1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
30 * Use is subject to license terms.
32 * Copyright (c) 2011, 2012, Whamcloud, Inc.
35 * This file is part of Lustre, http://www.lustre.org/
36 * Lustre is a trademark of Sun Microsystems, Inc.
39 #ifndef LLITE_INTERNAL_H
40 #define LLITE_INTERNAL_H
42 #include <lustre_acl.h>
44 #include <lustre_debug.h>
45 #include <lustre_ver.h>
46 #include <lustre_disk.h> /* for s2sbi */
47 #include <lustre_eacl.h>
49 /* for struct cl_lock_descr and struct cl_io */
50 #include <cl_object.h>
52 #include <lustre_mdc.h>
53 #include <linux/lustre_intent.h>
59 #ifndef VM_FAULT_RETRY
60 #define VM_FAULT_RETRY 0
63 /** Only used on client-side for indicating the tail of dir hash/offset. */
64 #define LL_DIR_END_OFF 0x7fffffffffffffffULL
65 #define LL_DIR_END_OFF_32BIT 0x7fffffffUL
67 #ifndef DCACHE_LUSTRE_INVALID
68 #define DCACHE_LUSTRE_INVALID 0x4000000
71 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
72 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
74 struct ll_dentry_data {
77 struct obd_client_handle lld_cwd_och;
78 struct obd_client_handle lld_mnt_och;
79 struct lookup_intent *lld_it;
80 unsigned int lld_sa_generation;
83 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
85 extern struct file_operations ll_pgcache_seq_fops;
87 #define LLI_INODE_MAGIC 0x111d0de5
88 #define LLI_INODE_DEAD 0xdeadd00d
90 /* remote client permission cache */
91 #define REMOTE_PERM_HASHSIZE 16
93 struct ll_getname_data {
94 char *lgd_name; /* points to a buffer with NAME_MAX+1 size */
95 struct lu_fid lgd_fid; /* target fid we are looking for */
96 int lgd_found; /* inode matched? */
99 /* llite setxid/access permission for user on remote client */
100 struct ll_remote_perm {
101 cfs_hlist_node_t lrp_list;
106 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
107 is access permission with
108 lrp_fsuid/lrp_fsgid. */
112 /* MDS has an authority for the Size-on-MDS attributes. */
113 LLIF_MDS_SIZE_LOCK = (1 << 0),
114 /* Epoch close is postponed. */
115 LLIF_EPOCH_PENDING = (1 << 1),
116 /* DONE WRITING is allowed. */
117 LLIF_DONE_WRITING = (1 << 2),
118 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
120 LLIF_SOM_DIRTY = (1 << 3),
121 /* File is contented */
122 LLIF_CONTENDED = (1 << 4),
123 /* Truncate uses server lock for this file */
124 LLIF_SRVLOCK = (1 << 5),
128 struct ll_inode_info {
129 __u32 lli_inode_magic;
133 cfs_spinlock_t lli_lock;
134 struct posix_acl *lli_posix_acl;
136 cfs_hlist_head_t *lli_remote_perms;
137 cfs_semaphore_t lli_rmtperm_sem;
139 /* identifying fields for both metadata and data stacks. */
140 struct lu_fid lli_fid;
141 /* Parent fid for accessing default stripe data on parent directory
142 * for allocating OST objects after a mknod() and later open-by-FID. */
143 struct lu_fid lli_pfid;
145 cfs_list_t lli_close_list;
146 cfs_list_t lli_oss_capas;
147 /* open count currently used by capability only, indicate whether
148 * capability needs renewal */
149 cfs_atomic_t lli_open_count;
150 struct obd_capa *lli_mds_capa;
151 cfs_time_t lli_rmtperm_time;
153 /* handle is to be sent to MDS later on done_writing and setattr.
