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, 2013, 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>
44 /* for struct cl_lock_descr and struct cl_io */
45 #include <cl_object.h>
47 #include <lustre_lmv.h>
48 #include <lustre_mdc.h>
49 #include <linux/lustre_intent.h>
50 #include <linux/compat.h>
56 #ifndef VM_FAULT_RETRY
57 #define VM_FAULT_RETRY 0
60 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
61 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
62 #ifndef LOOKUP_CONTINUE
63 #define LOOKUP_CONTINUE LOOKUP_PARENT
66 /** Only used on client-side for indicating the tail of dir hash/offset. */
67 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
68 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
70 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
71 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
73 struct ll_dentry_data {
76 struct obd_client_handle lld_cwd_och;
77 struct obd_client_handle lld_mnt_och;
78 struct lookup_intent *lld_it;
79 unsigned int lld_sa_generation;
80 unsigned int lld_invalid:1;
81 struct rcu_head lld_rcu_head;
84 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
86 extern struct file_operations ll_pgcache_seq_fops;
88 #define LLI_INODE_MAGIC 0x111d0de5
89 #define LLI_INODE_DEAD 0xdeadd00d
91 /* remote client permission cache */
92 #define REMOTE_PERM_HASHSIZE 16
94 struct ll_getname_data {
95 #ifdef HAVE_DIR_CONTEXT
96 struct dir_context ctx;
98 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
99 struct lu_fid lgd_fid; /* target fid we are looking for */
100 int lgd_found; /* inode matched? */
103 /* llite setxid/access permission for user on remote client */
104 struct ll_remote_perm {
105 cfs_hlist_node_t lrp_list;
110 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
111 is access permission with
112 lrp_fsuid/lrp_fsgid. */
116 /* MDS has an authority for the Size-on-MDS attributes. */
117 LLIF_MDS_SIZE_LOCK = (1 << 0),
118 /* Epoch close is postponed. */
119 LLIF_EPOCH_PENDING = (1 << 1),
120 /* DONE WRITING is allowed. */
121 LLIF_DONE_WRITING = (1 << 2),
122 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
124 LLIF_SOM_DIRTY = (1 << 3),
125 /* File is contented */
126 LLIF_CONTENDED = (1 << 4),
127 /* Truncate uses server lock for this file */
128 LLIF_SRVLOCK = (1 << 5),
129 /* File data is modified. */
130 LLIF_DATA_MODIFIED = (1 << 6),
131 /* File is being restored */
132 LLIF_FILE_RESTORING = (1 << 7),
133 /* Xattr cache is attached to the file */
134 LLIF_XATTR_CACHE = (1 << 8),
137 struct ll_inode_info {
138 __u32 lli_inode_magic;
143 struct posix_acl *lli_posix_acl;
145 cfs_hlist_head_t *lli_remote_perms;
146 struct mutex lli_rmtperm_mutex;
148 /* identifying fields for both metadata and data stacks. */
149 struct lu_fid lli_fid;
150 /* Parent fid for accessing default stripe data on parent directory
151 * for allocating OST objects after a mknod() and later open-by-FID. */
152 struct lu_fid lli_pfid;
154 cfs_list_t lli_close_list;
155 cfs_list_t lli_oss_capas;
156 /* open count currently used by capability only, indicate whether
157 * capability needs renewal */
158 cfs_atomic_t lli_open_count;
159 struct obd_capa *lli_mds_capa;
160 cfs_time_t lli_rmtperm_time;
162 /* handle is to be sent to MDS later on done_writing and setattr.
163 * Open handle data are needed for the recovery to reconstruct
164 * the inode state on the MDS. XXX: recovery is not ready yet. */
165 struct obd_client_handle *lli_pending_och;
167 /* We need all three because every inode may be opened in different
169 struct obd_client_handle *lli_mds_read_och;
170 struct obd_client_handle *lli_mds_write_och;
171 struct obd_client_handle *lli_mds_exec_och;
172 __u64 lli_open_fd_read_count;
173 __u64 lli_open_fd_write_count;
174 __u64 lli_open_fd_exec_count;
175 /* Protects access to och pointers and their usage counters */
176 struct mutex lli_och_mutex;
178 struct inode lli_vfs_inode;
180 /* the most recent timestamps obtained from mds */
181 struct ost_lvb lli_lvb;
182 spinlock_t lli_agl_lock;
184 /* Try to make the d::member and f::member are aligned. Before using
185 * these members, make clear whether it is directory or not. */
189 /* serialize normal readdir and statahead-readdir. */
190 struct mutex d_readdir_mutex;
192 /* metadata statahead */
193 /* since parent-child threads can share the same @file
194 * struct, "opendir_key" is the token when dir close for
195 * case of parent exit before child -- it is me should
196 * cleanup the dir readahead. */
198 struct ll_statahead_info *d_sai;
199 struct posix_acl *d_def_acl;
200 /* protect statahead stuff. */
201 spinlock_t d_sa_lock;
202 /* "opendir_pid" is the token when lookup/revalid
203 * -- I am the owner of dir statahead. */
205 /* directory stripe information */
206 struct lmv_stripe_md *d_lsm_md;
207 /* striped directory size */
208 loff_t d_stripe_size;
209 /* striped directory nlink */
210 __u64 d_stripe_nlink;
213 #define lli_readdir_mutex u.d.d_readdir_mutex
214 #define lli_opendir_key u.d.d_opendir_key
215 #define lli_sai u.d.d_sai
216 #define lli_def_acl u.d.d_def_acl
217 #define lli_sa_lock u.d.d_sa_lock
218 #define lli_opendir_pid u.d.d_opendir_pid
219 #define lli_lsm_md u.d.d_lsm_md
220 #define lli_stripe_dir_size u.d.d_stripe_size
221 #define lli_stripe_dir_nlink u.d.d_stripe_nlink
223 /* for non-directory */
225 struct semaphore f_size_sem;
226 void *f_size_sem_owner;
227 char *f_symlink_name;
230 * struct rw_semaphore {
231 * signed long count; // align d.d_def_acl
232 * spinlock_t wait_lock; // align d.d_sa_lock
233 * struct list_head wait_list;
236 struct rw_semaphore f_trunc_sem;
237 struct mutex f_write_mutex;
239 struct rw_semaphore f_glimpse_sem;
240 cfs_time_t f_glimpse_time;
241 cfs_list_t f_agl_list;
244 /* for writepage() only to communicate to fsync */
247 /* volatile file criteria is based on file name, this
248 * flag is used to keep the test result, so the strcmp
253 * whenever a process try to read/write the file, the
254 * jobid of the process will be saved here, and it'll
255 * be packed into the write PRC when flush later.
