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 <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 {
74 struct lookup_intent *lld_it;
75 unsigned int lld_sa_generation;
76 unsigned int lld_invalid:1;
77 struct rcu_head lld_rcu_head;
80 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
82 #define LLI_INODE_MAGIC 0x111d0de5
83 #define LLI_INODE_DEAD 0xdeadd00d
85 /* remote client permission cache */
86 #define REMOTE_PERM_HASHSIZE 16
88 struct ll_getname_data {
89 #ifdef HAVE_DIR_CONTEXT
90 struct dir_context ctx;
92 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
93 struct lu_fid lgd_fid; /* target fid we are looking for */
94 int lgd_found; /* inode matched? */
97 /* llite setxid/access permission for user on remote client */
98 struct ll_remote_perm {
99 struct hlist_node lrp_list;
104 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
105 is access permission with
106 lrp_fsuid/lrp_fsgid. */
110 /* MDS has an authority for the Size-on-MDS attributes. */
111 LLIF_MDS_SIZE_LOCK = (1 << 0),
112 /* Epoch close is postponed. */
113 LLIF_EPOCH_PENDING = (1 << 1),
114 /* DONE WRITING is allowed. */
115 LLIF_DONE_WRITING = (1 << 2),
116 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
118 LLIF_SOM_DIRTY = (1 << 3),
119 /* File data is modified. */
120 LLIF_DATA_MODIFIED = (1 << 4),
121 /* File is being restored */
122 LLIF_FILE_RESTORING = (1 << 5),
123 /* Xattr cache is attached to the file */
124 LLIF_XATTR_CACHE = (1 << 6),
127 struct ll_inode_info {
128 __u32 lli_inode_magic;
133 struct posix_acl *lli_posix_acl;
135 struct hlist_head *lli_remote_perms;
136 struct mutex lli_rmtperm_mutex;
138 /* identifying fields for both metadata and data stacks. */
139 struct lu_fid lli_fid;
140 /* master inode fid for stripe directory */
141 struct lu_fid lli_pfid;
143 struct list_head lli_close_list;
144 struct list_head lli_oss_capas;
145 /* open count currently used by capability only, indicate whether
146 * capability needs renewal */
147 atomic_t lli_open_count;
148 struct obd_capa *lli_mds_capa;
149 cfs_time_t lli_rmtperm_time;
151 /* handle is to be sent to MDS later on done_writing and setattr.
152 * Open handle data are needed for the recovery to reconstruct
153 * the inode state on the MDS. XXX: recovery is not ready yet. */
154 struct obd_client_handle *lli_pending_och;
156 /* We need all three because every inode may be opened in different
158 struct obd_client_handle *lli_mds_read_och;
159 struct obd_client_handle *lli_mds_write_och;
160 struct obd_client_handle *lli_mds_exec_och;
161 __u64 lli_open_fd_read_count;
162 __u64 lli_open_fd_write_count;
163 __u64 lli_open_fd_exec_count;
164 /* Protects access to och pointers and their usage counters */
165 struct mutex lli_och_mutex;
167 struct inode lli_vfs_inode;
169 /* the most recent timestamps obtained from mds */
170 struct ost_lvb lli_lvb;
171 spinlock_t lli_agl_lock;
173 /* Try to make the d::member and f::member are aligned. Before using
174 * these members, make clear whether it is directory or not. */
178 /* serialize normal readdir and statahead-readdir. */
179 struct mutex d_readdir_mutex;
181 /* metadata statahead */
182 /* since parent-child threads can share the same @file
183 * struct, "opendir_key" is the token when dir close for
184 * case of parent exit before child -- it is me should
185 * cleanup the dir readahead. */
187 struct ll_statahead_info *d_sai;
188 /* protect statahead stuff. */
189 spinlock_t d_sa_lock;
190 /* "opendir_pid" is the token when lookup/revalid
191 * -- I am the owner of dir statahead. */
193 /* stat will try to access statahead entries or start
194 * statahead if this flag is set, and this flag will be
195 * set upon dir open, and cleared when dir is closed,
196 * statahead hit ratio is too low, or start statahead
198 unsigned int d_sa_enabled:1;
199 /* directory stripe information */
200 struct lmv_stripe_md *d_lsm_md;
201 /* striped directory size */
202 loff_t d_stripe_size;
203 /* striped directory nlink */
204 __u64 d_stripe_nlink;
207 #define lli_readdir_mutex u.d.d_readdir_mutex
208 #define lli_opendir_key u.d.d_opendir_key
209 #define lli_sai u.d.d_sai
210 #define lli_sa_lock u.d.d_sa_lock
211 #define lli_sa_enabled u.d.d_sa_enabled
212 #define lli_opendir_pid u.d.d_opendir_pid
213 #define lli_lsm_md u.d.d_lsm_md
214 #define lli_stripe_dir_size u.d.d_stripe_size
215 #define lli_stripe_dir_nlink u.d.d_stripe_nlink
217 /* for non-directory */
219 struct mutex f_size_mutex;
220 char *f_symlink_name;
223 * struct rw_semaphore {
224 * signed long count; // align d.d_def_acl
225 * spinlock_t wait_lock; // align d.d_sa_lock
226 * struct list_head wait_list;
229 struct rw_semaphore f_trunc_sem;
230 struct mutex f_write_mutex;
232 struct rw_semaphore f_glimpse_sem;
233 cfs_time_t f_glimpse_time;
234 struct list_head f_agl_list;
237 /* for writepage() only to communicate to fsync */
241 * whenever a process try to read/write the file, the
242 * jobid of the process will be saved here, and it'll
243 * be packed into the write PRC when flush later.
