4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2014, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
37 #ifndef LLITE_INTERNAL_H
38 #define LLITE_INTERNAL_H
39 #include <lustre_debug.h>
40 #include <lustre_ver.h>
41 #include <lustre_disk.h> /* for s2sbi */
42 #include <lustre_eacl.h>
43 #include <lustre_linkea.h>
45 /* for struct cl_lock_descr and struct cl_io */
46 #include <cl_object.h>
47 #include <lustre_lmv.h>
48 #include <lustre_mdc.h>
49 #include <lustre_intent.h>
50 #include <linux/compat.h>
51 #include <linux/aio.h>
53 #include "vvp_internal.h"
54 #include "range_lock.h"
60 #ifndef VM_FAULT_RETRY
61 #define VM_FAULT_RETRY 0
64 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
65 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
66 #ifndef LOOKUP_CONTINUE
67 #define LOOKUP_CONTINUE LOOKUP_PARENT
70 /** Only used on client-side for indicating the tail of dir hash/offset. */
71 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
72 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
74 /* 4UL * 1024 * 1024 */
75 #define LL_MAX_BLKSIZE_BITS 22
77 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
78 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
80 struct ll_dentry_data {
81 struct lookup_intent *lld_it;
82 unsigned int lld_sa_generation;
83 unsigned int lld_invalid:1;
84 struct rcu_head lld_rcu_head;
87 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
89 #define LLI_INODE_MAGIC 0x111d0de5
90 #define LLI_INODE_DEAD 0xdeadd00d
92 /* remote client permission cache */
93 #define REMOTE_PERM_HASHSIZE 16
95 struct ll_getname_data {
96 #ifdef HAVE_DIR_CONTEXT
97 struct dir_context ctx;
99 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
100 struct lu_fid lgd_fid; /* target fid we are looking for */
101 int lgd_found; /* inode matched? */
104 /* llite setxid/access permission for user on remote client */
105 struct ll_remote_perm {
106 struct hlist_node lrp_list;
111 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
112 is access permission with
113 lrp_fsuid/lrp_fsgid. */
116 struct ll_grouplock {
117 struct lu_env *lg_env;
119 struct cl_lock *lg_lock;
120 unsigned long lg_gid;
124 /* File data is modified. */
125 LLIF_DATA_MODIFIED = 1 << 0,
126 /* File is being restored */
127 LLIF_FILE_RESTORING = 1 << 1,
128 /* Xattr cache is attached to the file */
129 LLIF_XATTR_CACHE = 1 << 2,
132 struct ll_inode_info {
133 __u32 lli_inode_magic;
137 struct posix_acl *lli_posix_acl;
139 struct hlist_head *lli_remote_perms;
140 struct mutex lli_rmtperm_mutex;
142 /* identifying fields for both metadata and data stacks. */
143 struct lu_fid lli_fid;
144 /* master inode fid for stripe directory */
145 struct lu_fid lli_pfid;
147 struct list_head lli_oss_capas;
148 /* open count currently used by capability only, indicate whether
149 * capability needs renewal */
150 atomic_t lli_open_count;
151 struct obd_capa *lli_mds_capa;
152 cfs_time_t lli_rmtperm_time;
154 /* We need all three because every inode may be opened in different
156 struct obd_client_handle *lli_mds_read_och;
157 struct obd_client_handle *lli_mds_write_och;
158 struct obd_client_handle *lli_mds_exec_och;
159 __u64 lli_open_fd_read_count;
160 __u64 lli_open_fd_write_count;
161 __u64 lli_open_fd_exec_count;
162 /* Protects access to och pointers and their usage counters */
163 struct mutex lli_och_mutex;
165 struct inode lli_vfs_inode;
167 /* the most recent timestamps obtained from mds */
171 spinlock_t lli_agl_lock;
173 /* Try to make the d::member and f::member are aligned. Before using
174 * these members, make clear whether it is directory or not. */
178 /* serialize normal readdir and statahead-readdir. */
179 struct mutex lli_readdir_mutex;
181 /* metadata statahead */
182 /* since parent-child threads can share the same @file
183 * struct, "opendir_key" is the token when dir close for
184 * case of parent exit before child -- it is me should
185 * cleanup the dir readahead. */
186 void *lli_opendir_key;
187 struct ll_statahead_info *lli_sai;
188 /* protect statahead stuff. */
189 spinlock_t lli_sa_lock;
190 /* "opendir_pid" is the token when lookup/revalid
191 * -- I am the owner of dir statahead. */
192 pid_t lli_opendir_pid;
193 /* stat will try to access statahead entries or start
194 * statahead if this flag is set, and this flag will be
195 * set upon dir open, and cleared when dir is closed,
196 * statahead hit ratio is too low, or start statahead
198 unsigned int lli_sa_enabled:1;
199 /* generation for statahead */
200 unsigned int lli_sa_generation;
201 /* directory stripe information */
202 struct lmv_stripe_md *lli_lsm_md;
203 /* default directory stripe offset. This is extracted
204 * from the "dmv" xattr in order to decide which MDT to
205 * create a subdirectory on. The MDS itself fetches
206 * "dmv" and gets the rest of the default layout itself
207 * (count, hash, etc). */
208 __u32 lli_def_stripe_offset;
211 /* for non-directory */
213 struct mutex lli_size_mutex;
214 char *lli_symlink_name;
216 * struct rw_semaphore {
217 * signed long count; // align d.d_def_acl
218 * spinlock_t wait_lock; // align d.d_sa_lock
219 * struct list_head wait_list;
222 struct rw_semaphore lli_trunc_sem;
223 struct range_lock_tree lli_write_tree;
225 struct rw_semaphore lli_glimpse_sem;
226 cfs_time_t lli_glimpse_time;
227 struct list_head lli_agl_list;
230 /* for writepage() only to communicate to fsync */
234 * whenever a process try to read/write the file, the
235 * jobid of the process will be saved here, and it'll
236 * be packed into the write PRC when flush later.
