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, 2012, 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_mdc.h>
48 #include <linux/lustre_intent.h>
49 #include <linux/compat.h>
50 #include <linux/posix_acl_xattr.h>
56 #ifndef VM_FAULT_RETRY
57 #define VM_FAULT_RETRY 0
60 /* Kernel 3.1 kills LOOKUP_CONTINUE, LOOKUP_PARENT is equivalent to it.
61 * seem kernel commit 49084c3bb2055c401f3493c13edae14d49128ca0 */
62 #ifndef LOOKUP_CONTINUE
63 #define LOOKUP_CONTINUE LOOKUP_PARENT
66 /** Only used on client-side for indicating the tail of dir hash/offset. */
67 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
68 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
70 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
71 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
73 struct ll_dentry_data {
76 struct obd_client_handle lld_cwd_och;
77 struct obd_client_handle lld_mnt_och;
78 struct lookup_intent *lld_it;
79 unsigned int lld_sa_generation;
80 unsigned int lld_invalid:1;
81 struct rcu_head lld_rcu_head;
84 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
86 extern struct file_operations ll_pgcache_seq_fops;
88 #define LLI_INODE_MAGIC 0x111d0de5
89 #define LLI_INODE_DEAD 0xdeadd00d
91 /* remote client permission cache */
92 #define REMOTE_PERM_HASHSIZE 16
94 struct ll_getname_data {
95 struct dir_context ctx;
96 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
97 struct lu_fid lgd_fid; /* target fid we are looking for */
98 int lgd_found; /* inode matched? */
101 /* llite setxid/access permission for user on remote client */
102 struct ll_remote_perm {
103 struct hlist_node lrp_list;
108 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
109 is access permission with
110 lrp_fsuid/lrp_fsgid. */
114 /* MDS has an authority for the Size-on-MDS attributes. */
115 LLIF_MDS_SIZE_LOCK = (1 << 0),
116 /* Epoch close is postponed. */
117 LLIF_EPOCH_PENDING = (1 << 1),
118 /* DONE WRITING is allowed. */
119 LLIF_DONE_WRITING = (1 << 2),
120 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
122 LLIF_SOM_DIRTY = (1 << 3),
123 /* File is contented */
124 LLIF_CONTENDED = (1 << 4),
125 /* Truncate uses server lock for this file */
126 LLIF_SRVLOCK = (1 << 5),
127 /* File data is modified. */
128 LLIF_DATA_MODIFIED = (1 << 6),
129 /* File is being restored */
130 LLIF_FILE_RESTORING = (1 << 7),
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 /* Parent fid for accessing default stripe data on parent directory
147 * for allocating OST objects after a mknod() and later open-by-FID. */
148 struct lu_fid lli_pfid;
150 struct list_head lli_close_list;
151 struct list_head lli_oss_capas;
152 /* open count currently used by capability only, indicate whether
153 * capability needs renewal */
154 atomic_t lli_open_count;
155 struct obd_capa *lli_mds_capa;
156 cfs_time_t lli_rmtperm_time;
158 /* handle is to be sent to MDS later on done_writing and setattr.
159 * Open handle data are needed for the recovery to reconstruct
160 * the inode state on the MDS. XXX: recovery is not ready yet. */
161 struct obd_client_handle *lli_pending_och;
163 /* We need all three because every inode may be opened in different
165 struct obd_client_handle *lli_mds_read_och;
166 struct obd_client_handle *lli_mds_write_och;
167 struct obd_client_handle *lli_mds_exec_och;
168 __u64 lli_open_fd_read_count;
169 __u64 lli_open_fd_write_count;
170 __u64 lli_open_fd_exec_count;
171 /* Protects access to och pointers and their usage counters */
172 struct mutex lli_och_mutex;
174 struct inode lli_vfs_inode;
176 /* the most recent timestamps obtained from mds */
177 struct ost_lvb lli_lvb;
178 spinlock_t lli_agl_lock;
180 /* Try to make the d::member and f::member are aligned. Before using
181 * these members, make clear whether it is directory or not. */
185 /* serialize normal readdir and statahead-readdir. */
186 struct mutex d_readdir_mutex;
188 /* metadata statahead */
189 /* since parent-child threads can share the same @file
190 * struct, "opendir_key" is the token when dir close for
191 * case of parent exit before child -- it is me should
192 * cleanup the dir readahead. */
194 struct ll_statahead_info *d_sai;
195 struct posix_acl *d_def_acl;
196 /* protect statahead stuff. */
197 spinlock_t d_sa_lock;
198 /* "opendir_pid" is the token when lookup/revalid
199 * -- I am the owner of dir statahead. */
203 #define lli_readdir_mutex u.d.d_readdir_mutex
204 #define lli_opendir_key u.d.d_opendir_key
205 #define lli_sai u.d.d_sai
206 #define lli_def_acl u.d.d_def_acl
207 #define lli_sa_lock u.d.d_sa_lock
208 #define lli_opendir_pid u.d.d_opendir_pid
210 /* for non-directory */
212 struct semaphore f_size_sem;
213 void *f_size_sem_owner;
214 char *f_symlink_name;
217 * struct rw_semaphore {
218 * signed long count; // align d.d_def_acl
219 * spinlock_t wait_lock; // align d.d_sa_lock
220 * struct list_head wait_list;
223 struct rw_semaphore f_trunc_sem;
224 struct mutex f_write_mutex;
226 struct rw_semaphore f_glimpse_sem;
227 cfs_time_t f_glimpse_time;
228 struct list_head f_agl_list;
231 /* for writepage() only to communicate to fsync */
234 /* volatile file criteria is based on file name, this
235 * flag is used to keep the test result, so the strcmp
240 * whenever a process try to read/write the file, the
241 * jobid of the process will be saved here, and it'll
242 * be packed into the write PRC when flush later.