245 * so the read/write statistics for jobid will not be
246 * accurate if the file is shared by different jobs.
248 char f_jobid[JOBSTATS_JOBID_SIZE];
251 #define lli_size_mutex u.f.f_size_mutex
252 #define lli_symlink_name u.f.f_symlink_name
253 #define lli_maxbytes u.f.f_maxbytes
254 #define lli_trunc_sem u.f.f_trunc_sem
255 #define lli_write_mutex u.f.f_write_mutex
256 #define lli_glimpse_sem u.f.f_glimpse_sem
257 #define lli_glimpse_time u.f.f_glimpse_time
258 #define lli_agl_list u.f.f_agl_list
259 #define lli_agl_index u.f.f_agl_index
260 #define lli_async_rc u.f.f_async_rc
261 #define lli_jobid u.f.f_jobid
265 /* XXX: For following frequent used members, although they maybe special
266 * used for non-directory object, it is some time-wasting to check
267 * whether the object is directory or not before using them. On the
268 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
269 * the "ll_inode_info" size even if moving those members into u.f.
270 * So keep them out side.
272 * In the future, if more members are added only for directory,
273 * some of the following members can be moved into u.f.
276 struct cl_object *lli_clob;
278 /* mutex to request for layout lock exclusively. */
279 struct mutex lli_layout_mutex;
280 /* Layout version, protected by lli_layout_lock */
281 __u32 lli_layout_gen;
282 spinlock_t lli_layout_lock;
284 struct rw_semaphore lli_xattrs_list_rwsem;
285 struct mutex lli_xattrs_enq_lock;
286 struct list_head lli_xattrs; /* ll_xattr_entry->xe_list */
289 static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
293 spin_lock(&lli->lli_layout_lock);
294 gen = lli->lli_layout_gen;
295 spin_unlock(&lli->lli_layout_lock);
300 static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
302 spin_lock(&lli->lli_layout_lock);
303 lli->lli_layout_gen = gen;
304 spin_unlock(&lli->lli_layout_lock);
307 int ll_xattr_cache_destroy(struct inode *inode);
309 int ll_xattr_cache_get(struct inode *inode,
316 * Locking to guarantee consistency of non-atomic updates to long long i_size,
317 * consistency between file size and KMS.
319 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
322 void ll_inode_size_lock(struct inode *inode);
323 void ll_inode_size_unlock(struct inode *inode);
325 // FIXME: replace the name of this with LL_I to conform to kernel stuff
326 // static inline struct ll_inode_info *LL_I(struct inode *inode)
327 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
329 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
332 /* default to about 40meg of readahead on a given system. That much tied
333 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
334 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT))
336 /* default to read-ahead full files smaller than 2MB on the second read */
337 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
342 RA_STAT_DISTANT_READPAGE,
343 RA_STAT_MISS_IN_WINDOW,
344 RA_STAT_FAILED_GRAB_PAGE,
345 RA_STAT_FAILED_MATCH,
350 RA_STAT_MAX_IN_FLIGHT,
351 RA_STAT_WRONG_GRAB_PAGE,
352 RA_STAT_FAILED_REACH_END,
357 atomic_t ra_cur_pages;
358 unsigned long ra_max_pages;
359 unsigned long ra_max_pages_per_file;
360 unsigned long ra_max_read_ahead_whole_pages;
363 /* ra_io_arg will be filled in the beginning of ll_readahead with
364 * ras_lock, then the following ll_read_ahead_pages will read RA
365 * pages according to this arg, all the items in this structure are
366 * counted by page index.
