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>
43 #include <lustre_linkea.h>
45 /* for struct cl_lock_descr and struct cl_io */
46 #include <cl_object.h>
48 #include <lustre_lmv.h>
49 #include <lustre_mdc.h>
50 #include <lustre_intent.h>
51 #include <linux/compat.h>
53 #include "range_lock.h"
59 #ifndef VM_FAULT_RETRY
60 #define VM_FAULT_RETRY 0
63 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
64 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
65 #ifndef LOOKUP_CONTINUE
66 #define LOOKUP_CONTINUE LOOKUP_PARENT
69 /** Only used on client-side for indicating the tail of dir hash/offset. */
70 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
71 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
73 /* 4UL * 1024 * 1024 */
74 #define LL_MAX_BLKSIZE_BITS 22
76 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
77 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
79 struct ll_dentry_data {
80 struct lookup_intent *lld_it;
81 unsigned int lld_sa_generation;
82 unsigned int lld_invalid:1;
83 struct rcu_head lld_rcu_head;
86 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
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 struct hlist_node 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 data is modified. */
126 LLIF_DATA_MODIFIED = (1 << 4),
127 /* File is being restored */
128 LLIF_FILE_RESTORING = (1 << 5),
129 /* Xattr cache is attached to the file */
130 LLIF_XATTR_CACHE = (1 << 6),
133 struct ll_inode_info {
134 __u32 lli_inode_magic;
139 struct posix_acl *lli_posix_acl;
141 struct hlist_head *lli_remote_perms;
142 struct mutex lli_rmtperm_mutex;
144 /* identifying fields for both metadata and data stacks. */
145 struct lu_fid lli_fid;
146 /* master inode fid for stripe directory */
147 struct lu_fid lli_pfid;
149 struct list_head lli_close_list;
150 struct list_head lli_oss_capas;
151 /* open count currently used by capability only, indicate whether
152 * capability needs renewal */
153 atomic_t lli_open_count;
154 struct obd_capa *lli_mds_capa;
155 cfs_time_t lli_rmtperm_time;
157 /* handle is to be sent to MDS later on done_writing and setattr.
158 * Open handle data are needed for the recovery to reconstruct
159 * the inode state on the MDS. XXX: recovery is not ready yet. */
160 struct obd_client_handle *lli_pending_och;
162 /* We need all three because every inode may be opened in different
164 struct obd_client_handle *lli_mds_read_och;
165 struct obd_client_handle *lli_mds_write_och;
166 struct obd_client_handle *lli_mds_exec_och;
167 __u64 lli_open_fd_read_count;
168 __u64 lli_open_fd_write_count;
169 __u64 lli_open_fd_exec_count;
170 /* Protects access to och pointers and their usage counters */
171 struct mutex lli_och_mutex;
173 struct inode lli_vfs_inode;
175 /* the most recent timestamps obtained from mds */
176 struct ost_lvb lli_lvb;
177 spinlock_t lli_agl_lock;
179 /* Try to make the d::member and f::member are aligned. Before using
180 * these members, make clear whether it is directory or not. */
184 /* serialize normal readdir and statahead-readdir. */
185 struct mutex d_readdir_mutex;
187 /* metadata statahead */
188 /* since parent-child threads can share the same @file
189 * struct, "opendir_key" is the token when dir close for
190 * case of parent exit before child -- it is me should
191 * cleanup the dir readahead. */
193 struct ll_statahead_info *d_sai;
194 /* protect statahead stuff. */
195 spinlock_t d_sa_lock;
196 /* "opendir_pid" is the token when lookup/revalid
197 * -- I am the owner of dir statahead. */
199 /* stat will try to access statahead entries or start
200 * statahead if this flag is set, and this flag will be
201 * set upon dir open, and cleared when dir is closed,
202 * statahead hit ratio is too low, or start statahead
204 unsigned int d_sa_enabled:1;
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_sa_lock u.d.d_sa_lock
217 #define lli_sa_enabled u.d.d_sa_enabled
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 mutex f_size_mutex;
226 char *f_symlink_name;
229 * struct rw_semaphore {
230 * signed long count; // align d.d_def_acl
231 * spinlock_t wait_lock; // align d.d_sa_lock
232 * struct list_head wait_list;
235 struct rw_semaphore f_trunc_sem;
236 struct range_lock_tree f_write_tree;
238 struct rw_semaphore f_glimpse_sem;
239 cfs_time_t f_glimpse_time;
240 struct list_head f_agl_list;
243 /* for writepage() only to communicate to fsync */
247 * whenever a process try to read/write the file, the
248 * jobid of the process will be saved here, and it'll
249 * be packed into the write PRC when flush later.
251 * so the read/write statistics for jobid will not be
252 * accurate if the file is shared by different jobs.
254 char f_jobid[LUSTRE_JOBID_SIZE];
257 #define lli_size_mutex u.f.f_size_mutex
258 #define lli_symlink_name u.f.f_symlink_name
259 #define lli_maxbytes u.f.f_maxbytes
260 #define lli_trunc_sem u.f.f_trunc_sem
261 #define lli_write_tree u.f.f_write_tree
262 #define lli_glimpse_sem u.f.f_glimpse_sem
263 #define lli_glimpse_time u.f.f_glimpse_time
264 #define lli_agl_list u.f.f_agl_list
265 #define lli_agl_index u.f.f_agl_index
266 #define lli_async_rc u.f.f_async_rc
267 #define lli_jobid u.f.f_jobid
271 /* XXX: For following frequent used members, although they maybe special
272 * used for non-directory object, it is some time-wasting to check
273 * whether the object is directory or not before using them. On the
274 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
275 * the "ll_inode_info" size even if moving those members into u.f.
