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 */
202 /* protect the file heat fields */
203 spinlock_t lli_heat_lock;
204 __u32 lli_heat_flags;
205 struct obd_heat_instance lli_heat_instances[OBD_HEAT_COUNT];
208 * Whenever a process try to read/write the file, the
209 * jobid of the process will be saved here, and it'll
210 * be packed into the write PRC when flush later.
212 * So the read/write statistics for jobid will not be
213 * accurate if the file is shared by different jobs.
215 char lli_jobid[LUSTRE_JOBID_SIZE];
219 /* XXX: For following frequent used members, although they maybe special
220 * used for non-directory object, it is some time-wasting to check
221 * whether the object is directory or not before using them. On the
222 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
223 * the "ll_inode_info" size even if moving those members into u.f.
224 * So keep them out side.
226 * In the future, if more members are added only for directory,
227 * some of the following members can be moved into u.f.
229 struct cl_object *lli_clob;
231 /* mutex to request for layout lock exclusively. */
232 struct mutex lli_layout_mutex;
233 /* Layout version, protected by lli_layout_lock */
234 __u32 lli_layout_gen;
235 spinlock_t lli_layout_lock;
237 __u32 lli_projid; /* project id */
239 struct rw_semaphore lli_xattrs_list_rwsem;
240 struct mutex lli_xattrs_enq_lock;
241 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
244 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
248 spin_lock(&lli->lli_layout_lock);
249 gen = lli->lli_layout_gen;
250 spin_unlock(&lli->lli_layout_lock);
255 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
257 spin_lock(&lli->lli_layout_lock);
258 lli->lli_layout_gen = gen;
259 spin_unlock(&lli->lli_layout_lock);
263 /* File data is modified. */
264 LLIF_DATA_MODIFIED = 0,
265 /* File is being restored */
266 LLIF_FILE_RESTORING = 1,
267 /* Xattr cache is attached to the file */
268 LLIF_XATTR_CACHE = 2,
269 /* Project inherit */
270 LLIF_PROJECT_INHERIT = 3,
273 static inline void ll_file_set_flag(struct ll_inode_info *lli,
274 enum ll_file_flags flag)
276 set_bit(flag, &lli->lli_flags);
279 static inline void ll_file_clear_flag(struct ll_inode_info *lli,
280 enum ll_file_flags flag)
282 clear_bit(flag, &lli->lli_flags);
285 static inline bool ll_file_test_flag(struct ll_inode_info *lli,
286 enum ll_file_flags flag)
288 return test_bit(flag, &lli->lli_flags);
291 static inline bool ll_file_test_and_clear_flag(struct ll_inode_info *lli,
292 enum ll_file_flags flag)
294 return test_and_clear_bit(flag, &lli->lli_flags);
297 int ll_xattr_cache_destroy(struct inode *inode);
299 int ll_xattr_cache_get(struct inode *inode,
305 static inline bool obd_connect_has_secctx(struct obd_connect_data *data)
307 #if defined(HAVE_SECURITY_DENTRY_INIT_SECURITY) && defined(CONFIG_SECURITY)
308 return data->ocd_connect_flags & OBD_CONNECT_FLAGS2 &&
309 data->ocd_connect_flags2 & OBD_CONNECT2_FILE_SECCTX;
312 #endif /* HAVE_SECURITY_DENTRY_INIT_SECURITY */
315 static inline void obd_connect_set_secctx(struct obd_connect_data *data)
317 #if defined(HAVE_SECURITY_DENTRY_INIT_SECURITY) && defined(CONFIG_SECURITY)
318 data->ocd_connect_flags2 |= OBD_CONNECT2_FILE_SECCTX;
319 #endif /* HAVE_SECURITY_DENTRY_INIT_SECURITY */
322 int ll_dentry_init_security(struct dentry *dentry, int mode, struct qstr *name,
323 const char **secctx_name, void **secctx,
325 int ll_inode_init_security(struct dentry *dentry, struct inode *inode,
328 int ll_listsecurity(struct inode *inode, char *secctx_name,
329 size_t secctx_name_size);
332 * Locking to guarantee consistency of non-atomic updates to long long i_size,
333 * consistency between file size and KMS.
