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_mutex_t lli_rmtperm_mutex;
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_mutex_t lli_och_mutex;
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_mutex_t d_readdir_mutex;
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_mutex u.d.d_readdir_mutex
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_mutex_t f_write_mutex;
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_mutex u.f.f_write_mutex
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 */
382 #define LL_SBI_VERBOSE 0x10000 /* verbose mount/umount */
384 /* default value for ll_sb_info->contention_time */
385 #define SBI_DEFAULT_CONTENTION_SECONDS 60
386 /* default value for lockless_truncate_enable */
387 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
388 #define RCE_HASHES 32
390 struct rmtacl_ctl_entry {
392 pid_t rce_key; /* hash key */
393 int rce_ops; /* acl operation type */
396 struct rmtacl_ctl_table {
397 cfs_spinlock_t rct_lock;
398 cfs_list_t rct_entries[RCE_HASHES];
405 pid_t ee_key; /* hash key */
406 struct lu_fid ee_fid;
407 int ee_type; /* ACL type for ACCESS or DEFAULT */
408 ext_acl_xattr_header *ee_acl;
412 cfs_spinlock_t et_lock;
413 cfs_list_t et_entries[EE_HASHES];
418 /* this protects pglist and ra_info. It isn't safe to
419 * grab from interrupt contexts */
420 cfs_spinlock_t ll_lock;
421 cfs_spinlock_t ll_pp_extent_lock; /* Lock for pp_extent entries */
422 cfs_spinlock_t ll_process_lock; /* Lock for ll_rw_process_info */
423 struct obd_uuid ll_sb_uuid;
424 struct obd_export *ll_md_exp;
425 struct obd_export *ll_dt_exp;
426 struct proc_dir_entry* ll_proc_root;
427 struct lu_fid ll_root_fid; /* root object fid */
430 cfs_list_t ll_conn_chain; /* per-conn chain of SBs */
431 struct lustre_client_ocd ll_lco;
433 cfs_list_t ll_orphan_dentry_list; /*please don't ask -p*/
434 struct ll_close_queue *ll_lcq;
436 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
438 unsigned long ll_async_page_max;
439 unsigned long ll_async_page_count;
441 struct lprocfs_stats *ll_ra_stats;
443 struct ll_ra_info ll_ra_info;
444 unsigned int ll_namelen;
445 struct file_operations *ll_fop;
447 /* =0 - hold lock over whole read/write
448 * >0 - max. chunk to be read/written w/o lock re-acquiring */
449 unsigned long ll_max_rw_chunk;
450 unsigned int ll_md_brw_size; /* used by readdir */
452 struct lu_site *ll_site;
453 struct cl_device *ll_cl;
455 struct ll_rw_extents_info ll_rw_extents_info;
456 int ll_extent_process_count;
457 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
458 unsigned int ll_offset_process_count;
459 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
460 unsigned int ll_rw_offset_entry_count;
461 int ll_stats_track_id;
462 enum stats_track_type ll_stats_track_type;
465 /* metadata stat-ahead */
466 unsigned int ll_sa_max; /* max statahead RPCs */
467 atomic_t ll_sa_total; /* statahead thread started
469 atomic_t ll_sa_wrong; /* statahead thread stopped for
471 atomic_t ll_agl_total; /* AGL thread started count */
473 dev_t ll_sdev_orig; /* save s_dev before assign for
475 struct rmtacl_ctl_table ll_rct;
476 struct eacl_table ll_et;
477 struct vfsmount *ll_mnt;
480 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
485 struct task_struct *lrr_reader;
486 cfs_list_t lrr_linkage;
490 * per file-descriptor read-ahead data.
492 struct ll_readahead_state {
493 cfs_spinlock_t ras_lock;
495 * index of the last page that read(2) needed and that wasn't in the
496 * cache. Used by ras_update() to detect seeks.
498 * XXX nikita: if access seeks into cached region, Lustre doesn't see
501 unsigned long ras_last_readpage;
503 * number of pages read after last read-ahead window reset. As window
504 * is reset on each seek, this is effectively a number of consecutive
505 * accesses. Maybe ->ras_accessed_in_window is better name.