276 * So keep them out side.
278 * In the future, if more members are added only for directory,
279 * some of the following members can be moved into u.f.
282 struct cl_object *lli_clob;
284 /* mutex to request for layout lock exclusively. */
285 struct mutex lli_layout_mutex;
286 /* Layout version, protected by lli_layout_lock */
287 __u32 lli_layout_gen;
288 spinlock_t lli_layout_lock;
290 struct rw_semaphore lli_xattrs_list_rwsem;
291 struct mutex lli_xattrs_enq_lock;
292 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
295 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
299 spin_lock(&lli->lli_layout_lock);
300 gen = lli->lli_layout_gen;
301 spin_unlock(&lli->lli_layout_lock);
306 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
308 spin_lock(&lli->lli_layout_lock);
309 lli->lli_layout_gen = gen;
310 spin_unlock(&lli->lli_layout_lock);
313 int ll_xattr_cache_destroy(struct inode *inode);
315 int ll_xattr_cache_get(struct inode *inode,
322 * Locking to guarantee consistency of non-atomic updates to long long i_size,
323 * consistency between file size and KMS.
325 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
328 void ll_inode_size_lock(struct inode *inode);
329 void ll_inode_size_unlock(struct inode *inode);
331 // FIXME: replace the name of this with LL_I to conform to kernel stuff
332 // static inline struct ll_inode_info *LL_I(struct inode *inode)
333 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
335 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
338 /* default to about 40meg of readahead on a given system. That much tied
339 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
340 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT))
342 /* default to read-ahead full files smaller than 2MB on the second read */
343 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
348 RA_STAT_DISTANT_READPAGE,
349 RA_STAT_MISS_IN_WINDOW,
350 RA_STAT_FAILED_GRAB_PAGE,
351 RA_STAT_FAILED_MATCH,
356 RA_STAT_MAX_IN_FLIGHT,
357 RA_STAT_WRONG_GRAB_PAGE,
358 RA_STAT_FAILED_REACH_END,
363 atomic_t ra_cur_pages;
364 unsigned long ra_max_pages;
365 unsigned long ra_max_pages_per_file;
366 unsigned long ra_max_read_ahead_whole_pages;
369 /* ra_io_arg will be filled in the beginning of ll_readahead with
370 * ras_lock, then the following ll_read_ahead_pages will read RA
371 * pages according to this arg, all the items in this structure are
372 * counted by page index.
375 unsigned long ria_start; /* start offset of read-ahead*/
376 unsigned long ria_end; /* end offset of read-ahead*/
377 /* If stride read pattern is detected, ria_stoff means where
378 * stride read is started. Note: for normal read-ahead, the
379 * value here is meaningless, and also it will not be accessed*/
381 /* ria_length and ria_pages are the length and pages length in the
382 * stride I/O mode. And they will also be used to check whether
383 * it is stride I/O read-ahead in the read-ahead pages*/
384 unsigned long ria_length;
385 unsigned long ria_pages;
388 /* LL_HIST_MAX=32 causes an overflow */
389 #define LL_HIST_MAX 28
390 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
391 #define LL_PROCESS_HIST_MAX 10
392 struct per_process_info {
394 struct obd_histogram pp_r_hist;
395 struct obd_histogram pp_w_hist;
398 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
399 struct ll_rw_extents_info {
400 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
403 #define LL_OFFSET_HIST_MAX 100
404 struct ll_rw_process_info {
407 loff_t rw_range_start;
409 loff_t rw_last_file_pos;
411 size_t rw_smallest_extent;
412 size_t rw_largest_extent;
413 struct ll_file_data *rw_last_file;
416 enum stats_track_type {
417 STATS_TRACK_ALL = 0, /* track all processes */
418 STATS_TRACK_PID, /* track process with this pid */
419 STATS_TRACK_PPID, /* track processes with this ppid */
420 STATS_TRACK_GID, /* track processes with this gid */
424 /* flags for sbi->ll_flags */
425 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
426 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
427 #define LL_SBI_FLOCK 0x04
428 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
429 #define LL_SBI_ACL 0x10 /* support ACL */
430 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
431 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
432 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
433 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
434 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
435 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
436 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
437 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
438 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
439 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
440 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
441 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
442 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
443 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
444 #define LL_SBI_NOROOTSQUASH 0x100000 /* do not apply root squash */
446 #define LL_SBI_FLAGS { \
470 #define RCE_HASHES 32
472 struct rmtacl_ctl_entry {
473 struct list_head rce_list;
474 pid_t rce_key; /* hash key */
475 int rce_ops; /* acl operation type */
478 struct rmtacl_ctl_table {
480 struct list_head rct_entries[RCE_HASHES];
486 struct list_head ee_list;
487 pid_t ee_key; /* hash key */
488 struct lu_fid ee_fid;
489 int ee_type; /* ACL type for ACCESS or DEFAULT */
490 ext_acl_xattr_header *ee_acl;
495 struct list_head et_entries[EE_HASHES];