335 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
338 void ll_inode_size_lock(struct inode *inode);
339 void ll_inode_size_unlock(struct inode *inode);
341 // FIXME: replace the name of this with LL_I to conform to kernel stuff
342 // static inline struct ll_inode_info *LL_I(struct inode *inode)
343 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
345 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
348 /* default to about 64M of readahead on a given system. */
349 #define SBI_DEFAULT_READAHEAD_MAX (64UL << (20 - PAGE_SHIFT))
351 /* default to read-ahead full files smaller than 2MB on the second read */
352 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_SHIFT))
357 RA_STAT_DISTANT_READPAGE,
358 RA_STAT_MISS_IN_WINDOW,
359 RA_STAT_FAILED_GRAB_PAGE,
360 RA_STAT_FAILED_MATCH,
365 RA_STAT_MAX_IN_FLIGHT,
366 RA_STAT_WRONG_GRAB_PAGE,
367 RA_STAT_FAILED_REACH_END,
372 atomic_t ra_cur_pages;
373 unsigned long ra_max_pages;
374 unsigned long ra_max_pages_per_file;
375 unsigned long ra_max_read_ahead_whole_pages;
378 /* ra_io_arg will be filled in the beginning of ll_readahead with
379 * ras_lock, then the following ll_read_ahead_pages will read RA
380 * pages according to this arg, all the items in this structure are
381 * counted by page index.
384 unsigned long ria_start; /* start offset of read-ahead*/
385 unsigned long ria_end; /* end offset of read-ahead*/
386 unsigned long ria_reserved; /* reserved pages for read-ahead */
387 unsigned long ria_end_min; /* minimum end to cover current read */
388 bool ria_eof; /* reach end of file */
389 /* If stride read pattern is detected, ria_stoff means where
390 * stride read is started. Note: for normal read-ahead, the
391 * value here is meaningless, and also it will not be accessed*/
393 /* ria_length and ria_pages are the length and pages length in the
394 * stride I/O mode. And they will also be used to check whether
395 * it is stride I/O read-ahead in the read-ahead pages*/
396 unsigned long ria_length;
397 unsigned long ria_pages;
400 /* LL_HIST_MAX=32 causes an overflow */
401 #define LL_HIST_MAX 28
402 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
403 #define LL_PROCESS_HIST_MAX 10
404 struct per_process_info {
406 struct obd_histogram pp_r_hist;
407 struct obd_histogram pp_w_hist;
410 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
411 struct ll_rw_extents_info {
412 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
415 #define LL_OFFSET_HIST_MAX 100
416 struct ll_rw_process_info {
419 loff_t rw_range_start;
421 loff_t rw_last_file_pos;
423 size_t rw_smallest_extent;
424 size_t rw_largest_extent;
425 struct ll_file_data *rw_last_file;
428 enum stats_track_type {
429 STATS_TRACK_ALL = 0, /* track all processes */
430 STATS_TRACK_PID, /* track process with this pid */
431 STATS_TRACK_PPID, /* track processes with this ppid */
432 STATS_TRACK_GID, /* track processes with this gid */
436 /* flags for sbi->ll_flags */
437 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
438 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
439 #define LL_SBI_FLOCK 0x04
440 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
441 #define LL_SBI_ACL 0x10 /* support ACL */
442 /* LL_SBI_RMT_CLIENT 0x40 remote client */
443 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa, obsolete */
444 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa, obsolete */
445 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
446 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
447 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
448 /* LL_SBI_SOM_PREVIEW 0x1000 SOM preview mount option, obsolete */
449 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
450 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
451 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
452 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
453 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
454 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
455 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
456 #define LL_SBI_NOROOTSQUASH 0x100000 /* do not apply root squash */
457 #define LL_SBI_ALWAYS_PING 0x200000 /* always ping even if server
459 #define LL_SBI_FAST_READ 0x400000 /* fast read support */
460 #define LL_SBI_FILE_SECCTX 0x800000 /* set file security context at create */
461 /* LL_SBI_PIO 0x1000000 parallel IO support, introduced in
463 #define LL_SBI_TINY_WRITE 0x2000000 /* tiny write support */
464 #define LL_SBI_FILE_HEAT 0x4000000 /* file heat support */
465 #define LL_SBI_FLAGS { \
495 /* This is embedded into llite super-blocks to keep track of connect
496 * flags (capabilities) supported by all imports given mount is
498 struct lustre_client_ocd {
499 /* This is conjunction of connect_flags across all imports
500 * (LOVs) this mount is connected to. This field is updated by
501 * cl_ocd_update() under ->lco_lock. */
503 struct mutex lco_lock;
504 struct obd_export *lco_md_exp;
505 struct obd_export *lco_dt_exp;
509 /* this protects pglist and ra_info. It isn't safe to
510 * grab from interrupt contexts */
512 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
513 spinlock_t ll_process_lock; /* ll_rw_process_info */