257 * so the read/write statistics for jobid will not be
258 * accurate if the file is shared by different jobs.
260 char f_jobid[JOBSTATS_JOBID_SIZE];
263 #define lli_size_sem u.f.f_size_sem
264 #define lli_size_sem_owner u.f.f_size_sem_owner
265 #define lli_symlink_name u.f.f_symlink_name
266 #define lli_maxbytes u.f.f_maxbytes
267 #define lli_trunc_sem u.f.f_trunc_sem
268 #define lli_write_mutex u.f.f_write_mutex
269 #define lli_glimpse_sem u.f.f_glimpse_sem
270 #define lli_glimpse_time u.f.f_glimpse_time
271 #define lli_agl_list u.f.f_agl_list
272 #define lli_agl_index u.f.f_agl_index
273 #define lli_async_rc u.f.f_async_rc
274 #define lli_jobid u.f.f_jobid
275 #define lli_volatile u.f.f_volatile
279 /* XXX: For following frequent used members, although they maybe special
280 * used for non-directory object, it is some time-wasting to check
281 * whether the object is directory or not before using them. On the
282 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
283 * the "ll_inode_info" size even if moving those members into u.f.
284 * So keep them out side.
286 * In the future, if more members are added only for directory,
287 * some of the following members can be moved into u.f.
290 struct cl_object *lli_clob;
292 /* mutex to request for layout lock exclusively. */
293 struct mutex lli_layout_mutex;
294 /* Layout version, protected by lli_layout_lock */
295 __u32 lli_layout_gen;
296 spinlock_t lli_layout_lock;
298 struct rw_semaphore lli_xattrs_list_rwsem;
299 struct mutex lli_xattrs_enq_lock;
300 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
303 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
307 spin_lock(&lli->lli_layout_lock);
308 gen = lli->lli_layout_gen;
309 spin_unlock(&lli->lli_layout_lock);
314 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
316 spin_lock(&lli->lli_layout_lock);
317 lli->lli_layout_gen = gen;
318 spin_unlock(&lli->lli_layout_lock);
321 int ll_xattr_cache_destroy(struct inode *inode);
323 int ll_xattr_cache_get(struct inode *inode,
330 * Locking to guarantee consistency of non-atomic updates to long long i_size,
331 * consistency between file size and KMS.
333 * Implemented by ->lli_size_sem and ->lsm_lock, nested in that order.
336 void ll_inode_size_lock(struct inode *inode);
337 void ll_inode_size_unlock(struct inode *inode);
339 // FIXME: replace the name of this with LL_I to conform to kernel stuff
340 // static inline struct ll_inode_info *LL_I(struct inode *inode)
341 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
343 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
346 /* default to about 40meg of readahead on a given system. That much tied
347 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
348 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT))
350 /* default to read-ahead full files smaller than 2MB on the second read */
351 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
356 RA_STAT_DISTANT_READPAGE,
357 RA_STAT_MISS_IN_WINDOW,
358 RA_STAT_FAILED_GRAB_PAGE,
359 RA_STAT_FAILED_MATCH,
364 RA_STAT_MAX_IN_FLIGHT,
365 RA_STAT_WRONG_GRAB_PAGE,
366 RA_STAT_FAILED_REACH_END,
371 cfs_atomic_t ra_cur_pages;
372 unsigned long ra_max_pages;
373 unsigned long ra_max_pages_per_file;
374 unsigned long ra_max_read_ahead_whole_pages;
377 /* ra_io_arg will be filled in the beginning of ll_readahead with
378 * ras_lock, then the following ll_read_ahead_pages will read RA
379 * pages according to this arg, all the items in this structure are
380 * counted by page index.
