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;
217 * struct rw_semaphore {
218 * signed long count; // align d.d_def_acl
219 * spinlock_t wait_lock; // align d.d_sa_lock
220 * struct list_head wait_list;
223 struct rw_semaphore lli_trunc_sem;
224 struct range_lock_tree lli_write_tree;
226 struct rw_semaphore lli_glimpse_sem;
227 cfs_time_t lli_glimpse_time;
228 struct list_head lli_agl_list;
231 /* for writepage() only to communicate to fsync */
235 * whenever a process try to read/write the file, the
236 * jobid of the process will be saved here, and it'll
237 * be packed into the write PRC when flush later.
239 * so the read/write statistics for jobid will not be
240 * accurate if the file is shared by different jobs.
242 char lli_jobid[LUSTRE_JOBID_SIZE];
246 /* XXX: For following frequent used members, although they maybe special
247 * used for non-directory object, it is some time-wasting to check
248 * whether the object is directory or not before using them. On the
249 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
250 * the "ll_inode_info" size even if moving those members into u.f.
251 * So keep them out side.
253 * In the future, if more members are added only for directory,
254 * some of the following members can be moved into u.f.
256 struct cl_object *lli_clob;
258 /* mutex to request for layout lock exclusively. */
259 struct mutex lli_layout_mutex;
260 /* Layout version, protected by lli_layout_lock */
261 __u32 lli_layout_gen;
262 spinlock_t lli_layout_lock;
264 struct rw_semaphore lli_xattrs_list_rwsem;
265 struct mutex lli_xattrs_enq_lock;
266 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
269 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
273 spin_lock(&lli->lli_layout_lock);
274 gen = lli->lli_layout_gen;
275 spin_unlock(&lli->lli_layout_lock);
280 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
282 spin_lock(&lli->lli_layout_lock);
283 lli->lli_layout_gen = gen;
284 spin_unlock(&lli->lli_layout_lock);
287 int ll_xattr_cache_destroy(struct inode *inode);
289 int ll_xattr_cache_get(struct inode *inode,
296 * Locking to guarantee consistency of non-atomic updates to long long i_size,
297 * consistency between file size and KMS.
299 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
302 void ll_inode_size_lock(struct inode *inode);
303 void ll_inode_size_unlock(struct inode *inode);
305 // FIXME: replace the name of this with LL_I to conform to kernel stuff
306 // static inline struct ll_inode_info *LL_I(struct inode *inode)
307 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
309 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
312 /* default to about 40meg of readahead on a given system. That much tied
313 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
314 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT))
316 /* default to read-ahead full files smaller than 2MB on the second read */
317 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
322 RA_STAT_DISTANT_READPAGE,
323 RA_STAT_MISS_IN_WINDOW,
324 RA_STAT_FAILED_GRAB_PAGE,
325 RA_STAT_FAILED_MATCH,
330 RA_STAT_MAX_IN_FLIGHT,
331 RA_STAT_WRONG_GRAB_PAGE,
332 RA_STAT_FAILED_REACH_END,
337 atomic_t ra_cur_pages;
338 unsigned long ra_max_pages;
339 unsigned long ra_max_pages_per_file;
340 unsigned long ra_max_read_ahead_whole_pages;
343 /* ra_io_arg will be filled in the beginning of ll_readahead with
344 * ras_lock, then the following ll_read_ahead_pages will read RA
345 * pages according to this arg, all the items in this structure are
346 * counted by page index.