499 /* This is embedded into llite super-blocks to keep track of connect
500 * flags (capabilities) supported by all imports given mount is
502 struct lustre_client_ocd {
503 /* This is conjunction of connect_flags across all imports
504 * (LOVs) this mount is connected to. This field is updated by
505 * cl_ocd_update() under ->lco_lock. */
507 struct mutex lco_lock;
508 struct obd_export *lco_md_exp;
509 struct obd_export *lco_dt_exp;
513 struct list_head ll_list;
514 /* this protects pglist and ra_info. It isn't safe to
515 * grab from interrupt contexts */
517 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
518 spinlock_t ll_process_lock; /* ll_rw_process_info */
519 struct obd_uuid ll_sb_uuid;
520 struct obd_export *ll_md_exp;
521 struct obd_export *ll_dt_exp;
522 struct proc_dir_entry* ll_proc_root;
523 struct lu_fid ll_root_fid; /* root object fid */
526 unsigned int ll_umounting:1,
527 ll_xattr_cache_enabled:1;
528 /* per-conn chain of SBs */
529 struct list_head ll_conn_chain;
530 struct lustre_client_ocd ll_lco;
532 /*please don't ask -p*/
533 struct list_head ll_orphan_dentry_list;
534 struct ll_close_queue *ll_lcq;
536 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
538 /* Used to track "unstable" pages on a client, and maintain a
539 * LRU list of clean pages. An "unstable" page is defined as
540 * any page which is sent to a server as part of a bulk request,
541 * but is uncommitted to stable storage. */
542 struct cl_client_cache ll_cache;
544 struct lprocfs_stats *ll_ra_stats;
546 struct ll_ra_info ll_ra_info;
547 unsigned int ll_namelen;
548 struct file_operations *ll_fop;
550 /* =0 - hold lock over whole read/write
551 * >0 - max. chunk to be read/written w/o lock re-acquiring */
552 unsigned long ll_max_rw_chunk;
553 unsigned int ll_md_brw_pages; /* readdir pages per RPC */
555 struct lu_site *ll_site;
556 struct cl_device *ll_cl;
558 struct ll_rw_extents_info ll_rw_extents_info;
559 int ll_extent_process_count;
560 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
561 unsigned int ll_offset_process_count;
562 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
563 unsigned int ll_rw_offset_entry_count;
564 int ll_stats_track_id;
565 enum stats_track_type ll_stats_track_type;
568 /* metadata stat-ahead */
569 unsigned int ll_sa_max; /* max statahead RPCs */
570 atomic_t ll_sa_total; /* statahead thread started
572 atomic_t ll_sa_wrong; /* statahead thread stopped for
574 atomic_t ll_sa_running; /* running statahead thread
576 atomic_t ll_agl_total; /* AGL thread started count */
578 dev_t ll_sdev_orig; /* save s_dev before assign for
580 struct rmtacl_ctl_table ll_rct;
581 struct eacl_table ll_et;
584 struct root_squash_info ll_squash;
587 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
592 struct task_struct *lrr_reader;
593 struct list_head lrr_linkage;
597 * per file-descriptor read-ahead data.
599 struct ll_readahead_state {
602 * index of the last page that read(2) needed and that wasn't in the
603 * cache. Used by ras_update() to detect seeks.
605 * XXX nikita: if access seeks into cached region, Lustre doesn't see
608 unsigned long ras_last_readpage;
610 * number of pages read after last read-ahead window reset. As window
611 * is reset on each seek, this is effectively a number of consecutive
612 * accesses. Maybe ->ras_accessed_in_window is better name.
614 * XXX nikita: window is also reset (by ras_update()) when Lustre
615 * believes that memory pressure evicts read-ahead pages. In that
616 * case, it probably doesn't make sense to expand window to
617 * PTLRPC_MAX_BRW_PAGES on the third access.
619 unsigned long ras_consecutive_pages;
621 * number of read requests after the last read-ahead window reset
622 * As window is reset on each seek, this is effectively the number
623 * on consecutive read request and is used to trigger read-ahead.
625 unsigned long ras_consecutive_requests;
627 * Parameters of current read-ahead window. Handled by
628 * ras_update(). On the initial access to the file or after a seek,
629 * window is reset to 0. After 3 consecutive accesses, window is
630 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
631 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
633 unsigned long ras_window_start, ras_window_len;
635 * Where next read-ahead should start at. This lies within read-ahead
636 * window. Read-ahead window is read in pieces rather than at once
637 * because: 1. lustre limits total number of pages under read-ahead by
638 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
639 * not covered by DLM lock.
641 unsigned long ras_next_readahead;
643 * Total number of ll_file_read requests issued, reads originating
644 * due to mmap are not counted in this total. This value is used to
645 * trigger full file read-ahead after multiple reads to a small file.
647 unsigned long ras_requests;
649 * Page index with respect to the current request, these value
650 * will not be accurate when dealing with reads issued via mmap.
652 unsigned long ras_request_index;
654 * list of struct ll_ra_read's one per read(2) call current in
655 * progress against this file descriptor. Used by read-ahead code,
656 * protected by ->ras_lock.
658 struct list_head ras_read_beads;
660 * The following 3 items are used for detecting the stride I/O
662 * In stride I/O mode,
663 * ...............|-----data-----|****gap*****|--------|******|....
664 * offset |-stride_pages-|-stride_gap-|
665 * ras_stride_offset = offset;
666 * ras_stride_length = stride_pages + stride_gap;
667 * ras_stride_pages = stride_pages;
668 * Note: all these three items are counted by pages.