369 unsigned long ria_start; /* start offset of read-ahead*/
370 unsigned long ria_end; /* end offset of read-ahead*/
371 /* If stride read pattern is detected, ria_stoff means where
372 * stride read is started. Note: for normal read-ahead, the
373 * value here is meaningless, and also it will not be accessed*/
375 /* ria_length and ria_pages are the length and pages length in the
376 * stride I/O mode. And they will also be used to check whether
377 * it is stride I/O read-ahead in the read-ahead pages*/
378 unsigned long ria_length;
379 unsigned long ria_pages;
382 /* LL_HIST_MAX=32 causes an overflow */
383 #define LL_HIST_MAX 28
384 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
385 #define LL_PROCESS_HIST_MAX 10
386 struct per_process_info {
388 struct obd_histogram pp_r_hist;
389 struct obd_histogram pp_w_hist;
392 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
393 struct ll_rw_extents_info {
394 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
397 #define LL_OFFSET_HIST_MAX 100
398 struct ll_rw_process_info {
401 loff_t rw_range_start;
403 loff_t rw_last_file_pos;
405 size_t rw_smallest_extent;
406 size_t rw_largest_extent;
407 struct ll_file_data *rw_last_file;
410 enum stats_track_type {
411 STATS_TRACK_ALL = 0, /* track all processes */
412 STATS_TRACK_PID, /* track process with this pid */
413 STATS_TRACK_PPID, /* track processes with this ppid */
414 STATS_TRACK_GID, /* track processes with this gid */
418 /* flags for sbi->ll_flags */
419 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
420 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
421 #define LL_SBI_FLOCK 0x04
422 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
423 #define LL_SBI_ACL 0x10 /* support ACL */
424 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
425 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
426 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
427 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
428 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
429 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
430 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
431 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
432 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
433 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
434 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
435 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
436 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
437 #define LL_SBI_XATTR_CACHE 0x80000 /* support for xattr cache */
438 #define LL_SBI_NOROOTSQUASH 0x100000 /* do not apply root squash */
440 #define LL_SBI_FLAGS { \
464 #define RCE_HASHES 32
466 struct rmtacl_ctl_entry {
467 struct list_head rce_list;
468 pid_t rce_key; /* hash key */
469 int rce_ops; /* acl operation type */
472 struct rmtacl_ctl_table {
474 struct list_head rct_entries[RCE_HASHES];
480 struct list_head ee_list;
481 pid_t ee_key; /* hash key */
482 struct lu_fid ee_fid;
483 int ee_type; /* ACL type for ACCESS or DEFAULT */
484 ext_acl_xattr_header *ee_acl;
489 struct list_head et_entries[EE_HASHES];
493 struct list_head ll_list;
494 /* this protects pglist and ra_info. It isn't safe to
495 * grab from interrupt contexts */
497 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
498 spinlock_t ll_process_lock; /* ll_rw_process_info */
499 struct obd_uuid ll_sb_uuid;
500 struct obd_export *ll_md_exp;
501 struct obd_export *ll_dt_exp;
502 struct proc_dir_entry* ll_proc_root;
503 struct lu_fid ll_root_fid; /* root object fid */
506 unsigned int ll_umounting:1,
507 ll_xattr_cache_enabled:1;
508 /* per-conn chain of SBs */
509 struct list_head ll_conn_chain;
510 struct lustre_client_ocd ll_lco;
512 /*please don't ask -p*/
513 struct list_head ll_orphan_dentry_list;
514 struct ll_close_queue *ll_lcq;
516 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
518 /* Used to track "unstable" pages on a client, and maintain a
519 * LRU list of clean pages. An "unstable" page is defined as
520 * any page which is sent to a server as part of a bulk request,
521 * but is uncommitted to stable storage. */
522 struct cl_client_cache ll_cache;
524 struct lprocfs_stats *ll_ra_stats;
526 struct ll_ra_info ll_ra_info;
527 unsigned int ll_namelen;
528 struct file_operations *ll_fop;
530 /* =0 - hold lock over whole read/write
531 * >0 - max. chunk to be read/written w/o lock re-acquiring */
532 unsigned long ll_max_rw_chunk;
533 unsigned int ll_md_brw_pages; /* readdir pages per RPC */
535 struct lu_site *ll_site;
536 struct cl_device *ll_cl;
538 struct ll_rw_extents_info ll_rw_extents_info;
539 int ll_extent_process_count;
540 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
541 unsigned int ll_offset_process_count;
542 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
543 unsigned int ll_rw_offset_entry_count;
544 int ll_stats_track_id;
545 enum stats_track_type ll_stats_track_type;
548 /* metadata stat-ahead */
549 unsigned int ll_sa_max; /* max statahead RPCs */
550 atomic_t ll_sa_total; /* statahead thread started
552 atomic_t ll_sa_wrong; /* statahead thread stopped for
554 atomic_t ll_sa_running; /* running statahead thread
556 atomic_t ll_agl_total; /* AGL thread started count */
558 dev_t ll_sdev_orig; /* save s_dev before assign for
560 struct rmtacl_ctl_table ll_rct;
561 struct eacl_table ll_et;
564 struct root_squash_info ll_squash;
567 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
572 struct task_struct *lrr_reader;
573 struct list_head lrr_linkage;
577 * per file-descriptor read-ahead data.
579 struct ll_readahead_state {
582 * index of the last page that read(2) needed and that wasn't in the
583 * cache. Used by ras_update() to detect seeks.
585 * XXX nikita: if access seeks into cached region, Lustre doesn't see
588 unsigned long ras_last_readpage;
590 * number of pages read after last read-ahead window reset. As window
591 * is reset on each seek, this is effectively a number of consecutive
592 * accesses. Maybe ->ras_accessed_in_window is better name.
594 * XXX nikita: window is also reset (by ras_update()) when Lustre
595 * believes that memory pressure evicts read-ahead pages. In that
596 * case, it probably doesn't make sense to expand window to
597 * PTLRPC_MAX_BRW_PAGES on the third access.
599 unsigned long ras_consecutive_pages;
601 * number of read requests after the last read-ahead window reset
602 * As window is reset on each seek, this is effectively the number
603 * on consecutive read request and is used to trigger read-ahead.
605 unsigned long ras_consecutive_requests;
607 * Parameters of current read-ahead window. Handled by
608 * ras_update(). On the initial access to the file or after a seek,
609 * window is reset to 0. After 3 consecutive accesses, window is
610 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
611 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
613 unsigned long ras_window_start, ras_window_len;
615 * Where next read-ahead should start at. This lies within read-ahead
616 * window. Read-ahead window is read in pieces rather than at once
617 * because: 1. lustre limits total number of pages under read-ahead by
618 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
619 * not covered by DLM lock.
621 unsigned long ras_next_readahead;
623 * Total number of ll_file_read requests issued, reads originating
624 * due to mmap are not counted in this total. This value is used to
625 * trigger full file read-ahead after multiple reads to a small file.