383 unsigned long ria_start; /* start offset of read-ahead*/
384 unsigned long ria_end; /* end offset of read-ahead*/
385 /* If stride read pattern is detected, ria_stoff means where
386 * stride read is started. Note: for normal read-ahead, the
387 * value here is meaningless, and also it will not be accessed*/
389 /* ria_length and ria_pages are the length and pages length in the
390 * stride I/O mode. And they will also be used to check whether
391 * it is stride I/O read-ahead in the read-ahead pages*/
392 unsigned long ria_length;
393 unsigned long ria_pages;
396 /* LL_HIST_MAX=32 causes an overflow */
397 #define LL_HIST_MAX 28
398 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
399 #define LL_PROCESS_HIST_MAX 10
400 struct per_process_info {
402 struct obd_histogram pp_r_hist;
403 struct obd_histogram pp_w_hist;
406 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
407 struct ll_rw_extents_info {
408 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
411 #define LL_OFFSET_HIST_MAX 100
412 struct ll_rw_process_info {
415 loff_t rw_range_start;
417 loff_t rw_last_file_pos;
419 size_t rw_smallest_extent;
420 size_t rw_largest_extent;
421 struct ll_file_data *rw_last_file;
424 enum stats_track_type {
425 STATS_TRACK_ALL = 0, /* track all processes */
426 STATS_TRACK_PID, /* track process with this pid */
427 STATS_TRACK_PPID, /* track processes with this ppid */
428 STATS_TRACK_GID, /* track processes with this gid */
432 /* flags for sbi->ll_flags */
433 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
434 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
435 #define LL_SBI_FLOCK 0x04
436 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
437 #define LL_SBI_ACL 0x10 /* support ACL */
438 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
439 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
440 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
441 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
442 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
443 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
444 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
445 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
446 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
447 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
448 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
449 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
450 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
451 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
453 #define LL_SBI_FLAGS { \
476 /* default value for ll_sb_info->contention_time */
477 #define SBI_DEFAULT_CONTENTION_SECONDS 60
478 /* default value for lockless_truncate_enable */
479 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
480 #define RCE_HASHES 32
482 struct rmtacl_ctl_entry {
484 pid_t rce_key; /* hash key */
485 int rce_ops; /* acl operation type */
488 struct rmtacl_ctl_table {
490 cfs_list_t rct_entries[RCE_HASHES];
497 pid_t ee_key; /* hash key */
498 struct lu_fid ee_fid;
499 int ee_type; /* ACL type for ACCESS or DEFAULT */
500 ext_acl_xattr_header *ee_acl;
505 cfs_list_t et_entries[EE_HASHES];
510 /* this protects pglist and ra_info. It isn't safe to
511 * grab from interrupt contexts */
513 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
514 spinlock_t ll_process_lock; /* ll_rw_process_info */
515 struct obd_uuid ll_sb_uuid;
516 struct obd_export *ll_md_exp;
517 struct obd_export *ll_dt_exp;
518 struct proc_dir_entry* ll_proc_root;
519 struct lu_fid ll_root_fid; /* root object fid */
522 unsigned int ll_umounting:1,
523 ll_xattr_cache_enabled:1;
524 cfs_list_t ll_conn_chain; /* per-conn chain of SBs */
525 struct lustre_client_ocd ll_lco;
527 cfs_list_t ll_orphan_dentry_list; /*please don't ask -p*/
528 struct ll_close_queue *ll_lcq;
530 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
532 /* Used to track "unstable" pages on a client, and maintain a
533 * LRU list of clean pages. An "unstable" page is defined as
534 * any page which is sent to a server as part of a bulk request,
535 * but is uncommitted to stable storage. */
536 struct cl_client_cache ll_cache;
538 struct lprocfs_stats *ll_ra_stats;
540 struct ll_ra_info ll_ra_info;
541 unsigned int ll_namelen;
542 struct file_operations *ll_fop;
544 /* =0 - hold lock over whole read/write
545 * >0 - max. chunk to be read/written w/o lock re-acquiring */
546 unsigned long ll_max_rw_chunk;
547 unsigned int ll_md_brw_size; /* used by readdir */
549 struct lu_site *ll_site;
550 struct cl_device *ll_cl;
552 struct ll_rw_extents_info ll_rw_extents_info;
553 int ll_extent_process_count;
554 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
555 unsigned int ll_offset_process_count;
556 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
557 unsigned int ll_rw_offset_entry_count;
558 int ll_stats_track_id;
559 enum stats_track_type ll_stats_track_type;
562 /* metadata stat-ahead */
563 unsigned int ll_sa_max; /* max statahead RPCs */
564 atomic_t ll_sa_total; /* statahead thread started
566 atomic_t ll_sa_wrong; /* statahead thread stopped for
568 atomic_t ll_agl_total; /* AGL thread started count */
570 dev_t ll_sdev_orig; /* save s_dev before assign for
572 struct rmtacl_ctl_table ll_rct;
573 struct eacl_table ll_et;
576 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
581 struct task_struct *lrr_reader;
582 cfs_list_t lrr_linkage;
586 * per file-descriptor read-ahead data.
588 struct ll_readahead_state {
591 * index of the last page that read(2) needed and that wasn't in the
592 * cache. Used by ras_update() to detect seeks.
594 * XXX nikita: if access seeks into cached region, Lustre doesn't see
597 unsigned long ras_last_readpage;
599 * number of pages read after last read-ahead window reset. As window
600 * is reset on each seek, this is effectively a number of consecutive
601 * accesses. Maybe ->ras_accessed_in_window is better name.
603 * XXX nikita: window is also reset (by ras_update()) when Lustre
604 * believes that memory pressure evicts read-ahead pages. In that
605 * case, it probably doesn't make sense to expand window to
606 * PTLRPC_MAX_BRW_PAGES on the third access.
608 unsigned long ras_consecutive_pages;
610 * number of read requests after the last read-ahead window reset
611 * As window is reset on each seek, this is effectively the number
612 * on consecutive read request and is used to trigger read-ahead.
614 unsigned long ras_consecutive_requests;
616 * Parameters of current read-ahead window. Handled by
617 * ras_update(). On the initial access to the file or after a seek,
618 * window is reset to 0. After 3 consecutive accesses, window is
619 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
620 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
622 unsigned long ras_window_start, ras_window_len;
624 * Where next read-ahead should start at. This lies within read-ahead
625 * window. Read-ahead window is read in pieces rather than at once
626 * because: 1. lustre limits total number of pages under read-ahead by
627 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
628 * not covered by DLM lock.
630 unsigned long ras_next_readahead;
632 * Total number of ll_file_read requests issued, reads originating
633 * due to mmap are not counted in this total. This value is used to
634 * trigger full file read-ahead after multiple reads to a small file.
636 unsigned long ras_requests;
638 * Page index with respect to the current request, these value
639 * will not be accurate when dealing with reads issued via mmap.