507 * XXX nikita: window is also reset (by ras_update()) when Lustre
508 * believes that memory pressure evicts read-ahead pages. In that
509 * case, it probably doesn't make sense to expand window to
510 * PTLRPC_MAX_BRW_PAGES on the third access.
512 unsigned long ras_consecutive_pages;
514 * number of read requests after the last read-ahead window reset
515 * As window is reset on each seek, this is effectively the number
516 * on consecutive read request and is used to trigger read-ahead.
518 unsigned long ras_consecutive_requests;
520 * Parameters of current read-ahead window. Handled by
521 * ras_update(). On the initial access to the file or after a seek,
522 * window is reset to 0. After 3 consecutive accesses, window is
523 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
524 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
526 unsigned long ras_window_start, ras_window_len;
528 * Where next read-ahead should start at. This lies within read-ahead
529 * window. Read-ahead window is read in pieces rather than at once
530 * because: 1. lustre limits total number of pages under read-ahead by
531 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
532 * not covered by DLM lock.
534 unsigned long ras_next_readahead;
536 * Total number of ll_file_read requests issued, reads originating
537 * due to mmap are not counted in this total. This value is used to
538 * trigger full file read-ahead after multiple reads to a small file.
540 unsigned long ras_requests;
542 * Page index with respect to the current request, these value
543 * will not be accurate when dealing with reads issued via mmap.
545 unsigned long ras_request_index;
547 * list of struct ll_ra_read's one per read(2) call current in
548 * progress against this file descriptor. Used by read-ahead code,
549 * protected by ->ras_lock.
551 cfs_list_t ras_read_beads;
553 * The following 3 items are used for detecting the stride I/O
555 * In stride I/O mode,
556 * ...............|-----data-----|****gap*****|--------|******|....
557 * offset |-stride_pages-|-stride_gap-|
558 * ras_stride_offset = offset;
559 * ras_stride_length = stride_pages + stride_gap;
560 * ras_stride_pages = stride_pages;
561 * Note: all these three items are counted by pages.
563 unsigned long ras_stride_length;
564 unsigned long ras_stride_pages;
565 pgoff_t ras_stride_offset;
567 * number of consecutive stride request count, and it is similar as
568 * ras_consecutive_requests, but used for stride I/O mode.
569 * Note: only more than 2 consecutive stride request are detected,
570 * stride read-ahead will be enable
572 unsigned long ras_consecutive_stride_requests;
580 extern cfs_mem_cache_t *ll_file_data_slab;
581 struct lustre_handle;
582 struct ll_file_data {
583 struct ll_readahead_state fd_ras;
585 struct ccc_grouplock fd_grouplock;
586 struct ll_file_dir fd_dir;
588 struct file *fd_file;
591 struct lov_stripe_md;
593 extern cfs_spinlock_t inode_lock;
595 extern struct proc_dir_entry *proc_lustre_fs_root;
597 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
599 return &lli->lli_vfs_inode;
603 struct inode *icbd_parent;
604 struct dentry **icbd_childp;
608 __u32 ll_i2suppgid(struct inode *i);
609 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
611 static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
613 #if BITS_PER_LONG == 32
616 return unlikely(cfs_curproc_is_32bit() || (sbi->ll_flags & LL_SBI_32BIT_API));
620 #define LLAP_MAGIC 98764321
622 extern cfs_mem_cache_t *ll_async_page_slab;
623 extern size_t ll_async_page_slab_size;
625 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
626 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
627 struct ll_ra_read *ll_ra_read_get(struct file *f);
629 /* llite/lproc_llite.c */
631 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
632 struct super_block *sb, char *osc, char *mdc);
633 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
634 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
635 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
637 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
638 struct super_block *sb, char *osc, char *mdc){return 0;}
639 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
640 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
641 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
643 memset(lvars, 0, sizeof(*lvars));
649 void ll_release_page(struct page *page, int remove);
650 extern struct file_operations ll_dir_operations;
651 extern struct inode_operations ll_dir_inode_operations;
652 struct page *ll_get_dir_page(struct file *filp, struct inode *dir, __u64 hash,