670 unsigned long ras_stride_length;
671 unsigned long ras_stride_pages;
672 pgoff_t ras_stride_offset;
674 * number of consecutive stride request count, and it is similar as
675 * ras_consecutive_requests, but used for stride I/O mode.
676 * Note: only more than 2 consecutive stride request are detected,
677 * stride read-ahead will be enable
679 unsigned long ras_consecutive_stride_requests;
682 extern struct kmem_cache *ll_file_data_slab;
683 struct lustre_handle;
684 struct ll_file_data {
685 struct ll_readahead_state fd_ras;
686 struct ccc_grouplock fd_grouplock;
690 /* openhandle if lease exists for this file.
691 * Borrow lli->lli_och_mutex to protect assignment */
692 struct obd_client_handle *fd_lease_och;
693 struct obd_client_handle *fd_och;
694 struct file *fd_file;
695 /* Indicate whether need to report failure when close.
696 * true: failure is known, not report again.
697 * false: unknown failure, should report. */
698 bool fd_write_failed;
699 rwlock_t fd_lock; /* protect lcc list */
700 struct list_head fd_lccs; /* list of ll_cl_context */
703 struct lov_stripe_md;
705 extern spinlock_t inode_lock;
707 extern struct proc_dir_entry *proc_lustre_fs_root;
709 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
711 return &lli->lli_vfs_inode;
714 __u32 ll_i2suppgid(struct inode *i);
715 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
717 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
719 #if BITS_PER_LONG == 32
721 #elif defined(CONFIG_COMPAT)
722 return unlikely(is_compat_task() || (sbi->ll_flags & LL_SBI_32BIT_API));
724 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
728 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
729 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
730 struct ll_ra_read *ll_ra_read_get(struct file *f);
732 /* llite/lproc_llite.c */
734 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
735 struct super_block *sb, char *osc, char *mdc);
736 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
737 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
738 extern struct lprocfs_seq_vars lprocfs_llite_obd_vars[];
740 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
741 struct super_block *sb, char *osc, char *mdc){return 0;}
742 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
743 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
748 LPROC_LL_DIRTY_MISSES,
750 LPROC_LL_WRITE_BYTES,
775 LPROC_LL_ALLOC_INODE,
778 LPROC_LL_GETXATTR_HITS,
780 LPROC_LL_REMOVEXATTR,
782 LPROC_LL_FILE_OPCODES
786 struct ll_dir_chain {
789 static inline void ll_dir_chain_init(struct ll_dir_chain *chain)
793 static inline void ll_dir_chain_fini(struct ll_dir_chain *chain)
797 extern const struct file_operations ll_dir_operations;
798 extern const struct inode_operations ll_dir_inode_operations;
799 #ifdef HAVE_DIR_CONTEXT
800 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
801 struct dir_context *ctx);
803 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
804 void *cookie, filldir_t filldir);
806 int ll_get_mdt_idx(struct inode *inode);
807 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
808 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
809 __u64 offset, struct ll_dir_chain *chain);
810 void ll_release_page(struct inode *inode, struct page *page, bool remove);
813 extern const struct inode_operations ll_special_inode_operations;
815 int ll_objects_destroy(struct ptlrpc_request *request,
817 struct inode *ll_iget(struct super_block *sb, ino_t hash,
818 struct lustre_md *lic);
819 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
820 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
821 void *data, int flag);
822 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
823 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
824 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
827 int ll_writepage(struct page *page, struct writeback_control *wbc);
828 int ll_writepages(struct address_space *, struct writeback_control *wbc);
829 int ll_readpage(struct file *file, struct page *page);
830 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
831 int ll_readahead(const struct lu_env *env, struct cl_io *io,
832 struct cl_page_list *queue, struct ll_readahead_state *ras,
834 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
835 struct ll_cl_context *ll_cl_find(struct file *file);
836 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io);
837 void ll_cl_remove(struct file *file, const struct lu_env *env);
839 #ifndef MS_HAS_NEW_AOPS
840 extern const struct address_space_operations ll_aops;
842 extern const struct address_space_operations_ext ll_aops;
846 extern struct file_operations ll_file_operations;
847 extern struct file_operations ll_file_operations_flock;
848 extern struct file_operations ll_file_operations_noflock;
849 extern struct inode_operations ll_file_inode_operations;
850 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
851 ldlm_mode_t l_req_mode);
852 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
853 struct lustre_handle *lockh, __u64 flags,
856 int ll_file_open(struct inode *inode, struct file *file);
857 int ll_file_release(struct inode *inode, struct file *file);
858 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
859 struct lov_stripe_md *lsm, lstat_t *st);
860 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
861 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
862 int ll_md_real_close(struct inode *inode, fmode_t fmode);
863 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
864 struct obd_client_handle **och, unsigned long flags);
865 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
866 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
867 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
868 __u64 ioepoch, int sync);
869 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
870 struct lustre_handle *fh);
871 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
872 struct ll_file_data *file, loff_t pos,
873 size_t count, int rw);
874 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
875 struct posix_acl *ll_get_acl(struct inode *inode, int type);