641 unsigned long ras_request_index;
643 * list of struct ll_ra_read's one per read(2) call current in
644 * progress against this file descriptor. Used by read-ahead code,
645 * protected by ->ras_lock.
647 cfs_list_t ras_read_beads;
649 * The following 3 items are used for detecting the stride I/O
651 * In stride I/O mode,
652 * ...............|-----data-----|****gap*****|--------|******|....
653 * offset |-stride_pages-|-stride_gap-|
654 * ras_stride_offset = offset;
655 * ras_stride_length = stride_pages + stride_gap;
656 * ras_stride_pages = stride_pages;
657 * Note: all these three items are counted by pages.
659 unsigned long ras_stride_length;
660 unsigned long ras_stride_pages;
661 pgoff_t ras_stride_offset;
663 * number of consecutive stride request count, and it is similar as
664 * ras_consecutive_requests, but used for stride I/O mode.
665 * Note: only more than 2 consecutive stride request are detected,
666 * stride read-ahead will be enable
668 unsigned long ras_consecutive_stride_requests;
671 extern struct kmem_cache *ll_file_data_slab;
672 struct lustre_handle;
673 struct ll_file_data {
674 struct ll_readahead_state fd_ras;
675 struct ccc_grouplock fd_grouplock;
679 /* openhandle if lease exists for this file.
680 * Borrow lli->lli_och_mutex to protect assignment */
681 struct obd_client_handle *fd_lease_och;
682 struct obd_client_handle *fd_och;
683 struct file *fd_file;
684 /* Indicate whether need to report failure when close.
685 * true: failure is known, not report again.
686 * false: unknown failure, should report. */
687 bool fd_write_failed;
690 struct lov_stripe_md;
692 extern spinlock_t inode_lock;
694 extern struct proc_dir_entry *proc_lustre_fs_root;
696 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
698 return &lli->lli_vfs_inode;
702 struct inode *icbd_parent;
703 struct dentry **icbd_childp;
707 __u32 ll_i2suppgid(struct inode *i);
708 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
710 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
712 #if BITS_PER_LONG == 32
719 (sbi->ll_flags & LL_SBI_32BIT_API)
724 #define LLAP_MAGIC 98764321
726 extern struct kmem_cache *ll_async_page_slab;
727 extern size_t ll_async_page_slab_size;
729 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
730 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
731 struct ll_ra_read *ll_ra_read_get(struct file *f);
733 /* llite/lproc_llite.c */
735 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
736 struct super_block *sb, char *osc, char *mdc);
737 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
738 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
739 extern struct lprocfs_seq_vars lprocfs_llite_obd_vars[];
741 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
742 struct super_block *sb, char *osc, char *mdc){return 0;}
743 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
744 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
749 extern struct file_operations ll_dir_operations;
750 extern struct inode_operations ll_dir_inode_operations;
751 #ifdef HAVE_DIR_CONTEXT
752 int ll_dir_read(struct inode *inode, struct md_op_data *op_data,
753 struct dir_context *ctx);
755 int ll_dir_read(struct inode *inode, struct md_op_data *op_data,
756 void *cookie, filldir_t filldir);
758 int ll_get_mdt_idx(struct inode *inode);
759 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
761 struct lu_dirent *ll_dir_entry_start(struct inode *dir,
762 struct md_op_data *op_data,
763 struct page **ppage);
765 struct lu_dirent *ll_dir_entry_next(struct inode *dir,
766 struct md_op_data *op_data,
767 struct lu_dirent *ent,
768 struct page **ppage);
770 int ll_objects_destroy(struct ptlrpc_request *request,
772 struct inode *ll_iget(struct super_block *sb, ino_t hash,
773 struct lustre_md *lic);
774 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
775 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
776 void *data, int flag);
777 #ifndef HAVE_IOP_ATOMIC_OPEN
778 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
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 int ll_d_mountpoint(struct dentry *dparent, struct dentry *dchild,
785 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
788 int ll_writepage(struct page *page, struct writeback_control *wbc);
789 int ll_writepages(struct address_space *, struct writeback_control *wbc);
790 void ll_removepage(struct page *page);
791 int ll_readpage(struct file *file, struct page *page);
792 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
793 int ll_file_punch(struct inode *, loff_t, int);
794 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
795 void ll_clear_file_contended(struct inode*);
796 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
797 int ll_readahead(const struct lu_env *env, struct cl_io *io,
798 struct cl_page_list *queue, struct ll_readahead_state *ras,
800 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
801 struct ll_cl_context *ll_cl_init(struct file *file, struct page *vmpage);
802 void ll_cl_fini(struct ll_cl_context *lcc);
805 extern struct file_operations ll_file_operations;
806 extern struct file_operations ll_file_operations_flock;
807 extern struct file_operations ll_file_operations_noflock;
808 extern struct inode_operations ll_file_inode_operations;
809 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
811 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
812 ldlm_mode_t l_req_mode);
813 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
814 struct lustre_handle *lockh, __u64 flags,
816 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
818 #ifdef HAVE_IOP_ATOMIC_OPEN
819 int ll_revalidate_nd(struct dentry *dentry, unsigned int flags);
821 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
823 int ll_file_open(struct inode *inode, struct file *file);
824 int ll_file_release(struct inode *inode, struct file *file);
825 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
826 struct lov_stripe_md *lsm, lstat_t *st);
827 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
828 int ll_local_open(struct file *file,
829 struct lookup_intent *it, struct ll_file_data *fd,
830 struct obd_client_handle *och);
831 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
832 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
834 int ll_md_real_close(struct inode *inode, fmode_t fmode);
835 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
836 struct obd_client_handle **och, unsigned long flags);
837 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
838 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
839 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
840 __u64 ioepoch, int sync);
841 int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data,