238 * so the read/write statistics for jobid will not be
239 * accurate if the file is shared by different jobs.
241 char lli_jobid[LUSTRE_JOBID_SIZE];
245 /* XXX: For following frequent used members, although they maybe special
246 * used for non-directory object, it is some time-wasting to check
247 * whether the object is directory or not before using them. On the
248 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
249 * the "ll_inode_info" size even if moving those members into u.f.
250 * So keep them out side.
252 * In the future, if more members are added only for directory,
253 * some of the following members can be moved into u.f.
255 struct cl_object *lli_clob;
257 /* mutex to request for layout lock exclusively. */
258 struct mutex lli_layout_mutex;
259 /* Layout version, protected by lli_layout_lock */
260 __u32 lli_layout_gen;
261 spinlock_t lli_layout_lock;
263 struct rw_semaphore lli_xattrs_list_rwsem;
264 struct mutex lli_xattrs_enq_lock;
265 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
268 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
272 spin_lock(&lli->lli_layout_lock);
273 gen = lli->lli_layout_gen;
274 spin_unlock(&lli->lli_layout_lock);
279 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
281 spin_lock(&lli->lli_layout_lock);
282 lli->lli_layout_gen = gen;
283 spin_unlock(&lli->lli_layout_lock);
286 int ll_xattr_cache_destroy(struct inode *inode);
288 int ll_xattr_cache_get(struct inode *inode,
295 * Locking to guarantee consistency of non-atomic updates to long long i_size,
296 * consistency between file size and KMS.
298 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
301 void ll_inode_size_lock(struct inode *inode);
302 void ll_inode_size_unlock(struct inode *inode);
304 // FIXME: replace the name of this with LL_I to conform to kernel stuff
305 // static inline struct ll_inode_info *LL_I(struct inode *inode)
306 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
308 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
311 /* default to about 40meg of readahead on a given system. That much tied
312 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
313 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT))
315 /* default to read-ahead full files smaller than 2MB on the second read */
316 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
321 RA_STAT_DISTANT_READPAGE,
322 RA_STAT_MISS_IN_WINDOW,
323 RA_STAT_FAILED_GRAB_PAGE,
324 RA_STAT_FAILED_MATCH,
329 RA_STAT_MAX_IN_FLIGHT,
330 RA_STAT_WRONG_GRAB_PAGE,
331 RA_STAT_FAILED_REACH_END,
336 atomic_t ra_cur_pages;
337 unsigned long ra_max_pages;
338 unsigned long ra_max_pages_per_file;
339 unsigned long ra_max_read_ahead_whole_pages;
342 /* ra_io_arg will be filled in the beginning of ll_readahead with
343 * ras_lock, then the following ll_read_ahead_pages will read RA
344 * pages according to this arg, all the items in this structure are
345 * counted by page index.