244 * so the read/write statistics for jobid will not be
245 * accurate if the file is shared by different jobs.
247 char f_jobid[JOBSTATS_JOBID_SIZE];
250 #define lli_size_sem u.f.f_size_sem
251 #define lli_size_sem_owner u.f.f_size_sem_owner
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
262 #define lli_volatile u.f.f_volatile
266 /* XXX: For following frequent used members, although they maybe special
267 * used for non-directory object, it is some time-wasting to check
268 * whether the object is directory or not before using them. On the
269 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
270 * the "ll_inode_info" size even if moving those members into u.f.
271 * So keep them out side.
273 * In the future, if more members are added only for directory,
274 * some of the following members can be moved into u.f.
277 struct cl_object *lli_clob;
279 /* mutex to request for layout lock exclusively. */
280 struct mutex lli_layout_mutex;
281 /* valid only inside LAYOUT ibits lock, protected by lli_layout_mutex */
282 __u32 lli_layout_gen;
286 * Locking to guarantee consistency of non-atomic updates to long long i_size,
287 * consistency between file size and KMS.
289 * Implemented by ->lli_size_sem and ->lsm_lock, nested in that order.
292 void ll_inode_size_lock(struct inode *inode);
293 void ll_inode_size_unlock(struct inode *inode);
295 // FIXME: replace the name of this with LL_I to conform to kernel stuff
296 // static inline struct ll_inode_info *LL_I(struct inode *inode)
297 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
299 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
302 /* default to about 40meg of readahead on a given system. That much tied
303 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
304 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - PAGE_CACHE_SHIFT))
306 /* default to read-ahead full files smaller than 2MB on the second read */
307 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_CACHE_SHIFT))
312 RA_STAT_DISTANT_READPAGE,
313 RA_STAT_MISS_IN_WINDOW,
314 RA_STAT_FAILED_GRAB_PAGE,
315 RA_STAT_FAILED_MATCH,
320 RA_STAT_MAX_IN_FLIGHT,
321 RA_STAT_WRONG_GRAB_PAGE,
326 atomic_t ra_cur_pages;
327 unsigned long ra_max_pages;
328 unsigned long ra_max_pages_per_file;
329 unsigned long ra_max_read_ahead_whole_pages;
332 /* ra_io_arg will be filled in the beginning of ll_readahead with
333 * ras_lock, then the following ll_read_ahead_pages will read RA
334 * pages according to this arg, all the items in this structure are
335 * counted by page index.
338 unsigned long ria_start; /* start offset of read-ahead*/
339 unsigned long ria_end; /* end offset of read-ahead*/
340 /* If stride read pattern is detected, ria_stoff means where
341 * stride read is started. Note: for normal read-ahead, the
342 * value here is meaningless, and also it will not be accessed*/
344 /* ria_length and ria_pages are the length and pages length in the
345 * stride I/O mode. And they will also be used to check whether
346 * it is stride I/O read-ahead in the read-ahead pages*/
347 unsigned long ria_length;
348 unsigned long ria_pages;
351 /* LL_HIST_MAX=32 causes an overflow */
352 #define LL_HIST_MAX 28
353 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
354 #define LL_PROCESS_HIST_MAX 10
355 struct per_process_info {
357 struct obd_histogram pp_r_hist;
358 struct obd_histogram pp_w_hist;
361 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
362 struct ll_rw_extents_info {
363 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
366 #define LL_OFFSET_HIST_MAX 100
367 struct ll_rw_process_info {
370 loff_t rw_range_start;
372 loff_t rw_last_file_pos;
374 size_t rw_smallest_extent;
375 size_t rw_largest_extent;
376 struct ll_file_data *rw_last_file;
379 enum stats_track_type {
380 STATS_TRACK_ALL = 0, /* track all processes */
381 STATS_TRACK_PID, /* track process with this pid */
382 STATS_TRACK_PPID, /* track processes with this ppid */
383 STATS_TRACK_GID, /* track processes with this gid */
387 /* flags for sbi->ll_flags */
388 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
389 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
390 #define LL_SBI_FLOCK 0x04
391 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
392 #define LL_SBI_ACL 0x10 /* support ACL */
393 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
394 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
395 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
396 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
397 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
398 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
399 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
400 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
401 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
402 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
403 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
404 #define LL_SBI_LAYOUT_LOCK 0x20000 /* layout lock support */
405 #define LL_SBI_USER_FID2PATH 0x40000 /* allow fid2path by unprivileged users */
407 #define LL_SBI_FLAGS { \
427 /* default value for ll_sb_info->contention_time */
428 #define SBI_DEFAULT_CONTENTION_SECONDS 60