154 * Open handle data are needed for the recovery to reconstruct
155 * the inode state on the MDS. XXX: recovery is not ready yet. */
156 struct obd_client_handle *lli_pending_och;
158 /* We need all three because every inode may be opened in different
160 struct obd_client_handle *lli_mds_read_och;
161 struct obd_client_handle *lli_mds_write_och;
162 struct obd_client_handle *lli_mds_exec_och;
163 __u64 lli_open_fd_read_count;
164 __u64 lli_open_fd_write_count;
165 __u64 lli_open_fd_exec_count;
166 /* Protects access to och pointers and their usage counters, also
167 * atomicity of check-update of lli_smd */
168 cfs_semaphore_t lli_och_sem;
170 struct inode lli_vfs_inode;
172 /* the most recent timestamps obtained from mds */
173 struct ost_lvb lli_lvb;
174 cfs_spinlock_t lli_agl_lock;
176 /* Try to make the d::member and f::member are aligned. Before using
177 * these members, make clear whether it is directory or not. */
181 /* serialize normal readdir and statahead-readdir. */
182 cfs_semaphore_t d_readdir_sem;
184 /* metadata statahead */
185 /* since parent-child threads can share the same @file
186 * struct, "opendir_key" is the token when dir close for
187 * case of parent exit before child -- it is me should
188 * cleanup the dir readahead. */
190 struct ll_statahead_info *d_sai;
192 struct posix_acl *d_def_acl;
193 /* protect statahead stuff. */
194 cfs_spinlock_t d_sa_lock;
195 /* "opendir_pid" is the token when lookup/revalid
196 * -- I am the owner of dir statahead. */
200 #define lli_readdir_sem u.d.d_readdir_sem
201 #define lli_opendir_key u.d.d_opendir_key
202 #define lli_sai u.d.d_sai
203 #define lli_sa_pos u.d.d_sa_pos
204 #define lli_def_acl u.d.d_def_acl
205 #define lli_sa_lock u.d.d_sa_lock
206 #define lli_opendir_pid u.d.d_opendir_pid
208 /* for non-directory */
210 cfs_semaphore_t f_size_sem;
211 void *f_size_sem_owner;
212 char *f_symlink_name;
215 * cfs_rw_semaphore_t {
216 * signed long count; // align u.d.d_def_acl
217 * cfs_spinlock_t wait_lock; // align u.d.d_sa_lock
218 * struct list_head wait_list;
221 cfs_rw_semaphore_t f_trunc_sem;
222 cfs_semaphore_t f_write_sem;
224 /* for writepage() only to communicate to fsync */
228 cfs_rw_semaphore_t f_glimpse_sem;
229 cfs_time_t f_glimpse_time;
230 cfs_list_t f_agl_list;
234 #define lli_size_sem u.f.f_size_sem
235 #define lli_size_sem_owner u.f.f_size_sem_owner
236 #define lli_symlink_name u.f.f_symlink_name
237 #define lli_maxbytes u.f.f_maxbytes
238 #define lli_trunc_sem u.f.f_trunc_sem
239 #define lli_write_sem u.f.f_write_sem
240 #define lli_async_rc u.f.f_async_rc
241 #define lli_write_rc u.f.f_write_rc
242 #define lli_glimpse_sem u.f.f_glimpse_sem
243 #define lli_glimpse_time u.f.f_glimpse_time
244 #define lli_agl_list u.f.f_agl_list
245 #define lli_agl_index u.f.f_agl_index
249 /* XXX: For following frequent used members, although they maybe special
250 * used for non-directory object, it is some time-wasting to check
251 * whether the object is directory or not before using them. On the
252 * other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
253 * the "ll_inode_info" size even if moving those members into u.f.
254 * So keep them out side.
256 * In the future, if more members are added only for directory,
257 * some of the following members can be moved into u.f.
259 struct lov_stripe_md *lli_smd;
260 struct cl_object *lli_clob;
264 * Locking to guarantee consistency of non-atomic updates to long long i_size,
265 * consistency between file size and KMS, and consistency within
266 * ->lli_smd->lsm_oinfo[]'s.
268 * Implemented by ->lli_size_sem and ->lsm_sem, nested in that order.
271 void ll_inode_size_lock(struct inode *inode, int lock_lsm);
272 void ll_inode_size_unlock(struct inode *inode, int unlock_lsm);
274 // FIXME: replace the name of this with LL_I to conform to kernel stuff
275 // static inline struct ll_inode_info *LL_I(struct inode *inode)
276 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
278 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
281 /* default to about 40meg of readahead on a given system. That much tied
282 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
283 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - CFS_PAGE_SHIFT))
285 /* default to read-ahead full files smaller than 2MB on the second read */
286 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - CFS_PAGE_SHIFT))
291 RA_STAT_DISTANT_READPAGE,
292 RA_STAT_MISS_IN_WINDOW,
293 RA_STAT_FAILED_GRAB_PAGE,
294 RA_STAT_FAILED_MATCH,
299 RA_STAT_MAX_IN_FLIGHT,
300 RA_STAT_WRONG_GRAB_PAGE,
305 cfs_atomic_t ra_cur_pages;
306 unsigned long ra_max_pages;
307 unsigned long ra_max_pages_per_file;
308 unsigned long ra_max_read_ahead_whole_pages;
311 /* ra_io_arg will be filled in the beginning of ll_readahead with
312 * ras_lock, then the following ll_read_ahead_pages will read RA
313 * pages according to this arg, all the items in this structure are
314 * counted by page index.