348 unsigned long ria_start; /* start offset of read-ahead*/
349 unsigned long ria_end; /* end offset of read-ahead*/
350 /* If stride read pattern is detected, ria_stoff means where
351 * stride read is started. Note: for normal read-ahead, the
352 * value here is meaningless, and also it will not be accessed*/
354 /* ria_length and ria_pages are the length and pages length in the
355 * stride I/O mode. And they will also be used to check whether
356 * it is stride I/O read-ahead in the read-ahead pages*/
357 unsigned long ria_length;
358 unsigned long ria_pages;
361 /* LL_HIST_MAX=32 causes an overflow */
362 #define LL_HIST_MAX 28
363 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
364 #define LL_PROCESS_HIST_MAX 10
365 struct per_process_info {
367 struct obd_histogram pp_r_hist;
368 struct obd_histogram pp_w_hist;
371 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
372 struct ll_rw_extents_info {
373 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
376 #define LL_OFFSET_HIST_MAX 100
377 struct ll_rw_process_info {
380 loff_t rw_range_start;
382 loff_t rw_last_file_pos;
384 size_t rw_smallest_extent;
385 size_t rw_largest_extent;
386 struct ll_file_data *rw_last_file;
389 enum stats_track_type {
390 STATS_TRACK_ALL = 0, /* track all processes */
391 STATS_TRACK_PID, /* track process with this pid */
392 STATS_TRACK_PPID, /* track processes with this ppid */
393 STATS_TRACK_GID, /* track processes with this gid */
397 /* flags for sbi->ll_flags */
398 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
399 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
400 #define LL_SBI_FLOCK 0x04
401 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
402 #define LL_SBI_ACL 0x10 /* support ACL */
403 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
404 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
405 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
406 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
407 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
408 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
409 /* LL_SBI_SOM_PREVIEW 0x1000 SOM preview mount option, obsolete */
410 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
411 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
412 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
413 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
414 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
415 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
416 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
417 #define LL_SBI_NOROOTSQUASH 0x100000 /* do not apply root squash */
419 #define LL_SBI_FLAGS { \
443 #define RCE_HASHES 32
445 struct rmtacl_ctl_entry {
446 struct list_head rce_list;
447 pid_t rce_key; /* hash key */
448 int rce_ops; /* acl operation type */
451 struct rmtacl_ctl_table {
453 struct list_head rct_entries[RCE_HASHES];
459 struct list_head ee_list;
460 pid_t ee_key; /* hash key */
461 struct lu_fid ee_fid;
462 int ee_type; /* ACL type for ACCESS or DEFAULT */
463 ext_acl_xattr_header *ee_acl;
468 struct list_head et_entries[EE_HASHES];
472 /* This is embedded into llite super-blocks to keep track of connect
473 * flags (capabilities) supported by all imports given mount is
475 struct lustre_client_ocd {
476 /* This is conjunction of connect_flags across all imports
477 * (LOVs) this mount is connected to. This field is updated by
478 * cl_ocd_update() under ->lco_lock. */
480 struct mutex lco_lock;
481 struct obd_export *lco_md_exp;
482 struct obd_export *lco_dt_exp;
486 struct list_head ll_list;
487 /* this protects pglist and ra_info. It isn't safe to
488 * grab from interrupt contexts */
490 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
491 spinlock_t ll_process_lock; /* ll_rw_process_info */
492 struct obd_uuid ll_sb_uuid;
493 struct obd_export *ll_md_exp;
494 struct obd_export *ll_dt_exp;
495 struct proc_dir_entry* ll_proc_root;
496 struct lu_fid ll_root_fid; /* root object fid */
499 unsigned int ll_umounting:1,
500 ll_xattr_cache_enabled:1,
501 ll_client_common_fill_super_succeeded:1;
503 /* per-conn chain of SBs */
504 struct list_head ll_conn_chain;
505 struct lustre_client_ocd ll_lco;
507 /*please don't ask -p*/
508 struct list_head ll_orphan_dentry_list;
510 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
512 /* Used to track "unstable" pages on a client, and maintain a
513 * LRU list of clean pages. An "unstable" page is defined as
514 * any page which is sent to a server as part of a bulk request,
515 * but is uncommitted to stable storage. */
516 struct cl_client_cache *ll_cache;
518 struct lprocfs_stats *ll_ra_stats;
520 struct ll_ra_info ll_ra_info;
521 unsigned int ll_namelen;
522 struct file_operations *ll_fop;
524 unsigned int ll_md_brw_pages; /* readdir pages per RPC */
526 struct lu_site *ll_site;
527 struct cl_device *ll_cl;
529 struct ll_rw_extents_info ll_rw_extents_info;
530 int ll_extent_process_count;
531 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
532 unsigned int ll_offset_process_count;
533 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
534 unsigned int ll_rw_offset_entry_count;
535 int ll_stats_track_id;
536 enum stats_track_type ll_stats_track_type;
539 /* metadata stat-ahead */
540 unsigned int ll_sa_max; /* max statahead RPCs */
541 atomic_t ll_sa_total; /* statahead thread started
543 atomic_t ll_sa_wrong; /* statahead thread stopped for
545 atomic_t ll_sa_running; /* running statahead thread
547 atomic_t ll_agl_total; /* AGL thread started count */
549 dev_t ll_sdev_orig; /* save s_dev before assign for
551 struct rmtacl_ctl_table ll_rct;
552 struct eacl_table ll_et;
555 struct root_squash_info ll_squash;
559 * per file-descriptor read-ahead data.
561 struct ll_readahead_state {
564 * index of the last page that read(2) needed and that wasn't in the
565 * cache. Used by ras_update() to detect seeks.
567 * XXX nikita: if access seeks into cached region, Lustre doesn't see
570 unsigned long ras_last_readpage;
572 * number of pages read after last read-ahead window reset. As window
573 * is reset on each seek, this is effectively a number of consecutive
574 * accesses. Maybe ->ras_accessed_in_window is better name.
576 * XXX nikita: window is also reset (by ras_update()) when Lustre
577 * believes that memory pressure evicts read-ahead pages. In that
578 * case, it probably doesn't make sense to expand window to
579 * PTLRPC_MAX_BRW_PAGES on the third access.