514 struct obd_uuid ll_sb_uuid;
515 struct obd_export *ll_md_exp;
516 struct obd_export *ll_dt_exp;
517 struct obd_device *ll_md_obd;
518 struct obd_device *ll_dt_obd;
519 struct dentry *ll_debugfs_entry;
520 struct lu_fid ll_root_fid; /* root object fid */
523 unsigned int ll_umounting:1,
524 ll_xattr_cache_enabled:1,
525 ll_xattr_cache_set:1, /* already set to 0/1 */
526 ll_client_common_fill_super_succeeded:1,
529 struct lustre_client_ocd ll_lco;
531 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
533 /* Used to track "unstable" pages on a client, and maintain a
534 * LRU list of clean pages. An "unstable" page is defined as
535 * any page which is sent to a server as part of a bulk request,
536 * but is uncommitted to stable storage. */
537 struct cl_client_cache *ll_cache;
539 struct lprocfs_stats *ll_ra_stats;
541 struct ll_ra_info ll_ra_info;
542 unsigned int ll_namelen;
543 struct file_operations *ll_fop;
545 struct lu_site *ll_site;
546 struct cl_device *ll_cl;
548 struct ll_rw_extents_info ll_rw_extents_info;
549 int ll_extent_process_count;
550 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
551 unsigned int ll_offset_process_count;
552 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
553 unsigned int ll_rw_offset_entry_count;
554 int ll_stats_track_id;
555 enum stats_track_type ll_stats_track_type;
558 /* metadata stat-ahead */
559 unsigned int ll_sa_running_max;/* max concurrent
560 * statahead instances */
561 unsigned int ll_sa_max; /* max statahead RPCs */
562 atomic_t ll_sa_total; /* statahead thread started
564 atomic_t ll_sa_wrong; /* statahead thread stopped for
566 atomic_t ll_sa_running; /* running statahead thread
568 atomic_t ll_agl_total; /* AGL thread started count */
570 dev_t ll_sdev_orig; /* save s_dev before assign for
573 struct root_squash_info ll_squash;
576 /* st_blksize returned by stat(2), when non-zero */
577 unsigned int ll_stat_blksize;
579 struct kset ll_kset; /* sysfs object */
580 struct completion ll_kobj_unregister;
583 unsigned int ll_heat_decay_weight;
584 unsigned int ll_heat_period_second;
586 /* filesystem fsname */
587 char ll_fsname[LUSTRE_MAXFSNAME + 1];
590 #define SBI_DEFAULT_HEAT_DECAY_WEIGHT ((80 * 256 + 50) / 100)
591 #define SBI_DEFAULT_HEAT_PERIOD_SECOND (60)
593 * per file-descriptor read-ahead data.
595 struct ll_readahead_state {
598 * index of the last page that read(2) needed and that wasn't in the
599 * cache. Used by ras_update() to detect seeks.
601 * XXX nikita: if access seeks into cached region, Lustre doesn't see
604 unsigned long ras_last_readpage;
606 * number of pages read after last read-ahead window reset. As window
607 * is reset on each seek, this is effectively a number of consecutive
608 * accesses. Maybe ->ras_accessed_in_window is better name.
610 * XXX nikita: window is also reset (by ras_update()) when Lustre
611 * believes that memory pressure evicts read-ahead pages. In that
612 * case, it probably doesn't make sense to expand window to
613 * PTLRPC_MAX_BRW_PAGES on the third access.
615 unsigned long ras_consecutive_pages;
617 * number of read requests after the last read-ahead window reset
618 * As window is reset on each seek, this is effectively the number
619 * on consecutive read request and is used to trigger read-ahead.
621 unsigned long ras_consecutive_requests;
623 * Parameters of current read-ahead window. Handled by
624 * ras_update(). On the initial access to the file or after a seek,
625 * window is reset to 0. After 3 consecutive accesses, window is
626 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
627 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
629 unsigned long ras_window_start, ras_window_len;
631 * Optimal RPC size. It decides how many pages will be sent
632 * for each read-ahead.
634 unsigned long ras_rpc_size;
636 * Where next read-ahead should start at. This lies within read-ahead
637 * window. Read-ahead window is read in pieces rather than at once
638 * because: 1. lustre limits total number of pages under read-ahead by
639 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
640 * not covered by DLM lock.
642 unsigned long ras_next_readahead;
644 * Total number of ll_file_read requests issued, reads originating
645 * due to mmap are not counted in this total. This value is used to
646 * trigger full file read-ahead after multiple reads to a small file.
648 unsigned long ras_requests;
650 * Page index with respect to the current request, these value
651 * will not be accurate when dealing with reads issued via mmap.
653 unsigned long ras_request_index;
655 * The following 3 items are used for detecting the stride I/O
657 * In stride I/O mode,
658 * ...............|-----data-----|****gap*****|--------|******|....
659 * offset |-stride_pages-|-stride_gap-|
660 * ras_stride_offset = offset;
661 * ras_stride_length = stride_pages + stride_gap;
662 * ras_stride_pages = stride_pages;
663 * Note: all these three items are counted by pages.
665 unsigned long ras_stride_length;
666 unsigned long ras_stride_pages;
667 pgoff_t ras_stride_offset;
669 * number of consecutive stride request count, and it is similar as
670 * ras_consecutive_requests, but used for stride I/O mode.