317 unsigned long ria_start; /* start offset of read-ahead*/
318 unsigned long ria_end; /* end offset of read-ahead*/
319 /* If stride read pattern is detected, ria_stoff means where
320 * stride read is started. Note: for normal read-ahead, the
321 * value here is meaningless, and also it will not be accessed*/
323 /* ria_length and ria_pages are the length and pages length in the
324 * stride I/O mode. And they will also be used to check whether
325 * it is stride I/O read-ahead in the read-ahead pages*/
326 unsigned long ria_length;
327 unsigned long ria_pages;
330 /* LL_HIST_MAX=32 causes an overflow */
331 #define LL_HIST_MAX 28
332 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
333 #define LL_PROCESS_HIST_MAX 10
334 struct per_process_info {
336 struct obd_histogram pp_r_hist;
337 struct obd_histogram pp_w_hist;
340 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
341 struct ll_rw_extents_info {
342 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
345 #define LL_OFFSET_HIST_MAX 100
346 struct ll_rw_process_info {
349 loff_t rw_range_start;
351 loff_t rw_last_file_pos;
353 size_t rw_smallest_extent;
354 size_t rw_largest_extent;
355 struct ll_file_data *rw_last_file;
358 enum stats_track_type {
359 STATS_TRACK_ALL = 0, /* track all processes */
360 STATS_TRACK_PID, /* track process with this pid */
361 STATS_TRACK_PPID, /* track processes with this ppid */
362 STATS_TRACK_GID, /* track processes with this gid */
366 /* flags for sbi->ll_flags */
367 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
368 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
369 #define LL_SBI_FLOCK 0x04
370 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
371 #define LL_SBI_ACL 0x10 /* support ACL */
372 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
373 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
374 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
375 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
376 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
377 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
378 #define LL_SBI_SOM_PREVIEW 0x1000 /* SOM preview mount option */
379 #define LL_SBI_32BIT_API 0x2000 /* generate 32 bit inodes. */
380 #define LL_SBI_64BIT_HASH 0x4000 /* support 64-bits dir hash/offset */
381 #define LL_SBI_AGL_ENABLED 0x8000 /* enable agl */
383 /* default value for ll_sb_info->contention_time */
384 #define SBI_DEFAULT_CONTENTION_SECONDS 60
385 /* default value for lockless_truncate_enable */
386 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
387 #define RCE_HASHES 32
389 struct rmtacl_ctl_entry {
391 pid_t rce_key; /* hash key */
392 int rce_ops; /* acl operation type */
395 struct rmtacl_ctl_table {
396 cfs_spinlock_t rct_lock;
397 cfs_list_t rct_entries[RCE_HASHES];
404 pid_t ee_key; /* hash key */
405 struct lu_fid ee_fid;
406 int ee_type; /* ACL type for ACCESS or DEFAULT */
407 ext_acl_xattr_header *ee_acl;
411 cfs_spinlock_t et_lock;
412 cfs_list_t et_entries[EE_HASHES];
417 /* this protects pglist and ra_info. It isn't safe to
418 * grab from interrupt contexts */
419 cfs_spinlock_t ll_lock;
420 cfs_spinlock_t ll_pp_extent_lock; /* Lock for pp_extent entries */
421 cfs_spinlock_t ll_process_lock; /* Lock for ll_rw_process_info */
422 struct obd_uuid ll_sb_uuid;
423 struct obd_export *ll_md_exp;
424 struct obd_export *ll_dt_exp;
425 struct proc_dir_entry* ll_proc_root;
426 struct lu_fid ll_root_fid; /* root object fid */
429 cfs_list_t ll_conn_chain; /* per-conn chain of SBs */
430 struct lustre_client_ocd ll_lco;
432 cfs_list_t ll_orphan_dentry_list; /*please don't ask -p*/
433 struct ll_close_queue *ll_lcq;
435 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
437 unsigned long ll_async_page_max;
438 unsigned long ll_async_page_count;
440 struct lprocfs_stats *ll_ra_stats;
442 struct ll_ra_info ll_ra_info;
443 unsigned int ll_namelen;
444 struct file_operations *ll_fop;
446 /* =0 - hold lock over whole read/write
447 * >0 - max. chunk to be read/written w/o lock re-acquiring */
448 unsigned long ll_max_rw_chunk;
449 unsigned int ll_md_brw_size; /* used by readdir */
451 struct lu_site *ll_site;
452 struct cl_device *ll_cl;
454 struct ll_rw_extents_info ll_rw_extents_info;
455 int ll_extent_process_count;
456 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
457 unsigned int ll_offset_process_count;
458 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
459 unsigned int ll_rw_offset_entry_count;
460 int ll_stats_track_id;
461 enum stats_track_type ll_stats_track_type;
464 /* metadata stat-ahead */
465 unsigned int ll_sa_max; /* max statahead RPCs */
466 atomic_t ll_sa_total; /* statahead thread started
468 atomic_t ll_sa_wrong; /* statahead thread stopped for
470 atomic_t ll_agl_total; /* AGL thread started count */
472 dev_t ll_sdev_orig; /* save s_dev before assign for
474 struct rmtacl_ctl_table ll_rct;
475 struct eacl_table ll_et;
476 struct vfsmount *ll_mnt;
479 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
484 struct task_struct *lrr_reader;
485 cfs_list_t lrr_linkage;
489 * per file-descriptor read-ahead data.
491 struct ll_readahead_state {
492 cfs_spinlock_t ras_lock;
494 * index of the last page that read(2) needed and that wasn't in the
495 * cache. Used by ras_update() to detect seeks.
497 * XXX nikita: if access seeks into cached region, Lustre doesn't see
500 unsigned long ras_last_readpage;
502 * number of pages read after last read-ahead window reset. As window
503 * is reset on each seek, this is effectively a number of consecutive
504 * accesses. Maybe ->ras_accessed_in_window is better name.
506 * XXX nikita: window is also reset (by ras_update()) when Lustre
507 * believes that memory pressure evicts read-ahead pages. In that
508 * case, it probably doesn't make sense to expand window to
509 * PTLRPC_MAX_BRW_PAGES on the third access.
511 unsigned long ras_consecutive_pages;
513 * number of read requests after the last read-ahead window reset
514 * As window is reset on each seek, this is effectively the number
515 * on consecutive read request and is used to trigger read-ahead.
