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 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
37 #ifndef LLITE_INTERNAL_H
38 #define LLITE_INTERNAL_H
40 #include <lustre_acl.h>
42 #include <lustre_debug.h>
43 #include <lustre_ver.h>
44 #include <lustre_disk.h> /* for s2sbi */
45 #include <lustre_eacl.h>
47 /* for struct cl_lock_descr and struct cl_io */
48 #include <cl_object.h>
55 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
56 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
58 #ifdef HAVE_VFS_INTENT_PATCHES
59 static inline struct lookup_intent *ll_nd2it(struct nameidata *nd)
65 struct ll_dentry_data {
68 struct obd_client_handle lld_cwd_och;
69 struct obd_client_handle lld_mnt_och;
70 #ifndef HAVE_VFS_INTENT_PATCHES
71 struct lookup_intent *lld_it;
73 unsigned int lld_sa_generation;
76 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
78 extern struct file_operations ll_pgcache_seq_fops;
80 #define LLI_INODE_MAGIC 0x111d0de5
81 #define LLI_INODE_DEAD 0xdeadd00d
83 /* remote client permission cache */
84 #define REMOTE_PERM_HASHSIZE 16
86 /* llite setxid/access permission for user on remote client */
87 struct ll_remote_perm {
88 struct hlist_node lrp_list;
93 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
94 is access permission with
95 lrp_fsuid/lrp_fsgid. */
99 /* MDS has an authority for the Size-on-MDS attributes. */
100 LLIF_MDS_SIZE_LOCK = (1 << 0),
101 /* Epoch close is postponed. */
102 LLIF_EPOCH_PENDING = (1 << 1),
103 /* DONE WRITING is allowed. */
104 LLIF_DONE_WRITING = (1 << 2),
105 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
107 LLIF_SOM_DIRTY = (1 << 3),
108 /* File is contented */
109 LLIF_CONTENDED = (1 << 4),
110 /* Truncate uses server lock for this file */
111 LLIF_SRVLOCK = (1 << 5)
115 struct ll_inode_info {
117 struct semaphore lli_size_sem; /* protect open and change size */
118 void *lli_size_sem_owner;
119 struct semaphore lli_write_sem;
120 char *lli_symlink_name;
123 unsigned long lli_flags;
124 cfs_time_t lli_contention_time;
126 /* this lock protects posix_acl, pending_write_llaps, mmap_cnt */
128 struct list_head lli_pending_write_llaps;
129 struct list_head lli_close_list;
130 /* handle is to be sent to MDS later on done_writing and setattr.
131 * Open handle data are needed for the recovery to reconstruct
132 * the inode state on the MDS. XXX: recovery is not ready yet. */
133 struct obd_client_handle *lli_pending_och;
135 /* for writepage() only to communicate to fsync */
138 struct posix_acl *lli_posix_acl;
140 /* remote permission hash */
141 struct hlist_head *lli_remote_perms;
142 unsigned long lli_rmtperm_utime;
143 struct semaphore lli_rmtperm_sem;
145 struct list_head lli_dead_list;
147 struct semaphore lli_och_sem; /* Protects access to och pointers
148 and their usage counters */
149 /* We need all three because every inode may be opened in different
151 struct obd_client_handle *lli_mds_read_och;
152 __u64 lli_open_fd_read_count;
153 struct obd_client_handle *lli_mds_write_och;
154 __u64 lli_open_fd_write_count;
155 struct obd_client_handle *lli_mds_exec_och;
156 __u64 lli_open_fd_exec_count;
158 struct inode lli_vfs_inode;
160 /* identifying fields for both metadata and data stacks. */
161 struct lu_fid lli_fid;
162 struct lov_stripe_md *lli_smd;
165 /* open count currently used by capability only, indicate whether
166 * capability needs renewal */
167 atomic_t lli_open_count;
168 struct obd_capa *lli_mds_capa;
169 struct list_head lli_oss_capas;
171 /* metadata stat-ahead */
173 * "opendir_pid" is the token when lookup/revalid -- I am the owner of
176 pid_t lli_opendir_pid;
178 * since parent-child threads can share the same @file struct,
179 * "opendir_key" is the token when dir close for case of parent exit
180 * before child -- it is me should cleanup the dir readahead. */
181 void *lli_opendir_key;
182 struct ll_statahead_info *lli_sai;
183 struct cl_object *lli_clob;
187 * Locking to guarantee consistency of non-atomic updates to long long i_size,
188 * consistency between file size and KMS, and consistency within
189 * ->lli_smd->lsm_oinfo[]'s.