842 struct md_open_data **mod);
843 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
844 struct lustre_handle *fh);
845 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
846 struct ll_file_data *file, loff_t pos,
847 size_t count, int rw);
848 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
849 struct lookup_intent *it, struct kstat *stat);
850 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
851 struct ll_file_data *ll_file_data_get(void);
852 struct posix_acl * ll_get_acl(struct inode *inode, int type);
853 int ll_migrate(struct inode *parent, struct file *file, int mdtidx,
854 const char *name, int namelen);
855 int ll_get_fid_by_name(struct inode *parent, const char *name,
856 int namelen, struct lu_fid *fid);
857 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
858 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
860 # ifndef HAVE_INODE_PERMISION_2ARGS
861 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
863 int ll_inode_permission(struct inode *inode, int mask);
867 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
868 __u64 flags, struct lov_user_md *lum,
870 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
871 struct lov_mds_md **lmm, int *lmm_size,
872 struct ptlrpc_request **request);
873 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
875 int ll_dir_getstripe(struct inode *inode, void **lmmp,
876 int *lmm_size, struct ptlrpc_request **request,
878 #ifdef HAVE_FILE_FSYNC_4ARGS
879 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
880 #elif defined(HAVE_FILE_FSYNC_2ARGS)
881 int ll_fsync(struct file *file, int data);
883 int ll_fsync(struct file *file, struct dentry *dentry, int data);
885 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
887 int ll_merge_lvb(const struct lu_env *env, struct inode *inode);
888 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
889 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
890 int ll_fid2path(struct inode *inode, void *arg);
891 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
892 int ll_hsm_release(struct inode *inode);
894 struct obd_client_handle *ll_lease_open(struct inode *inode, struct file *file,
895 fmode_t mode, __u64 flags);
896 int ll_lease_close(struct obd_client_handle *och, struct inode *inode,
901 int ll_d_init(struct dentry *de);
902 extern struct dentry_operations ll_d_ops;
903 void ll_intent_drop_lock(struct lookup_intent *);
904 void ll_intent_release(struct lookup_intent *);
905 void ll_invalidate_aliases(struct inode *);
906 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
907 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
908 int ll_revalidate_it_finish(struct ptlrpc_request *request,
909 struct lookup_intent *it, struct dentry *de);
911 /* llite/llite_lib.c */
912 extern struct super_operations lustre_super_operations;
914 char *ll_read_opt(const char *opt, char *data);
915 void ll_lli_init(struct ll_inode_info *lli);
916 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
917 void ll_put_super(struct super_block *sb);
918 void ll_kill_super(struct super_block *sb);
919 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
920 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
921 void ll_clear_inode(struct inode *inode);
922 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import);
923 int ll_setattr(struct dentry *de, struct iattr *attr);
924 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
925 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
926 __u64 max_age, __u32 flags);
927 void ll_update_inode(struct inode *inode, struct lustre_md *md);
928 void ll_read_inode2(struct inode *inode, void *opaque);
929 void ll_delete_inode(struct inode *inode);
930 int ll_iocontrol(struct inode *inode, struct file *file,
931 unsigned int cmd, unsigned long arg);
932 int ll_flush_ctx(struct inode *inode);
933 void ll_umount_begin(struct super_block *sb);
934 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
935 #ifdef HAVE_SUPEROPS_USE_DENTRY
936 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
938 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
940 void ll_dirty_page_discard_warn(struct page *page, int ioret);
941 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
942 struct super_block *, struct lookup_intent *);
943 void lustre_dump_dentry(struct dentry *, int recur);
944 void lustre_dump_inode(struct inode *);
945 int ll_obd_statfs(struct inode *inode, void *arg);
946 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
947 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
948 int ll_get_max_cookiesize(struct ll_sb_info *sbi, int *max_cookiesize);
949 int ll_get_default_cookiesize(struct ll_sb_info *sbi, int *default_cookiesize);
950 int ll_process_config(struct lustre_cfg *lcfg);
951 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
952 struct inode *i1, struct inode *i2,
953 const char *name, int namelen,
954 int mode, __u32 opc, void *data);
955 void ll_finish_md_op_data(struct md_op_data *op_data);
956 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
957 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
959 /* llite/llite_nfs.c */
960 extern struct export_operations lustre_export_operations;
961 __u32 get_uuid2int(const char *name, int len);
962 struct inode *search_inode_for_lustre(struct super_block *sb,
963 const struct lu_fid *fid);
965 /* llite/special.c */
966 extern struct inode_operations ll_special_inode_operations;
967 extern struct file_operations ll_special_chr_inode_fops;
968 extern struct file_operations ll_special_chr_file_fops;
969 extern struct file_operations ll_special_blk_inode_fops;
970 extern struct file_operations ll_special_fifo_inode_fops;
971 extern struct file_operations ll_special_fifo_file_fops;
972 extern struct file_operations ll_special_sock_inode_fops;
974 /* llite/symlink.c */
975 extern struct inode_operations ll_fast_symlink_inode_operations;
977 /* llite/llite_close.c */
978 struct ll_close_queue {
981 wait_queue_head_t lcq_waitq;
982 struct completion lcq_comp;
983 cfs_atomic_t lcq_stop;
986 struct ccc_object *cl_inode2ccc(struct inode *inode);
989 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
990 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
992 /* specific achitecture can implement only part of this list */
993 enum vvp_io_subtype {
996 /** io started from splice_{read|write} */
1003 enum vvp_io_subtype cui_io_subtype;
1007 struct pipe_inode_info *cui_pipe;
1008 unsigned int cui_flags;
1010 struct vvp_fault_io {
1012 * Inode modification time that is checked across DLM
1016 struct vm_area_struct *ft_vma;
1018 * locked page returned from vvp_io
1020 struct page *ft_vmpage;
1021 struct vm_fault_api {
1025 struct vm_fault *ft_vmf;
1027 * fault API used bitflags for return code.