517 unsigned long ras_consecutive_requests;
519 * Parameters of current read-ahead window. Handled by
520 * ras_update(). On the initial access to the file or after a seek,
521 * window is reset to 0. After 3 consecutive accesses, window is
522 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
523 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
525 unsigned long ras_window_start, ras_window_len;
527 * Where next read-ahead should start at. This lies within read-ahead
528 * window. Read-ahead window is read in pieces rather than at once
529 * because: 1. lustre limits total number of pages under read-ahead by
530 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
531 * not covered by DLM lock.
533 unsigned long ras_next_readahead;
535 * Total number of ll_file_read requests issued, reads originating
536 * due to mmap are not counted in this total. This value is used to
537 * trigger full file read-ahead after multiple reads to a small file.
539 unsigned long ras_requests;
541 * Page index with respect to the current request, these value
542 * will not be accurate when dealing with reads issued via mmap.
544 unsigned long ras_request_index;
546 * list of struct ll_ra_read's one per read(2) call current in
547 * progress against this file descriptor. Used by read-ahead code,
548 * protected by ->ras_lock.
550 cfs_list_t ras_read_beads;
552 * The following 3 items are used for detecting the stride I/O
554 * In stride I/O mode,
555 * ...............|-----data-----|****gap*****|--------|******|....
556 * offset |-stride_pages-|-stride_gap-|
557 * ras_stride_offset = offset;
558 * ras_stride_length = stride_pages + stride_gap;
559 * ras_stride_pages = stride_pages;
560 * Note: all these three items are counted by pages.
562 unsigned long ras_stride_length;
563 unsigned long ras_stride_pages;
564 pgoff_t ras_stride_offset;
566 * number of consecutive stride request count, and it is similar as
567 * ras_consecutive_requests, but used for stride I/O mode.
568 * Note: only more than 2 consecutive stride request are detected,
569 * stride read-ahead will be enable
571 unsigned long ras_consecutive_stride_requests;
579 extern cfs_mem_cache_t *ll_file_data_slab;
580 struct lustre_handle;
581 struct ll_file_data {
582 struct ll_readahead_state fd_ras;
584 struct ccc_grouplock fd_grouplock;
585 struct ll_file_dir fd_dir;
587 struct file *fd_file;
590 struct lov_stripe_md;
592 extern cfs_spinlock_t inode_lock;
594 extern struct proc_dir_entry *proc_lustre_fs_root;
596 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
598 return &lli->lli_vfs_inode;
602 struct inode *icbd_parent;
603 struct dentry **icbd_childp;
607 __u32 ll_i2suppgid(struct inode *i);
608 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
610 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
612 #if BITS_PER_LONG == 32
615 return unlikely(cfs_curproc_is_32bit() || (sbi->ll_flags & LL_SBI_32BIT_API));
619 #define LLAP_MAGIC 98764321
621 extern cfs_mem_cache_t *ll_async_page_slab;
622 extern size_t ll_async_page_slab_size;
624 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
625 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
626 struct ll_ra_read *ll_ra_read_get(struct file *f);
628 /* llite/lproc_llite.c */
630 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
631 struct super_block *sb, char *osc, char *mdc);
632 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
633 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
634 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
636 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
637 struct super_block *sb, char *osc, char *mdc){return 0;}
638 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
639 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
640 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
642 memset(lvars, 0, sizeof(*lvars));
648 void ll_release_page(struct page *page, int remove);
649 extern struct file_operations ll_dir_operations;
650 extern struct inode_operations ll_dir_inode_operations;
651 struct page *ll_get_dir_page(struct file *filp, struct inode *dir, __u64 hash,
652 struct ll_dir_chain *chain);
653 int ll_readdir(struct file *filp, void *cookie, filldir_t filldir);
655 int ll_get_mdt_idx(struct inode *inode);
657 int ll_objects_destroy(struct ptlrpc_request *request,
659 struct inode *ll_iget(struct super_block *sb, ino_t hash,
660 struct lustre_md *lic);
661 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
662 void *data, int flag);
663 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
665 int ll_lookup_it_finish(struct ptlrpc_request *request,
666 struct lookup_intent *it, void *data);
667 struct dentry *ll_find_alias(struct inode *inode, struct dentry *de);
670 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
671 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
672 int ll_writepage(struct page *page, struct writeback_control *wbc);
673 void ll_removepage(struct page *page);
674 int ll_readpage(struct file *file, struct page *page);
675 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
676 void ll_truncate(struct inode *inode);
677 int ll_file_punch(struct inode *, loff_t, int);
678 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
679 void ll_clear_file_contended(struct inode*);
680 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
681 int ll_readahead(const struct lu_env *env, struct cl_io *io,
682 struct ll_readahead_state *ras, struct address_space *mapping,
683 struct cl_page_list *queue, int flags);
686 extern struct file_operations ll_file_operations;
687 extern struct file_operations ll_file_operations_flock;
688 extern struct file_operations ll_file_operations_noflock;
689 extern struct inode_operations ll_file_inode_operations;
690 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
692 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
693 ldlm_mode_t l_req_mode);
694 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
695 struct lustre_handle *lockh);
696 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
698 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
699 int ll_file_open(struct inode *inode, struct file *file);
700 int ll_file_release(struct inode *inode, struct file *file);
701 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
702 struct lov_stripe_md *lsm, lstat_t *st);
703 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
704 int ll_local_open(struct file *file,
705 struct lookup_intent *it, struct ll_file_data *fd,