349 unsigned long ria_start; /* start offset of read-ahead*/
350 unsigned long ria_end; /* end offset of read-ahead*/
351 /* If stride read pattern is detected, ria_stoff means where
352 * stride read is started. Note: for normal read-ahead, the
353 * value here is meaningless, and also it will not be accessed*/
355 /* ria_length and ria_pages are the length and pages length in the
356 * stride I/O mode. And they will also be used to check whether
357 * it is stride I/O read-ahead in the read-ahead pages*/
358 unsigned long ria_length;
359 unsigned long ria_pages;
362 /* LL_HIST_MAX=32 causes an overflow */
363 #define LL_HIST_MAX 28
364 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
365 #define LL_PROCESS_HIST_MAX 10
366 struct per_process_info {
368 struct obd_histogram pp_r_hist;
369 struct obd_histogram pp_w_hist;
372 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
373 struct ll_rw_extents_info {
374 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
377 #define LL_OFFSET_HIST_MAX 100
378 struct ll_rw_process_info {
381 loff_t rw_range_start;
383 loff_t rw_last_file_pos;
385 size_t rw_smallest_extent;
386 size_t rw_largest_extent;
387 struct ll_file_data *rw_last_file;
390 enum stats_track_type {
391 STATS_TRACK_ALL = 0, /* track all processes */
392 STATS_TRACK_PID, /* track process with this pid */
393 STATS_TRACK_PPID, /* track processes with this ppid */
394 STATS_TRACK_GID, /* track processes with this gid */
398 /* flags for sbi->ll_flags */
399 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
400 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
401 #define LL_SBI_FLOCK 0x04
402 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
403 #define LL_SBI_ACL 0x10 /* support ACL */
404 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
405 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
406 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
407 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
408 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
409 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
410 /* LL_SBI_SOM_PREVIEW 0x1000 SOM preview mount option, obsolete */
411 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
412 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
413 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
414 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
415 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
416 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
417 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
418 #define LL_SBI_NOROOTSQUASH 0x100000 /* do not apply root squash */
420 #define LL_SBI_FLAGS { \
444 #define RCE_HASHES 32
446 struct rmtacl_ctl_entry {
447 struct list_head rce_list;
448 pid_t rce_key; /* hash key */
449 int rce_ops; /* acl operation type */
452 struct rmtacl_ctl_table {
454 struct list_head rct_entries[RCE_HASHES];
460 struct list_head ee_list;
461 pid_t ee_key; /* hash key */
462 struct lu_fid ee_fid;
463 int ee_type; /* ACL type for ACCESS or DEFAULT */
464 ext_acl_xattr_header *ee_acl;
469 struct list_head et_entries[EE_HASHES];
473 /* This is embedded into llite super-blocks to keep track of connect
474 * flags (capabilities) supported by all imports given mount is
476 struct lustre_client_ocd {
477 /* This is conjunction of connect_flags across all imports
478 * (LOVs) this mount is connected to. This field is updated by
479 * cl_ocd_update() under ->lco_lock. */
481 struct mutex lco_lock;
482 struct obd_export *lco_md_exp;
483 struct obd_export *lco_dt_exp;
487 struct list_head ll_list;
488 /* this protects pglist and ra_info. It isn't safe to
489 * grab from interrupt contexts */
491 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
492 spinlock_t ll_process_lock; /* ll_rw_process_info */
493 struct obd_uuid ll_sb_uuid;
494 struct obd_export *ll_md_exp;
495 struct obd_export *ll_dt_exp;
496 struct proc_dir_entry* ll_proc_root;
497 struct lu_fid ll_root_fid; /* root object fid */
500 unsigned int ll_umounting:1,
501 ll_xattr_cache_enabled:1,
502 ll_client_common_fill_super_succeeded:1;
504 /* per-conn chain of SBs */
505 struct list_head ll_conn_chain;
506 struct lustre_client_ocd ll_lco;
508 /*please don't ask -p*/
509 struct list_head ll_orphan_dentry_list;
511 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
513 /* Used to track "unstable" pages on a client, and maintain a
514 * LRU list of clean pages. An "unstable" page is defined as
515 * any page which is sent to a server as part of a bulk request,
516 * but is uncommitted to stable storage. */
517 struct cl_client_cache *ll_cache;
519 struct lprocfs_stats *ll_ra_stats;
521 struct ll_ra_info ll_ra_info;
522 unsigned int ll_namelen;
523 struct file_operations *ll_fop;
525 /* =0 - hold lock over whole read/write
526 * >0 - max. chunk to be read/written w/o lock re-acquiring */
527 unsigned long ll_max_rw_chunk;
528 unsigned int ll_md_brw_pages; /* readdir pages per RPC */
530 struct lu_site *ll_site;
531 struct cl_device *ll_cl;
533 struct ll_rw_extents_info ll_rw_extents_info;
534 int ll_extent_process_count;
535 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
536 unsigned int ll_offset_process_count;
537 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
538 unsigned int ll_rw_offset_entry_count;
539 int ll_stats_track_id;
540 enum stats_track_type ll_stats_track_type;
543 /* metadata stat-ahead */
544 unsigned int ll_sa_max; /* max statahead RPCs */
545 atomic_t ll_sa_total; /* statahead thread started
547 atomic_t ll_sa_wrong; /* statahead thread stopped for
549 atomic_t ll_sa_running; /* running statahead thread
551 atomic_t ll_agl_total; /* AGL thread started count */
553 dev_t ll_sdev_orig; /* save s_dev before assign for
555 struct rmtacl_ctl_table ll_rct;
556 struct eacl_table ll_et;
559 struct root_squash_info ll_squash;
562 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
565 * per file-descriptor read-ahead data.