653 struct ll_dir_chain *chain);
654 int ll_readdir(struct file *filp, void *cookie, filldir_t filldir);
656 int ll_get_mdt_idx(struct inode *inode);
658 int ll_objects_destroy(struct ptlrpc_request *request,
660 struct inode *ll_iget(struct super_block *sb, ino_t hash,
661 struct lustre_md *lic);
662 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
663 void *data, int flag);
664 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
666 int ll_lookup_it_finish(struct ptlrpc_request *request,
667 struct lookup_intent *it, void *data);
668 struct dentry *ll_find_alias(struct inode *inode, struct dentry *de);
671 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
672 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
673 int ll_writepage(struct page *page, struct writeback_control *wbc);
674 void ll_removepage(struct page *page);
675 int ll_readpage(struct file *file, struct page *page);
676 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
677 void ll_truncate(struct inode *inode);
678 int ll_file_punch(struct inode *, loff_t, int);
679 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
680 void ll_clear_file_contended(struct inode*);
681 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
682 int ll_readahead(const struct lu_env *env, struct cl_io *io,
683 struct ll_readahead_state *ras, struct address_space *mapping,
684 struct cl_page_list *queue, int flags);
687 extern struct file_operations ll_file_operations;
688 extern struct file_operations ll_file_operations_flock;
689 extern struct file_operations ll_file_operations_noflock;
690 extern struct inode_operations ll_file_inode_operations;
691 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
693 extern int ll_have_md_lock(struct inode *inode, __u64 *bits,
694 ldlm_mode_t l_req_mode);
695 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
696 struct lustre_handle *lockh);
697 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *,
699 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
700 int ll_file_open(struct inode *inode, struct file *file);
701 int ll_file_release(struct inode *inode, struct file *file);
702 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
703 struct lov_stripe_md *lsm, lstat_t *st);
704 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
705 int ll_local_open(struct file *file,
706 struct lookup_intent *it, struct ll_file_data *fd,
707 struct obd_client_handle *och);
708 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
709 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
711 int ll_md_real_close(struct inode *inode, int flags);
712 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
713 struct obd_client_handle **och, unsigned long flags);
714 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
715 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
716 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
717 __u64 ioepoch, int sync);
718 int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data,
719 struct md_open_data **mod);
720 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
721 struct lustre_handle *fh);
722 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
723 struct ll_file_data *file, loff_t pos,
724 size_t count, int rw);
725 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
726 struct lookup_intent *it, struct kstat *stat);
727 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
728 struct ll_file_data *ll_file_data_get(void);
729 #ifdef HAVE_GENERIC_PERMISSION_4ARGS
730 int ll_inode_permission(struct inode *inode, int mask, unsigned int flags);
732 # ifndef HAVE_INODE_PERMISION_2ARGS
733 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
735 int ll_inode_permission(struct inode *inode, int mask);
738 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
739 int flags, struct lov_user_md *lum,
741 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
742 struct lov_mds_md **lmm, int *lmm_size,
743 struct ptlrpc_request **request);
744 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
746 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmm,
747 int *lmm_size, struct ptlrpc_request **request);
748 #ifdef HAVE_FILE_FSYNC_4ARGS
749 int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
750 #elif defined(HAVE_FILE_FSYNC_2ARGS)
751 int ll_fsync(struct file *file, int data);
753 int ll_fsync(struct file *file, struct dentry *dentry, int data);
755 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
757 int ll_merge_lvb(struct inode *inode);
758 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
759 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
760 int ll_fid2path(struct obd_export *exp, void *arg);