876 int ll_migrate(struct inode *parent, struct file *file, int mdtidx,
877 const char *name, int namelen);
878 int ll_get_fid_by_name(struct inode *parent, const char *name,
879 int namelen, struct lu_fid *fid);
880 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
881 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
883 # ifndef HAVE_INODE_PERMISION_2ARGS
884 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
886 int ll_inode_permission(struct inode *inode, int mask);
890 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
891 __u64 flags, struct lov_user_md *lum,
893 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
894 struct lov_mds_md **lmm, int *lmm_size,
895 struct ptlrpc_request **request);
896 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
898 int ll_dir_getstripe(struct inode *inode, void **lmmp,
899 int *lmm_size, struct ptlrpc_request **request,
901 #ifdef HAVE_FILE_FSYNC_4ARGS
902 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
903 #elif defined(HAVE_FILE_FSYNC_2ARGS)
904 int ll_fsync(struct file *file, int data);
906 int ll_fsync(struct file *file, struct dentry *dentry, int data);
908 int ll_merge_lvb(const struct lu_env *env, struct inode *inode);
909 int ll_fid2path(struct inode *inode, void __user *arg);
910 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
911 int ll_hsm_release(struct inode *inode);
915 int ll_d_init(struct dentry *de);
916 extern const struct dentry_operations ll_d_ops;
917 void ll_intent_drop_lock(struct lookup_intent *);
918 void ll_intent_release(struct lookup_intent *);
919 void ll_invalidate_aliases(struct inode *);
920 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
921 int ll_revalidate_it_finish(struct ptlrpc_request *request,
922 struct lookup_intent *it, struct dentry *de);
924 /* llite/llite_lib.c */
925 extern struct super_operations lustre_super_operations;
927 void ll_lli_init(struct ll_inode_info *lli);
928 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
929 void ll_put_super(struct super_block *sb);
930 void ll_kill_super(struct super_block *sb);
931 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
932 void ll_clear_inode(struct inode *inode);
933 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import);
934 int ll_setattr(struct dentry *de, struct iattr *attr);
935 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
936 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
937 __u64 max_age, __u32 flags);
938 int ll_update_inode(struct inode *inode, struct lustre_md *md);
939 int ll_read_inode2(struct inode *inode, void *opaque);
940 void ll_delete_inode(struct inode *inode);
941 int ll_iocontrol(struct inode *inode, struct file *file,
942 unsigned int cmd, unsigned long arg);
943 int ll_flush_ctx(struct inode *inode);
944 void ll_umount_begin(struct super_block *sb);
945 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
946 #ifdef HAVE_SUPEROPS_USE_DENTRY
947 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
949 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
951 void ll_dirty_page_discard_warn(struct page *page, int ioret);
952 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
953 struct super_block *, struct lookup_intent *);
954 void lustre_dump_dentry(struct dentry *, int recur);
955 int ll_obd_statfs(struct inode *inode, void __user *arg);
956 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
957 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
958 int ll_get_max_cookiesize(struct ll_sb_info *sbi, int *max_cookiesize);
959 int ll_get_default_cookiesize(struct ll_sb_info *sbi, int *default_cookiesize);
960 int ll_process_config(struct lustre_cfg *lcfg);
961 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
962 struct inode *i1, struct inode *i2,
963 const char *name, int namelen,
964 int mode, __u32 opc, void *data);
965 void ll_finish_md_op_data(struct md_op_data *op_data);
966 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
967 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
968 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
969 ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
970 struct lov_user_md **kbuf);
972 /* Compute expected user md size when passing in a md from user space */
973 static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
975 switch (lum->lmm_magic) {
976 case LOV_USER_MAGIC_V1:
977 return sizeof(struct lov_user_md_v1);
978 case LOV_USER_MAGIC_V3:
979 return sizeof(struct lov_user_md_v3);
980 case LOV_USER_MAGIC_SPECIFIC:
981 if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
984 return lov_user_md_size(lum->lmm_stripe_count,
985 LOV_USER_MAGIC_SPECIFIC);
991 /* llite/llite_nfs.c */
992 extern struct export_operations lustre_export_operations;
993 __u32 get_uuid2int(const char *name, int len);
994 struct inode *search_inode_for_lustre(struct super_block *sb,
995 const struct lu_fid *fid);
996 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
998 /* llite/symlink.c */
999 extern struct inode_operations ll_fast_symlink_inode_operations;
1001 /* llite/llite_close.c */
1002 struct ll_close_queue {
1003 spinlock_t lcq_lock;
1004 struct list_head lcq_head;
1005 wait_queue_head_t lcq_waitq;
1006 struct completion lcq_comp;
1010 struct ccc_object *cl_inode2ccc(struct inode *inode);
1013 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
1014 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
1016 /* specific achitecture can implement only part of this list */
1017 enum vvp_io_subtype {
1020 /** io started from splice_{read|write} */
1027 enum vvp_io_subtype cui_io_subtype;
1031 struct pipe_inode_info *cui_pipe;
1032 unsigned int cui_flags;
1034 struct vvp_fault_io {
1036 * Inode modification time that is checked across DLM
1040 struct vm_area_struct *ft_vma;
1042 * locked page returned from vvp_io
1044 struct page *ft_vmpage;
1045 struct vm_fault_api {
1049 struct vm_fault *ft_vmf;
1051 * fault API used bitflags for return code.
1053 unsigned int ft_flags;
1055 * check that flags are from filemap_fault
1057 bool ft_flags_valid;
1062 * Read-ahead state used by read and page-fault IO contexts.
1064 struct ll_ra_read cui_bead;
1066 * Set when cui_bead has been initialized.