567 struct ll_readahead_state {
570 * index of the last page that read(2) needed and that wasn't in the
571 * cache. Used by ras_update() to detect seeks.
573 * XXX nikita: if access seeks into cached region, Lustre doesn't see
576 unsigned long ras_last_readpage;
578 * number of pages read after last read-ahead window reset. As window
579 * is reset on each seek, this is effectively a number of consecutive
580 * accesses. Maybe ->ras_accessed_in_window is better name.
582 * XXX nikita: window is also reset (by ras_update()) when Lustre
583 * believes that memory pressure evicts read-ahead pages. In that
584 * case, it probably doesn't make sense to expand window to
585 * PTLRPC_MAX_BRW_PAGES on the third access.
587 unsigned long ras_consecutive_pages;
589 * number of read requests after the last read-ahead window reset
590 * As window is reset on each seek, this is effectively the number
591 * on consecutive read request and is used to trigger read-ahead.
593 unsigned long ras_consecutive_requests;
595 * Parameters of current read-ahead window. Handled by
596 * ras_update(). On the initial access to the file or after a seek,
597 * window is reset to 0. After 3 consecutive accesses, window is
598 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
599 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
601 unsigned long ras_window_start, ras_window_len;
603 * Where next read-ahead should start at. This lies within read-ahead
604 * window. Read-ahead window is read in pieces rather than at once
605 * because: 1. lustre limits total number of pages under read-ahead by
606 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
607 * not covered by DLM lock.
609 unsigned long ras_next_readahead;
611 * Total number of ll_file_read requests issued, reads originating
612 * due to mmap are not counted in this total. This value is used to
613 * trigger full file read-ahead after multiple reads to a small file.
615 unsigned long ras_requests;
617 * Page index with respect to the current request, these value
618 * will not be accurate when dealing with reads issued via mmap.
620 unsigned long ras_request_index;
622 * The following 3 items are used for detecting the stride I/O
624 * In stride I/O mode,
625 * ...............|-----data-----|****gap*****|--------|******|....
626 * offset |-stride_pages-|-stride_gap-|
627 * ras_stride_offset = offset;
628 * ras_stride_length = stride_pages + stride_gap;
629 * ras_stride_pages = stride_pages;
630 * Note: all these three items are counted by pages.
632 unsigned long ras_stride_length;
633 unsigned long ras_stride_pages;
634 pgoff_t ras_stride_offset;
636 * number of consecutive stride request count, and it is similar as
637 * ras_consecutive_requests, but used for stride I/O mode.
638 * Note: only more than 2 consecutive stride request are detected,
639 * stride read-ahead will be enable
641 unsigned long ras_consecutive_stride_requests;
644 extern struct kmem_cache *ll_file_data_slab;
645 struct lustre_handle;
646 struct ll_file_data {
647 struct ll_readahead_state fd_ras;
648 struct ll_grouplock fd_grouplock;
652 /* openhandle if lease exists for this file.
653 * Borrow lli->lli_och_mutex to protect assignment */
654 struct obd_client_handle *fd_lease_och;
655 struct obd_client_handle *fd_och;
656 struct file *fd_file;
657 /* Indicate whether need to report failure when close.
658 * true: failure is known, not report again.