627 unsigned long ras_requests;
629 * Page index with respect to the current request, these value
630 * will not be accurate when dealing with reads issued via mmap.
632 unsigned long ras_request_index;
634 * list of struct ll_ra_read's one per read(2) call current in
635 * progress against this file descriptor. Used by read-ahead code,
636 * protected by ->ras_lock.
638 struct list_head ras_read_beads;
640 * The following 3 items are used for detecting the stride I/O
642 * In stride I/O mode,
643 * ...............|-----data-----|****gap*****|--------|******|....
644 * offset |-stride_pages-|-stride_gap-|
645 * ras_stride_offset = offset;
646 * ras_stride_length = stride_pages + stride_gap;
647 * ras_stride_pages = stride_pages;
648 * Note: all these three items are counted by pages.
650 unsigned long ras_stride_length;
651 unsigned long ras_stride_pages;
652 pgoff_t ras_stride_offset;
654 * number of consecutive stride request count, and it is similar as
655 * ras_consecutive_requests, but used for stride I/O mode.
656 * Note: only more than 2 consecutive stride request are detected,
657 * stride read-ahead will be enable
659 unsigned long ras_consecutive_stride_requests;
662 extern struct kmem_cache *ll_file_data_slab;
663 struct lustre_handle;
664 struct ll_file_data {
665 struct ll_readahead_state fd_ras;
666 struct ccc_grouplock fd_grouplock;
670 /* openhandle if lease exists for this file.
671 * Borrow lli->lli_och_mutex to protect assignment */
672 struct obd_client_handle *fd_lease_och;
673 struct obd_client_handle *fd_och;
674 struct file *fd_file;
675 /* Indicate whether need to report failure when close.
676 * true: failure is known, not report again.
677 * false: unknown failure, should report. */
678 bool fd_write_failed;
679 rwlock_t fd_lock; /* protect lcc list */
680 struct list_head fd_lccs; /* list of ll_cl_context */
683 struct lov_stripe_md;
685 extern spinlock_t inode_lock;
687 extern struct proc_dir_entry *proc_lustre_fs_root;
689 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
691 return &lli->lli_vfs_inode;
694 __u32 ll_i2suppgid(struct inode *i);
695 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
697 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
699 #if BITS_PER_LONG == 32
701 #elif defined(CONFIG_COMPAT)
702 return unlikely(is_compat_task() || (sbi->ll_flags & LL_SBI_32BIT_API));
704 return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
708 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
709 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
710 struct ll_ra_read *ll_ra_read_get(struct file *f);
712 /* llite/lproc_llite.c */
714 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
715 struct super_block *sb, char *osc, char *mdc);
716 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
717 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
718 extern struct lprocfs_seq_vars lprocfs_llite_obd_vars[];
720 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
721 struct super_block *sb, char *osc, char *mdc){return 0;}
722 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
723 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
728 extern const struct file_operations ll_dir_operations;
729 extern const struct inode_operations ll_dir_inode_operations;
730 #ifdef HAVE_DIR_CONTEXT
731 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
732 struct dir_context *ctx);
734 int ll_dir_read(struct inode *inode, __u64 *pos, struct md_op_data *op_data,
735 void *cookie, filldir_t filldir);
737 int ll_get_mdt_idx(struct inode *inode);
738 int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
739 struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
740 __u64 offset, struct ll_dir_chain *chain);
741 void ll_release_page(struct inode *inode, struct page *page, bool remove);
744 extern const struct inode_operations ll_special_inode_operations;
746 int ll_objects_destroy(struct ptlrpc_request *request,
748 struct inode *ll_iget(struct super_block *sb, ino_t hash,
749 struct lustre_md *lic);
750 int ll_test_inode_by_fid(struct inode *inode, void *opaque);
751 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
752 void *data, int flag);
753 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
754 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
755 void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
758 int ll_writepage(struct page *page, struct writeback_control *wbc);
759 int ll_writepages(struct address_space *, struct writeback_control *wbc);
760 int ll_readpage(struct file *file, struct page *page);
761 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
762 int ll_readahead(const struct lu_env *env, struct cl_io *io,
763 struct cl_page_list *queue, struct ll_readahead_state *ras,
765 int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
766 struct ll_cl_context *ll_cl_find(struct file *file);
767 void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io);
768 void ll_cl_remove(struct file *file, const struct lu_env *env);
770 #ifndef MS_HAS_NEW_AOPS
771 extern const struct address_space_operations ll_aops;
773 extern const struct address_space_operations_ext ll_aops;
777 extern struct file_operations ll_file_operations;
778 extern struct file_operations ll_file_operations_flock;
779 extern struct file_operations ll_file_operations_noflock;
780 extern struct inode_operations ll_file_inode_operations;
781 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
782 ldlm_mode_t l_req_mode);
783 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
784 struct lustre_handle *lockh, __u64 flags,
787 int ll_file_open(struct inode *inode, struct file *file);
788 int ll_file_release(struct inode *inode, struct file *file);
789 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
790 struct lov_stripe_md *lsm, lstat_t *st);
791 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
792 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
793 int ll_md_real_close(struct inode *inode, fmode_t fmode);
794 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
795 struct obd_client_handle **och, unsigned long flags);
796 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