671 * Note: only more than 2 consecutive stride request are detected,
672 * stride read-ahead will be enable
674 unsigned long ras_consecutive_stride_requests;
677 extern struct kmem_cache *ll_file_data_slab;
678 struct lustre_handle;
679 struct ll_file_data {
680 struct ll_readahead_state fd_ras;
681 struct ll_grouplock fd_grouplock;
685 /* openhandle if lease exists for this file.
686 * Borrow lli->lli_och_mutex to protect assignment */
687 struct obd_client_handle *fd_lease_och;
688 struct obd_client_handle *fd_och;
689 struct file *fd_file;
690 /* Indicate whether need to report failure when close.
691 * true: failure is known, not report again.
692 * false: unknown failure, should report. */
693 bool fd_write_failed;
694 bool ll_lock_no_expand;
695 rwlock_t fd_lock; /* protect lcc list */
696 struct list_head fd_lccs; /* list of ll_cl_context */
697 /* Used by mirrored file to lead IOs to a specific mirror, usually
698 * for mirror resync. 0 means default. */
699 __u32 fd_designated_mirror;
700 /* The layout version when resync starts. Resync I/O should carry this
701 * layout version for verification to OST objects */
702 __u32 fd_layout_version;
705 void llite_tunables_unregister(void);
706 int llite_tunables_register(void);
708 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
710 return &lli->lli_vfs_inode;
713 __u32 ll_i2suppgid(struct inode *i);
714 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
716 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
718 #if BITS_PER_LONG == 32
720 #elif defined(CONFIG_COMPAT)
721 if (unlikely(sbi->ll_flags & LL_SBI_32BIT_API))
724 # ifdef CONFIG_X86_X32
725 /* in_compat_syscall() returns true when called from a kthread
726 * and CONFIG_X86_X32 is enabled, which is wrong. So check
727 * whether the caller comes from a syscall (ie. not a kthread)
728 * before calling in_compat_syscall(). */
729 if (current->flags & PF_KTHREAD)
733 return unlikely(in_compat_syscall());
735 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
739 static inline bool ll_sbi_has_fast_read(struct ll_sb_info *sbi)
741 return !!(sbi->ll_flags & LL_SBI_FAST_READ);
744 static inline bool ll_sbi_has_tiny_write(struct ll_sb_info *sbi)
746 return !!(sbi->ll_flags & LL_SBI_TINY_WRITE);
749 static inline bool ll_sbi_has_file_heat(struct ll_sb_info *sbi)
751 return !!(sbi->ll_flags & LL_SBI_FILE_HEAT);
754 void ll_ras_enter(struct file *f);
756 /* llite/lcommon_misc.c */
757 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
758 enum obd_notify_event ev, void *owner);
759 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
760 struct ll_grouplock *lg);
761 void cl_put_grouplock(struct ll_grouplock *lg);
763 /* llite/lproc_llite.c */
764 int ll_debugfs_register_super(struct super_block *sb, const char *name);
765 void ll_debugfs_unregister_super(struct super_block *sb);
766 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
770 LPROC_LL_DIRTY_MISSES,
772 LPROC_LL_WRITE_BYTES,
797 LPROC_LL_ALLOC_INODE,
800 LPROC_LL_GETXATTR_HITS,
802 LPROC_LL_REMOVEXATTR,
804 LPROC_LL_FILE_OPCODES
808 struct ll_dir_chain {
811 static inline void ll_dir_chain_init(struct ll_dir_chain *chain)
815 static inline void ll_dir_chain_fini(struct ll_dir_chain *chain)
819 extern const struct file_operations ll_dir_operations;
820 extern const struct inode_operations ll_dir_inode_operations;
821 #ifdef HAVE_DIR_CONTEXT
822 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
823 struct dir_context *ctx);
825 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
826 void *cookie, filldir_t filldir);
828 int ll_get_mdt_idx(struct inode *inode);
829 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
830 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
831 __u64 offset, struct ll_dir_chain *chain);
832 void ll_release_page(struct inode *inode, struct page *page, bool remove);
835 extern const struct inode_operations ll_special_inode_operations;
837 struct inode *ll_iget(struct super_block *sb, ino_t hash,
838 struct lustre_md *lic);
839 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
840 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
841 void *data, int flag);
842 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
843 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
844 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
847 int ll_writepage(struct page *page, struct writeback_control *wbc);
848 int ll_writepages(struct address_space *, struct writeback_control *wbc);
849 int ll_readpage(struct file *file, struct page *page);
850 int ll_io_read_page(const struct lu_env *env, struct cl_io *io,
851 struct cl_page *page, struct file *file);
852 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
853 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
856 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io,
858 void ll_cl_remove(struct file *file, const struct lu_env *env);
859 struct ll_cl_context *ll_cl_find(struct file *file);
861 extern const struct address_space_operations ll_aops;
864 extern struct file_operations ll_file_operations;
865 extern struct file_operations ll_file_operations_flock;
866 extern struct file_operations ll_file_operations_noflock;