191 * Implemented by ->lli_size_sem and ->lsm_sem, nested in that order.
194 void ll_inode_size_lock(struct inode *inode, int lock_lsm);
195 void ll_inode_size_unlock(struct inode *inode, int unlock_lsm);
197 // FIXME: replace the name of this with LL_I to conform to kernel stuff
198 // static inline struct ll_inode_info *LL_I(struct inode *inode)
199 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
201 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
204 /* default to about 40meg of readahead on a given system. That much tied
205 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
206 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - CFS_PAGE_SHIFT))
208 /* default to read-ahead full files smaller than 2MB on the second read */
209 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - CFS_PAGE_SHIFT))
214 RA_STAT_DISTANT_READPAGE,
215 RA_STAT_MISS_IN_WINDOW,
216 RA_STAT_FAILED_GRAB_PAGE,
217 RA_STAT_FAILED_MATCH,
222 RA_STAT_MAX_IN_FLIGHT,
223 RA_STAT_WRONG_GRAB_PAGE,
228 atomic_t ra_cur_pages;
229 unsigned long ra_max_pages;
230 unsigned long ra_max_pages_per_file;
231 unsigned long ra_max_read_ahead_whole_pages;
234 /* ra_io_arg will be filled in the beginning of ll_readahead with
235 * ras_lock, then the following ll_read_ahead_pages will read RA
236 * pages according to this arg, all the items in this structure are
237 * counted by page index.
240 unsigned long ria_start; /* start offset of read-ahead*/
241 unsigned long ria_end; /* end offset of read-ahead*/
242 /* If stride read pattern is detected, ria_stoff means where
243 * stride read is started. Note: for normal read-ahead, the
244 * value here is meaningless, and also it will not be accessed*/
246 /* ria_length and ria_pages are the length and pages length in the
247 * stride I/O mode. And they will also be used to check whether
248 * it is stride I/O read-ahead in the read-ahead pages*/
249 unsigned long ria_length;
250 unsigned long ria_pages;
253 /* LL_HIST_MAX=32 causes an overflow */
254 #define LL_HIST_MAX 28
255 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
256 #define LL_PROCESS_HIST_MAX 10
257 struct per_process_info {
259 struct obd_histogram pp_r_hist;
260 struct obd_histogram pp_w_hist;
263 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
264 struct ll_rw_extents_info {
265 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
268 #define LL_OFFSET_HIST_MAX 100
269 struct ll_rw_process_info {
272 loff_t rw_range_start;
274 loff_t rw_last_file_pos;
276 size_t rw_smallest_extent;
277 size_t rw_largest_extent;
278 struct ll_file_data *rw_last_file;
281 enum stats_track_type {
282 STATS_TRACK_ALL = 0, /* track all processes */
283 STATS_TRACK_PID, /* track process with this pid */
284 STATS_TRACK_PPID, /* track processes with this ppid */
285 STATS_TRACK_GID, /* track processes with this gid */
289 /* flags for sbi->ll_flags */
290 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
291 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
292 #define LL_SBI_FLOCK 0x04
293 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
294 #define LL_SBI_ACL 0x10 /* support ACL */
295 #define LL_SBI_JOIN 0x20 /* support JOIN */
296 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
297 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
298 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
299 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
300 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
301 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
303 /* default value for ll_sb_info->contention_time */
304 #define SBI_DEFAULT_CONTENTION_SECONDS 60
305 /* default value for lockless_truncate_enable */
306 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
307 #define RCE_HASHES 32
309 struct rmtacl_ctl_entry {
310 struct list_head rce_list;
311 pid_t rce_key; /* hash key */
312 int rce_ops; /* acl operation type */
315 struct rmtacl_ctl_table {
317 struct list_head rct_entries[RCE_HASHES];
323 struct list_head ee_list;
324 pid_t ee_key; /* hash key */
325 struct lu_fid ee_fid;
326 int ee_type; /* ACL type for ACCESS or DEFAULT */
327 ext_acl_xattr_header *ee_acl;
332 struct list_head et_entries[EE_HASHES];
336 struct list_head ll_list;
337 /* this protects pglist and ra_info. It isn't safe to
338 * grab from interrupt contexts */
340 spinlock_t ll_pp_extent_lock; /* Lock for pp_extent entries */
341 spinlock_t ll_process_lock; /* Lock for ll_rw_process_info */
342 struct obd_uuid ll_sb_uuid;
343 struct obd_export *ll_md_exp;
344 struct obd_export *ll_dt_exp;
345 struct proc_dir_entry* ll_proc_root;
346 struct lu_fid ll_root_fid; /* root object fid */