429 /* default value for lockless_truncate_enable */
430 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
431 #define RCE_HASHES 32
433 struct rmtacl_ctl_entry {
434 struct list_head rce_list;
435 pid_t rce_key; /* hash key */
436 int rce_ops; /* acl operation type */
439 struct rmtacl_ctl_table {
441 struct list_head rct_entries[RCE_HASHES];
448 struct list_head et_entries[EE_HASHES];
452 struct list_head ll_list;
453 /* this protects pglist and ra_info. It isn't safe to
454 * grab from interrupt contexts */
456 spinlock_t ll_pp_extent_lock; /* pp_extent entry*/
457 spinlock_t ll_process_lock; /* ll_rw_process_info */
458 struct obd_uuid ll_sb_uuid;
459 struct obd_export *ll_md_exp;
460 struct obd_export *ll_dt_exp;
461 struct proc_dir_entry* ll_proc_root;
462 struct lu_fid ll_root_fid; /* root object fid */
465 unsigned int ll_umounting:1;
466 struct list_head ll_conn_chain; /* per-conn chain of SBs */
467 struct lustre_client_ocd ll_lco;
469 struct list_head ll_orphan_dentry_list; /*please don't ask -p*/
470 struct ll_close_queue *ll_lcq;
472 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
474 struct cl_client_cache ll_cache;
476 struct lprocfs_stats *ll_ra_stats;
478 struct ll_ra_info ll_ra_info;
479 unsigned int ll_namelen;
480 struct file_operations *ll_fop;
482 /* =0 - hold lock over whole read/write
483 * >0 - max. chunk to be read/written w/o lock re-acquiring */
484 unsigned long ll_max_rw_chunk;
485 unsigned int ll_md_brw_size; /* used by readdir */
487 struct lu_site *ll_site;
488 struct cl_device *ll_cl;
490 struct ll_rw_extents_info ll_rw_extents_info;
491 int ll_extent_process_count;
492 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
493 unsigned int ll_offset_process_count;
494 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
495 unsigned int ll_rw_offset_entry_count;
496 int ll_stats_track_id;
497 enum stats_track_type ll_stats_track_type;
500 /* metadata stat-ahead */
501 unsigned int ll_sa_max; /* max statahead RPCs */
502 atomic_t ll_sa_total; /* statahead thread started
504 atomic_t ll_sa_wrong; /* statahead thread stopped for
506 atomic_t ll_agl_total; /* AGL thread started count */
508 dev_t ll_sdev_orig; /* save s_dev before assign for
510 struct rmtacl_ctl_table ll_rct;
511 struct eacl_table ll_et;
512 __kernel_fsid_t ll_fsid;
515 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
520 struct task_struct *lrr_reader;
521 struct list_head lrr_linkage;
525 * per file-descriptor read-ahead data.
527 struct ll_readahead_state {
530 * index of the last page that read(2) needed and that wasn't in the
531 * cache. Used by ras_update() to detect seeks.
533 * XXX nikita: if access seeks into cached region, Lustre doesn't see
536 unsigned long ras_last_readpage;
538 * number of pages read after last read-ahead window reset. As window
539 * is reset on each seek, this is effectively a number of consecutive
540 * accesses. Maybe ->ras_accessed_in_window is better name.
542 * XXX nikita: window is also reset (by ras_update()) when Lustre
543 * believes that memory pressure evicts read-ahead pages. In that
544 * case, it probably doesn't make sense to expand window to
545 * PTLRPC_MAX_BRW_PAGES on the third access.
547 unsigned long ras_consecutive_pages;
549 * number of read requests after the last read-ahead window reset
550 * As window is reset on each seek, this is effectively the number
551 * on consecutive read request and is used to trigger read-ahead.
553 unsigned long ras_consecutive_requests;
555 * Parameters of current read-ahead window. Handled by
556 * ras_update(). On the initial access to the file or after a seek,
557 * window is reset to 0. After 3 consecutive accesses, window is
558 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
559 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
561 unsigned long ras_window_start, ras_window_len;
563 * Where next read-ahead should start at. This lies within read-ahead
564 * window. Read-ahead window is read in pieces rather than at once
565 * because: 1. lustre limits total number of pages under read-ahead by
566 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
567 * not covered by DLM lock.
569 unsigned long ras_next_readahead;
571 * Total number of ll_file_read requests issued, reads originating
572 * due to mmap are not counted in this total. This value is used to
573 * trigger full file read-ahead after multiple reads to a small file.
575 unsigned long ras_requests;
577 * Page index with respect to the current request, these value
578 * will not be accurate when dealing with reads issued via mmap.
580 unsigned long ras_request_index;
582 * list of struct ll_ra_read's one per read(2) call current in
583 * progress against this file descriptor. Used by read-ahead code,
584 * protected by ->ras_lock.
586 struct list_head ras_read_beads;
588 * The following 3 items are used for detecting the stride I/O
590 * In stride I/O mode,
591 * ...............|-----data-----|****gap*****|--------|******|....
592 * offset |-stride_pages-|-stride_gap-|
593 * ras_stride_offset = offset;
594 * ras_stride_length = stride_pages + stride_gap;
595 * ras_stride_pages = stride_pages;
596 * Note: all these three items are counted by pages.
