */
1226 __u64 sai_index_wait; /* index of entry which is the
1227 * caller is waiting for */
1228 __u64 sai_hit; /* hit count */
1229 __u64 sai_miss; /* miss count:
1230 * for "ls -al" case, includes
1231 * hidden dentry miss;
1232 * for "ls -l" case, it does not
1233 * include hidden dentry miss.
1234 * "sai_miss_hidden" is used for
1237 unsigned int sai_consecutive_miss; /* consecutive miss */
1238 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1239 * is not a hidden one */
1240 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1242 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1244 sai_agl_valid:1,/* AGL is valid for the dir */
1245 sai_in_readpage:1;/* statahead is in readdir()*/
1246 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1247 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1248 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1249 struct list_head sai_interim_entries; /* entries which got async
1250 * stat reply, but not
1252 struct list_head sai_entries; /* completed entries */
1253 struct list_head sai_agls; /* AGLs to be sent */
1254 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1255 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1256 atomic_t sai_cache_count; /* entry count in cache */
1259 int ll_statahead(struct inode *dir, struct dentry **dentry, bool unplug);
1260 void ll_authorize_statahead(struct inode *dir, void *key);
1261 void ll_deauthorize_statahead(struct inode *dir, void *key);
1264 blkcnt_t dirty_cnt(struct inode *inode);
1266 int cl_glimpse_size0(struct inode *inode, int agl);
1267 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1268 struct inode *inode, struct cl_object *clob, int agl);
1270 static inline int cl_glimpse_size(struct inode *inode)
1272 return cl_glimpse_size0(inode, 0);
1275 /* AGL is 'asychronous glimpse lock', which is a speculative lock taken as
1276 * part of statahead */
1277 static inline int cl_agl(struct inode *inode)
1279 return cl_glimpse_size0(inode, 1);
1282 int ll_file_lock_ahead(struct file *file, struct llapi_lu_ladvise *ladvise);
1284 int cl_io_get(struct inode *inode, struct lu_env **envout,
1285 struct cl_io **ioout, __u16 *refcheck);
1287 static inline int ll_glimpse_size(struct inode *inode)
1289 struct ll_inode_info *lli = ll_i2info(inode);
1292 down_read(&lli->lli_glimpse_sem);
1293 rc = cl_glimpse_size(inode);
1294 lli->lli_glimpse_time = ktime_get();
1295 up_read(&lli->lli_glimpse_sem);
1299 /* dentry may statahead when statahead is enabled and current process has opened
1300 * parent directory, and this dentry hasn't accessed statahead cache before */
1302 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1304 struct ll_inode_info *lli;
1305 struct ll_dentry_data *ldd;
1307 if (ll_i2sbi(dir)->ll_sa_max == 0)
1310 lli = ll_i2info(dir);
1312 /* statahead is not allowed for this dir, there may be three causes:
1313 * 1. dir is not opened.
1314 * 2. statahead hit ratio is too low.
1315 * 3. previous stat started statahead thread failed. */
1316 if (!lli->lli_sa_enabled)
1319 /* not the same process, don't statahead */
1320 if (lli->lli_opendir_pid != current_pid())
1324 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1325 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1326 * For patchless client, lookup intent is not accurate, which may
1327 * misguide statahead. For example:
1328 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1329 * have the same intent -- IT_GETATTR, while one dentry should access
1330 * statahead cache once, otherwise statahead windows is messed up.
1331 * The solution is as following:
1332 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1333 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1334 * bypass interacting with statahead cache by checking
1335 * 'lld_sa_generation == lli->lli_sa_generation'.