659 * false: unknown failure, should report. */
660 bool fd_write_failed;
661 rwlock_t fd_lock; /* protect lcc list */
662 struct list_head fd_lccs; /* list of ll_cl_context */
665 struct lov_stripe_md;
667 extern spinlock_t inode_lock;
669 extern struct proc_dir_entry *proc_lustre_fs_root;
671 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
673 return &lli->lli_vfs_inode;
676 __u32 ll_i2suppgid(struct inode *i);
677 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
679 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
681 #if BITS_PER_LONG == 32
683 #elif defined(CONFIG_COMPAT)
684 return unlikely(is_compat_task() || (sbi->ll_flags & LL_SBI_32BIT_API));
686 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
690 void ll_ras_enter(struct file *f);
692 /* llite/lcommon_misc.c */
693 int cl_ocd_update(struct obd_device *host, struct obd_device *watched,
694 enum obd_notify_event ev, void *owner, void *data);
695 int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
696 struct ll_grouplock *lg);
697 void cl_put_grouplock(struct ll_grouplock *lg);
699 /* llite/lproc_llite.c */
700 #ifdef CONFIG_PROC_FS
701 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
702 struct super_block *sb, char *osc, char *mdc);
703 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
704 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
705 extern struct lprocfs_vars lprocfs_llite_obd_vars[];
707 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
708 struct super_block *sb, char *osc, char *mdc){return 0;}
709 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
710 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
715 LPROC_LL_DIRTY_MISSES,
717 LPROC_LL_WRITE_BYTES,
742 LPROC_LL_ALLOC_INODE,
745 LPROC_LL_GETXATTR_HITS,
747 LPROC_LL_REMOVEXATTR,
749 LPROC_LL_FILE_OPCODES
753 struct ll_dir_chain {
756 static inline void ll_dir_chain_init(struct ll_dir_chain *chain)
760 static inline void ll_dir_chain_fini(struct ll_dir_chain *chain)
764 extern const struct file_operations ll_dir_operations;
765 extern const struct inode_operations ll_dir_inode_operations;
766 #ifdef HAVE_DIR_CONTEXT
767 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
768 struct dir_context *ctx);
770 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
771 void *cookie, filldir_t filldir);
773 int ll_get_mdt_idx(struct inode *inode);
774 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
775 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
776 __u64 offset, struct ll_dir_chain *chain);
777 void ll_release_page(struct inode *inode, struct page *page, bool remove);
780 extern const struct inode_operations ll_special_inode_operations;
782 struct inode *ll_iget(struct super_block *sb, ino_t hash,
783 struct lustre_md *lic);
784 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
785 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
786 void *data, int flag);
787 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
788 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
789 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
792 int ll_writepage(struct page *page, struct writeback_control *wbc);
793 int ll_writepages(struct address_space *, struct writeback_control *wbc);
794 int ll_readpage(struct file *file, struct page *page);
795 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
796 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
797 struct ll_cl_context *ll_cl_find(struct file *file);
798 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io);
799 void ll_cl_remove(struct file *file, const struct lu_env *env);
801 #ifndef MS_HAS_NEW_AOPS
802 extern const struct address_space_operations ll_aops;
804 extern const struct address_space_operations_ext ll_aops;
808 extern struct file_operations ll_file_operations;
809 extern struct file_operations ll_file_operations_flock;
810 extern struct file_operations ll_file_operations_noflock;
811 extern struct inode_operations ll_file_inode_operations;
812 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
813 ldlm_mode_t l_req_mode);
814 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
815 struct lustre_handle *lockh, __u64 flags,
818 int ll_file_open(struct inode *inode, struct file *file);
819 int ll_file_release(struct inode *inode, struct file *file);
820 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
821 int ll_md_real_close(struct inode *inode, fmode_t fmode);
822 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
823 struct lustre_handle *fh);
824 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
825 struct ll_file_data *file, loff_t pos,
826 size_t count, int rw);
827 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
828 struct posix_acl *ll_get_acl(struct inode *inode, int type);
829 int ll_migrate(struct inode *parent, struct file *file, int mdtidx,
830 const char *name, int namelen);
831 int ll_get_fid_by_name(struct inode *parent, const char *name,
832 int namelen, struct lu_fid *fid);
833 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