867 extern struct inode_operations ll_file_inode_operations;
868 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
869 enum ldlm_mode l_req_mode);
870 extern enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
871 struct lustre_handle *lockh, __u64 flags,
872 enum ldlm_mode mode);
874 int ll_file_open(struct inode *inode, struct file *file);
875 int ll_file_release(struct inode *inode, struct file *file);
876 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
877 int ll_md_real_close(struct inode *inode, fmode_t fmode);
878 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
879 struct ll_file_data *file, loff_t pos,
880 size_t count, int rw);
881 #ifdef HAVE_INODEOPS_ENHANCED_GETATTR
882 int ll_getattr(const struct path *path, struct kstat *stat,
883 u32 request_mask, unsigned int flags);
885 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
887 struct posix_acl *ll_get_acl(struct inode *inode, int type);
888 #ifdef HAVE_IOP_SET_ACL
889 #ifdef CONFIG_FS_POSIX_ACL
890 int ll_set_acl(struct inode *inode, struct posix_acl *acl, int type);
891 #else /* !CONFIG_FS_POSIX_ACL */
892 #define ll_set_acl NULL
893 #endif /* CONFIG_FS_POSIX_ACL */
896 int ll_migrate(struct inode *parent, struct file *file,
897 struct lmv_user_md *lum, const char *name);
898 int ll_get_fid_by_name(struct inode *parent, const char *name,
899 int namelen, struct lu_fid *fid, struct inode **inode);
900 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
901 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
903 # ifndef HAVE_INODE_PERMISION_2ARGS
904 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
906 int ll_inode_permission(struct inode *inode, int mask);
909 int ll_ioctl_check_project(struct inode *inode, struct fsxattr *fa);
910 int ll_ioctl_fsgetxattr(struct inode *inode, unsigned int cmd,
912 int ll_ioctl_fssetxattr(struct inode *inode, unsigned int cmd,
915 int ll_lov_setstripe_ea_info(struct inode *inode, struct dentry *dentry,
916 __u64 flags, struct lov_user_md *lum,
918 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
919 struct lov_mds_md **lmm, int *lmm_size,
920 struct ptlrpc_request **request);
921 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
923 int ll_dir_getstripe(struct inode *inode, void **lmmp,
924 int *lmm_size, struct ptlrpc_request **request,
926 #ifdef HAVE_FILE_FSYNC_4ARGS
927 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
928 #elif defined(HAVE_FILE_FSYNC_2ARGS)
929 int ll_fsync(struct file *file, int data);
931 int ll_fsync(struct file *file, struct dentry *dentry, int data);
933 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
934 int ll_fid2path(struct inode *inode, void __user *arg);
935 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
936 int ll_hsm_release(struct inode *inode);
937 int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
938 void ll_io_set_mirror(struct cl_io *io, const struct file *file);
942 int ll_d_init(struct dentry *de);
943 extern const struct dentry_operations ll_d_ops;
944 void ll_intent_drop_lock(struct lookup_intent *);
945 void ll_intent_release(struct lookup_intent *);
946 void ll_invalidate_aliases(struct inode *);
947 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
948 int ll_revalidate_it_finish(struct ptlrpc_request *request,
949 struct lookup_intent *it, struct dentry *de);
951 /* llite/llite_lib.c */
952 extern struct super_operations lustre_super_operations;
954 void ll_lli_init(struct ll_inode_info *lli);
955 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
956 void ll_put_super(struct super_block *sb);
957 void ll_kill_super(struct super_block *sb);
958 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
959 void ll_dir_clear_lsm_md(struct inode *inode);
960 void ll_clear_inode(struct inode *inode);
961 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr,
962 enum op_xvalid xvalid, bool hsm_import);
963 int ll_setattr(struct dentry *de, struct iattr *attr);
964 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
965 int ll_statfs_internal(struct ll_sb_info *sbi, struct obd_statfs *osfs,
967 int ll_update_inode(struct inode *inode, struct lustre_md *md);
968 void ll_update_inode_flags(struct inode *inode, int ext_flags);
969 int ll_read_inode2(struct inode *inode, void *opaque);
970 void ll_delete_inode(struct inode *inode);
971 int ll_iocontrol(struct inode *inode, struct file *file,
972 unsigned int cmd, unsigned long arg);
973 int ll_flush_ctx(struct inode *inode);
974 void ll_umount_begin(struct super_block *sb);
975 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
976 #ifdef HAVE_SUPEROPS_USE_DENTRY
977 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
979 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
981 void ll_dirty_page_discard_warn(struct page *page, int ioret);
982 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
983 struct super_block *, struct lookup_intent *);
984 int ll_obd_statfs(struct inode *inode, void __user *arg);
985 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
986 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
987 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
990 LUSTRE_OPC_MKDIR = 0,
991 LUSTRE_OPC_SYMLINK = 1,
992 LUSTRE_OPC_MKNOD = 2,
993 LUSTRE_OPC_CREATE = 3,