1337 ldd = ll_d2d(dentry);
1338 if (ldd != NULL && ldd->lld_sa_generation == lli->lli_sa_generation)
1344 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1345 enum cl_fsync_mode mode, int ignore_layout);
1347 static inline int ll_file_nolock(const struct file *file)
1349 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1350 struct inode *inode = file_inode((struct file *)file);
1352 LASSERT(fd != NULL);
1353 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1354 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1357 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1358 struct lookup_intent *it, __u64 *bits)
1360 if (!it->it_lock_set) {
1361 struct lustre_handle handle;
1363 /* If this inode is a remote object, it will get two
1364 * separate locks in different namespaces, Master MDT,
1365 * where the name entry is, will grant LOOKUP lock,
1366 * remote MDT, where the object is, will grant
1367 * UPDATE|PERM lock. The inode will be attched to both
1368 * LOOKUP and PERM locks, so revoking either locks will
1369 * case the dcache being cleared */
1370 if (it->it_remote_lock_mode) {
1371 handle.cookie = it->it_remote_lock_handle;
1372 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1373 "(%p) for remote lock %#llx\n",
1374 PFID(ll_inode2fid(inode)), inode,
1376 md_set_lock_data(exp, &handle, inode, NULL);
1379 handle.cookie = it->it_lock_handle;
1381 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1382 " for lock %#llx\n",
1383 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1385 md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1386 it->it_lock_set = 1;
1390 *bits = it->it_lock_bits;
1393 static inline void ll_lock_dcache(struct inode *inode)
1395 #ifdef HAVE_DCACHE_LOCK
1396 spin_lock(&dcache_lock);
1398 spin_lock(&inode->i_lock);
1402 static inline void ll_unlock_dcache(struct inode *inode)
1404 #ifdef HAVE_DCACHE_LOCK
1405 spin_unlock(&dcache_lock);
1407 spin_unlock(&inode->i_lock);
1411 static inline int d_lustre_invalid(const struct dentry *dentry)
1413 struct ll_dentry_data *lld = ll_d2d(dentry);
1415 return (lld == NULL) || lld->lld_invalid;
1418 static inline void __d_lustre_invalidate(struct dentry *dentry)
1420 struct ll_dentry_data *lld = ll_d2d(dentry);
1423 lld->lld_invalid = 1;
1427 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1428 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1429 * else dput() of the last refcount will unhash this dentry and kill it.
1431 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1433 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1434 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1435 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1437 spin_lock_nested(&dentry->d_lock,
1438 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1439 __d_lustre_invalidate(dentry);
1441 * We should be careful about dentries created by d_obtain_alias().
1442 * These dentries are not put in the dentry tree, instead they are
1443 * linked to sb->s_anon through dentry->d_hash.
1444 * shrink_dcache_for_umount() shrinks the tree and sb->s_anon list.
1445 * If we unhashed such a dentry, unmount would not be able to find
1446 * it and busy inodes would be reported.
1448 if (ll_d_count(dentry) == 0 && !(dentry->d_flags & DCACHE_DISCONNECTED))
1450 spin_unlock(&dentry->d_lock);
1453 static inline void d_lustre_revalidate(struct dentry *dentry)
1455 spin_lock(&dentry->d_lock);
1456 LASSERT(ll_d2d(dentry) != NULL);
1457 ll_d2d(dentry)->lld_invalid = 0;
1458 spin_unlock(&dentry->d_lock);
1461 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1462 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1463 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1464 int ll_layout_write_intent(struct inode *inode, enum layout_intent_opc opc,
1465 struct lu_extent *ext);
1467 int ll_xattr_init(void);
1468 void ll_xattr_fini(void);
1470 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1471 struct cl_page *page, enum cl_req_type crt);
1473 int ll_getparent(struct file *file, struct getparent __user *arg);
1476 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1477 enum op_xvalid xvalid, unsigned int attr_flags);
1479 extern struct lu_env *cl_inode_fini_env;
1480 extern __u16 cl_inode_fini_refcheck;
1482 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1483 void cl_inode_fini(struct inode *inode);
1485 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1486 u32 cl_fid_build_gen(const struct lu_fid *fid);
1488 #endif /* LLITE_INTERNAL_H */