581 unsigned long ras_consecutive_pages;
583 * number of read requests after the last read-ahead window reset
584 * As window is reset on each seek, this is effectively the number
585 * on consecutive read request and is used to trigger read-ahead.
587 unsigned long ras_consecutive_requests;
589 * Parameters of current read-ahead window. Handled by
590 * ras_update(). On the initial access to the file or after a seek,
591 * window is reset to 0. After 3 consecutive accesses, window is
592 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
593 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
595 unsigned long ras_window_start, ras_window_len;
597 * Where next read-ahead should start at. This lies within read-ahead
598 * window. Read-ahead window is read in pieces rather than at once
599 * because: 1. lustre limits total number of pages under read-ahead by
600 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
601 * not covered by DLM lock.
603 unsigned long ras_next_readahead;
605 * Total number of ll_file_read requests issued, reads originating
606 * due to mmap are not counted in this total. This value is used to
607 * trigger full file read-ahead after multiple reads to a small file.
609 unsigned long ras_requests;
611 * Page index with respect to the current request, these value
612 * will not be accurate when dealing with reads issued via mmap.
614 unsigned long ras_request_index;
616 * The following 3 items are used for detecting the stride I/O
618 * In stride I/O mode,
619 * ...............|-----data-----|****gap*****|--------|******|....
620 * offset |-stride_pages-|-stride_gap-|
621 * ras_stride_offset = offset;
622 * ras_stride_length = stride_pages + stride_gap;
623 * ras_stride_pages = stride_pages;
624 * Note: all these three items are counted by pages.
626 unsigned long ras_stride_length;
627 unsigned long ras_stride_pages;
628 pgoff_t ras_stride_offset;
630 * number of consecutive stride request count, and it is similar as
631 * ras_consecutive_requests, but used for stride I/O mode.
632 * Note: only more than 2 consecutive stride request are detected,
633 * stride read-ahead will be enable
635 unsigned long ras_consecutive_stride_requests;
638 extern struct kmem_cache *ll_file_data_slab;
639 struct lustre_handle;
640 struct ll_file_data {
641 struct ll_readahead_state fd_ras;
642 struct ll_grouplock fd_grouplock;
646 /* openhandle if lease exists for this file.
647 * Borrow lli->lli_och_mutex to protect assignment */
648 struct obd_client_handle *fd_lease_och;
649 struct obd_client_handle *fd_och;
650 struct file *fd_file;
651 /* Indicate whether need to report failure when close.
652 * true: failure is known, not report again.
653 * false: unknown failure, should report. */
654 bool fd_write_failed;
655 rwlock_t fd_lock; /* protect lcc list */
656 struct list_head fd_lccs; /* list of ll_cl_context */
659 struct lov_stripe_md;
661 extern spinlock_t inode_lock;
663 extern struct proc_dir_entry *proc_lustre_fs_root;
665 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
667 return &lli->lli_vfs_inode;
670 __u32 ll_i2suppgid(struct inode *i);
671 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
673 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
675 #if BITS_PER_LONG == 32
677 #elif defined(CONFIG_COMPAT)
678 return unlikely(is_compat_task() || (sbi->ll_flags & LL_SBI_32BIT_API));
680 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
684 void ll_ras_enter(struct file *f);
686 /* llite/lcommon_misc.c */
687 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
688 enum obd_notify_event ev, void *owner, void *data);
689 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
690 struct ll_grouplock *lg);
691 void cl_put_grouplock(struct ll_grouplock *lg);
693 /* llite/lproc_llite.c */
694 #ifdef CONFIG_PROC_FS
695 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
696 struct super_block *sb, char *osc, char *mdc);
697 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
698 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
699 extern struct lprocfs_vars lprocfs_llite_obd_vars[];
701 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
702 struct super_block *sb, char *osc, char *mdc){return 0;}
703 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
704 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
709 LPROC_LL_DIRTY_MISSES,
711 LPROC_LL_WRITE_BYTES,
736 LPROC_LL_ALLOC_INODE,
739 LPROC_LL_GETXATTR_HITS,
741 LPROC_LL_REMOVEXATTR,
743 LPROC_LL_FILE_OPCODES
747 struct ll_dir_chain {
750 static inline void ll_dir_chain_init(struct ll_dir_chain *chain)
754 static inline void ll_dir_chain_fini(struct ll_dir_chain *chain)
758 extern const struct file_operations ll_dir_operations;
759 extern const struct inode_operations ll_dir_inode_operations;
760 #ifdef HAVE_DIR_CONTEXT
761 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
762 struct dir_context *ctx);
764 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
765 void *cookie, filldir_t filldir);
767 int ll_get_mdt_idx(struct inode *inode);
768 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
769 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
770 __u64 offset, struct ll_dir_chain *chain);
771 void ll_release_page(struct inode *inode, struct page *page, bool remove);
774 extern const struct inode_operations ll_special_inode_operations;
776 struct inode *ll_iget(struct super_block *sb, ino_t hash,
777 struct lustre_md *lic);
778 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
779 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
780 void *data, int flag);
781 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
782 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
783 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
786 int ll_writepage(struct page *page, struct writeback_control *wbc);
787 int ll_writepages(struct address_space *, struct writeback_control *wbc);
788 int ll_readpage(struct file *file, struct page *page);
789 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
790 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
791 struct ll_cl_context *ll_cl_find(struct file *file);
792 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io);
793 void ll_cl_remove(struct file *file, const struct lu_env *env);
795 #ifndef MS_HAS_NEW_AOPS