997 void ll_unlock_md_op_lsm(struct md_op_data *op_data);
998 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
999 struct inode *i1, struct inode *i2,
1000 const char *name, size_t namelen,
1001 __u32 mode, __u32 opc, void *data);
1002 void ll_finish_md_op_data(struct md_op_data *op_data);
1003 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
1004 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
1005 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
1006 struct lov_user_md **kbuf);
1007 void ll_open_cleanup(struct super_block *sb, struct ptlrpc_request *open_req);
1009 void ll_dom_finish_open(struct inode *inode, struct ptlrpc_request *req,
1010 struct lookup_intent *it);
1012 /* Compute expected user md size when passing in a md from user space */
1013 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
1015 switch (lum->lmm_magic) {
1016 case LOV_USER_MAGIC_V1:
1017 return sizeof(struct lov_user_md_v1);
1018 case LOV_USER_MAGIC_V3:
1019 return sizeof(struct lov_user_md_v3);
1020 case LOV_USER_MAGIC_SPECIFIC:
1021 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
1024 return lov_user_md_size(lum->lmm_stripe_count,
1025 LOV_USER_MAGIC_SPECIFIC);
1026 case LOV_USER_MAGIC_COMP_V1:
1027 return ((struct lov_comp_md_v1 *)lum)->lcm_size;
1033 /* llite/llite_nfs.c */
1034 extern struct export_operations lustre_export_operations;
1035 __u32 get_uuid2int(const char *name, int len);
1036 struct inode *search_inode_for_lustre(struct super_block *sb,
1037 const struct lu_fid *fid);
1038 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
1040 /* llite/symlink.c */
1041 extern struct inode_operations ll_fast_symlink_inode_operations;
1044 * IO arguments for various VFS I/O interfaces.
1046 struct vvp_io_args {
1047 /** normal/sendfile/splice */
1048 enum vvp_io_subtype via_io_subtype;
1052 struct kiocb *via_iocb;
1053 struct iov_iter *via_iter;
1056 struct pipe_inode_info *via_pipe;
1057 unsigned int via_flags;
1067 struct ll_cl_context {
1068 struct list_head lcc_list;
1070 const struct lu_env *lcc_env;
1071 struct cl_io *lcc_io;
1072 struct cl_page *lcc_page;
1073 enum lcc_type lcc_type;
1076 struct ll_thread_info {
1077 struct iov_iter lti_iter;
1078 struct vvp_io_args lti_args;
1079 struct ra_io_arg lti_ria;
1080 struct ll_cl_context lti_io_ctx;
1083 extern struct lu_context_key ll_thread_key;
1085 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
1087 struct ll_thread_info *lti;
1089 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
1090 LASSERT(lti != NULL);
1095 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env,
1096 enum vvp_io_subtype type)
1098 struct vvp_io_args *via = &ll_env_info(env)->lti_args;
1100 via->via_io_subtype = type;
1105 /* llite/llite_mmap.c */
1107 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1108 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1109 void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
1110 unsigned long addr, size_t count);
1111 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1114 static inline void ll_invalidate_page(struct page *vmpage)
1116 struct address_space *mapping = vmpage->mapping;
1117 loff_t offset = vmpage->index << PAGE_SHIFT;
1119 LASSERT(PageLocked(vmpage));
1120 if (mapping == NULL)
1124 * truncate_complete_page() calls
1125 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
1127 ll_teardown_mmaps(mapping, offset, offset + PAGE_SIZE);
1128 truncate_complete_page(mapping, vmpage);
1131 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1133 /* don't need an addref as the sb_info should be holding one */
1134 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1136 return ll_s2sbi(sb)->ll_dt_exp;
1139 /* don't need an addref as the sb_info should be holding one */
1140 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1142 return ll_s2sbi(sb)->ll_md_exp;
1145 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1147 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1153 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1154 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1156 return ll_s2sbi(inode->i_sb);
1159 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1161 return ll_s2dtexp(inode->i_sb);
1164 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1166 return ll_s2mdexp(inode->i_sb);
1169 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1173 LASSERT(inode != NULL);
1174 fid = &ll_i2info(inode)->lli_fid;
1179 static inline loff_t ll_file_maxbytes(struct inode *inode)
1181 struct cl_object *obj = ll_i2info(inode)->lli_clob;
1184 return MAX_LFS_FILESIZE;
1186 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
1190 extern const struct xattr_handler *ll_xattr_handlers[];
1192 #define XATTR_USER_T 1
1193 #define XATTR_TRUSTED_T 2
1194 #define XATTR_SECURITY_T 3
1195 #define XATTR_ACL_ACCESS_T 4
1196 #define XATTR_ACL_DEFAULT_T 5
1197 #define XATTR_LUSTRE_T 6
1198 #define XATTR_OTHER_T 7
1200 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1201 int ll_xattr_list(struct inode *inode, const char *name, int type,
1202 void *buffer, size_t size, u64 valid);
1203 const struct xattr_handler *get_xattr_type(const char *name);
1206 * Common IO arguments for various VFS I/O interfaces.