834 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
836 # ifndef HAVE_INODE_PERMISION_2ARGS
837 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
839 int ll_inode_permission(struct inode *inode, int mask);
843 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
844 __u64 flags, struct lov_user_md *lum,
846 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
847 struct lov_mds_md **lmm, int *lmm_size,
848 struct ptlrpc_request **request);
849 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
851 int ll_dir_getstripe(struct inode *inode, void **lmmp,
852 int *lmm_size, struct ptlrpc_request **request,
854 #ifdef HAVE_FILE_FSYNC_4ARGS
855 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
856 #elif defined(HAVE_FILE_FSYNC_2ARGS)
857 int ll_fsync(struct file *file, int data);
859 int ll_fsync(struct file *file, struct dentry *dentry, int data);
861 int ll_merge_attr(const struct lu_env *env, struct inode *inode);
862 int ll_fid2path(struct inode *inode, void __user *arg);
863 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
864 int ll_hsm_release(struct inode *inode);
868 int ll_d_init(struct dentry *de);
869 extern const struct dentry_operations ll_d_ops;
870 void ll_intent_drop_lock(struct lookup_intent *);
871 void ll_intent_release(struct lookup_intent *);
872 void ll_invalidate_aliases(struct inode *);
873 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
874 int ll_revalidate_it_finish(struct ptlrpc_request *request,
875 struct lookup_intent *it, struct dentry *de);
877 /* llite/llite_lib.c */
878 extern struct super_operations lustre_super_operations;
880 void ll_lli_init(struct ll_inode_info *lli);
881 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
882 void ll_put_super(struct super_block *sb);
883 void ll_kill_super(struct super_block *sb);
884 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
885 void ll_clear_inode(struct inode *inode);
886 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import);
887 int ll_setattr(struct dentry *de, struct iattr *attr);
888 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
889 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
890 __u64 max_age, __u32 flags);
891 int ll_update_inode(struct inode *inode, struct lustre_md *md);
892 int ll_read_inode2(struct inode *inode, void *opaque);
893 void ll_delete_inode(struct inode *inode);
894 int ll_iocontrol(struct inode *inode, struct file *file,
895 unsigned int cmd, unsigned long arg);
896 int ll_flush_ctx(struct inode *inode);
897 void ll_umount_begin(struct super_block *sb);
898 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
899 #ifdef HAVE_SUPEROPS_USE_DENTRY
900 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
902 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
904 void ll_dirty_page_discard_warn(struct page *page, int ioret);
905 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
906 struct super_block *, struct lookup_intent *);
907 void lustre_dump_dentry(struct dentry *, int recur);
908 int ll_obd_statfs(struct inode *inode, void __user *arg);
909 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
910 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
911 int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
912 int ll_get_max_cookiesize(struct ll_sb_info *sbi, int *max_cookiesize);
913 int ll_get_default_cookiesize(struct ll_sb_info *sbi, int *default_cookiesize);
914 int ll_process_config(struct lustre_cfg *lcfg);
917 LUSTRE_OPC_MKDIR = 0,
918 LUSTRE_OPC_SYMLINK = 1,
919 LUSTRE_OPC_MKNOD = 2,
920 LUSTRE_OPC_CREATE = 3,
924 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
925 struct inode *i1, struct inode *i2,
926 const char *name, size_t namelen,
927 __u32 mode, __u32 opc, void *data);
928 void ll_finish_md_op_data(struct md_op_data *op_data);
929 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
930 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
931 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
932 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
933 struct lov_user_md **kbuf);
934 void ll_open_cleanup(struct super_block *sb, struct ptlrpc_request *open_req);
936 /* Compute expected user md size when passing in a md from user space */
937 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
939 switch (lum->lmm_magic) {
940 case LOV_USER_MAGIC_V1:
941 return sizeof(struct lov_user_md_v1);
942 case LOV_USER_MAGIC_V3:
943 return sizeof(struct lov_user_md_v3);
944 case LOV_USER_MAGIC_SPECIFIC:
945 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
948 return lov_user_md_size(lum->lmm_stripe_count,
949 LOV_USER_MAGIC_SPECIFIC);
955 /* llite/llite_nfs.c */
956 extern struct export_operations lustre_export_operations;
957 __u32 get_uuid2int(const char *name, int len);
958 struct inode *search_inode_for_lustre(struct super_block *sb,
959 const struct lu_fid *fid);
960 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
962 /* llite/symlink.c */
963 extern struct inode_operations ll_fast_symlink_inode_operations;
966 * IO arguments for various VFS I/O interfaces.