797 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
798 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
799 __u64 ioepoch, int sync);
800 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
801 struct lustre_handle *fh);
802 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
803 struct ll_file_data *file, loff_t pos,
804 size_t count, int rw);
805 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
806 struct posix_acl *ll_get_acl(struct inode *inode, int type);
807 int ll_migrate(struct inode *parent, struct file *file, int mdtidx,
808 const char *name, int namelen);
809 int ll_get_fid_by_name(struct inode *parent, const char *name,
810 int namelen, struct lu_fid *fid);
811 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
812 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
814 # ifndef HAVE_INODE_PERMISION_2ARGS
815 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
817 int ll_inode_permission(struct inode *inode, int mask);
821 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
822 __u64 flags, struct lov_user_md *lum,
824 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
825 struct lov_mds_md **lmm, int *lmm_size,
826 struct ptlrpc_request **request);
827 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
829 int ll_dir_getstripe(struct inode *inode, void **lmmp,
830 int *lmm_size, struct ptlrpc_request **request,
832 #ifdef HAVE_FILE_FSYNC_4ARGS
833 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
834 #elif defined(HAVE_FILE_FSYNC_2ARGS)
835 int ll_fsync(struct file *file, int data);
837 int ll_fsync(struct file *file, struct dentry *dentry, int data);
839 int ll_merge_lvb(const struct lu_env *env, struct inode *inode);
840 int ll_fid2path(struct inode *inode, void __user *arg);
841 int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
842 int ll_hsm_release(struct inode *inode);
846 int ll_d_init(struct dentry *de);
847 extern const struct dentry_operations ll_d_ops;
848 void ll_intent_drop_lock(struct lookup_intent *);
849 void ll_intent_release(struct lookup_intent *);
850 void ll_invalidate_aliases(struct inode *);
851 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
852 int ll_revalidate_it_finish(struct ptlrpc_request *request,
853 struct lookup_intent *it, struct dentry *de);
855 /* llite/llite_lib.c */
856 extern struct super_operations lustre_super_operations;
858 void ll_lli_init(struct ll_inode_info *lli);
859 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
860 void ll_put_super(struct super_block *sb);
861 void ll_kill_super(struct super_block *sb);
862 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
863 void ll_clear_inode(struct inode *inode);
864 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import);
865 int ll_setattr(struct dentry *de, struct iattr *attr);
866 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
867 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
868 __u64 max_age, __u32 flags);
869 int ll_update_inode(struct inode *inode, struct lustre_md *md);
870 int ll_read_inode2(struct inode *inode, void *opaque);
871 void ll_delete_inode(struct inode *inode);
872 int ll_iocontrol(struct inode *inode, struct file *file,
873 unsigned int cmd, unsigned long arg);
874 int ll_flush_ctx(struct inode *inode);
875 void ll_umount_begin(struct super_block *sb);
876 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
877 #ifdef HAVE_SUPEROPS_USE_DENTRY
878 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
880 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
882 void ll_dirty_page_discard_warn(struct page *page, int ioret);
883 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
884 struct super_block *, struct lookup_intent *);
885 void lustre_dump_dentry(struct dentry *, int recur);
886 int ll_obd_statfs(struct inode *inode, void *arg);
887 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
888 int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
889 int ll_get_max_cookiesize(struct ll_sb_info *sbi, int *max_cookiesize);
890 int ll_get_default_cookiesize(struct ll_sb_info *sbi, int *default_cookiesize);
891 int ll_process_config(struct lustre_cfg *lcfg);
892 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
893 struct inode *i1, struct inode *i2,
894 const char *name, int namelen,
895 int mode, __u32 opc, void *data);
896 void ll_finish_md_op_data(struct md_op_data *op_data);
897 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
898 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
899 void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
901 /* llite/llite_nfs.c */
902 extern struct export_operations lustre_export_operations;
903 __u32 get_uuid2int(const char *name, int len);
904 struct inode *search_inode_for_lustre(struct super_block *sb,
905 const struct lu_fid *fid);
906 int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
908 /* llite/symlink.c */
909 extern struct inode_operations ll_fast_symlink_inode_operations;
911 /* llite/llite_close.c */
912 struct ll_close_queue {
914 struct list_head lcq_head;
915 wait_queue_head_t lcq_waitq;
916 struct completion lcq_comp;
920 struct ccc_object *cl_inode2ccc(struct inode *inode);
923 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
924 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
926 /* specific achitecture can implement only part of this list */
927 enum vvp_io_subtype {
930 /** io started from splice_{read|write} */
937 enum vvp_io_subtype cui_io_subtype;
941 struct pipe_inode_info *cui_pipe;
942 unsigned int cui_flags;
944 struct vvp_fault_io {
946 * Inode modification time that is checked across DLM
950 struct vm_area_struct *ft_vma;
952 * locked page returned from vvp_io
954 struct page *ft_vmpage;
955 struct vm_fault_api {
959 struct vm_fault *ft_vmf;
961 * fault API used bitflags for return code.
963 unsigned int ft_flags;
965 * check that flags are from filemap_fault
972 * Read-ahead state used by read and page-fault IO contexts.
974 struct ll_ra_read cui_bead;
976 * Set when cui_bead has been initialized.
978 int cui_ra_window_set;
982 * IO arguments for various VFS I/O interfaces.