763 int ll_dops_init(struct dentry *de, int block, int init_sa);
764 extern cfs_spinlock_t ll_lookup_lock;
765 extern struct dentry_operations ll_d_ops;
766 void ll_intent_drop_lock(struct lookup_intent *);
767 void ll_intent_release(struct lookup_intent *);
768 int ll_drop_dentry(struct dentry *dentry);
769 void ll_unhash_aliases(struct inode *);
770 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
771 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
772 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
773 int ll_revalidate_it_finish(struct ptlrpc_request *request,
774 struct lookup_intent *it, struct dentry *de);
776 /* llite/llite_lib.c */
777 extern struct super_operations lustre_super_operations;
779 char *ll_read_opt(const char *opt, char *data);
780 void ll_lli_init(struct ll_inode_info *lli);
781 int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
782 void ll_put_super(struct super_block *sb);
783 void ll_kill_super(struct super_block *sb);
784 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
785 void ll_clear_inode(struct inode *inode);
786 int ll_setattr_raw(struct dentry *dentry, struct iattr *attr);
787 int ll_setattr(struct dentry *de, struct iattr *attr);
788 #ifndef HAVE_STATFS_DENTRY_PARAM
789 int ll_statfs(struct super_block *sb, struct kstatfs *sfs);
791 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
793 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
794 __u64 max_age, __u32 flags);
795 void ll_update_inode(struct inode *inode, struct lustre_md *md);
796 void ll_read_inode2(struct inode *inode, void *opaque);
797 void ll_delete_inode(struct inode *inode);
798 int ll_iocontrol(struct inode *inode, struct file *file,
799 unsigned int cmd, unsigned long arg);
800 int ll_flush_ctx(struct inode *inode);
801 #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT
802 void ll_umount_begin(struct vfsmount *vfsmnt, int flags);
804 void ll_umount_begin(struct super_block *sb);
806 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
807 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
808 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
809 struct super_block *);
810 void lustre_dump_dentry(struct dentry *, int recur);
811 void lustre_dump_inode(struct inode *);
812 int ll_obd_statfs(struct inode *inode, void *arg);
813 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
814 int ll_process_config(struct lustre_cfg *lcfg);
815 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
816 struct inode *i1, struct inode *i2,
817 const char *name, int namelen,
818 int mode, __u32 opc, void *data);
819 void ll_finish_md_op_data(struct md_op_data *op_data);
820 int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
822 /* llite/llite_nfs.c */
823 extern struct export_operations lustre_export_operations;
824 __u32 get_uuid2int(const char *name, int len);
826 /* llite/special.c */
827 extern struct inode_operations ll_special_inode_operations;
828 extern struct file_operations ll_special_chr_inode_fops;
829 extern struct file_operations ll_special_chr_file_fops;
830 extern struct file_operations ll_special_blk_inode_fops;
831 extern struct file_operations ll_special_fifo_inode_fops;
832 extern struct file_operations ll_special_fifo_file_fops;
833 extern struct file_operations ll_special_sock_inode_fops;
835 /* llite/symlink.c */
836 extern struct inode_operations ll_fast_symlink_inode_operations;
838 /* llite/llite_close.c */
839 struct ll_close_queue {
840 cfs_spinlock_t lcq_lock;
842 cfs_waitq_t lcq_waitq;
843 cfs_completion_t lcq_comp;
844 cfs_atomic_t lcq_stop;
847 struct ccc_object *cl_inode2ccc(struct inode *inode);
850 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
851 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
853 /* specific achitecture can implement only part of this list */
854 enum vvp_io_subtype {
857 /** io called from .sendfile */
859 /** io started from splice_{read|write} */
866 enum vvp_io_subtype cui_io_subtype;
870 read_actor_t cui_actor;
874 struct pipe_inode_info *cui_pipe;
875 unsigned int cui_flags;
877 struct vvp_fault_io {
879 * Inode modification time that is checked across DLM
883 struct vm_area_struct *ft_vma;
885 * locked page returned from vvp_io
887 cfs_page_t *ft_vmpage;
888 #ifndef HAVE_VM_OP_FAULT
889 struct vm_nopage_api {
891 * Virtual address at which fault occurred.
893 unsigned long ft_address;
895 * Fault type, as to be supplied to
901 struct vm_fault_api {
905 struct vm_fault *ft_vmf;
907 * fault API used bitflags for return code.
909 unsigned int ft_flags;
915 * Read-ahead state used by read and page-fault IO contexts.