706 struct obd_client_handle *och);
707 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
708 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
710 int ll_md_real_close(struct inode *inode, int flags);
711 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
712 struct obd_client_handle **och, unsigned long flags);
713 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
714 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
715 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
716 __u64 ioepoch, int sync);
717 int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data,
718 struct md_open_data **mod);
719 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
720 struct lustre_handle *fh);
721 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
722 struct ll_file_data *file, loff_t pos,
723 size_t count, int rw);
724 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
725 struct lookup_intent *it, struct kstat *stat);
726 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
727 struct ll_file_data *ll_file_data_get(void);
728 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
729 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
731 # ifndef HAVE_INODE_PERMISION_2ARGS
732 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
734 int ll_inode_permission(struct inode *inode, int mask);
737 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
738 int flags, struct lov_user_md *lum,
740 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
741 struct lov_mds_md **lmm, int *lmm_size,
742 struct ptlrpc_request **request);
743 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
745 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmm,
746 int *lmm_size, struct ptlrpc_request **request);
747 #ifndef HAVE_FILE_FSYNC_2ARGS
748 int ll_fsync(struct file *file, struct dentry *dentry, int data);
750 int ll_fsync(struct file *file, int data);
752 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
754 int ll_merge_lvb(struct inode *inode);
755 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
756 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
757 int ll_fid2path(struct obd_export *exp, void *arg);
760 int ll_dops_init(struct dentry *de, int block, int init_sa);
761 extern cfs_spinlock_t ll_lookup_lock;
762 extern struct dentry_operations ll_d_ops;
763 void ll_intent_drop_lock(struct lookup_intent *);
764 void ll_intent_release(struct lookup_intent *);
765 int ll_drop_dentry(struct dentry *dentry);
766 void ll_unhash_aliases(struct inode *);
767 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
768 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
769 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
770 int ll_revalidate_it_finish(struct ptlrpc_request *request,
771 struct lookup_intent *it, struct dentry *de);
773 /* llite/llite_lib.c */
774 extern struct super_operations lustre_super_operations;
776 char *ll_read_opt(const char *opt, char *data);
777 void ll_lli_init(struct ll_inode_info *lli);
778 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
779 void ll_put_super(struct super_block *sb);
780 void ll_kill_super(struct super_block *sb);
781 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
782 void ll_clear_inode(struct inode *inode);
783 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr);
784 int ll_setattr(struct dentry *de, struct iattr *attr);
785 #ifndef HAVE_STATFS_DENTRY_PARAM
786 int ll_statfs(struct super_block *sb, struct kstatfs *sfs);
788 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
790 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
791 __u64 max_age, __u32 flags);
792 void ll_update_inode(struct inode *inode, struct lustre_md *md);
793 void ll_read_inode2(struct inode *inode, void *opaque);
794 void ll_delete_inode(struct inode *inode);
795 int ll_iocontrol(struct inode *inode, struct file *file,
796 unsigned int cmd, unsigned long arg);
797 int ll_flush_ctx(struct inode *inode);
798 #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT
799 void ll_umount_begin(struct vfsmount *vfsmnt, int flags);
801 void ll_umount_begin(struct super_block *sb);
803 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
804 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
805 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
806 struct super_block *);
807 void lustre_dump_dentry(struct dentry *, int recur);
808 void lustre_dump_inode(struct inode *);
809 int ll_obd_statfs(struct inode *inode, void *arg);
810 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
811 int ll_process_config(struct lustre_cfg *lcfg);
812 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
813 struct inode *i1, struct inode *i2,
814 const char *name, int namelen,
815 int mode, __u32 opc, void *data);
816 void ll_finish_md_op_data(struct md_op_data *op_data);
817 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
819 /* llite/llite_nfs.c */
820 extern struct export_operations lustre_export_operations;
821 __u32 get_uuid2int(const char *name, int len);
823 /* llite/special.c */
824 extern struct inode_operations ll_special_inode_operations;
825 extern struct file_operations ll_special_chr_inode_fops;
826 extern struct file_operations ll_special_chr_file_fops;
827 extern struct file_operations ll_special_blk_inode_fops;
828 extern struct file_operations ll_special_fifo_inode_fops;
829 extern struct file_operations ll_special_fifo_file_fops;
830 extern struct file_operations ll_special_sock_inode_fops;
832 /* llite/symlink.c */
833 extern struct inode_operations ll_fast_symlink_inode_operations;
835 /* llite/llite_close.c */
836 struct ll_close_queue {
837 cfs_spinlock_t lcq_lock;
839 cfs_waitq_t lcq_waitq;
840 cfs_completion_t lcq_comp;
841 cfs_atomic_t lcq_stop;
844 struct ccc_object *cl_inode2ccc(struct inode *inode);
847 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
848 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
850 /* specific achitecture can implement only part of this list */
851 enum vvp_io_subtype {
854 /** io called from .sendfile */
856 /** io started from splice_{read|write} */
863 enum vvp_io_subtype cui_io_subtype;
867 read_actor_t cui_actor;
871 struct pipe_inode_info *cui_pipe;
872 unsigned int cui_flags;
874 struct vvp_fault_io {
876 * Inode modification time that is checked across DLM
880 struct vm_area_struct *ft_vma;
882 * locked page returned from vvp_io
884 cfs_page_t *ft_vmpage;
885 #ifndef HAVE_VM_OP_FAULT
886 struct vm_nopage_api {
888 * Virtual address at which fault occurred.