985 /** normal/sendfile/splice */
986 enum vvp_io_subtype via_io_subtype;
990 struct kiocb *via_iocb;
991 struct iovec *via_iov;
992 unsigned long via_nrsegs;
995 struct pipe_inode_info *via_pipe;
996 unsigned int via_flags;
1001 struct ll_cl_context {
1002 struct list_head lcc_list;
1004 const struct lu_env *lcc_env;
1005 struct cl_io *lcc_io;
1006 struct cl_page *lcc_page;
1009 struct vvp_thread_info {
1010 struct iovec vti_local_iov;
1011 struct vvp_io_args vti_args;
1012 struct ra_io_arg vti_ria;
1013 struct kiocb vti_kiocb;
1014 struct ll_cl_context vti_io_ctx;
1017 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
1019 extern struct lu_context_key vvp_key;
1020 struct vvp_thread_info *info;
1022 info = lu_context_key_get(&env->le_ctx, &vvp_key);
1023 LASSERT(info != NULL);
1027 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
1028 enum vvp_io_subtype type)
1030 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
1032 ret->via_io_subtype = type;
1037 struct vvp_session {
1038 struct vvp_io vs_ios;
1041 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1043 extern struct lu_context_key vvp_session_key;
1044 struct vvp_session *ses;
1046 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1047 LASSERT(ses != NULL);
1051 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1053 return &vvp_env_session(env)->vs_ios;
1056 int vvp_global_init(void);
1057 void vvp_global_fini(void);
1059 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1060 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1061 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1063 /* llite/llite_mmap.c */
1065 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1066 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1067 void policy_from_vma(ldlm_policy_data_t *policy,
1068 struct vm_area_struct *vma, unsigned long addr, size_t count);
1069 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1072 static inline void ll_invalidate_page(struct page *vmpage)
1074 struct address_space *mapping = vmpage->mapping;
1075 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1077 LASSERT(PageLocked(vmpage));
1078 if (mapping == NULL)
1082 * truncate_complete_page() calls
1083 * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
1085 ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE);
1086 truncate_complete_page(mapping, vmpage);
1089 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1091 /* don't need an addref as the sb_info should be holding one */
1092 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1094 return ll_s2sbi(sb)->ll_dt_exp;
1097 /* don't need an addref as the sb_info should be holding one */
1098 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1100 return ll_s2sbi(sb)->ll_md_exp;
1103 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1105 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1111 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1112 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1114 return ll_s2sbi(inode->i_sb);
1117 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1119 return ll_s2dtexp(inode->i_sb);
1122 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1124 return ll_s2mdexp(inode->i_sb);
1127 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1131 LASSERT(inode != NULL);
1132 fid = &ll_i2info(inode)->lli_fid;
1137 static inline __u64 ll_file_maxbytes(struct inode *inode)
1139 return ll_i2info(inode)->lli_maxbytes;
1143 int ll_setxattr(struct dentry *dentry, const char *name,
1144 const void *value, size_t size, int flags);
1145 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1146 void *buffer, size_t size);
1147 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1148 int ll_removexattr(struct dentry *dentry, const char *name);
1150 /* llite/remote_perm.c */
1151 extern struct kmem_cache *ll_remote_perm_cachep;
1152 extern struct kmem_cache *ll_rmtperm_hash_cachep;
1154 void free_rmtperm_hash(struct hlist_head *hash);
1155 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1156 int lustre_check_remote_perm(struct inode *inode, int mask);
1158 /* llite/llite_capa.c */
1159 extern struct timer_list ll_capa_timer;
1161 int ll_capa_thread_start(void);
1162 void ll_capa_thread_stop(void);
1163 void ll_capa_timer_callback(unsigned long unused);
1165 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1167 void ll_capa_open(struct inode *inode);
1168 void ll_capa_close(struct inode *inode);
1170 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1171 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1173 void ll_truncate_free_capa(struct obd_capa *ocapa);
1174 void ll_clear_inode_capas(struct inode *inode);
1175 void ll_print_capa_stat(struct ll_sb_info *sbi);
1177 /* llite/llite_cl.c */
1178 extern struct lu_device_type vvp_device_type;
1181 * Common IO arguments for various VFS I/O interfaces.
1183 int cl_sb_init(struct super_block *sb);
1184 int cl_sb_fini(struct super_block *sb);
1186 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1187 struct ll_readahead_state *ras, unsigned long index,
1189 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1190 void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1192 /* llite/llite_rmtacl.c */
1193 #ifdef CONFIG_FS_POSIX_ACL
1194 obd_valid rce_ops2valid(int ops);
1195 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1196 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1197 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1198 void rct_init(struct rmtacl_ctl_table *rct);
1199 void rct_fini(struct rmtacl_ctl_table *rct);
1201 void ee_free(struct eacl_entry *ee);
1202 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1203 ext_acl_xattr_header *header);
1204 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1205 struct lu_fid *fid, int type);
1206 void et_search_free(struct eacl_table *et, pid_t key);
1207 void et_init(struct eacl_table *et);
1208 void et_fini(struct eacl_table *et);
1213 #define LL_SA_RPC_MIN 2
1214 #define LL_SA_RPC_DEF 32
1215 #define LL_SA_RPC_MAX 8192
1217 #define LL_SA_CACHE_BIT 5
1218 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1219 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1221 /* per inode struct, for dir only */
1222 struct ll_statahead_info {
1223 struct inode *sai_inode;
1224 atomic_t sai_refcount; /* when access this struct, hold
1226 unsigned int sai_generation; /* generation for statahead */
1227 unsigned int sai_max; /* max ahead of lookup */
1228 __u64 sai_sent; /* stat requests sent count */
1229 __u64 sai_replied; /* stat requests which received
1231 __u64 sai_index; /* index of statahead entry */
1232 __u64 sai_index_wait; /* index of entry which is the
1233 * caller is waiting for */
1234 __u64 sai_hit; /* hit count */
1235 __u64 sai_miss; /* miss count:
1236 * for "ls -al" case, it
1237 * includes hidden dentry miss;
1238 * for "ls -l" case, it does not
1239 * include hidden dentry miss.