1029 unsigned int ft_flags;
1034 * Read-ahead state used by read and page-fault IO contexts.
1036 struct ll_ra_read cui_bead;
1038 * Set when cui_bead has been initialized.
1040 int cui_ra_window_set;
1042 * Partially truncated page, that vvp_io_trunc_start() keeps locked
1045 struct cl_page *cui_partpage;
1049 * IO arguments for various VFS I/O interfaces.
1051 struct vvp_io_args {
1052 /** normal/sendfile/splice */
1053 enum vvp_io_subtype via_io_subtype;
1057 struct kiocb *via_iocb;
1058 struct iovec *via_iov;
1059 unsigned long via_nrsegs;
1062 struct pipe_inode_info *via_pipe;
1063 unsigned int via_flags;
1068 struct ll_cl_context {
1070 struct cl_io *lcc_io;
1071 struct cl_page *lcc_page;
1072 struct lu_env *lcc_env;
1077 struct vvp_thread_info {
1078 struct ost_lvb vti_lvb;
1079 struct cl_2queue vti_queue;
1080 struct iovec vti_local_iov;
1081 struct vvp_io_args vti_args;
1082 struct ra_io_arg vti_ria;
1083 struct kiocb vti_kiocb;
1084 struct ll_cl_context vti_io_ctx;
1087 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
1089 extern struct lu_context_key vvp_key;
1090 struct vvp_thread_info *info;
1092 info = lu_context_key_get(&env->le_ctx, &vvp_key);
1093 LASSERT(info != NULL);
1097 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
1098 enum vvp_io_subtype type)
1100 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
1102 ret->via_io_subtype = type;
1107 struct vvp_session {
1108 struct vvp_io vs_ios;
1111 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1113 extern struct lu_context_key vvp_session_key;
1114 struct vvp_session *ses;
1116 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1117 LASSERT(ses != NULL);
1121 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1123 return &vvp_env_session(env)->vs_ios;
1126 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1127 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1128 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1130 /* llite/llite_mmap.c */
1131 typedef struct rb_root rb_root_t;
1132 typedef struct rb_node rb_node_t;
1134 struct ll_lock_tree_node;
1135 struct ll_lock_tree {
1137 cfs_list_t lt_locked_list;
1138 struct ll_file_data *lt_fd;
1141 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1142 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1143 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
1144 __u64 end, ldlm_mode_t mode);
1145 void policy_from_vma(ldlm_policy_data_t *policy,
1146 struct vm_area_struct *vma, unsigned long addr, size_t count);
1147 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1150 static inline void ll_invalidate_page(struct page *vmpage)
1152 struct address_space *mapping = vmpage->mapping;
1153 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1155 LASSERT(PageLocked(vmpage));
1156 if (mapping == NULL)
1160 * truncate_complete_page() calls
1161 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
1163 ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE);
1164 truncate_complete_page(mapping, vmpage);
1167 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1169 /* don't need an addref as the sb_info should be holding one */
1170 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1172 return ll_s2sbi(sb)->ll_dt_exp;
1175 /* don't need an addref as the sb_info should be holding one */
1176 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1178 return ll_s2sbi(sb)->ll_md_exp;
1181 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1183 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1189 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1190 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1192 return ll_s2sbi(inode->i_sb);
1195 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1197 return ll_s2dtexp(inode->i_sb);
1200 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1202 return ll_s2mdexp(inode->i_sb);
1205 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1209 LASSERT(inode != NULL);
1210 fid = &ll_i2info(inode)->lli_fid;
1215 static inline __u64 ll_file_maxbytes(struct inode *inode)
1217 return ll_i2info(inode)->lli_maxbytes;
1221 int ll_setxattr(struct dentry *dentry, const char *name,
1222 const void *value, size_t size, int flags);
1223 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1224 void *buffer, size_t size);
1225 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1226 int ll_removexattr(struct dentry *dentry, const char *name);
1228 /* llite/remote_perm.c */
1229 extern struct kmem_cache *ll_remote_perm_cachep;
1230 extern struct kmem_cache *ll_rmtperm_hash_cachep;
1232 cfs_hlist_head_t *alloc_rmtperm_hash(void);
1233 void free_rmtperm_hash(cfs_hlist_head_t *hash);
1234 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1235 int lustre_check_remote_perm(struct inode *inode, int mask);
1237 /* llite/llite_capa.c */
1238 extern struct timer_list ll_capa_timer;
1240 int ll_capa_thread_start(void);
1241 void ll_capa_thread_stop(void);
1242 void ll_capa_timer_callback(unsigned long unused);
1244 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1245 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1247 void ll_capa_open(struct inode *inode);
1248 void ll_capa_close(struct inode *inode);
1250 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1251 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1253 void ll_truncate_free_capa(struct obd_capa *ocapa);
1254 void ll_clear_inode_capas(struct inode *inode);
1255 void ll_print_capa_stat(struct ll_sb_info *sbi);
1257 /* llite/llite_cl.c */
1258 extern struct lu_device_type vvp_device_type;
1261 * Common IO arguments for various VFS I/O interfaces.