917 struct ll_ra_read cui_bead;
919 * Set when cui_bead has been initialized.
921 int cui_ra_window_set;
923 * Partially truncated page, that vvp_io_trunc_start() keeps locked
926 struct cl_page *cui_partpage;
930 * IO arguments for various VFS I/O interfaces.
933 /** normal/sendfile/splice */
934 enum vvp_io_subtype via_io_subtype;
938 #ifndef HAVE_FILE_WRITEV
939 struct kiocb *via_iocb;
941 struct iovec *via_iov;
942 unsigned long via_nrsegs;
945 read_actor_t via_actor;
949 struct pipe_inode_info *via_pipe;
950 unsigned int via_flags;
955 struct ll_cl_context {
957 struct cl_io *lcc_io;
958 struct cl_page *lcc_page;
959 struct lu_env *lcc_env;
964 struct vvp_thread_info {
965 struct ost_lvb vti_lvb;
966 struct cl_2queue vti_queue;
967 struct iovec vti_local_iov;
968 struct vvp_io_args vti_args;
969 struct ra_io_arg vti_ria;
970 struct kiocb vti_kiocb;
971 struct ll_cl_context vti_io_ctx;
974 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
976 extern struct lu_context_key vvp_key;
977 struct vvp_thread_info *info;
979 info = lu_context_key_get(&env->le_ctx, &vvp_key);
980 LASSERT(info != NULL);
984 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
985 enum vvp_io_subtype type)
987 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
989 ret->via_io_subtype = type;
995 struct vvp_io vs_ios;
998 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
1000 extern struct lu_context_key vvp_session_key;
1001 struct vvp_session *ses;
1003 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
1004 LASSERT(ses != NULL);
1008 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
1010 return &vvp_env_session(env)->vs_ios;
1013 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
1014 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
1015 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
1017 /* llite/llite_mmap.c */
1018 typedef struct rb_root rb_root_t;
1019 typedef struct rb_node rb_node_t;
1021 struct ll_lock_tree_node;
1022 struct ll_lock_tree {
1024 cfs_list_t lt_locked_list;
1025 struct ll_file_data *lt_fd;
1028 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
1029 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
1030 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
1031 __u64 end, ldlm_mode_t mode);
1032 void policy_from_vma(ldlm_policy_data_t *policy,
1033 struct vm_area_struct *vma, unsigned long addr, size_t count);
1034 struct vm_area_struct *our_vma(unsigned long addr, size_t count);
1036 static inline void ll_invalidate_page(struct page *vmpage)
1038 struct address_space *mapping = vmpage->mapping;
1039 loff_t offset = vmpage->index << PAGE_CACHE_SHIFT;
1041 LASSERT(PageLocked(vmpage));
1042 if (mapping == NULL)
1045 ll_teardown_mmaps(mapping, offset, offset + CFS_PAGE_SIZE);
1046 truncate_complete_page(mapping, vmpage);
1049 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
1051 /* don't need an addref as the sb_info should be holding one */
1052 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
1054 return ll_s2sbi(sb)->ll_dt_exp;
1057 /* don't need an addref as the sb_info should be holding one */
1058 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
1060 return ll_s2sbi(sb)->ll_md_exp;
1063 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
1065 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
1071 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
1072 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
1074 return ll_s2sbi(inode->i_sb);
1077 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
1079 return ll_s2dtexp(inode->i_sb);
1082 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
1084 return ll_s2mdexp(inode->i_sb);
1087 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
1090 LASSERT(inode != NULL);
1091 fid = &ll_i2info(inode)->lli_fid;
1092 LASSERT(fid_is_igif(fid) || fid_ver(fid) == 0);
1096 static inline int ll_mds_max_easize(struct super_block *sb)