969 /** normal/sendfile/splice */
970 enum vvp_io_subtype via_io_subtype;
974 struct kiocb *via_iocb;
975 struct iovec *via_iov;
976 unsigned long via_nrsegs;
979 struct pipe_inode_info *via_pipe;
980 unsigned int via_flags;
985 struct ll_cl_context {
986 struct list_head lcc_list;
988 const struct lu_env *lcc_env;
989 struct cl_io *lcc_io;
990 struct cl_page *lcc_page;
993 struct ll_thread_info {
994 struct iovec lti_local_iov;
995 struct vvp_io_args lti_args;
996 struct ra_io_arg lti_ria;
997 struct kiocb lti_kiocb;
998 struct ll_cl_context lti_io_ctx;
1001 extern struct lu_context_key ll_thread_key;
1003 static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
1005 struct ll_thread_info *lti;
1007 lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
1008 LASSERT(lti != NULL);
1013 static inline struct vvp_io_args *ll_env_args(const struct lu_env *env,
1014 enum vvp_io_subtype type)
1016 struct vvp_io_args *via = &ll_env_info(env)->lti_args;
1018 via->via_io_subtype = type;
1023 /* llite/llite_mmap.c */
1025 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1026 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1027 void policy_from_vma(ldlm_policy_data_t *policy,
1028 struct vm_area_struct *vma, unsigned long addr, size_t count);
1029 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1032 static inline void ll_invalidate_page(struct page *vmpage)
1034 struct address_space *mapping = vmpage->mapping;
1035 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1037 LASSERT(PageLocked(vmpage));
1038 if (mapping == NULL)
1042 * truncate_complete_page() calls
1043 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
1045 ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE);
1046 truncate_complete_page(mapping, vmpage);
1049 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1051 /* don't need an addref as the sb_info should be holding one */
1052 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1054 return ll_s2sbi(sb)->ll_dt_exp;
1057 /* don't need an addref as the sb_info should be holding one */
1058 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1060 return ll_s2sbi(sb)->ll_md_exp;
1063 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1065 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1071 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1072 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1074 return ll_s2sbi(inode->i_sb);
1077 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1079 return ll_s2dtexp(inode->i_sb);
1082 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1084 return ll_s2mdexp(inode->i_sb);
1087 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1091 LASSERT(inode != NULL);
1092 fid = &ll_i2info(inode)->lli_fid;
1097 static inline __u64 ll_file_maxbytes(struct inode *inode)
1099 return ll_i2info(inode)->lli_maxbytes;
1103 int ll_setxattr(struct dentry *dentry, const char *name,
1104 const void *value, size_t size, int flags);
1105 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1106 void *buffer, size_t size);
1107 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1108 int ll_removexattr(struct dentry *dentry, const char *name);
1110 /* llite/remote_perm.c */
1111 extern struct kmem_cache *ll_remote_perm_cachep;
1112 extern struct kmem_cache *ll_rmtperm_hash_cachep;
1114 void free_rmtperm_hash(struct hlist_head *hash);
1115 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1116 int lustre_check_remote_perm(struct inode *inode, int mask);
1118 /* llite/llite_capa.c */
1119 extern struct timer_list ll_capa_timer;
1121 int ll_capa_thread_start(void);
1122 void ll_capa_thread_stop(void);
1123 void ll_capa_timer_callback(unsigned long unused);
1125 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1127 void ll_capa_open(struct inode *inode);
1128 void ll_capa_close(struct inode *inode);
1130 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1131 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1133 void ll_truncate_free_capa(struct obd_capa *ocapa);
1134 void ll_clear_inode_capas(struct inode *inode);
1135 void ll_print_capa_stat(struct ll_sb_info *sbi);
1138 * Common IO arguments for various VFS I/O interfaces.
1140 int cl_sb_init(struct super_block *sb);
1141 int cl_sb_fini(struct super_block *sb);
1143 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1144 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1146 /* llite/llite_rmtacl.c */
1147 #ifdef CONFIG_FS_POSIX_ACL
1148 u64 rce_ops2valid(int ops);
1149 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1150 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1151 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1152 void rct_init(struct rmtacl_ctl_table *rct);
1153 void rct_fini(struct rmtacl_ctl_table *rct);
1155 void ee_free(struct eacl_entry *ee);
1156 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1157 ext_acl_xattr_header *header);
1158 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1159 struct lu_fid *fid, int type);
1160 void et_search_free(struct eacl_table *et, pid_t key);
1161 void et_init(struct eacl_table *et);
1162 void et_fini(struct eacl_table *et);
1167 #define LL_SA_RPC_MIN 2
1168 #define LL_SA_RPC_DEF 32
1169 #define LL_SA_RPC_MAX 8192
1171 #define LL_SA_CACHE_BIT 5
1172 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1173 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1175 /* per inode struct, for dir only */
1176 struct ll_statahead_info {
1177 struct dentry *sai_dentry;
1178 atomic_t sai_refcount; /* when access this struct, hold
1180 unsigned int sai_max; /* max ahead of lookup */
1181 __u64 sai_sent; /* stat requests sent count */
1182 __u64 sai_replied; /* stat requests which received
1184 __u64 sai_index; /* index of statahead entry */
1185 __u64 sai_index_wait; /* index of entry which is the
1186 * caller is waiting for */
1187 __u64 sai_hit; /* hit count */
1188 __u64 sai_miss; /* miss count:
1189 * for "ls -al" case, includes
1190 * hidden dentry miss;
1191 * for "ls -l" case, it does not
1192 * include hidden dentry miss.