1068 int cui_ra_window_set;
1072 * IO arguments for various VFS I/O interfaces.
1074 struct vvp_io_args {
1075 /** normal/sendfile/splice */
1076 enum vvp_io_subtype via_io_subtype;
1080 struct kiocb *via_iocb;
1081 struct iovec *via_iov;
1082 unsigned long via_nrsegs;
1085 struct pipe_inode_info *via_pipe;
1086 unsigned int via_flags;
1091 struct ll_cl_context {
1092 struct list_head lcc_list;
1094 const struct lu_env *lcc_env;
1095 struct cl_io *lcc_io;
1096 struct cl_page *lcc_page;
1099 struct vvp_thread_info {
1100 struct iovec vti_local_iov;
1101 struct vvp_io_args vti_args;
1102 struct ra_io_arg vti_ria;
1103 struct kiocb vti_kiocb;
1104 struct ll_cl_context vti_io_ctx;
1107 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
1109 extern struct lu_context_key vvp_key;
1110 struct vvp_thread_info *info;
1112 info = lu_context_key_get(&env->le_ctx, &vvp_key);
1113 LASSERT(info != NULL);
1117 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
1118 enum vvp_io_subtype type)
1120 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
1122 ret->via_io_subtype = type;
1127 struct vvp_session {
1128 struct vvp_io vs_ios;
1131 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1133 extern struct lu_context_key vvp_session_key;
1134 struct vvp_session *ses;
1136 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1137 LASSERT(ses != NULL);
1141 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1143 return &vvp_env_session(env)->vs_ios;
1146 int vvp_global_init(void);
1147 void vvp_global_fini(void);
1149 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1150 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1151 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1153 /* llite/llite_mmap.c */
1155 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1156 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1157 void policy_from_vma(ldlm_policy_data_t *policy,
1158 struct vm_area_struct *vma, unsigned long addr, size_t count);
1159 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1162 static inline void ll_invalidate_page(struct page *vmpage)
1164 struct address_space *mapping = vmpage->mapping;
1165 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1167 LASSERT(PageLocked(vmpage));
1168 if (mapping == NULL)
1172 * truncate_complete_page() calls
1173 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
1175 ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE);
1176 truncate_complete_page(mapping, vmpage);
1179 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1181 /* don't need an addref as the sb_info should be holding one */
1182 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1184 return ll_s2sbi(sb)->ll_dt_exp;
1187 /* don't need an addref as the sb_info should be holding one */
1188 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1190 return ll_s2sbi(sb)->ll_md_exp;
1193 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1195 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1201 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1202 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1204 return ll_s2sbi(inode->i_sb);
1207 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1209 return ll_s2dtexp(inode->i_sb);
1212 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1214 return ll_s2mdexp(inode->i_sb);
1217 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1221 LASSERT(inode != NULL);
1222 fid = &ll_i2info(inode)->lli_fid;
1227 static inline __u64 ll_file_maxbytes(struct inode *inode)
1229 return ll_i2info(inode)->lli_maxbytes;
1233 int ll_setxattr(struct dentry *dentry, const char *name,
1234 const void *value, size_t size, int flags);
1235 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1236 void *buffer, size_t size);
1237 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1238 int ll_removexattr(struct dentry *dentry, const char *name);
1240 /* llite/remote_perm.c */
1241 extern struct kmem_cache *ll_remote_perm_cachep;
1242 extern struct kmem_cache *ll_rmtperm_hash_cachep;
1244 void free_rmtperm_hash(struct hlist_head *hash);
1245 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1246 int lustre_check_remote_perm(struct inode *inode, int mask);
1248 /* llite/llite_capa.c */
1249 extern struct timer_list ll_capa_timer;
1251 int ll_capa_thread_start(void);
1252 void ll_capa_thread_stop(void);
1253 void ll_capa_timer_callback(unsigned long unused);
1255 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1257 void ll_capa_open(struct inode *inode);
1258 void ll_capa_close(struct inode *inode);
1260 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1261 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1263 void ll_truncate_free_capa(struct obd_capa *ocapa);
1264 void ll_clear_inode_capas(struct inode *inode);
1265 void ll_print_capa_stat(struct ll_sb_info *sbi);
1267 /* llite/llite_cl.c */
1268 extern struct lu_device_type vvp_device_type;
1271 * Common IO arguments for various VFS I/O interfaces.
1273 int cl_sb_init(struct super_block *sb);
1274 int cl_sb_fini(struct super_block *sb);
1276 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1277 struct ll_readahead_state *ras, unsigned long index,
1279 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1280 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1282 /* llite/llite_rmtacl.c */
1283 #ifdef CONFIG_FS_POSIX_ACL
1284 obd_valid rce_ops2valid(int ops);
1285 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1286 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1287 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1288 void rct_init(struct rmtacl_ctl_table *rct);
1289 void rct_fini(struct rmtacl_ctl_table *rct);
1291 void ee_free(struct eacl_entry *ee);
1292 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1293 ext_acl_xattr_header *header);
1294 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1295 struct lu_fid *fid, int type);
1296 void et_search_free(struct eacl_table *et, pid_t key);
1297 void et_init(struct eacl_table *et);
1298 void et_fini(struct eacl_table *et);
1303 #define LL_SA_RPC_MIN 2
1304 #define LL_SA_RPC_DEF 32
1305 #define LL_SA_RPC_MAX 8192
1307 #define LL_SA_CACHE_BIT 5
1308 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1309 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1311 /* per inode struct, for dir only */
1312 struct ll_statahead_info {
1313 struct inode *sai_inode;
1314 atomic_t sai_refcount; /* when access this struct, hold
1316 unsigned int sai_generation; /* generation for statahead */
1317 unsigned int sai_max; /* max ahead of lookup */
1318 __u64 sai_sent; /* stat requests sent count */
1319 __u64 sai_replied; /* stat requests which received
1321 __u64 sai_index; /* index of statahead entry */
1322 __u64 sai_index_wait; /* index of entry which is the
1323 * caller is waiting for */
1324 __u64 sai_hit; /* hit count */
1325 __u64 sai_miss; /* miss count:
1326 * for "ls -al" case, it
1327 * includes hidden dentry miss;
1328 * for "ls -l" case, it does not
1329 * include hidden dentry miss.