598 unsigned long ras_stride_length;
599 unsigned long ras_stride_pages;
600 pgoff_t ras_stride_offset;
602 * number of consecutive stride request count, and it is similar as
603 * ras_consecutive_requests, but used for stride I/O mode.
604 * Note: only more than 2 consecutive stride request are detected,
605 * stride read-ahead will be enable
607 unsigned long ras_consecutive_stride_requests;
610 extern struct kmem_cache *ll_file_data_slab;
611 struct lustre_handle;
612 struct ll_file_data {
613 struct ll_readahead_state fd_ras;
614 struct ccc_grouplock fd_grouplock;
618 /* openhandle if lease exists for this file.
619 * Borrow lli->lli_och_mutex to protect assignment */
620 struct obd_client_handle *fd_lease_och;
621 struct obd_client_handle *fd_och;
622 struct file *fd_file;
623 /* Indicate whether need to report failure when close.
624 * true: failure is known, not report again.
625 * false: unknown failure, should report. */
626 bool fd_write_failed;
629 struct lov_stripe_md;
631 extern spinlock_t inode_lock;
633 extern struct proc_dir_entry *proc_lustre_fs_root;
635 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
637 return &lli->lli_vfs_inode;
641 struct inode *icbd_parent;
642 struct dentry **icbd_childp;
646 __u32 ll_i2suppgid(struct inode *i);
647 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
649 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
651 #if BITS_PER_LONG == 32
658 (sbi->ll_flags & LL_SBI_32BIT_API)
663 #define LLAP_MAGIC 98764321
665 extern struct kmem_cache *ll_async_page_slab;
666 extern size_t ll_async_page_slab_size;
668 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
669 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
670 struct ll_ra_read *ll_ra_read_get(struct file *f);
672 /* llite/lproc_llite.c */
674 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
675 struct super_block *sb, char *osc, char *mdc);
676 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
677 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
678 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
680 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
681 struct super_block *sb, char *osc, char *mdc){return 0;}
682 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
683 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
684 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
686 memset(lvars, 0, sizeof(*lvars));
692 void ll_release_page(struct page *page, int remove);
693 extern struct file_operations ll_dir_operations;
694 extern struct inode_operations ll_dir_inode_operations;
695 struct page *ll_get_dir_page(struct inode *dir, __u64 hash,
696 struct ll_dir_chain *chain);
697 int ll_dir_read(struct inode *inode, struct dir_context *ctx);
699 int ll_get_mdt_idx(struct inode *inode);
701 int ll_objects_destroy(struct ptlrpc_request *request,
703 struct inode *ll_iget(struct super_block *sb, ino_t hash,
704 struct lustre_md *lic);
705 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
706 void *data, int flag);
707 struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
708 int ll_rmdir_entry(struct inode *dir, char *name, int namelen);
711 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
712 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
713 int ll_writepage(struct page *page, struct writeback_control *wbc);
714 int ll_writepages(struct address_space *, struct writeback_control *wbc);
715 void ll_removepage(struct page *page);
716 int ll_readpage(struct file *file, struct page *page);
717 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
718 int ll_file_punch(struct inode *, loff_t, int);
719 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
720 void ll_clear_file_contended(struct inode*);
721 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
722 int ll_readahead(const struct lu_env *env, struct cl_io *io,
723 struct ll_readahead_state *ras, struct address_space *mapping,
724 struct cl_page_list *queue, int flags);
727 extern struct file_operations ll_file_operations;
728 extern struct file_operations ll_file_operations_flock;
729 extern struct file_operations ll_file_operations_noflock;
730 extern struct inode_operations ll_file_inode_operations;
731 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
733 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
734 ldlm_mode_t l_req_mode);
735 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
736 struct lustre_handle *lockh, __u64 flags);
737 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
739 int ll_revalidate_nd(struct dentry *dentry, unsigned int flags);
740 int ll_file_open(struct inode *inode, struct file *file);
741 int ll_file_release(struct inode *inode, struct file *file);
742 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
743 struct lov_stripe_md *lsm, lstat_t *st);
744 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
745 int ll_local_open(struct file *file,
746 struct lookup_intent *it, struct ll_file_data *fd,
747 struct obd_client_handle *och);
748 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
749 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
751 int ll_md_real_close(struct inode *inode, int flags);
752 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
753 struct obd_client_handle **och, unsigned long flags);
754 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
755 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
756 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
757 __u64 ioepoch, int sync);
758 int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data,
759 struct md_open_data **mod);
760 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
761 struct lustre_handle *fh);
762 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
763 struct ll_file_data *file, loff_t pos,
764 size_t count, int rw);
765 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
766 struct lookup_intent *it, struct kstat *stat);
767 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
768 struct ll_file_data *ll_file_data_get(void);