1193 * "sai_miss_hidden" is used for
1196 unsigned int sai_consecutive_miss; /* consecutive miss */
1197 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1198 * is not a hidden one */
1199 unsigned int sai_skip_hidden;/* skipped hidden dentry count
1201 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1203 sai_agl_valid:1,/* AGL is valid for the dir */
1204 sai_in_readpage:1;/* statahead is in readdir()*/
1205 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1206 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1207 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1208 struct list_head sai_interim_entries; /* entries which got async
1209 * stat reply, but not
1211 struct list_head sai_entries; /* completed entries */
1212 struct list_head sai_agls; /* AGLs to be sent */
1213 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1214 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1215 atomic_t sai_cache_count; /* entry count in cache */
1218 int ll_statahead(struct inode *dir, struct dentry **dentry, bool unplug);
1219 void ll_authorize_statahead(struct inode *dir, void *key);
1220 void ll_deauthorize_statahead(struct inode *dir, void *key);
1223 blkcnt_t dirty_cnt(struct inode *inode);
1225 int cl_glimpse_size0(struct inode *inode, int agl);
1226 int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1227 struct inode *inode, struct cl_object *clob, int agl);
1229 static inline int cl_glimpse_size(struct inode *inode)
1231 return cl_glimpse_size0(inode, 0);
1234 static inline int cl_agl(struct inode *inode)
1236 return cl_glimpse_size0(inode, 1);
1239 static inline int ll_glimpse_size(struct inode *inode)
1241 struct ll_inode_info *lli = ll_i2info(inode);
1244 down_read(&lli->lli_glimpse_sem);
1245 rc = cl_glimpse_size(inode);
1246 lli->lli_glimpse_time = cfs_time_current();
1247 up_read(&lli->lli_glimpse_sem);
1251 /* dentry may statahead when statahead is enabled and current process has opened
1252 * parent directory, and this dentry hasn't accessed statahead cache before */
1254 dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1256 struct ll_inode_info *lli;
1257 struct ll_dentry_data *ldd;
1259 if (ll_i2sbi(dir)->ll_sa_max == 0)
1262 lli = ll_i2info(dir);
1264 /* statahead is not allowed for this dir, there may be three causes:
1265 * 1. dir is not opened.
1266 * 2. statahead hit ratio is too low.
1267 * 3. previous stat started statahead thread failed. */
1268 if (!lli->lli_sa_enabled)
1271 /* not the same process, don't statahead */
1272 if (lli->lli_opendir_pid != current_pid())
1276 * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1277 * multiple times, eg. for 'getattr', 'getxattr' and etc.
1278 * For patchless client, lookup intent is not accurate, which may
1279 * misguide statahead. For example:
1280 * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1281 * have the same intent -- IT_GETATTR, while one dentry should access
1282 * statahead cache once, otherwise statahead windows is messed up.
1283 * The solution is as following:
1284 * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1285 * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1286 * bypass interacting with statahead cache by checking
1287 * 'lld_sa_generation == lli->lli_sa_generation'.
1289 ldd = ll_d2d(dentry);
1290 if (ldd != NULL && ldd->lld_sa_generation == lli->lli_sa_generation)
1296 /* llite ioctl register support rountine */
1302 #define LLIOC_MAX_CMD 256
1305 * Rules to write a callback function:
1308 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1309 * returned to ll_iocontrol_register. Callback functions should use this
1310 * data to check the potential collasion of ioctl cmd. If collasion is
1311 * found, callback function should return LLIOC_CONT.
1312 * @rcp: The result of ioctl command.
1315 * If @magic matches the pointer returned by ll_iocontrol_data, the
1316 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1318 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1319 struct file *file, unsigned int cmd, unsigned long arg,
1320 void *magic, int *rcp);
1322 /* export functions */
1323 /* Register ioctl block dynamatically for a regular file.
1325 * @cmd: the array of ioctl command set
1326 * @count: number of commands in the @cmd
1327 * @cb: callback function, it will be called if an ioctl command is found to
1328 * belong to the command list @cmd.
1331 * A magic pointer will be returned if success;
1332 * otherwise, NULL will be returned.