1098 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1101 static inline __u64 ll_file_maxbytes(struct inode *inode)
1103 return ll_i2info(inode)->lli_maxbytes;
1107 int ll_setxattr(struct dentry *dentry, const char *name,
1108 const void *value, size_t size, int flags);
1109 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1110 void *buffer, size_t size);
1111 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1112 int ll_removexattr(struct dentry *dentry, const char *name);
1114 /* llite/remote_perm.c */
1115 extern cfs_mem_cache_t *ll_remote_perm_cachep;
1116 extern cfs_mem_cache_t *ll_rmtperm_hash_cachep;
1118 cfs_hlist_head_t *alloc_rmtperm_hash(void);
1119 void free_rmtperm_hash(cfs_hlist_head_t *hash);
1120 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1121 int lustre_check_remote_perm(struct inode *inode, int mask);
1123 /* llite/llite_capa.c */
1124 extern cfs_timer_t ll_capa_timer;
1126 int ll_capa_thread_start(void);
1127 void ll_capa_thread_stop(void);
1128 void ll_capa_timer_callback(unsigned long unused);
1130 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1131 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1133 void ll_capa_open(struct inode *inode);
1134 void ll_capa_close(struct inode *inode);
1136 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1137 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1139 void ll_truncate_free_capa(struct obd_capa *ocapa);
1140 void ll_clear_inode_capas(struct inode *inode);
1141 void ll_print_capa_stat(struct ll_sb_info *sbi);
1143 /* llite/llite_cl.c */
1144 extern struct lu_device_type vvp_device_type;
1147 * Common IO arguments for various VFS I/O interfaces.
1149 int cl_sb_init(struct super_block *sb);
1150 int cl_sb_fini(struct super_block *sb);
1151 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1152 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1154 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1155 struct ll_readahead_state *ras, unsigned long index,
1157 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1158 int ll_is_file_contended(struct file *file);
1159 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1161 /* llite/llite_rmtacl.c */
1162 #ifdef CONFIG_FS_POSIX_ACL
1163 obd_valid rce_ops2valid(int ops);
1164 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1165 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1166 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1167 void rct_init(struct rmtacl_ctl_table *rct);
1168 void rct_fini(struct rmtacl_ctl_table *rct);
1170 void ee_free(struct eacl_entry *ee);
1171 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1172 ext_acl_xattr_header *header);
1173 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1174 struct lu_fid *fid, int type);
1175 void et_search_free(struct eacl_table *et, pid_t key);
1176 void et_init(struct eacl_table *et);
1177 void et_fini(struct eacl_table *et);
1182 #define LL_SA_RPC_MIN 2
1183 #define LL_SA_RPC_DEF 32
1184 #define LL_SA_RPC_MAX 8192
1186 #define LL_SA_CACHE_BIT 5
1187 #define LL_SA_CACHE_SIZE (1 << LL_SA_CACHE_BIT)
1188 #define LL_SA_CACHE_MASK (LL_SA_CACHE_SIZE - 1)
1190 /* per inode struct, for dir only */
1191 struct ll_statahead_info {
1192 struct inode *sai_inode;
1193 cfs_atomic_t sai_refcount; /* when access this struct, hold
1195 unsigned int sai_generation; /* generation for statahead */
1196 unsigned int sai_max; /* max ahead of lookup */
1197 __u64 sai_sent; /* stat requests sent count */
1198 __u64 sai_replied; /* stat requests which received
1200 __u64 sai_index; /* index of statahead entry */
1201 __u64 sai_index_wait; /* index of entry which is the
1202 * caller is waiting for */
1203 __u64 sai_hit; /* hit count */
1204 __u64 sai_miss; /* miss count:
1205 * for "ls -al" case, it includes
1206 * hidden dentry miss;
1207 * for "ls -l" case, it does not
1208 * include hidden dentry miss.