349 struct list_head ll_conn_chain; /* per-conn chain of SBs */
350 struct lustre_client_ocd ll_lco;
352 struct list_head ll_orphan_dentry_list; /*please don't ask -p*/
353 struct ll_close_queue *ll_lcq;
355 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
357 unsigned long ll_async_page_max;
358 unsigned long ll_async_page_count;
360 struct lprocfs_stats *ll_ra_stats;
362 struct ll_ra_info ll_ra_info;
363 unsigned int ll_namelen;
364 struct file_operations *ll_fop;
366 #ifdef HAVE_EXPORT___IGET
367 struct list_head ll_deathrow; /* inodes to be destroyed (b1443) */
368 spinlock_t ll_deathrow_lock;
370 /* =0 - hold lock over whole read/write
371 * >0 - max. chunk to be read/written w/o lock re-acquiring */
372 unsigned long ll_max_rw_chunk;
374 struct lu_site *ll_site;
375 struct cl_device *ll_cl;
377 struct ll_rw_extents_info ll_rw_extents_info;
378 int ll_extent_process_count;
379 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
380 unsigned int ll_offset_process_count;
381 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
382 unsigned int ll_rw_offset_entry_count;
383 enum stats_track_type ll_stats_track_type;
384 int ll_stats_track_id;
387 /* metadata stat-ahead */
388 unsigned int ll_sa_max; /* max statahead RPCs */
389 unsigned int ll_sa_wrong; /* statahead thread stopped for
391 unsigned int ll_sa_total; /* statahead thread started
393 unsigned long long ll_sa_blocked; /* ls count waiting for
395 unsigned long long ll_sa_cached; /* ls count got in cache */
396 unsigned long long ll_sa_hit; /* hit count */
397 unsigned long long ll_sa_miss; /* miss count */
399 dev_t ll_sdev_orig; /* save s_dev before assign for
401 struct rmtacl_ctl_table ll_rct;
402 struct eacl_table ll_et;
405 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
410 struct task_struct *lrr_reader;
411 struct list_head lrr_linkage;
415 * per file-descriptor read-ahead data.
417 struct ll_readahead_state {
420 * index of the last page that read(2) needed and that wasn't in the
421 * cache. Used by ras_update() to detect seeks.
423 * XXX nikita: if access seeks into cached region, Lustre doesn't see
426 unsigned long ras_last_readpage;
428 * number of pages read after last read-ahead window reset. As window
429 * is reset on each seek, this is effectively a number of consecutive
430 * accesses. Maybe ->ras_accessed_in_window is better name.
432 * XXX nikita: window is also reset (by ras_update()) when Lustre
433 * believes that memory pressure evicts read-ahead pages. In that
434 * case, it probably doesn't make sense to expand window to
435 * PTLRPC_MAX_BRW_PAGES on the third access.
437 unsigned long ras_consecutive_pages;
439 * number of read requests after the last read-ahead window reset
440 * As window is reset on each seek, this is effectively the number
441 * on consecutive read request and is used to trigger read-ahead.
443 unsigned long ras_consecutive_requests;
445 * Parameters of current read-ahead window. Handled by
446 * ras_update(). On the initial access to the file or after a seek,
447 * window is reset to 0. After 3 consecutive accesses, window is
448 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
449 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
451 unsigned long ras_window_start, ras_window_len;
453 * Where next read-ahead should start at. This lies within read-ahead
454 * window. Read-ahead window is read in pieces rather than at once
455 * because: 1. lustre limits total number of pages under read-ahead by
456 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
457 * not covered by DLM lock.
459 unsigned long ras_next_readahead;
461 * Total number of ll_file_read requests issued, reads originating
462 * due to mmap are not counted in this total. This value is used to
463 * trigger full file read-ahead after multiple reads to a small file.
465 unsigned long ras_requests;
467 * Page index with respect to the current request, these value
468 * will not be accurate when dealing with reads issued via mmap.
470 unsigned long ras_request_index;
472 * list of struct ll_ra_read's one per read(2) call current in
473 * progress against this file descriptor. Used by read-ahead code,
474 * protected by ->ras_lock.
476 struct list_head ras_read_beads;
478 * The following 3 items are used for detecting the stride I/O
480 * In stride I/O mode,
481 * ...............|-----data-----|****gap*****|--------|******|....
482 * offset |-stride_pages-|-stride_gap-|
483 * ras_stride_offset = offset;
484 * ras_stride_length = stride_pages + stride_gap;
485 * ras_stride_pages = stride_pages;
486 * Note: all these three items are counted by pages.