1208 int cl_sb_init(struct super_block *sb);
1209 int cl_sb_fini(struct super_block *sb);
1211 enum ras_update_flags {
1215 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1216 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1220 #define LL_SA_RPC_MIN 2
1221 #define LL_SA_RPC_DEF 32
1222 #define LL_SA_RPC_MAX 512
1224 /* XXX: If want to support more concurrent statahead instances,
1225 * please consider to decentralize the RPC lists attached
1226 * on related import, such as imp_{sending,delayed}_list.
1228 #define LL_SA_RUNNING_MAX 256
1229 #define LL_SA_RUNNING_DEF 16
1231 #define LL_SA_CACHE_BIT 5
1232 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1233 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1235 /* per inode struct, for dir only */
1236 struct ll_statahead_info {
1237 struct dentry *sai_dentry;
1238 atomic_t sai_refcount; /* when access this struct, hold
1240 unsigned int sai_max; /* max ahead of lookup */
1241 __u64 sai_sent; /* stat requests sent count */
1242 __u64 sai_replied; /* stat requests which received
1244 __u64 sai_index; /* index of statahead entry */
1245 __u64 sai_index_wait; /* index of entry which is the
1246 * caller is waiting for */
1247 __u64 sai_hit; /* hit count */
1248 __u64 sai_miss; /* miss count:
1249 * for "ls -al" case, includes
1250 * hidden dentry miss;
1251 * for "ls -l" case, it does not
1252 * include hidden dentry miss.
1253 * "sai_miss_hidden" is used for
1256 unsigned int sai_consecutive_miss; /* consecutive miss */
1257 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1258 * is not a hidden one */
1259 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1261 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1263 sai_agl_valid:1,/* AGL is valid for the dir */
1264 sai_in_readpage:1;/* statahead is in readdir()*/
1265 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1266 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1267 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1268 struct list_head sai_interim_entries; /* entries which got async
1269 * stat reply, but not
1271 struct list_head sai_entries; /* completed entries */
1272 struct list_head sai_agls; /* AGLs to be sent */
1273 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1274 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1275 atomic_t sai_cache_count; /* entry count in cache */
1278 int ll_statahead(struct inode *dir, struct dentry **dentry, bool unplug);
1279 void ll_authorize_statahead(struct inode *dir, void *key);
1280 void ll_deauthorize_statahead(struct inode *dir, void *key);
1283 blkcnt_t dirty_cnt(struct inode *inode);
1285 int cl_glimpse_size0(struct inode *inode, int agl);
1286 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1287 struct inode *inode, struct cl_object *clob, int agl);
1289 static inline int cl_glimpse_size(struct inode *inode)
1291 return cl_glimpse_size0(inode, 0);
1294 /* AGL is 'asychronous glimpse lock', which is a speculative lock taken as
1295 * part of statahead */
1296 static inline int cl_agl(struct inode *inode)
1298 return cl_glimpse_size0(inode, 1);
1301 int ll_file_lock_ahead(struct file *file, struct llapi_lu_ladvise *ladvise);
1303 int cl_io_get(struct inode *inode, struct lu_env **envout,
1304 struct cl_io **ioout, __u16 *refcheck);
1306 static inline int ll_glimpse_size(struct inode *inode)
1308 struct ll_inode_info *lli = ll_i2info(inode);
1311 down_read(&lli->lli_glimpse_sem);
1312 rc = cl_glimpse_size(inode);
1313 lli->lli_glimpse_time = ktime_get();
1314 up_read(&lli->lli_glimpse_sem);
1318 /* dentry may statahead when statahead is enabled and current process has opened
1319 * parent directory, and this dentry hasn't accessed statahead cache before */
1321 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1323 struct ll_inode_info *lli;
1324 struct ll_dentry_data *ldd;
1326 if (ll_i2sbi(dir)->ll_sa_max == 0)
1329 lli = ll_i2info(dir);
1331 /* statahead is not allowed for this dir, there may be three causes:
1332 * 1. dir is not opened.
1333 * 2. statahead hit ratio is too low.