1209 * "sai_miss_hidden" is used for
1212 unsigned int sai_consecutive_miss; /* consecutive miss */
1213 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1214 * is not a hidden one */
1215 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1216 unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
1218 sai_in_readpage:1,/* statahead is in readdir()*/
1219 sai_agl_valid:1;/* AGL is valid for the dir */
1220 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
1221 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1222 struct ptlrpc_thread sai_agl_thread; /* AGL thread */
1223 cfs_list_t sai_entries_sent; /* entries sent out */
1224 cfs_list_t sai_entries_received; /* entries returned */
1225 cfs_list_t sai_entries_stated; /* entries stated */
1226 cfs_list_t sai_entries_agl; /* AGL entries to be sent */
1227 cfs_list_t sai_cache[LL_SA_CACHE_SIZE];
1228 cfs_spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
1229 cfs_atomic_t sai_cache_count; /* entry count in cache */
1232 int do_statahead_enter(struct inode *dir, struct dentry **dentry,
1234 void ll_stop_statahead(struct inode *dir, void *key);
1236 static inline int ll_glimpse_size(struct inode *inode)
1238 struct ll_inode_info *lli = ll_i2info(inode);
1241 cfs_down_read(&lli->lli_glimpse_sem);
1242 rc = cl_glimpse_size(inode);
1243 lli->lli_glimpse_time = cfs_time_current();
1244 cfs_up_read(&lli->lli_glimpse_sem);
1249 ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1251 struct ll_inode_info *lli = ll_i2info(dir);
1252 struct ll_statahead_info *sai = lli->lli_sai;
1253 struct ll_dentry_data *ldd = ll_d2d(dentry);
1255 /* not the same process, don't mark */
1256 if (lli->lli_opendir_pid != cfs_curproc_pid())
1259 if (sai != NULL && ldd != NULL)
1260 ldd->lld_sa_generation = sai->sai_generation;
1264 ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
1266 struct ll_inode_info *lli;
1267 struct ll_dentry_data *ldd;
1269 if (ll_i2sbi(dir)->ll_sa_max == 0)
1272 lli = ll_i2info(dir);
1273 /* not the same process, don't statahead */
1274 if (lli->lli_opendir_pid != cfs_curproc_pid())
1277 ldd = ll_d2d(*dentryp);
1279 * When stats a dentry, the system trigger more than once "revalidate"
1280 * or "lookup", for "getattr", for "getxattr", and maybe for others.
1281 * Under patchless client mode, the operation intent is not accurate,
1282 * which maybe misguide the statahead thread. For example:
1283 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1284 * have the same operation intent -- "IT_GETATTR".
1285 * In fact, one dentry should has only one chance to interact with the
1286 * statahead thread, otherwise the statahead windows will be confused.
1287 * The solution is as following:
1288 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1289 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1290 * will bypass interacting with statahead thread for checking:
1291 * "lld_sa_generation == lli_sai->sai_generation"
1293 if (ldd && lli->lli_sai &&
1294 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1297 return do_statahead_enter(dir, dentryp, only_unplug);
1300 /* llite ioctl register support rountine */
1307 #define LLIOC_MAX_CMD 256
1310 * Rules to write a callback function:
1313 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1314 * returned to ll_iocontrol_register. Callback functions should use this
1315 * data to check the potential collasion of ioctl cmd. If collasion is
1316 * found, callback function should return LLIOC_CONT.
1317 * @rcp: The result of ioctl command.
1320 * If @magic matches the pointer returned by ll_iocontrol_data, the
1321 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1323 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1324 struct file *file, unsigned int cmd, unsigned long arg,
1325 void *magic, int *rcp);
1327 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1328 unsigned int cmd, unsigned long arg, int *rcp);
1330 /* export functions */
1331 /* Register ioctl block dynamatically for a regular file.
1333 * @cmd: the array of ioctl command set
1334 * @count: number of commands in the @cmd
1335 * @cb: callback function, it will be called if an ioctl command is found to
1336 * belong to the command list @cmd.
1339 * A magic pointer will be returned if success;
1340 * otherwise, NULL will be returned.