1263 int cl_sb_init(struct super_block *sb);
1264 int cl_sb_fini(struct super_block *sb);
1265 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1266 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1268 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1269 struct ll_readahead_state *ras, unsigned long index,
1271 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1272 int ll_is_file_contended(struct file *file);
1273 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1275 /* llite/llite_rmtacl.c */
1276 #ifdef CONFIG_FS_POSIX_ACL
1277 obd_valid rce_ops2valid(int ops);
1278 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1279 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1280 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1281 void rct_init(struct rmtacl_ctl_table *rct);
1282 void rct_fini(struct rmtacl_ctl_table *rct);
1284 void ee_free(struct eacl_entry *ee);
1285 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1286 ext_acl_xattr_header *header);
1287 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1288 struct lu_fid *fid, int type);
1289 void et_search_free(struct eacl_table *et, pid_t key);
1290 void et_init(struct eacl_table *et);
1291 void et_fini(struct eacl_table *et);
1296 #define LL_SA_RPC_MIN 2
1297 #define LL_SA_RPC_DEF 32
1298 #define LL_SA_RPC_MAX 8192
1300 #define LL_SA_CACHE_BIT 5
1301 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1302 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1304 /* per inode struct, for dir only */
1305 struct ll_statahead_info {
1306 struct inode *sai_inode;
1307 cfs_atomic_t sai_refcount; /* when access this struct, hold
1309 unsigned int sai_generation; /* generation for statahead */
1310 unsigned int sai_max; /* max ahead of lookup */
1311 __u64 sai_sent; /* stat requests sent count */
1312 __u64 sai_replied; /* stat requests which received
1314 __u64 sai_index; /* index of statahead entry */
1315 __u64 sai_index_wait; /* index of entry which is the
1316 * caller is waiting for */
1317 __u64 sai_hit; /* hit count */
1318 __u64 sai_miss; /* miss count:
1319 * for "ls -al" case, it includes
1320 * hidden dentry miss;
1321 * for "ls -l" case, it does not
1322 * include hidden dentry miss.
1323 * "sai_miss_hidden" is used for
1326 unsigned int sai_consecutive_miss; /* consecutive miss */
1327 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1328 * is not a hidden one */
1329 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1330 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1332 sai_in_readpage:1,/* statahead is in readdir()*/
1333 sai_agl_valid:1;/* AGL is valid for the dir */
1334 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1335 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1336 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1337 cfs_list_t sai_entries; /* entry list */
1338 cfs_list_t sai_entries_received; /* entries returned */
1339 cfs_list_t sai_entries_stated; /* entries stated */
1340 cfs_list_t sai_entries_agl; /* AGL entries to be sent */
1341 cfs_list_t sai_cache[LL_SA_CACHE_SIZE];
1342 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1343 cfs_atomic_t sai_cache_count; /* entry count in cache */
1346 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1348 void ll_stop_statahead(struct inode *dir, void *key);
1350 static inline int ll_glimpse_size(struct inode *inode)
1352 struct ll_inode_info *lli = ll_i2info(inode);
1355 down_read(&lli->lli_glimpse_sem);
1356 rc = cl_glimpse_size(inode);
1357 lli->lli_glimpse_time = cfs_time_current();
1358 up_read(&lli->lli_glimpse_sem);
1363 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1365 struct ll_inode_info *lli = ll_i2info(dir);
1366 struct ll_statahead_info *sai = lli->lli_sai;
1367 struct ll_dentry_data *ldd = ll_d2d(dentry);
1369 /* not the same process, don't mark */
1370 if (lli->lli_opendir_pid != current_pid())
1373 LASSERT(ldd != NULL);
1375 ldd->lld_sa_generation = sai->sai_generation;
1379 d_need_statahead(struct inode *dir, struct dentry *dentryp)
1381 struct ll_inode_info *lli;
1382 struct ll_dentry_data *ldd;
1384 if (ll_i2sbi(dir)->ll_sa_max == 0)
1387 lli = ll_i2info(dir);
1388 /* not the same process, don't statahead */
1389 if (lli->lli_opendir_pid != current_pid())
1392 /* statahead has been stopped */
1393 if (lli->lli_opendir_key == NULL)
1396 ldd = ll_d2d(dentryp);
1398 * When stats a dentry, the system trigger more than once "revalidate"
1399 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1400 * Under patchless client mode, the operation intent is not accurate,
1401 * which maybe misguide the statahead thread. For example:
1402 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1403 * have the same operation intent -- "IT_GETATTR".
1404 * In fact, one dentry should has only one chance to interact with the
1405 * statahead thread, otherwise the statahead windows will be confused.
1406 * The solution is as following:
1407 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1408 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1409 * will bypass interacting with statahead thread for checking:
1410 * "lld_sa_generation == lli_sai->sai_generation"
1412 if (ldd && lli->lli_sai &&
1413 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1420 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1424 ret = d_need_statahead(dir, *dentryp);
1428 return do_statahead_enter(dir, dentryp, only_unplug);
1431 /* llite ioctl register support rountine */
1438 #define LLIOC_MAX_CMD 256
1441 * Rules to write a callback function:
1444 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1445 * returned to ll_iocontrol_register. Callback functions should use this
1446 * data to check the potential collasion of ioctl cmd. If collasion is
1447 * found, callback function should return LLIOC_CONT.
1448 * @rcp: The result of ioctl command.
1451 * If @magic matches the pointer returned by ll_iocontrol_data, the
1452 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1454 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1455 struct file *file, unsigned int cmd, unsigned long arg,
1456 void *magic, int *rcp);
1458 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1459 unsigned int cmd, unsigned long arg, int *rcp);
1461 /* export functions */
1462 /* Register ioctl block dynamatically for a regular file.
1464 * @cmd: the array of ioctl command set
1465 * @count: number of commands in the @cmd
1466 * @cb: callback function, it will be called if an ioctl command is found to
1467 * belong to the command list @cmd.
1470 * A magic pointer will be returned if success;
1471 * otherwise, NULL will be returned.