796 extern const struct address_space_operations ll_aops;
798 extern const struct address_space_operations_ext ll_aops;
802 extern struct file_operations ll_file_operations;
803 extern struct file_operations ll_file_operations_flock;
804 extern struct file_operations ll_file_operations_noflock;
805 extern struct inode_operations ll_file_inode_operations;
806 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
807 ldlm_mode_t l_req_mode);
808 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
809 struct lustre_handle *lockh, __u64 flags,
812 int ll_file_open(struct inode *inode, struct file *file);
813 int ll_file_release(struct inode *inode, struct file *file);
814 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
815 int ll_md_real_close(struct inode *inode, fmode_t fmode);
816 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
817 struct lustre_handle *fh);
818 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
819 struct ll_file_data *file, loff_t pos,
820 size_t count, int rw);
821 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
822 struct posix_acl *ll_get_acl(struct inode *inode, int type);
823 int ll_migrate(struct inode *parent, struct file *file, int mdtidx,
824 const char *name, int namelen);
825 int ll_get_fid_by_name(struct inode *parent, const char *name,
826 int namelen, struct lu_fid *fid);
827 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
828 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
830 # ifndef HAVE_INODE_PERMISION_2ARGS
831 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
833 int ll_inode_permission(struct inode *inode, int mask);
837 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
838 __u64 flags, struct lov_user_md *lum,
840 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
841 struct lov_mds_md **lmm, int *lmm_size,
842 struct ptlrpc_request **request);
843 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
845 int ll_dir_getstripe(struct inode *inode, void **lmmp,
846 int *lmm_size, struct ptlrpc_request **request,
848 #ifdef HAVE_FILE_FSYNC_4ARGS
849 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
850 #elif defined(HAVE_FILE_FSYNC_2ARGS)
851 int ll_fsync(struct file *file, int data);
853 int ll_fsync(struct file *file, struct dentry *dentry, int data);
855 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
856 int ll_fid2path(struct inode *inode, void __user *arg);
857 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
858 int ll_hsm_release(struct inode *inode);
862 int ll_d_init(struct dentry *de);
863 extern const struct dentry_operations ll_d_ops;
864 void ll_intent_drop_lock(struct lookup_intent *);
865 void ll_intent_release(struct lookup_intent *);
866 void ll_invalidate_aliases(struct inode *);
867 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
868 int ll_revalidate_it_finish(struct ptlrpc_request *request,
869 struct lookup_intent *it, struct dentry *de);
871 /* llite/llite_lib.c */
872 extern struct super_operations lustre_super_operations;
874 void ll_lli_init(struct ll_inode_info *lli);
875 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
876 void ll_put_super(struct super_block *sb);
877 void ll_kill_super(struct super_block *sb);
878 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
879 void ll_clear_inode(struct inode *inode);
880 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import);
881 int ll_setattr(struct dentry *de, struct iattr *attr);
882 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
883 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
884 __u64 max_age, __u32 flags);
885 int ll_update_inode(struct inode *inode, struct lustre_md *md);
886 int ll_read_inode2(struct inode *inode, void *opaque);
887 void ll_delete_inode(struct inode *inode);
888 int ll_iocontrol(struct inode *inode, struct file *file,
889 unsigned int cmd, unsigned long arg);
890 int ll_flush_ctx(struct inode *inode);
891 void ll_umount_begin(struct super_block *sb);
892 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
893 #ifdef HAVE_SUPEROPS_USE_DENTRY
894 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
896 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
898 void ll_dirty_page_discard_warn(struct page *page, int ioret);
899 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
900 struct super_block *, struct lookup_intent *);
901 void lustre_dump_dentry(struct dentry *, int recur);
902 int ll_obd_statfs(struct inode *inode, void __user *arg);
903 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
904 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
905 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
906 int ll_process_config(struct lustre_cfg *lcfg);
909 LUSTRE_OPC_MKDIR = 0,
910 LUSTRE_OPC_SYMLINK = 1,
911 LUSTRE_OPC_MKNOD = 2,
912 LUSTRE_OPC_CREATE = 3,
916 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
917 struct inode *i1, struct inode *i2,
918 const char *name, size_t namelen,
919 __u32 mode, __u32 opc, void *data);
920 void ll_finish_md_op_data(struct md_op_data *op_data);
921 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
922 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
923 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
924 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
925 struct lov_user_md **kbuf);
926 void ll_open_cleanup(struct super_block *sb, struct ptlrpc_request *open_req);
928 /* Compute expected user md size when passing in a md from user space */
929 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
931 switch (lum->lmm_magic) {
932 case LOV_USER_MAGIC_V1:
933 return sizeof(struct lov_user_md_v1);
934 case LOV_USER_MAGIC_V3:
935 return sizeof(struct lov_user_md_v3);
936 case LOV_USER_MAGIC_SPECIFIC:
937 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
940 return lov_user_md_size(lum->lmm_stripe_count,
941 LOV_USER_MAGIC_SPECIFIC);
947 /* llite/llite_nfs.c */
948 extern struct export_operations lustre_export_operations;
949 __u32 get_uuid2int(const char *name, int len);
950 struct inode *search_inode_for_lustre(struct super_block *sb,
951 const struct lu_fid *fid);
952 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
954 /* llite/symlink.c */
955 extern struct inode_operations ll_fast_symlink_inode_operations;
958 * IO arguments for various VFS I/O interfaces.