1330 * "sai_miss_hidden" is used for
1333 unsigned int sai_consecutive_miss; /* consecutive miss */
1334 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1335 * is not a hidden one */
1336 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1337 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1339 sai_agl_valid:1,/* AGL is valid for the dir */
1340 sai_in_readpage:1;/* statahead is in readdir()*/
1341 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1342 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1343 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1344 struct list_head sai_interim_entries; /* entries which got async
1345 * stat reply, but not
1347 struct list_head sai_entries; /* completed entries */
1348 struct list_head sai_agls; /* AGLs to be sent */
1349 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1350 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1351 atomic_t sai_cache_count; /* entry count in cache */
1354 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1356 void ll_authorize_statahead(struct inode *dir, void *key);
1357 void ll_deauthorize_statahead(struct inode *dir, void *key);
1359 static inline int ll_glimpse_size(struct inode *inode)
1361 struct ll_inode_info *lli = ll_i2info(inode);
1364 down_read(&lli->lli_glimpse_sem);
1365 rc = cl_glimpse_size(inode);
1366 lli->lli_glimpse_time = cfs_time_current();
1367 up_read(&lli->lli_glimpse_sem);
1372 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1374 struct ll_inode_info *lli = ll_i2info(dir);
1375 struct ll_statahead_info *sai = lli->lli_sai;
1376 struct ll_dentry_data *ldd = ll_d2d(dentry);
1378 /* not the same process, don't mark */
1379 if (lli->lli_opendir_pid != current_pid())
1382 LASSERT(ldd != NULL);
1384 ldd->lld_sa_generation = sai->sai_generation;
1388 dentry_need_statahead(struct inode *dir, struct dentry *dentry)
1390 struct ll_inode_info *lli;
1391 struct ll_dentry_data *ldd;
1393 if (ll_i2sbi(dir)->ll_sa_max == 0)
1396 lli = ll_i2info(dir);
1398 /* statahead is not allowed for this dir, there may be three causes:
1399 * 1. dir is not opened.
1400 * 2. statahead hit ratio is too low.
1401 * 3. previous stat started statahead thread failed. */
1402 if (!lli->lli_sa_enabled)
1405 /* not the same process, don't statahead */
1406 if (lli->lli_opendir_pid != current_pid())
1409 ldd = ll_d2d(dentry);
1411 * When stats a dentry, the system trigger more than once "revalidate"
1412 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1413 * Under patchless client mode, the operation intent is not accurate,
1414 * which maybe misguide the statahead thread. For example:
1415 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1416 * have the same operation intent -- "IT_GETATTR".
1417 * In fact, one dentry should has only one chance to interact with the
1418 * statahead thread, otherwise the statahead windows will be confused.
1419 * The solution is as following:
1420 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1421 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1422 * will bypass interacting with statahead thread for checking:
1423 * "lld_sa_generation == lli_sai->sai_generation"
1425 if (ldd && lli->lli_sai &&
1426 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1433 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1435 if (!dentry_need_statahead(dir, *dentryp))
1438 return do_statahead_enter(dir, dentryp, only_unplug);
1441 /* llite ioctl register support rountine */
1447 #define LLIOC_MAX_CMD 256
1450 * Rules to write a callback function:
1453 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1454 * returned to ll_iocontrol_register. Callback functions should use this
1455 * data to check the potential collasion of ioctl cmd. If collasion is
1456 * found, callback function should return LLIOC_CONT.
1457 * @rcp: The result of ioctl command.
1460 * If @magic matches the pointer returned by ll_iocontrol_data, the
1461 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1463 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1464 struct file *file, unsigned int cmd, unsigned long arg,
1465 void *magic, int *rcp);
1467 /* export functions */
1468 /* Register ioctl block dynamatically for a regular file.
1470 * @cmd: the array of ioctl command set
1471 * @count: number of commands in the @cmd
1472 * @cb: callback function, it will be called if an ioctl command is found to
1473 * belong to the command list @cmd.
1476 * A magic pointer will be returned if success;
1477 * otherwise, NULL will be returned.