769 struct posix_acl * ll_get_acl(struct inode *inode, int type);
771 int ll_inode_permission(struct inode *inode, int mask);
773 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
774 int flags, struct lov_user_md *lum,
776 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
777 struct lov_mds_md **lmm, int *lmm_size,
778 struct ptlrpc_request **request);
779 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
781 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmmp,
782 int *lmm_size, struct ptlrpc_request **request);
783 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
784 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
786 int ll_merge_lvb(const struct lu_env *env, struct inode *inode);
787 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
788 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
789 int ll_fid2path(struct inode *inode, void *arg);
790 int ll_data_version(struct inode *inode, __u64 *data_version, int extent_lock);
792 struct obd_client_handle *ll_lease_open(struct inode *inode, struct file *file,
794 int ll_lease_close(struct obd_client_handle *och, struct inode *inode,
799 int ll_dops_init(struct dentry *de, int block, int init_sa);
800 extern struct dentry_operations ll_d_ops;
801 void ll_intent_drop_lock(struct lookup_intent *);
802 void ll_intent_release(struct lookup_intent *);
803 void ll_invalidate_aliases(struct inode *);
804 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
805 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
806 int ll_dcompare(const struct dentry *parent, const struct dentry *dentry,
807 unsigned int len, const char *str, const struct qstr *d_name);
808 int ll_revalidate_it_finish(struct ptlrpc_request *request,
809 struct lookup_intent *it, struct dentry *de);
811 /* llite/llite_lib.c */
812 extern struct super_operations lustre_super_operations;
814 char *ll_read_opt(const char *opt, char *data);
815 void ll_lli_init(struct ll_inode_info *lli);
816 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
817 void ll_put_super(struct super_block *sb);
818 void ll_kill_super(struct super_block *sb);
819 struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
820 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
821 void ll_clear_inode(struct inode *inode);
822 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr);
823 int ll_setattr(struct dentry *de, struct iattr *attr);
824 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
825 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
826 __u64 max_age, __u32 flags);
827 void ll_update_inode(struct inode *inode, struct lustre_md *md);
828 void ll_read_inode2(struct inode *inode, void *opaque);
829 void ll_delete_inode(struct inode *inode);
830 int ll_iocontrol(struct inode *inode, struct file *file,
831 unsigned int cmd, unsigned long arg);
832 int ll_flush_ctx(struct inode *inode);
833 void ll_umount_begin(struct super_block *sb);
834 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
835 int ll_show_options(struct seq_file *seq, struct dentry *dentry);
836 void ll_dirty_page_discard_warn(struct page *page, int ioret);
837 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
838 struct super_block *, struct lookup_intent *);
839 void lustre_dump_dentry(struct dentry *, int recur);
840 void lustre_dump_inode(struct inode *);
841 int ll_obd_statfs(struct inode *inode, void *arg);
842 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
843 int ll_process_config(struct lustre_cfg *lcfg);
844 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
845 struct inode *i1, struct inode *i2,
846 const char *name, int namelen,
847 int mode, __u32 opc, void *data);
848 void ll_finish_md_op_data(struct md_op_data *op_data);
849 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
850 char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
852 /* llite/llite_nfs.c */
853 extern struct export_operations lustre_export_operations;
854 __u32 get_uuid2int(const char *name, int len);
855 void get_uuid2fsid(const char *name, int len, __kernel_fsid_t *fsid);
856 struct inode *search_inode_for_lustre(struct super_block *sb,
857 const struct lu_fid *fid);
859 /* llite/special.c */
860 extern struct inode_operations ll_special_inode_operations;
861 extern struct file_operations ll_special_chr_inode_fops;
862 extern struct file_operations ll_special_chr_file_fops;
863 extern struct file_operations ll_special_blk_inode_fops;
864 extern struct file_operations ll_special_fifo_inode_fops;
865 extern struct file_operations ll_special_fifo_file_fops;
866 extern struct file_operations ll_special_sock_inode_fops;
868 /* llite/symlink.c */
869 extern struct inode_operations ll_fast_symlink_inode_operations;
871 /* llite/llite_close.c */
872 struct ll_close_queue {
874 struct list_head lcq_head;
875 wait_queue_head_t lcq_waitq;
876 struct completion lcq_comp;
880 struct ccc_object *cl_inode2ccc(struct inode *inode);
883 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
884 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
886 /* specific achitecture can implement only part of this list */
887 enum vvp_io_subtype {
890 /** io called from .sendfile */
892 /** io started from splice_{read|write} */
899 enum vvp_io_subtype cui_io_subtype;
903 read_actor_t cui_actor;
907 struct pipe_inode_info *cui_pipe;
908 unsigned int cui_flags;
910 struct vvp_fault_io {
912 * Inode modification time that is checked across DLM
916 struct vm_area_struct *ft_vma;
918 * locked page returned from vvp_io
920 struct page *ft_vmpage;
921 struct vm_fault_api {
925 struct vm_fault *ft_vmf;
927 * fault API used bitflags for return code.
929 unsigned int ft_flags;
934 * Read-ahead state used by read and page-fault IO contexts.
936 struct ll_ra_read cui_bead;
938 * Set when cui_bead has been initialized.
940 int cui_ra_window_set;
942 * Partially truncated page, that vvp_io_trunc_start() keeps locked
945 struct cl_page *cui_partpage;
949 * IO arguments for various VFS I/O interfaces.