1342 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1343 void ll_iocontrol_unregister(void *magic);
1347 /* lclient compat stuff */
1348 #define cl_inode_info ll_inode_info
1349 #define cl_i2info(info) ll_i2info(info)
1350 #define cl_inode_mode(inode) ((inode)->i_mode)
1351 #define cl_i2sbi ll_i2sbi
1353 static inline struct ll_file_data *cl_iattr2fd(struct inode *inode,
1354 const struct iattr *attr)
1356 LASSERT(attr->ia_valid & ATTR_FILE);
1357 return LUSTRE_FPRIVATE(attr->ia_file);
1360 static inline void cl_isize_lock(struct inode *inode, int lsmlock)
1362 ll_inode_size_lock(inode, lsmlock);
1365 static inline void cl_isize_unlock(struct inode *inode, int lsmlock)
1367 ll_inode_size_unlock(inode, lsmlock);
1370 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1372 LASSERT_SEM_LOCKED(&ll_i2info(inode)->lli_size_sem);
1373 i_size_write(inode, kms);
1376 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1378 ll_inode_size_lock(inode, 0);
1379 i_size_write(inode, kms);
1380 ll_inode_size_unlock(inode, 0);
1383 #define cl_isize_read(inode) i_size_read(inode)
1385 static inline int cl_merge_lvb(struct inode *inode)
1387 return ll_merge_lvb(inode);
1390 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1391 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1392 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1394 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1396 /** direct write pages */
1397 struct ll_dio_pages {
1398 /** page array to be written. we don't support
1399 * partial pages except the last one. */
1400 struct page **ldp_pages;
1401 /* offset of each page */
1402 loff_t *ldp_offsets;
1403 /** if ldp_offsets is NULL, it means a sequential
1404 * pages to be written, then this is the file offset
1405 * of the * first page. */
1406 loff_t ldp_start_offset;
1407 /** how many bytes are to be written. */
1409 /** # of pages in the array. */
1413 static inline void cl_stats_tally(struct cl_device *dev, enum cl_req_type crt,
1416 int opc = (crt == CRT_READ) ? LPROC_LL_OSC_READ :
1419 ll_stats_ops_tally(ll_s2sbi(cl2ccc_dev(dev)->cdv_sb), opc, rc);
1422 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1423 int rw, struct inode *inode,
1424 struct ll_dio_pages *pv);
1426 static inline int ll_file_nolock(const struct file *file)
1428 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1429 struct inode *inode = file->f_dentry->d_inode;
1431 LASSERT(fd != NULL);
1432 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1433 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1436 static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1437 struct lookup_intent *it, __u64 *bits)
1439 if (!it->d.lustre.it_lock_set) {
1440 CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)\n",
1441 inode, inode->i_ino, inode->i_generation);
1442 md_set_lock_data(exp, &it->d.lustre.it_lock_handle,
1443 inode, &it->d.lustre.it_lock_bits);
1444 it->d.lustre.it_lock_set = 1;
1448 *bits = it->d.lustre.it_lock_bits;
1451 static inline void ll_dentry_rehash(struct dentry *dentry, int locked)
1454 cfs_spin_lock(&ll_lookup_lock);
1455 spin_lock(&dcache_lock);
1457 if (d_unhashed(dentry))
1458 d_rehash_cond(dentry, 0);
1460 spin_unlock(&dcache_lock);
1461 cfs_spin_unlock(&ll_lookup_lock);
1465 static inline void ll_dentry_reset_flags(struct dentry *dentry, __u64 bits)
1467 if (bits & MDS_INODELOCK_LOOKUP &&
1468 dentry->d_flags & DCACHE_LUSTRE_INVALID) {
1469 lock_dentry(dentry);
1470 dentry->d_flags &= ~DCACHE_LUSTRE_INVALID;
1471 unlock_dentry(dentry);
1475 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2,7,50,0)
1476 /* Compatibility for old (1.8) compiled userspace quota code */
1477 struct if_quotactl_18 {
1482 struct obd_dqinfo qc_dqinfo;
1483 struct obd_dqblk qc_dqblk;
1485 struct obd_uuid obd_uuid;
1487 #define LL_IOC_QUOTACTL_18 _IOWR('f', 162, struct if_quotactl_18 *)
1488 /* End compatibility for old (1.8) compiled userspace quota code */
1490 #warning "remove old LL_IOC_QUOTACTL_18 compatibility code"
1491 #endif /* LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2,7,50,0) */
1493 #endif /* LLITE_INTERNAL_H */