1334 * 3. previous stat started statahead thread failed. */
1335 if (!lli->lli_sa_enabled)
1338 /* not the same process, don't statahead */
1339 if (lli->lli_opendir_pid != current_pid())
1343 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1344 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1345 * For patchless client, lookup intent is not accurate, which may
1346 * misguide statahead. For example:
1347 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1348 * have the same intent -- IT_GETATTR, while one dentry should access
1349 * statahead cache once, otherwise statahead windows is messed up.
1350 * The solution is as following:
1351 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1352 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1353 * bypass interacting with statahead cache by checking
1354 * 'lld_sa_generation == lli->lli_sa_generation'.
1356 ldd = ll_d2d(dentry);
1357 if (ldd != NULL && ldd->lld_sa_generation == lli->lli_sa_generation)
1363 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1364 enum cl_fsync_mode mode, int ignore_layout);
1366 static inline int ll_file_nolock(const struct file *file)
1368 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1369 struct inode *inode = file_inode((struct file *)file);
1371 LASSERT(fd != NULL);
1372 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1373 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1376 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1377 struct lookup_intent *it, __u64 *bits)
1379 if (!it->it_lock_set) {
1380 struct lustre_handle handle;
1382 /* If this inode is a remote object, it will get two
1383 * separate locks in different namespaces, Master MDT,
1384 * where the name entry is, will grant LOOKUP lock,
1385 * remote MDT, where the object is, will grant
1386 * UPDATE|PERM lock. The inode will be attched to both
1387 * LOOKUP and PERM locks, so revoking either locks will
1388 * case the dcache being cleared */
1389 if (it->it_remote_lock_mode) {
1390 handle.cookie = it->it_remote_lock_handle;
1391 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1392 "(%p) for remote lock %#llx\n",
1393 PFID(ll_inode2fid(inode)), inode,
1395 md_set_lock_data(exp, &handle, inode, NULL);
1398 handle.cookie = it->it_lock_handle;
1400 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1401 " for lock %#llx\n",
1402 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1404 md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1405 it->it_lock_set = 1;
1409 *bits = it->it_lock_bits;
1412 static inline void ll_lock_dcache(struct inode *inode)
1414 #ifdef HAVE_DCACHE_LOCK
1415 spin_lock(&dcache_lock);
1417 spin_lock(&inode->i_lock);
1421 static inline void ll_unlock_dcache(struct inode *inode)
1423 #ifdef HAVE_DCACHE_LOCK
1424 spin_unlock(&dcache_lock);
1426 spin_unlock(&inode->i_lock);
1430 static inline int d_lustre_invalid(const struct dentry *dentry)
1432 struct ll_dentry_data *lld = ll_d2d(dentry);
1434 return (lld == NULL) || lld->lld_invalid;
1437 static inline void __d_lustre_invalidate(struct dentry *dentry)
1439 struct ll_dentry_data *lld = ll_d2d(dentry);
1442 lld->lld_invalid = 1;
1446 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1447 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1448 * else dput() of the last refcount will unhash this dentry and kill it.
1450 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1452 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1453 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1454 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1456 spin_lock_nested(&dentry->d_lock,
1457 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1458 __d_lustre_invalidate(dentry);
1460 * We should be careful about dentries created by d_obtain_alias().
1461 * These dentries are not put in the dentry tree, instead they are
1462 * linked to sb->s_anon through dentry->d_hash.
1463 * shrink_dcache_for_umount() shrinks the tree and sb->s_anon list.
1464 * If we unhashed such a dentry, unmount would not be able to find
1465 * it and busy inodes would be reported.
1467 if (ll_d_count(dentry) == 0 && !(dentry->d_flags & DCACHE_DISCONNECTED))
1469 spin_unlock(&dentry->d_lock);
1472 static inline void d_lustre_revalidate(struct dentry *dentry)
1474 spin_lock(&dentry->d_lock);
1475 LASSERT(ll_d2d(dentry) != NULL);
1476 ll_d2d(dentry)->lld_invalid = 0;
1477 spin_unlock(&dentry->d_lock);
1480 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1481 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1482 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1483 int ll_layout_write_intent(struct inode *inode, enum layout_intent_opc opc,
1484 struct lu_extent *ext);
1486 int ll_xattr_init(void);
1487 void ll_xattr_fini(void);
1489 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1490 struct cl_page *page, enum cl_req_type crt);
1492 int ll_getparent(struct file *file, struct getparent __user *arg);
1495 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1496 enum op_xvalid xvalid, unsigned int attr_flags);
1498 extern struct lu_env *cl_inode_fini_env;
1499 extern __u16 cl_inode_fini_refcheck;
1501 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1502 void cl_inode_fini(struct inode *inode);
1504 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1505 u32 cl_fid_build_gen(const struct lu_fid *fid);
1507 #endif /* LLITE_INTERNAL_H */