890 unsigned long ft_address;
892 * Fault type, as to be supplied to
898 struct vm_fault_api {
902 struct vm_fault *ft_vmf;
904 * fault API used bitflags for return code.
906 unsigned int ft_flags;
912 * Read-ahead state used by read and page-fault IO contexts.
914 struct ll_ra_read cui_bead;
916 * Set when cui_bead has been initialized.
918 int cui_ra_window_set;
920 * Partially truncated page, that vvp_io_trunc_start() keeps locked
923 struct cl_page *cui_partpage;
927 * IO arguments for various VFS I/O interfaces.
930 /** normal/sendfile/splice */
931 enum vvp_io_subtype via_io_subtype;
935 #ifndef HAVE_FILE_WRITEV
936 struct kiocb *via_iocb;
938 struct iovec *via_iov;
939 unsigned long via_nrsegs;
942 read_actor_t via_actor;
946 struct pipe_inode_info *via_pipe;
947 unsigned int via_flags;
952 struct ll_cl_context {
954 struct cl_io *lcc_io;
955 struct cl_page *lcc_page;
956 struct lu_env *lcc_env;
961 struct vvp_thread_info {
962 struct ost_lvb vti_lvb;
963 struct cl_2queue vti_queue;
964 struct iovec vti_local_iov;
965 struct vvp_io_args vti_args;
966 struct ra_io_arg vti_ria;
967 struct kiocb vti_kiocb;
968 struct ll_cl_context vti_io_ctx;
971 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
973 extern struct lu_context_key vvp_key;
974 struct vvp_thread_info *info;
976 info = lu_context_key_get(&env->le_ctx, &vvp_key);
977 LASSERT(info != NULL);
981 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
982 enum vvp_io_subtype type)
984 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
986 ret->via_io_subtype = type;
992 struct vvp_io vs_ios;
995 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
997 extern struct lu_context_key vvp_session_key;
998 struct vvp_session *ses;
1000 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1001 LASSERT(ses != NULL);
1005 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1007 return &vvp_env_session(env)->vs_ios;
1010 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1011 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1012 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1014 /* llite/llite_mmap.c */
1015 typedef struct rb_root rb_root_t;
1016 typedef struct rb_node rb_node_t;
1018 struct ll_lock_tree_node;
1019 struct ll_lock_tree {
1021 cfs_list_t lt_locked_list;
1022 struct ll_file_data *lt_fd;
1025 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1026 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1027 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
1028 __u64 end, ldlm_mode_t mode);
1029 void policy_from_vma(ldlm_policy_data_t *policy,
1030 struct vm_area_struct *vma, unsigned long addr, size_t count);
1031 struct vm_area_struct *our_vma(unsigned long addr, size_t count);
1033 static inline void ll_invalidate_page(struct page *vmpage)
1035 struct address_space *mapping = vmpage->mapping;
1036 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1038 LASSERT(PageLocked(vmpage));
1039 if (mapping == NULL)
1042 ll_teardown_mmaps(mapping, offset, offset + CFS_PAGE_SIZE);
1043 truncate_complete_page(mapping, vmpage);
1046 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1048 /* don't need an addref as the sb_info should be holding one */
1049 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1051 return ll_s2sbi(sb)->ll_dt_exp;
1054 /* don't need an addref as the sb_info should be holding one */
1055 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1057 return ll_s2sbi(sb)->ll_md_exp;
1060 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1062 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1068 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1069 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1071 return ll_s2sbi(inode->i_sb);
1074 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1076 return ll_s2dtexp(inode->i_sb);
1079 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1081 return ll_s2mdexp(inode->i_sb);
1084 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1087 LASSERT(inode != NULL);
1088 fid = &ll_i2info(inode)->lli_fid;
1089 LASSERT(fid_is_igif(fid) || fid_ver(fid) == 0);
1093 static inline int ll_mds_max_easize(struct super_block *sb)
1095 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1098 static inline __u64 ll_file_maxbytes(struct inode *inode)
1100 return ll_i2info(inode)->lli_maxbytes;
1104 int ll_setxattr(struct dentry *dentry, const char *name,
1105 const void *value, size_t size, int flags);
1106 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1107 void *buffer, size_t size);
1108 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1109 int ll_removexattr(struct dentry *dentry, const char *name);
1111 /* llite/remote_perm.c */
1112 extern cfs_mem_cache_t *ll_remote_perm_cachep;
1113 extern cfs_mem_cache_t *ll_rmtperm_hash_cachep;
1115 cfs_hlist_head_t *alloc_rmtperm_hash(void);
1116 void free_rmtperm_hash(cfs_hlist_head_t *hash);
1117 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1118 int lustre_check_remote_perm(struct inode *inode, int mask);
1120 /* llite/llite_capa.c */
1121 extern cfs_timer_t ll_capa_timer;
1123 int ll_capa_thread_start(void);
1124 void ll_capa_thread_stop(void);
1125 void ll_capa_timer_callback(unsigned long unused);
1127 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1128 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1130 void ll_capa_open(struct inode *inode);
1131 void ll_capa_close(struct inode *inode);
1133 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1134 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1136 void ll_truncate_free_capa(struct obd_capa *ocapa);
1137 void ll_clear_inode_capas(struct inode *inode);
1138 void ll_print_capa_stat(struct ll_sb_info *sbi);
1140 /* llite/llite_cl.c */
1141 extern struct lu_device_type vvp_device_type;
1144 * Common IO arguments for various VFS I/O interfaces.