1240 * "sai_miss_hidden" is used for
1243 unsigned int sai_consecutive_miss; /* consecutive miss */
1244 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1245 * is not a hidden one */
1246 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1247 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1249 sai_agl_valid:1,/* AGL is valid for the dir */
1250 sai_in_readpage:1;/* statahead is in readdir()*/
1251 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1252 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1253 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1254 struct list_head sai_entries; /* entry list */
1255 struct list_head sai_entries_received; /* entries returned */
1256 struct list_head sai_entries_stated; /* entries stated */
1257 struct list_head sai_entries_agl; /* AGL entries to be sent */
1258 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1259 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1260 atomic_t sai_cache_count; /* entry count in cache */
1263 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1265 void ll_authorize_statahead(struct inode *dir, void *key);
1266 void ll_deauthorize_statahead(struct inode *dir, void *key);
1268 static inline int ll_glimpse_size(struct inode *inode)
1270 struct ll_inode_info *lli = ll_i2info(inode);
1273 down_read(&lli->lli_glimpse_sem);
1274 rc = cl_glimpse_size(inode);
1275 lli->lli_glimpse_time = cfs_time_current();
1276 up_read(&lli->lli_glimpse_sem);
1281 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1283 struct ll_inode_info *lli = ll_i2info(dir);
1284 struct ll_statahead_info *sai = lli->lli_sai;
1285 struct ll_dentry_data *ldd = ll_d2d(dentry);
1287 /* not the same process, don't mark */
1288 if (lli->lli_opendir_pid != current_pid())
1291 LASSERT(ldd != NULL);
1293 ldd->lld_sa_generation = sai->sai_generation;
1297 dentry_need_statahead(struct inode *dir, struct dentry *dentry)
1299 struct ll_inode_info *lli;
1300 struct ll_dentry_data *ldd;
1302 if (ll_i2sbi(dir)->ll_sa_max == 0)
1305 lli = ll_i2info(dir);
1307 /* statahead is not allowed for this dir, there may be three causes:
1308 * 1. dir is not opened.
1309 * 2. statahead hit ratio is too low.
1310 * 3. previous stat started statahead thread failed. */
1311 if (!lli->lli_sa_enabled)
1314 /* not the same process, don't statahead */
1315 if (lli->lli_opendir_pid != current_pid())
1318 ldd = ll_d2d(dentry);
1320 * When stats a dentry, the system trigger more than once "revalidate"
1321 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1322 * Under patchless client mode, the operation intent is not accurate,
1323 * which maybe misguide the statahead thread. For example:
1324 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1325 * have the same operation intent -- "IT_GETATTR".
1326 * In fact, one dentry should has only one chance to interact with the
1327 * statahead thread, otherwise the statahead windows will be confused.
1328 * The solution is as following:
1329 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1330 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1331 * will bypass interacting with statahead thread for checking:
1332 * "lld_sa_generation == lli_sai->sai_generation"
1334 if (ldd && lli->lli_sai &&
1335 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1342 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1344 if (!dentry_need_statahead(dir, *dentryp))
1347 return do_statahead_enter(dir, dentryp, only_unplug);
1350 /* llite ioctl register support rountine */
1356 #define LLIOC_MAX_CMD 256
1359 * Rules to write a callback function:
1362 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1363 * returned to ll_iocontrol_register. Callback functions should use this
1364 * data to check the potential collasion of ioctl cmd. If collasion is
1365 * found, callback function should return LLIOC_CONT.
1366 * @rcp: The result of ioctl command.
1369 * If @magic matches the pointer returned by ll_iocontrol_data, the
1370 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1372 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1373 struct file *file, unsigned int cmd, unsigned long arg,
1374 void *magic, int *rcp);
1376 /* export functions */
1377 /* Register ioctl block dynamatically for a regular file.
1379 * @cmd: the array of ioctl command set
1380 * @count: number of commands in the @cmd
1381 * @cb: callback function, it will be called if an ioctl command is found to
1382 * belong to the command list @cmd.
1385 * A magic pointer will be returned if success;
1386 * otherwise, NULL will be returned.