1473 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1474 void ll_iocontrol_unregister(void *magic);
1478 /* lclient compat stuff */
1479 #define cl_inode_info ll_inode_info
1480 #define cl_i2info(info) ll_i2info(info)
1481 #define cl_inode_mode(inode) ((inode)->i_mode)
1482 #define cl_i2sbi ll_i2sbi
1484 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1485 const struct iattr *attr)
1487 LASSERT(attr->ia_valid & ATTR_FILE);
1488 return LUSTRE_FPRIVATE(attr->ia_file);
1491 static inline void cl_isize_lock(struct inode *inode)
1493 ll_inode_size_lock(inode);
1496 static inline void cl_isize_unlock(struct inode *inode)
1498 ll_inode_size_unlock(inode);
1501 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1503 LASSERT(down_trylock(&ll_i2info(inode)->lli_size_sem) != 0);
1504 i_size_write(inode, kms);
1507 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1509 ll_inode_size_lock(inode);
1510 i_size_write(inode, kms);
1511 ll_inode_size_unlock(inode);
1514 #define cl_isize_read(inode) i_size_read(inode)
1516 static inline int cl_merge_lvb(const struct lu_env *env, struct inode *inode)
1518 return ll_merge_lvb(env, inode);
1521 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1522 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1523 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1525 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1527 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1528 enum cl_fsync_mode mode, int ignore_layout);
1530 /** direct write pages */
1531 struct ll_dio_pages {
1532 /** page array to be written. we don't support
1533 * partial pages except the last one. */
1534 struct page **ldp_pages;
1535 /* offset of each page */
1536 loff_t *ldp_offsets;
1537 /** if ldp_offsets is NULL, it means a sequential
1538 * pages to be written, then this is the file offset
1539 * of the * first page. */
1540 loff_t ldp_start_offset;
1541 /** how many bytes are to be written. */
1543 /** # of pages in the array. */
1547 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1550 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1553 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1556 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1557 int rw, struct inode *inode,
1558 struct ll_dio_pages *pv);
1560 static inline int ll_file_nolock(const struct file *file)
1562 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1563 struct inode *inode = file->f_dentry->d_inode;
1565 LASSERT(fd != NULL);
1566 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1567 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1570 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1571 struct lookup_intent *it, __u64 *bits)
1573 if (!it->d.lustre.it_lock_set) {
1574 struct lustre_handle handle;
1576 /* If this inode is a remote object, it will get two
1577 * separate locks in different namespaces, Master MDT,
1578 * where the name entry is, will grant LOOKUP lock,
1579 * remote MDT, where the object is, will grant
1580 * UPDATE|PERM lock. The inode will be attched to both
1581 * LOOKUP and PERM locks, so revoking either locks will
1582 * case the dcache being cleared */
1583 if (it->d.lustre.it_remote_lock_mode) {
1584 handle.cookie = it->d.lustre.it_remote_lock_handle;
1585 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1586 "(%p) for remote lock "LPX64"\n",
1587 PFID(ll_inode2fid(inode)), inode,
1589 md_set_lock_data(exp, &handle.cookie, inode, NULL);
1592 handle.cookie = it->d.lustre.it_lock_handle;
1594 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1595 " for lock "LPX64"\n",
1596 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1598 md_set_lock_data(exp, &handle.cookie, inode,
1599 &it->d.lustre.it_lock_bits);
1600 it->d.lustre.it_lock_set = 1;
1604 *bits = it->d.lustre.it_lock_bits;
1607 static inline void ll_lock_dcache(struct inode *inode)
1609 #ifdef HAVE_DCACHE_LOCK
1610 spin_lock(&dcache_lock);
1612 spin_lock(&inode->i_lock);
1616 static inline void ll_unlock_dcache(struct inode *inode)
1618 #ifdef HAVE_DCACHE_LOCK
1619 spin_unlock(&dcache_lock);
1621 spin_unlock(&inode->i_lock);
1625 static inline int d_lustre_invalid(const struct dentry *dentry)
1627 struct ll_dentry_data *lld = ll_d2d(dentry);
1629 return (lld == NULL) || lld->lld_invalid;
1632 static inline void __d_lustre_invalidate(struct dentry *dentry)
1634 struct ll_dentry_data *lld = ll_d2d(dentry);
1637 lld->lld_invalid = 1;
1641 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1642 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1643 * else dput() of the last refcount will unhash this dentry and kill it.
1645 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1647 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1648 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1649 dentry->d_parent, dentry->d_inode, d_count(dentry));
1651 spin_lock_nested(&dentry->d_lock,
1652 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1653 __d_lustre_invalidate(dentry);
1654 if (d_count(dentry) == 0)
1656 spin_unlock(&dentry->d_lock);
1659 static inline void d_lustre_revalidate(struct dentry *dentry)
1661 spin_lock(&dentry->d_lock);
1662 LASSERT(ll_d2d(dentry) != NULL);
1663 ll_d2d(dentry)->lld_invalid = 0;
1664 spin_unlock(&dentry->d_lock);
1667 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0)
1668 /* Compatibility for old (1.8) compiled userspace quota code */
1669 struct if_quotactl_18 {
1674 struct obd_dqinfo qc_dqinfo;
1675 struct obd_dqblk qc_dqblk;
1677 struct obd_uuid obd_uuid;
1679 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1680 /* End compatibility for old (1.8) compiled userspace quota code */
1682 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1683 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */
1686 LL_LAYOUT_GEN_NONE = ((__u32)-2), /* layout lock was cancelled */
1687 LL_LAYOUT_GEN_EMPTY = ((__u32)-1) /* for empty layout */
1690 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1691 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1692 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1694 int ll_xattr_init(void);
1695 void ll_xattr_fini(void);
1697 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1698 struct cl_page *page, enum cl_req_type crt);
1700 #endif /* LLITE_INTERNAL_H */