1146 int cl_sb_init(struct super_block *sb);
1147 int cl_sb_fini(struct super_block *sb);
1148 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1149 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1151 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1152 struct ll_readahead_state *ras, unsigned long index,
1154 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1155 int ll_is_file_contended(struct file *file);
1156 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1158 /* llite/llite_rmtacl.c */
1159 #ifdef CONFIG_FS_POSIX_ACL
1160 obd_valid rce_ops2valid(int ops);
1161 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1162 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1163 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1164 void rct_init(struct rmtacl_ctl_table *rct);
1165 void rct_fini(struct rmtacl_ctl_table *rct);
1167 void ee_free(struct eacl_entry *ee);
1168 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1169 ext_acl_xattr_header *header);
1170 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1171 struct lu_fid *fid, int type);
1172 void et_search_free(struct eacl_table *et, pid_t key);
1173 void et_init(struct eacl_table *et);
1174 void et_fini(struct eacl_table *et);
1179 #define LL_SA_RPC_MIN 2
1180 #define LL_SA_RPC_DEF 32
1181 #define LL_SA_RPC_MAX 8192
1183 #define LL_SA_CACHE_BIT 5
1184 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1185 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1187 /* per inode struct, for dir only */
1188 struct ll_statahead_info {
1189 struct inode *sai_inode;
1190 cfs_atomic_t sai_refcount; /* when access this struct, hold
1192 unsigned int sai_generation; /* generation for statahead */
1193 unsigned int sai_max; /* max ahead of lookup */
1194 __u64 sai_sent; /* stat requests sent count */
1195 __u64 sai_replied; /* stat requests which received
1197 __u64 sai_index; /* index of statahead entry */
1198 __u64 sai_index_wait; /* index of entry which is the
1199 * caller is waiting for */
1200 __u64 sai_hit; /* hit count */
1201 __u64 sai_miss; /* miss count:
1202 * for "ls -al" case, it includes
1203 * hidden dentry miss;
1204 * for "ls -l" case, it does not
1205 * include hidden dentry miss.
1206 * "sai_miss_hidden" is used for
1209 unsigned int sai_consecutive_miss; /* consecutive miss */
1210 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1211 * is not a hidden one */
1212 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1213 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1215 sai_in_readpage:1,/* statahead is in readdir()*/
1216 sai_agl_valid:1;/* AGL is valid for the dir */
1217 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
1218 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1219 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1220 cfs_list_t sai_entries_sent; /* entries sent out */
1221 cfs_list_t sai_entries_received; /* entries returned */
1222 cfs_list_t sai_entries_stated; /* entries stated */
1223 cfs_list_t sai_entries_agl; /* AGL entries to be sent */
1224 cfs_list_t sai_cache[LL_SA_CACHE_SIZE];
1225 cfs_spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1226 cfs_atomic_t sai_cache_count; /* entry count in cache */
1229 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1231 void ll_stop_statahead(struct inode *dir, void *key);
1233 static inline int ll_glimpse_size(struct inode *inode)
1235 struct ll_inode_info *lli = ll_i2info(inode);
1238 cfs_down_read(&lli->lli_glimpse_sem);
1239 rc = cl_glimpse_size(inode);
1240 lli->lli_glimpse_time = cfs_time_current();
1241 cfs_up_read(&lli->lli_glimpse_sem);
1246 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1248 struct ll_inode_info *lli = ll_i2info(dir);
1249 struct ll_statahead_info *sai = lli->lli_sai;
1250 struct ll_dentry_data *ldd = ll_d2d(dentry);
1252 /* not the same process, don't mark */
1253 if (lli->lli_opendir_pid != cfs_curproc_pid())
1256 if (sai != NULL && ldd != NULL)
1257 ldd->lld_sa_generation = sai->sai_generation;
1261 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1263 struct ll_inode_info *lli;
1264 struct ll_dentry_data *ldd;
1266 if (ll_i2sbi(dir)->ll_sa_max == 0)
1269 lli = ll_i2info(dir);
1270 /* not the same process, don't statahead */
1271 if (lli->lli_opendir_pid != cfs_curproc_pid())
1274 ldd = ll_d2d(*dentryp);
1276 * When stats a dentry, the system trigger more than once "revalidate"
1277 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1278 * Under patchless client mode, the operation intent is not accurate,
1279 * which maybe misguide the statahead thread. For example:
1280 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1281 * have the same operation intent -- "IT_GETATTR".