961 /** normal/sendfile/splice */
962 enum vvp_io_subtype via_io_subtype;
966 struct kiocb *via_iocb;
967 struct iovec *via_iov;
968 unsigned long via_nrsegs;
971 struct pipe_inode_info *via_pipe;
972 unsigned int via_flags;
977 struct ll_cl_context {
978 struct list_head lcc_list;
980 const struct lu_env *lcc_env;
981 struct cl_io *lcc_io;
982 struct cl_page *lcc_page;
985 struct ll_thread_info {
986 struct iovec lti_local_iov;
987 struct vvp_io_args lti_args;
988 struct ra_io_arg lti_ria;
989 struct kiocb lti_kiocb;
990 struct ll_cl_context lti_io_ctx;
993 extern struct lu_context_key ll_thread_key;
995 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
997 struct ll_thread_info *lti;
999 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
1000 LASSERT(lti != NULL);
1005 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env,
1006 enum vvp_io_subtype type)
1008 struct vvp_io_args *via = &ll_env_info(env)->lti_args;
1010 via->via_io_subtype = type;
1015 /* llite/llite_mmap.c */
1017 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1018 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1019 void policy_from_vma(ldlm_policy_data_t *policy,
1020 struct vm_area_struct *vma, unsigned long addr, size_t count);
1021 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1024 static inline void ll_invalidate_page(struct page *vmpage)
1026 struct address_space *mapping = vmpage->mapping;
1027 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1029 LASSERT(PageLocked(vmpage));
1030 if (mapping == NULL)
1034 * truncate_complete_page() calls
1035 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
1037 ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE);
1038 truncate_complete_page(mapping, vmpage);
1041 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1043 /* don't need an addref as the sb_info should be holding one */
1044 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1046 return ll_s2sbi(sb)->ll_dt_exp;
1049 /* don't need an addref as the sb_info should be holding one */
1050 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1052 return ll_s2sbi(sb)->ll_md_exp;
1055 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1057 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1063 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1064 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1066 return ll_s2sbi(inode->i_sb);
1069 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1071 return ll_s2dtexp(inode->i_sb);
1074 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1076 return ll_s2mdexp(inode->i_sb);
1079 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1083 LASSERT(inode != NULL);
1084 fid = &ll_i2info(inode)->lli_fid;
1089 static inline loff_t ll_file_maxbytes(struct inode *inode)
1091 struct cl_object *obj = ll_i2info(inode)->lli_clob;
1094 return MAX_LFS_FILESIZE;
1096 return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
1100 int ll_setxattr(struct dentry *dentry, const char *name,
1101 const void *value, size_t size, int flags);
1102 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1103 void *buffer, size_t size);
1104 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1105 int ll_removexattr(struct dentry *dentry, const char *name);
1107 /* llite/remote_perm.c */
1108 extern struct kmem_cache *ll_remote_perm_cachep;
1109 extern struct kmem_cache *ll_rmtperm_hash_cachep;
1111 void free_rmtperm_hash(struct hlist_head *hash);
1112 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1113 int lustre_check_remote_perm(struct inode *inode, int mask);
1115 /* llite/llite_capa.c */
1116 extern struct timer_list ll_capa_timer;
1118 int ll_capa_thread_start(void);
1119 void ll_capa_thread_stop(void);
1120 void ll_capa_timer_callback(unsigned long unused);
1122 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1124 void ll_capa_open(struct inode *inode);
1125 void ll_capa_close(struct inode *inode);
1127 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1128 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1130 void ll_truncate_free_capa(struct obd_capa *ocapa);
1131 void ll_clear_inode_capas(struct inode *inode);
1132 void ll_print_capa_stat(struct ll_sb_info *sbi);
1135 * Common IO arguments for various VFS I/O interfaces.