1479 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1480 void ll_iocontrol_unregister(void *magic);
1483 /* lclient compat stuff */
1484 #define cl_inode_info ll_inode_info
1485 #define cl_i2info(info) ll_i2info(info)
1486 #define cl_inode_mode(inode) ((inode)->i_mode)
1487 #define cl_i2sbi ll_i2sbi
1489 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1490 const struct iattr *attr)
1492 LASSERT(attr->ia_valid & ATTR_FILE);
1493 return LUSTRE_FPRIVATE(attr->ia_file);
1496 static inline void cl_isize_lock(struct inode *inode)
1498 ll_inode_size_lock(inode);
1501 static inline void cl_isize_unlock(struct inode *inode)
1503 ll_inode_size_unlock(inode);
1506 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1508 LASSERT(mutex_is_locked(&ll_i2info(inode)->lli_size_mutex));
1509 i_size_write(inode, kms);
1512 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1514 ll_inode_size_lock(inode);
1515 i_size_write(inode, kms);
1516 ll_inode_size_unlock(inode);
1519 #define cl_isize_read(inode) i_size_read(inode)
1521 static inline int cl_merge_lvb(const struct lu_env *env, struct inode *inode)
1523 return ll_merge_lvb(env, inode);
1526 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1527 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1528 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1530 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1532 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1533 enum cl_fsync_mode mode, int ignore_layout);
1535 /** direct write pages */
1536 struct ll_dio_pages {
1537 /** page array to be written. we don't support
1538 * partial pages except the last one. */
1539 struct page **ldp_pages;
1540 /* offset of each page */
1541 loff_t *ldp_offsets;
1542 /** if ldp_offsets is NULL, it means a sequential
1543 * pages to be written, then this is the file offset
1544 * of the * first page. */
1545 loff_t ldp_start_offset;
1546 /** how many bytes are to be written. */
1548 /** # of pages in the array. */
1552 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1555 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1558 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1561 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1562 int rw, struct inode *inode,
1563 struct ll_dio_pages *pv);
1565 static inline int ll_file_nolock(const struct file *file)
1567 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1568 struct inode *inode = file->f_dentry->d_inode;
1570 LASSERT(fd != NULL);
1571 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1572 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1575 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1576 struct lookup_intent *it, __u64 *bits)
1578 if (!it->d.lustre.it_lock_set) {
1579 struct lustre_handle handle;
1581 /* If this inode is a remote object, it will get two
1582 * separate locks in different namespaces, Master MDT,
1583 * where the name entry is, will grant LOOKUP lock,
1584 * remote MDT, where the object is, will grant
1585 * UPDATE|PERM lock. The inode will be attched to both
1586 * LOOKUP and PERM locks, so revoking either locks will
1587 * case the dcache being cleared */
1588 if (it->d.lustre.it_remote_lock_mode) {
1589 handle.cookie = it->d.lustre.it_remote_lock_handle;
1590 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1591 "(%p) for remote lock "LPX64"\n",
1592 PFID(ll_inode2fid(inode)), inode,
1594 md_set_lock_data(exp, &handle.cookie, inode, NULL);
1597 handle.cookie = it->d.lustre.it_lock_handle;
1599 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1600 " for lock "LPX64"\n",
1601 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1603 md_set_lock_data(exp, &handle.cookie, inode,
1604 &it->d.lustre.it_lock_bits);
1605 it->d.lustre.it_lock_set = 1;
1609 *bits = it->d.lustre.it_lock_bits;
1612 static inline void ll_lock_dcache(struct inode *inode)
1614 #ifdef HAVE_DCACHE_LOCK
1615 spin_lock(&dcache_lock);
1617 spin_lock(&inode->i_lock);
1621 static inline void ll_unlock_dcache(struct inode *inode)
1623 #ifdef HAVE_DCACHE_LOCK
1624 spin_unlock(&dcache_lock);
1626 spin_unlock(&inode->i_lock);
1630 static inline int d_lustre_invalid(const struct dentry *dentry)
1632 struct ll_dentry_data *lld = ll_d2d(dentry);
1634 return (lld == NULL) || lld->lld_invalid;
1637 static inline void __d_lustre_invalidate(struct dentry *dentry)
1639 struct ll_dentry_data *lld = ll_d2d(dentry);
1642 lld->lld_invalid = 1;
1646 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1647 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1648 * else dput() of the last refcount will unhash this dentry and kill it.
1650 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1652 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1653 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1654 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1656 spin_lock_nested(&dentry->d_lock,
1657 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1658 __d_lustre_invalidate(dentry);
1659 if (ll_d_count(dentry) == 0)
1661 spin_unlock(&dentry->d_lock);
1664 static inline void d_lustre_revalidate(struct dentry *dentry)
1666 spin_lock(&dentry->d_lock);
1667 LASSERT(ll_d2d(dentry) != NULL);
1668 ll_d2d(dentry)->lld_invalid = 0;
1669 spin_unlock(&dentry->d_lock);
1672 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 53, 0)
1673 /* Compatibility for old (1.8) compiled userspace quota code */
1674 struct if_quotactl_18 {
1679 struct obd_dqinfo qc_dqinfo;
1680 struct obd_dqblk qc_dqblk;
1682 struct obd_uuid obd_uuid;
1684 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1685 /* End compatibility for old (1.8) compiled userspace quota code */
1686 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 53, 0) */
1689 LL_LAYOUT_GEN_NONE = ((__u32)-2), /* layout lock was cancelled */
1690 LL_LAYOUT_GEN_EMPTY = ((__u32)-1) /* for empty layout */
1693 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1694 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1695 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1697 int ll_xattr_init(void);
1698 void ll_xattr_fini(void);
1700 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1701 struct cl_page *page, enum cl_req_type crt);
1703 int ll_getparent(struct file *file, struct getparent __user *arg);
1705 #endif /* LLITE_INTERNAL_H */