952 /** normal/sendfile/splice */
953 enum vvp_io_subtype via_io_subtype;
957 struct kiocb *via_iocb;
958 struct iovec *via_iov;
959 unsigned long via_nrsegs;
962 read_actor_t via_actor;
966 struct pipe_inode_info *via_pipe;
967 unsigned int via_flags;
972 struct ll_cl_context {
974 struct cl_io *lcc_io;
975 struct cl_page *lcc_page;
976 struct lu_env *lcc_env;
981 struct vvp_thread_info {
982 struct ost_lvb vti_lvb;
983 struct cl_2queue vti_queue;
984 struct iovec vti_local_iov;
985 struct vvp_io_args vti_args;
986 struct ra_io_arg vti_ria;
987 struct kiocb vti_kiocb;
988 struct ll_cl_context vti_io_ctx;
991 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
993 extern struct lu_context_key vvp_key;
994 struct vvp_thread_info *info;
996 info = lu_context_key_get(&env->le_ctx, &vvp_key);
997 LASSERT(info != NULL);
1001 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
1002 enum vvp_io_subtype type)
1004 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
1006 ret->via_io_subtype = type;
1011 struct vvp_session {
1012 struct vvp_io vs_ios;
1015 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1017 extern struct lu_context_key vvp_session_key;
1018 struct vvp_session *ses;
1020 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1021 LASSERT(ses != NULL);
1025 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1027 return &vvp_env_session(env)->vs_ios;
1030 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1031 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1032 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1034 /* llite/llite_mmap.c */
1035 typedef struct rb_root rb_root_t;
1036 typedef struct rb_node rb_node_t;
1038 struct ll_lock_tree_node;
1039 struct ll_lock_tree {
1041 struct list_head lt_locked_list;
1042 struct ll_file_data *lt_fd;
1045 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1046 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1047 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
1048 __u64 end, ldlm_mode_t mode);
1049 void policy_from_vma(ldlm_policy_data_t *policy,
1050 struct vm_area_struct *vma, unsigned long addr, size_t count);
1051 struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
1054 static inline void ll_invalidate_page(struct page *vmpage)
1056 struct address_space *mapping = vmpage->mapping;
1057 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1059 LASSERT(PageLocked(vmpage));
1060 if (mapping == NULL)
1063 ll_teardown_mmaps(mapping, offset, offset + PAGE_CACHE_SIZE);
1064 truncate_complete_page(mapping, vmpage);
1067 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1069 /* don't need an addref as the sb_info should be holding one */
1070 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1072 return ll_s2sbi(sb)->ll_dt_exp;
1075 /* don't need an addref as the sb_info should be holding one */
1076 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1078 return ll_s2sbi(sb)->ll_md_exp;
1081 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1083 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1089 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1090 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1092 return ll_s2sbi(inode->i_sb);
1095 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1097 return ll_s2dtexp(inode->i_sb);
1100 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1102 return ll_s2mdexp(inode->i_sb);
1105 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1109 LASSERT(inode != NULL);
1110 fid = &ll_i2info(inode)->lli_fid;
1115 static inline int ll_mds_max_easize(struct super_block *sb)
1117 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1120 static inline __u64 ll_file_maxbytes(struct inode *inode)
1122 return ll_i2info(inode)->lli_maxbytes;
1126 int ll_setxattr(struct dentry *dentry, const char *name,
1127 const void *value, size_t size, int flags);
1128 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1129 void *buffer, size_t size);
1130 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1131 int ll_removexattr(struct dentry *dentry, const char *name);
1133 /* llite/remote_perm.c */
1134 extern struct kmem_cache *ll_remote_perm_cachep;
1135 extern struct kmem_cache *ll_rmtperm_hash_cachep;
1137 struct hlist_head *alloc_rmtperm_hash(void);
1138 void free_rmtperm_hash(struct hlist_head *hash);
1139 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1140 int lustre_check_remote_perm(struct inode *inode, int mask);
1142 /* llite/llite_capa.c */
1143 extern struct timer_list ll_capa_timer;
1145 int ll_capa_thread_start(void);
1146 void ll_capa_thread_stop(void);
1147 void ll_capa_timer_callback(unsigned long unused);
1149 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1150 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1152 void ll_capa_open(struct inode *inode);
1153 void ll_capa_close(struct inode *inode);
1155 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1156 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1158 void ll_truncate_free_capa(struct obd_capa *ocapa);
1159 void ll_clear_inode_capas(struct inode *inode);
1160 void ll_print_capa_stat(struct ll_sb_info *sbi);
1162 /* llite/llite_cl.c */
1163 extern struct lu_device_type vvp_device_type;
1166 * Common IO arguments for various VFS I/O interfaces.