1137 int cl_sb_init(struct super_block *sb);
1138 int cl_sb_fini(struct super_block *sb);
1140 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1141 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1143 /* llite/llite_rmtacl.c */
1144 #ifdef CONFIG_FS_POSIX_ACL
1145 u64 rce_ops2valid(int ops);
1146 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1147 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1148 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1149 void rct_init(struct rmtacl_ctl_table *rct);
1150 void rct_fini(struct rmtacl_ctl_table *rct);
1152 void ee_free(struct eacl_entry *ee);
1153 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1154 ext_acl_xattr_header *header);
1155 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1156 struct lu_fid *fid, int type);
1157 void et_search_free(struct eacl_table *et, pid_t key);
1158 void et_init(struct eacl_table *et);
1159 void et_fini(struct eacl_table *et);
1164 #define LL_SA_RPC_MIN 2
1165 #define LL_SA_RPC_DEF 32
1166 #define LL_SA_RPC_MAX 8192
1168 #define LL_SA_CACHE_BIT 5
1169 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1170 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1172 /* per inode struct, for dir only */
1173 struct ll_statahead_info {
1174 struct dentry *sai_dentry;
1175 atomic_t sai_refcount; /* when access this struct, hold
1177 unsigned int sai_max; /* max ahead of lookup */
1178 __u64 sai_sent; /* stat requests sent count */
1179 __u64 sai_replied; /* stat requests which received
1181 __u64 sai_index; /* index of statahead entry */
1182 __u64 sai_index_wait; /* index of entry which is the
1183 * caller is waiting for */
1184 __u64 sai_hit; /* hit count */
1185 __u64 sai_miss; /* miss count:
1186 * for "ls -al" case, includes
1187 * hidden dentry miss;
1188 * for "ls -l" case, it does not
1189 * include hidden dentry miss.
1190 * "sai_miss_hidden" is used for
1193 unsigned int sai_consecutive_miss; /* consecutive miss */
1194 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1195 * is not a hidden one */
1196 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1198 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1200 sai_agl_valid:1,/* AGL is valid for the dir */
1201 sai_in_readpage:1;/* statahead is in readdir()*/
1202 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1203 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1204 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1205 struct list_head sai_interim_entries; /* entries which got async
1206 * stat reply, but not
1208 struct list_head sai_entries; /* completed entries */
1209 struct list_head sai_agls; /* AGLs to be sent */
1210 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1211 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1212 atomic_t sai_cache_count; /* entry count in cache */
1215 int ll_statahead(struct inode *dir, struct dentry **dentry, bool unplug);
1216 void ll_authorize_statahead(struct inode *dir, void *key);
1217 void ll_deauthorize_statahead(struct inode *dir, void *key);
1220 blkcnt_t dirty_cnt(struct inode *inode);
1222 int cl_glimpse_size0(struct inode *inode, int agl);
1223 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1224 struct inode *inode, struct cl_object *clob, int agl);
1226 static inline int cl_glimpse_size(struct inode *inode)
1228 return cl_glimpse_size0(inode, 0);
1231 static inline int cl_agl(struct inode *inode)
1233 return cl_glimpse_size0(inode, 1);
1236 static inline int ll_glimpse_size(struct inode *inode)
1238 struct ll_inode_info *lli = ll_i2info(inode);
1241 down_read(&lli->lli_glimpse_sem);
1242 rc = cl_glimpse_size(inode);
1243 lli->lli_glimpse_time = cfs_time_current();
1244 up_read(&lli->lli_glimpse_sem);
1248 /* dentry may statahead when statahead is enabled and current process has opened
1249 * parent directory, and this dentry hasn't accessed statahead cache before */
1251 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1253 struct ll_inode_info *lli;
1254 struct ll_dentry_data *ldd;
1256 if (ll_i2sbi(dir)->ll_sa_max == 0)
1259 lli = ll_i2info(dir);
1261 /* statahead is not allowed for this dir, there may be three causes:
1262 * 1. dir is not opened.
1263 * 2. statahead hit ratio is too low.
1264 * 3. previous stat started statahead thread failed. */
1265 if (!lli->lli_sa_enabled)
1268 /* not the same process, don't statahead */
1269 if (lli->lli_opendir_pid != current_pid())
1273 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1274 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1275 * For patchless client, lookup intent is not accurate, which may
1276 * misguide statahead. For example:
1277 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1278 * have the same intent -- IT_GETATTR, while one dentry should access
1279 * statahead cache once, otherwise statahead windows is messed up.
1280 * The solution is as following:
1281 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1282 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1283 * bypass interacting with statahead cache by checking
1284 * 'lld_sa_generation == lli->lli_sa_generation'.
1286 ldd = ll_d2d(dentry);
1287 if (ldd != NULL && ldd->lld_sa_generation == lli->lli_sa_generation)
1293 /* llite ioctl register support rountine */
1299 #define LLIOC_MAX_CMD 256
1302 * Rules to write a callback function:
1305 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1306 * returned to ll_iocontrol_register. Callback functions should use this
1307 * data to check the potential collasion of ioctl cmd. If collasion is
1308 * found, callback function should return LLIOC_CONT.
1309 * @rcp: The result of ioctl command.
1312 * If @magic matches the pointer returned by ll_iocontrol_data, the
1313 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1315 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1316 struct file *file, unsigned int cmd, unsigned long arg,
1317 void *magic, int *rcp);
1319 /* export functions */
1320 /* Register ioctl block dynamatically for a regular file.