1282 * In fact, one dentry should has only one chance to interact with the
1283 * statahead thread, otherwise the statahead windows will be confused.
1284 * The solution is as following:
1285 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1286 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1287 * will bypass interacting with statahead thread for checking:
1288 * "lld_sa_generation == lli_sai->sai_generation"
1290 if (ldd && lli->lli_sai &&
1291 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1294 return do_statahead_enter(dir, dentryp, only_unplug);
1297 /* llite ioctl register support rountine */
1304 #define LLIOC_MAX_CMD 256
1307 * Rules to write a callback function:
1310 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1311 * returned to ll_iocontrol_register. Callback functions should use this
1312 * data to check the potential collasion of ioctl cmd. If collasion is
1313 * found, callback function should return LLIOC_CONT.
1314 * @rcp: The result of ioctl command.
1317 * If @magic matches the pointer returned by ll_iocontrol_data, the
1318 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1320 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1321 struct file *file, unsigned int cmd, unsigned long arg,
1322 void *magic, int *rcp);
1324 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1325 unsigned int cmd, unsigned long arg, int *rcp);
1327 /* export functions */
1328 /* Register ioctl block dynamatically for a regular file.
1330 * @cmd: the array of ioctl command set
1331 * @count: number of commands in the @cmd
1332 * @cb: callback function, it will be called if an ioctl command is found to
1333 * belong to the command list @cmd.
1336 * A magic pointer will be returned if success;
1337 * otherwise, NULL will be returned.
1339 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1340 void ll_iocontrol_unregister(void *magic);
1344 /* lclient compat stuff */
1345 #define cl_inode_info ll_inode_info
1346 #define cl_i2info(info) ll_i2info(info)
1347 #define cl_inode_mode(inode) ((inode)->i_mode)
1348 #define cl_i2sbi ll_i2sbi
1350 static inline void cl_isize_lock(struct inode *inode, int lsmlock)
1352 ll_inode_size_lock(inode, lsmlock);
1355 static inline void cl_isize_unlock(struct inode *inode, int lsmlock)
1357 ll_inode_size_unlock(inode, lsmlock);
1360 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1362 LASSERT_SEM_LOCKED(&ll_i2info(inode)->lli_size_sem);
1363 i_size_write(inode, kms);
1366 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1368 ll_inode_size_lock(inode, 0);
1369 i_size_write(inode, kms);
1370 ll_inode_size_unlock(inode, 0);
1373 #define cl_isize_read(inode) i_size_read(inode)
1375 static inline int cl_merge_lvb(struct inode *inode)
1377 return ll_merge_lvb(inode);
1380 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1381 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1382 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1384 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1386 /** direct write pages */
1387 struct ll_dio_pages {
1388 /** page array to be written. we don't support
1389 * partial pages except the last one. */
1390 struct page **ldp_pages;
1391 /* offset of each page */
1392 loff_t *ldp_offsets;
1393 /** if ldp_offsets is NULL, it means a sequential
1394 * pages to be written, then this is the file offset
1395 * of the * first page. */
1396 loff_t ldp_start_offset;
1397 /** how many bytes are to be written. */
1399 /** # of pages in the array. */
1403 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1406 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1409 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1412 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1413 int rw, struct inode *inode,
1414 struct ll_dio_pages *pv);
1416 static inline int ll_file_nolock(const struct file *file)
1418 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1419 struct inode *inode = file->f_dentry->d_inode;
1421 LASSERT(fd != NULL);
1422 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1423 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1426 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1427 struct lookup_intent *it, __u64 *bits)
1429 if (!it->d.lustre.it_lock_set) {
1430 CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)\n",
1431 inode, inode->i_ino, inode->i_generation);
1432 md_set_lock_data(exp, &it->d.lustre.it_lock_handle,
1433 inode, &it->d.lustre.it_lock_bits);
1434 it->d.lustre.it_lock_set = 1;
1438 *bits = it->d.lustre.it_lock_bits;
1441 static inline void ll_dentry_rehash(struct dentry *dentry, int locked)
1444 cfs_spin_lock(&ll_lookup_lock);
1445 spin_lock(&dcache_lock);
1447 if (d_unhashed(dentry))
1448 d_rehash_cond(dentry, 0);
1450 spin_unlock(&dcache_lock);
1451 cfs_spin_unlock(&ll_lookup_lock);
1455 static inline void ll_dentry_reset_flags(struct dentry *dentry, __u64 bits)
1457 if (bits & MDS_INODELOCK_LOOKUP &&
1458 dentry->d_flags & DCACHE_LUSTRE_INVALID) {
1459 lock_dentry(dentry);
1460 dentry->d_flags &= ~DCACHE_LUSTRE_INVALID;
1461 unlock_dentry(dentry);
1465 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2,7,50,0)
1466 /* Compatibility for old (1.8) compiled userspace quota code */
1467 struct if_quotactl_18 {
1472 struct obd_dqinfo qc_dqinfo;
1473 struct obd_dqblk qc_dqblk;
1475 struct obd_uuid obd_uuid;
1477 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1478 /* End compatibility for old (1.8) compiled userspace quota code */
1480 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1481 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2,7,50,0) */
1483 #endif /* LLITE_INTERNAL_H */