1168 int cl_sb_init(struct super_block *sb);
1169 int cl_sb_fini(struct super_block *sb);
1170 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1171 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1173 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1174 struct ll_readahead_state *ras, unsigned long index,
1176 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1177 int ll_is_file_contended(struct file *file);
1178 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1180 /* llite/llite_rmtacl.c */
1181 #ifdef CONFIG_FS_POSIX_ACL
1183 struct list_head ee_list;
1184 pid_t ee_key; /* hash key */
1185 struct lu_fid ee_fid;
1186 int ee_type; /* ACL type for ACCESS or DEFAULT */
1187 ext_acl_xattr_header *ee_acl;
1190 obd_valid rce_ops2valid(int ops);
1191 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1192 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1193 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1194 void rct_init(struct rmtacl_ctl_table *rct);
1195 void rct_fini(struct rmtacl_ctl_table *rct);
1197 void ee_free(struct eacl_entry *ee);
1198 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1199 ext_acl_xattr_header *header);
1200 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1201 struct lu_fid *fid, int type);
1202 void et_search_free(struct eacl_table *et, pid_t key);
1203 void et_init(struct eacl_table *et);
1204 void et_fini(struct eacl_table *et);
1206 static inline obd_valid rce_ops2valid(int ops)
1214 #define LL_SA_RPC_MIN 2
1215 #define LL_SA_RPC_DEF 32
1216 #define LL_SA_RPC_MAX 8192
1218 #define LL_SA_CACHE_BIT 5
1219 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1220 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1222 /* per inode struct, for dir only */
1223 struct ll_statahead_info {
1224 struct inode *sai_inode;
1225 atomic_t sai_refcount; /* when access this struct, hold
1227 unsigned int sai_generation; /* generation for statahead */
1228 unsigned int sai_max; /* max ahead of lookup */
1229 __u64 sai_sent; /* stat requests sent count */
1230 __u64 sai_replied; /* stat requests which received
1232 __u64 sai_index; /* index of statahead entry */
1233 __u64 sai_index_wait; /* index of entry which is the
1234 * caller is waiting for */
1235 __u64 sai_hit; /* hit count */
1236 __u64 sai_miss; /* miss count:
1237 * for "ls -al" case, it includes
1238 * hidden dentry miss;
1239 * for "ls -l" case, it does not
1240 * include hidden dentry miss.
1241 * "sai_miss_hidden" is used for
1244 unsigned int sai_consecutive_miss; /* consecutive miss */
1245 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1246 * is not a hidden one */
1247 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1248 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1250 sai_in_readpage:1,/* statahead is in readdir()*/
1251 sai_agl_valid:1;/* AGL is valid for the dir */
1252 wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
1253 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1254 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1255 struct list_head sai_entries; /* entry list */
1256 struct list_head sai_entries_received; /* entries returned */
1257 struct list_head sai_entries_stated; /* entries stated */
1258 struct list_head sai_entries_agl; /* AGL entries to be sent */
1259 struct list_head sai_cache[LL_SA_CACHE_SIZE];
1260 spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1261 atomic_t sai_cache_count; /* entry count in cache */
1264 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1266 void ll_stop_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 if (sai != NULL && ldd != NULL)
1292 ldd->lld_sa_generation = sai->sai_generation;
1296 ll_need_statahead(struct inode *dir, struct dentry *dentryp)
1298 struct ll_inode_info *lli;
1299 struct ll_dentry_data *ldd;
1301 if (ll_i2sbi(dir)->ll_sa_max == 0)
1304 lli = ll_i2info(dir);
1305 /* not the same process, don't statahead */
1306 if (lli->lli_opendir_pid != current_pid())
1309 /* statahead has been stopped */
1310 if (lli->lli_opendir_key == NULL)
1313 ldd = ll_d2d(dentryp);
1315 * When stats a dentry, the system trigger more than once "revalidate"
1316 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1317 * Under patchless client mode, the operation intent is not accurate,
1318 * which maybe misguide the statahead thread. For example:
1319 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1320 * have the same operation intent -- "IT_GETATTR".
1321 * In fact, one dentry should has only one chance to interact with the
1322 * statahead thread, otherwise the statahead windows will be confused.
1323 * The solution is as following:
1324 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1325 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1326 * will bypass interacting with statahead thread for checking:
1327 * "lld_sa_generation == lli_sai->sai_generation"
1329 if (ldd && lli->lli_sai &&
1330 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1337 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1341 ret = ll_need_statahead(dir, *dentryp);
1345 return do_statahead_enter(dir, dentryp, only_unplug);
1348 /* llite ioctl register support rountine */
1354 #define LLIOC_MAX_CMD 256
1357 * Rules to write a callback function:
1360 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1361 * returned to ll_iocontrol_register. Callback functions should use this
1362 * data to check the potential collasion of ioctl cmd. If collasion is
1363 * found, callback function should return LLIOC_CONT.
1364 * @rcp: The result of ioctl command.
1367 * If @magic matches the pointer returned by ll_iocontrol_data, the
1368 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1370 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1371 struct file *file, unsigned int cmd, unsigned long arg,
1372 void *magic, int *rcp);
1374 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1375 unsigned int cmd, unsigned long arg, int *rcp);
1377 /* export functions */
1378 /* Register ioctl block dynamatically for a regular file.
1380 * @cmd: the array of ioctl command set
1381 * @count: number of commands in the @cmd
1382 * @cb: callback function, it will be called if an ioctl command is found to
1383 * belong to the command list @cmd.
1386 * A magic pointer will be returned if success;
1387 * otherwise, NULL will be returned.