1322 * @cmd: the array of ioctl command set
1323 * @count: number of commands in the @cmd
1324 * @cb: callback function, it will be called if an ioctl command is found to
1325 * belong to the command list @cmd.
1328 * A magic pointer will be returned if success;
1329 * otherwise, NULL will be returned.
1331 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1332 void ll_iocontrol_unregister(void *magic);
1334 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1336 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1337 enum cl_fsync_mode mode, int ignore_layout);
1339 /** direct write pages */
1340 struct ll_dio_pages {
1341 /** page array to be written. we don't support
1342 * partial pages except the last one. */
1343 struct page **ldp_pages;
1344 /* offset of each page */
1345 loff_t *ldp_offsets;
1346 /** if ldp_offsets is NULL, it means a sequential
1347 * pages to be written, then this is the file offset
1348 * of the * first page. */
1349 loff_t ldp_start_offset;
1350 /** how many bytes are to be written. */
1352 /** # of pages in the array. */
1356 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1359 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1362 ll_stats_ops_tally(ll_s2sbi(cl2vvp_dev(dev)->vdv_sb), opc, rc);
1365 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1366 int rw, struct inode *inode,
1367 struct ll_dio_pages *pv);
1369 static inline int ll_file_nolock(const struct file *file)
1371 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1372 struct inode *inode = file->f_dentry->d_inode;
1374 LASSERT(fd != NULL);
1375 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1376 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1379 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1380 struct lookup_intent *it, __u64 *bits)
1382 if (!it->d.lustre.it_lock_set) {
1383 struct lustre_handle handle;
1385 /* If this inode is a remote object, it will get two
1386 * separate locks in different namespaces, Master MDT,
1387 * where the name entry is, will grant LOOKUP lock,
1388 * remote MDT, where the object is, will grant
1389 * UPDATE|PERM lock. The inode will be attched to both
1390 * LOOKUP and PERM locks, so revoking either locks will
1391 * case the dcache being cleared */
1392 if (it->d.lustre.it_remote_lock_mode) {
1393 handle.cookie = it->d.lustre.it_remote_lock_handle;
1394 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1395 "(%p) for remote lock "LPX64"\n",
1396 PFID(ll_inode2fid(inode)), inode,
1398 md_set_lock_data(exp, &handle.cookie, inode, NULL);
1401 handle.cookie = it->d.lustre.it_lock_handle;
1403 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1404 " for lock "LPX64"\n",
1405 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1407 md_set_lock_data(exp, &handle.cookie, inode,
1408 &it->d.lustre.it_lock_bits);
1409 it->d.lustre.it_lock_set = 1;
1413 *bits = it->d.lustre.it_lock_bits;
1416 static inline void ll_lock_dcache(struct inode *inode)
1418 #ifdef HAVE_DCACHE_LOCK
1419 spin_lock(&dcache_lock);
1421 spin_lock(&inode->i_lock);
1425 static inline void ll_unlock_dcache(struct inode *inode)
1427 #ifdef HAVE_DCACHE_LOCK
1428 spin_unlock(&dcache_lock);
1430 spin_unlock(&inode->i_lock);
1434 static inline int d_lustre_invalid(const struct dentry *dentry)
1436 struct ll_dentry_data *lld = ll_d2d(dentry);
1438 return (lld == NULL) || lld->lld_invalid;
1441 static inline void __d_lustre_invalidate(struct dentry *dentry)
1443 struct ll_dentry_data *lld = ll_d2d(dentry);
1446 lld->lld_invalid = 1;
1450 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1451 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1452 * else dput() of the last refcount will unhash this dentry and kill it.
1454 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1456 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1457 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1458 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1460 spin_lock_nested(&dentry->d_lock,
1461 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1462 __d_lustre_invalidate(dentry);
1463 if (ll_d_count(dentry) == 0)
1465 spin_unlock(&dentry->d_lock);
1468 static inline void d_lustre_revalidate(struct dentry *dentry)
1470 spin_lock(&dentry->d_lock);
1471 LASSERT(ll_d2d(dentry) != NULL);
1472 ll_d2d(dentry)->lld_invalid = 0;
1473 spin_unlock(&dentry->d_lock);
1476 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1477 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1478 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1480 int ll_xattr_init(void);
1481 void ll_xattr_fini(void);
1483 int ll_getxattr_common(struct inode *inode, const char *name,
1484 void *buffer, size_t size, __u64 valid);
1485 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1486 struct cl_page *page, enum cl_req_type crt);
1488 int ll_getparent(struct file *file, struct getparent __user *arg);
1491 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1492 unsigned int attr_flags, struct obd_capa *capa);
1494 extern struct lu_env *cl_inode_fini_env;
1495 extern int cl_inode_fini_refcheck;
1497 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1498 void cl_inode_fini(struct inode *inode);
1500 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1501 u32 cl_fid_build_gen(const struct lu_fid *fid);
1503 #endif /* LLITE_INTERNAL_H */