1334 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1335 void ll_iocontrol_unregister(void *magic);
1337 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1339 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1340 enum cl_fsync_mode mode, int ignore_layout);
1342 /** direct write pages */
1343 struct ll_dio_pages {
1344 /** page array to be written. we don't support
1345 * partial pages except the last one. */
1346 struct page **ldp_pages;
1347 /* offset of each page */
1348 loff_t *ldp_offsets;
1349 /** if ldp_offsets is NULL, it means a sequential
1350 * pages to be written, then this is the file offset
1351 * of the * first page. */
1352 loff_t ldp_start_offset;
1353 /** how many bytes are to be written. */
1355 /** # of pages in the array. */
1359 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1362 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1365 ll_stats_ops_tally(ll_s2sbi(cl2vvp_dev(dev)->vdv_sb), opc, rc);
1368 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1369 int rw, struct inode *inode,
1370 struct ll_dio_pages *pv);
1372 static inline int ll_file_nolock(const struct file *file)
1374 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1375 struct inode *inode = file->f_dentry->d_inode;
1377 LASSERT(fd != NULL);
1378 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1379 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1382 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1383 struct lookup_intent *it, __u64 *bits)
1385 if (!it->d.lustre.it_lock_set) {
1386 struct lustre_handle handle;
1388 /* If this inode is a remote object, it will get two
1389 * separate locks in different namespaces, Master MDT,
1390 * where the name entry is, will grant LOOKUP lock,
1391 * remote MDT, where the object is, will grant
1392 * UPDATE|PERM lock. The inode will be attched to both
1393 * LOOKUP and PERM locks, so revoking either locks will
1394 * case the dcache being cleared */
1395 if (it->d.lustre.it_remote_lock_mode) {
1396 handle.cookie = it->d.lustre.it_remote_lock_handle;
1397 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1398 "(%p) for remote lock "LPX64"\n",
1399 PFID(ll_inode2fid(inode)), inode,
1401 md_set_lock_data(exp, &handle.cookie, inode, NULL);
1404 handle.cookie = it->d.lustre.it_lock_handle;
1406 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1407 " for lock "LPX64"\n",
1408 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1410 md_set_lock_data(exp, &handle.cookie, inode,
1411 &it->d.lustre.it_lock_bits);
1412 it->d.lustre.it_lock_set = 1;
1416 *bits = it->d.lustre.it_lock_bits;
1419 static inline void ll_lock_dcache(struct inode *inode)
1421 #ifdef HAVE_DCACHE_LOCK
1422 spin_lock(&dcache_lock);
1424 spin_lock(&inode->i_lock);
1428 static inline void ll_unlock_dcache(struct inode *inode)
1430 #ifdef HAVE_DCACHE_LOCK
1431 spin_unlock(&dcache_lock);
1433 spin_unlock(&inode->i_lock);
1437 static inline int d_lustre_invalid(const struct dentry *dentry)
1439 struct ll_dentry_data *lld = ll_d2d(dentry);
1441 return (lld == NULL) || lld->lld_invalid;
1444 static inline void __d_lustre_invalidate(struct dentry *dentry)
1446 struct ll_dentry_data *lld = ll_d2d(dentry);
1449 lld->lld_invalid = 1;
1453 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1454 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1455 * else dput() of the last refcount will unhash this dentry and kill it.
1457 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1459 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1460 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1461 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1463 spin_lock_nested(&dentry->d_lock,
1464 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1465 __d_lustre_invalidate(dentry);
1466 if (ll_d_count(dentry) == 0)
1468 spin_unlock(&dentry->d_lock);
1471 static inline void d_lustre_revalidate(struct dentry *dentry)
1473 spin_lock(&dentry->d_lock);
1474 LASSERT(ll_d2d(dentry) != NULL);
1475 ll_d2d(dentry)->lld_invalid = 0;
1476 spin_unlock(&dentry->d_lock);
1479 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 53, 0)
1480 /* Compatibility for old (1.8) compiled userspace quota code */
1481 struct if_quotactl_18 {
1486 struct obd_dqinfo qc_dqinfo;
1487 struct obd_dqblk qc_dqblk;
1489 struct obd_uuid obd_uuid;
1491 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1492 /* End compatibility for old (1.8) compiled userspace quota code */
1493 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 53, 0) */
1495 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1496 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1497 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1499 int ll_xattr_init(void);
1500 void ll_xattr_fini(void);
1502 int ll_getxattr_common(struct inode *inode, const char *name,
1503 void *buffer, size_t size, __u64 valid);
1504 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1505 struct cl_page *page, enum cl_req_type crt);
1507 int ll_getparent(struct file *file, struct getparent __user *arg);
1510 int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1511 unsigned int attr_flags, struct obd_capa *capa);
1513 extern struct lu_env *cl_inode_fini_env;
1514 extern int cl_inode_fini_refcheck;
1516 int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1517 void cl_inode_fini(struct inode *inode);
1519 u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1520 u32 cl_fid_build_gen(const struct lu_fid *fid);
1522 #endif /* LLITE_INTERNAL_H */