1389 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1390 void ll_iocontrol_unregister(void *magic);
1393 /* lclient compat stuff */
1394 #define cl_inode_info ll_inode_info
1395 #define cl_i2info(info) ll_i2info(info)
1396 #define cl_inode_mode(inode) ((inode)->i_mode)
1397 #define cl_i2sbi ll_i2sbi
1399 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1400 const struct iattr *attr)
1402 LASSERT(attr->ia_valid & ATTR_FILE);
1403 return LUSTRE_FPRIVATE(attr->ia_file);
1406 static inline void cl_isize_lock(struct inode *inode)
1408 ll_inode_size_lock(inode);
1411 static inline void cl_isize_unlock(struct inode *inode)
1413 ll_inode_size_unlock(inode);
1416 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1418 LASSERT(down_trylock(&ll_i2info(inode)->lli_size_sem) != 0);
1419 i_size_write(inode, kms);
1422 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1424 ll_inode_size_lock(inode);
1425 i_size_write(inode, kms);
1426 ll_inode_size_unlock(inode);
1429 #define cl_isize_read(inode) i_size_read(inode)
1431 static inline int cl_merge_lvb(const struct lu_env *env, struct inode *inode)
1433 return ll_merge_lvb(env, inode);
1436 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1437 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1438 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1440 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1442 int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1443 enum cl_fsync_mode mode, int ignore_layout);
1445 /** direct write pages */
1446 struct ll_dio_pages {
1447 /** page array to be written. we don't support
1448 * partial pages except the last one. */
1449 struct page **ldp_pages;
1450 /* offset of each page */
1451 loff_t *ldp_offsets;
1452 /** if ldp_offsets is NULL, it means a sequential
1453 * pages to be written, then this is the file offset
1454 * of the * first page. */
1455 loff_t ldp_start_offset;
1456 /** how many bytes are to be written. */
1458 /** # of pages in the array. */
1462 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1465 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1468 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1471 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1472 int rw, struct inode *inode,
1473 struct ll_dio_pages *pv);
1475 static inline int ll_file_nolock(const struct file *file)
1477 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1478 struct inode *inode = file->f_dentry->d_inode;
1480 LASSERT(fd != NULL);
1481 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1482 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1485 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1486 struct lookup_intent *it, __u64 *bits)
1488 if (!it->d.lustre.it_lock_set) {
1489 struct lustre_handle handle;
1491 /* If this inode is a remote object, it will get two
1492 * separate locks in different namespaces, Master MDT,
1493 * where the name entry is, will grant LOOKUP lock,
1494 * remote MDT, where the object is, will grant
1495 * UPDATE|PERM lock. The inode will be attched to both
1496 * LOOKUP and PERM locks, so revoking either locks will
1497 * case the dcache being cleared */
1498 if (it->d.lustre.it_remote_lock_mode) {
1499 handle.cookie = it->d.lustre.it_remote_lock_handle;
1500 CDEBUG(D_DLMTRACE, "setting l_data to inode %p"
1501 "(%lu/%u) for remote lock "LPX64"\n", inode,
1502 inode->i_ino, inode->i_generation,
1504 md_set_lock_data(exp, &handle.cookie, inode, NULL);
1507 handle.cookie = it->d.lustre.it_lock_handle;
1509 CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)"
1510 " for lock "LPX64"\n", inode, inode->i_ino,
1511 inode->i_generation, handle.cookie);
1513 md_set_lock_data(exp, &handle.cookie, inode,
1514 &it->d.lustre.it_lock_bits);
1515 it->d.lustre.it_lock_set = 1;
1519 *bits = it->d.lustre.it_lock_bits;
1522 static inline void ll_lock_dcache(struct inode *inode)
1524 spin_lock(&inode->i_lock);
1527 static inline void ll_unlock_dcache(struct inode *inode)
1529 spin_unlock(&inode->i_lock);
1532 static inline int d_lustre_invalid(const struct dentry *dentry)
1534 struct ll_dentry_data *lld = ll_d2d(dentry);
1536 return (lld == NULL) || lld->lld_invalid;
1539 static inline void __d_lustre_invalidate(struct dentry *dentry)
1541 struct ll_dentry_data *lld = ll_d2d(dentry);
1544 lld->lld_invalid = 1;
1548 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1549 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1550 * else dput() of the last refcount will unhash this dentry and kill it.
1552 static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1554 CDEBUG(D_DENTRY, "invalidate dentry %.*s (%p) parent %p inode %p "
1555 "refc %d\n", dentry->d_name.len, dentry->d_name.name, dentry,
1556 dentry->d_parent, dentry->d_inode, d_count(dentry));
1558 spin_lock_nested(&dentry->d_lock,
1559 nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1560 __d_lustre_invalidate(dentry);
1561 if (d_count(dentry) == 0)
1563 spin_unlock(&dentry->d_lock);
1566 static inline void d_lustre_revalidate(struct dentry *dentry)
1568 spin_lock(&dentry->d_lock);
1569 LASSERT(ll_d2d(dentry) != NULL);
1570 ll_d2d(dentry)->lld_invalid = 0;
1571 spin_unlock(&dentry->d_lock);
1574 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0)
1575 /* Compatibility for old (1.8) compiled userspace quota code */
1576 struct if_quotactl_18 {
1581 struct obd_dqinfo qc_dqinfo;
1582 struct obd_dqblk qc_dqblk;
1584 struct obd_uuid obd_uuid;
1586 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1587 /* End compatibility for old (1.8) compiled userspace quota code */
1589 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1590 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0) */
1593 LL_LAYOUT_GEN_NONE = ((__u32)-2), /* layout lock was cancelled */
1594 LL_LAYOUT_GEN_EMPTY = ((__u32)-1) /* for empty layout */
1597 int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1598 int ll_layout_refresh(struct inode *inode, __u32 *gen);
1599 int ll_layout_restore(struct inode *inode);
1601 #endif /* LLITE_INTERNAL_H */