1388 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1389 void ll_iocontrol_unregister(void *magic);
1392 /* lclient compat stuff */
1393 #define cl_inode_info ll_inode_info
1394 #define cl_i2info(info) ll_i2info(info)
1395 #define cl_inode_mode(inode) ((inode)->i_mode)
1396 #define cl_i2sbi ll_i2sbi
1398 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1399 const struct iattr *attr)
1401 LASSERT(attr->ia_valid & ATTR_FILE);
1402 return LUSTRE_FPRIVATE(attr->ia_file);
1405 static inline void cl_isize_lock(struct inode *inode)
1407 ll_inode_size_lock(inode);
1410 static inline void cl_isize_unlock(struct inode *inode)
1412 ll_inode_size_unlock(inode);
1415 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1417 LASSERT(mutex_is_locked(&ll_i2info(inode)->lli_size_mutex));
1418 i_size_write(inode, kms);
1421 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1423 ll_inode_size_lock(inode);
1424 i_size_write(inode, kms);
1425 ll_inode_size_unlock(inode);
1428 #define cl_isize_read(inode) i_size_read(inode)
1430 static inline int cl_merge_lvb(const struct lu_env *env, struct inode *inode)
1432 return ll_merge_lvb(env, inode);
1435 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1436 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1437 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1439 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1441 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1442 enum cl_fsync_mode mode, int ignore_layout);
1444 /** direct write pages */
1445 struct ll_dio_pages {
1446 /** page array to be written. we don't support
1447 * partial pages except the last one. */
1448 struct page **ldp_pages;
1449 /* offset of each page */
1450 loff_t *ldp_offsets;
1451 /** if ldp_offsets is NULL, it means a sequential
1452 * pages to be written, then this is the file offset
1453 * of the * first page. */
1454 loff_t ldp_start_offset;
1455 /** how many bytes are to be written. */
1457 /** # of pages in the array. */
1461 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1464 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1467 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1470 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1471 int rw, struct inode *inode,
1472 struct ll_dio_pages *pv);
1474 static inline int ll_file_nolock(const struct file *file)
1476 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1477 struct inode *inode = file->f_dentry->d_inode;
1479 LASSERT(fd != NULL);
1480 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1481 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1484 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1485 struct lookup_intent *it, __u64 *bits)
1487 if (!it->d.lustre.it_lock_set) {
1488 struct lustre_handle handle;
1490 /* If this inode is a remote object, it will get two
1491 * separate locks in different namespaces, Master MDT,
1492 * where the name entry is, will grant LOOKUP lock,
1493 * remote MDT, where the object is, will grant
1494 * UPDATE|PERM lock. The inode will be attched to both
1495 * LOOKUP and PERM locks, so revoking either locks will
1496 * case the dcache being cleared */
1497 if (it->d.lustre.it_remote_lock_mode) {
1498 handle.cookie = it->d.lustre.it_remote_lock_handle;
1499 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID
1500 "(%p) for remote lock "LPX64"\n",
1501 PFID(ll_inode2fid(inode)), inode,
1503 md_set_lock_data(exp, &handle.cookie, inode, NULL);
1506 handle.cookie = it->d.lustre.it_lock_handle;
1508 CDEBUG(D_DLMTRACE, "setting l_data to inode "DFID"(%p)"
1509 " for lock "LPX64"\n",
1510 PFID(ll_inode2fid(inode)), inode, handle.cookie);
1512 md_set_lock_data(exp, &handle.cookie, inode,
1513 &it->d.lustre.it_lock_bits);
1514 it->d.lustre.it_lock_set = 1;
1518 *bits = it->d.lustre.it_lock_bits;
1521 static inline void ll_lock_dcache(struct inode *inode)
1523 #ifdef HAVE_DCACHE_LOCK
1524 spin_lock(&dcache_lock);
1526 spin_lock(&inode->i_lock);
1530 static inline void ll_unlock_dcache(struct inode *inode)
1532 #ifdef HAVE_DCACHE_LOCK
1533 spin_unlock(&dcache_lock);
1535 spin_unlock(&inode->i_lock);
1539 static inline int d_lustre_invalid(const struct dentry *dentry)
1541 struct ll_dentry_data *lld = ll_d2d(dentry);
1543 return (lld == NULL) || lld->lld_invalid;
1546 static inline void __d_lustre_invalidate(struct dentry *dentry)
1548 struct ll_dentry_data *lld = ll_d2d(dentry);
1551 lld->lld_invalid = 1;
1555 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1556 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1557 * else dput() of the last refcount will unhash this dentry and kill it.
1559 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1561 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1562 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1563 dentry->d_parent, dentry->d_inode, ll_d_count(dentry));
1565 spin_lock_nested(&dentry->d_lock,
1566 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1567 __d_lustre_invalidate(dentry);
1568 if (ll_d_count(dentry) == 0)
1570 spin_unlock(&dentry->d_lock);
1573 static inline void d_lustre_revalidate(struct dentry *dentry)
1575 spin_lock(&dentry->d_lock);
1576 LASSERT(ll_d2d(dentry) != NULL);
1577 ll_d2d(dentry)->lld_invalid = 0;
1578 spin_unlock(&dentry->d_lock);
1581 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 53, 0)
1582 /* Compatibility for old (1.8) compiled userspace quota code */
1583 struct if_quotactl_18 {
1588 struct obd_dqinfo qc_dqinfo;
1589 struct obd_dqblk qc_dqblk;
1591 struct obd_uuid obd_uuid;
1593 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1594 /* End compatibility for old (1.8) compiled userspace quota code */
1595 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 53, 0) */
1598 LL_LAYOUT_GEN_NONE = ((__u32)-2), /* layout lock was cancelled */
1599 LL_LAYOUT_GEN_EMPTY = ((__u32)-1) /* for empty layout */
1602 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1603 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1604 int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1606 int ll_xattr_init(void);
1607 void ll_xattr_fini(void);
1609 int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1610 struct cl_page *page, enum cl_req_type crt);
1612 #endif /* LLITE_INTERNAL_H */