488 unsigned long ras_stride_length;
489 unsigned long ras_stride_pages;
490 pgoff_t ras_stride_offset;
492 * number of consecutive stride request count, and it is similar as
493 * ras_consecutive_requests, but used for stride I/O mode.
494 * Note: only more than 2 consecutive stride request are detected,
495 * stride read-ahead will be enable
497 unsigned long ras_consecutive_stride_requests;
503 extern cfs_mem_cache_t *ll_file_data_slab;
504 struct lustre_handle;
505 struct ll_file_data {
506 struct ll_readahead_state fd_ras;
508 struct ccc_grouplock fd_grouplock;
509 struct ll_file_dir fd_dir;
511 struct file *fd_file;
514 struct lov_stripe_md;
516 extern spinlock_t inode_lock;
518 extern struct proc_dir_entry *proc_lustre_fs_root;
520 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
522 return &lli->lli_vfs_inode;
526 struct inode *icbd_parent;
527 struct dentry **icbd_childp;
531 __u32 ll_i2suppgid(struct inode *i);
532 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
534 #define LLAP_MAGIC 98764321
536 extern cfs_mem_cache_t *ll_async_page_slab;
537 extern size_t ll_async_page_slab_size;
539 #ifdef HAVE_REGISTER_CACHE
540 #include <linux/cache_def.h>
541 #define ll_register_cache(cache) register_cache(cache)
542 #define ll_unregister_cache(cache) unregister_cache(cache)
543 #elif defined(HAVE_SHRINKER_CACHE)
544 struct cache_definition {
547 struct shrinker *shrinker;
550 #define ll_register_cache(cache) do { \
551 struct cache_definition *c = (cache); \
552 c->shrinker = set_shrinker(DEFAULT_SEEKS, c->shrink); \
555 #define ll_unregister_cache(cache) do { \
556 remove_shrinker((cache)->shrinker); \
557 (cache)->shrinker = NULL; \
560 #define ll_register_cache(cache) do {} while (0)
561 #define ll_unregister_cache(cache) do {} while (0)
564 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
565 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
566 struct ll_ra_read *ll_ra_read_get(struct file *f);
568 /* llite/lproc_llite.c */
570 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
571 struct super_block *sb, char *osc, char *mdc);
572 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
573 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
574 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
576 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
577 struct super_block *sb, char *osc, char *mdc){return 0;}
578 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
579 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
580 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
582 memset(lvars, 0, sizeof(*lvars));
588 static inline void ll_put_page(struct page *page)
591 page_cache_release(page);
594 extern struct file_operations ll_dir_operations;
595 extern struct inode_operations ll_dir_inode_operations;
596 struct page *ll_get_dir_page(struct inode *dir, __u64 hash, int exact,
597 struct ll_dir_chain *chain);
599 int ll_objects_destroy(struct ptlrpc_request *request,
601 struct inode *ll_iget(struct super_block *sb, ino_t hash,
602 struct lustre_md *lic);
603 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
604 void *data, int flag);
605 #ifndef HAVE_VFS_INTENT_PATCHES
606 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
609 int ll_lookup_it_finish(struct ptlrpc_request *request,
610 struct lookup_intent *it, void *data);
611 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
614 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
615 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
616 int ll_writepage(struct page *page, struct writeback_control *wbc);
617 void ll_removepage(struct page *page);
618 int ll_readpage(struct file *file, struct page *page);
619 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
620 void ll_truncate(struct inode *inode);
621 int ll_file_punch(struct inode *, loff_t, int);
622 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
623 void ll_clear_file_contended(struct inode*);
624 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
625 int ll_readahead(const struct lu_env *env, struct cl_io *io, struct ll_readahead_state *ras,
626 struct address_space *mapping, struct cl_page_list *queue, int flags);
629 extern struct file_operations ll_file_operations;
630 extern struct file_operations ll_file_operations_flock;
631 extern struct file_operations ll_file_operations_noflock;
632 extern struct inode_operations ll_file_inode_operations;
633 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *);
634 extern int ll_have_md_lock(struct inode *inode, __u64 bits);
635 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
636 struct lustre_handle *lockh);
637 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *, __u64 bits);
638 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
639 int ll_file_open(struct inode *inode, struct file *file);
640 int ll_file_release(struct inode *inode, struct file *file);
641 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
642 struct lov_stripe_md *lsm, lstat_t *st);
643 int ll_local_open(struct file *file,
644 struct lookup_intent *it, struct ll_file_data *fd,
645 struct obd_client_handle *och);
646 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
647 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
649 int ll_md_real_close(struct inode *inode, int flags);
650 void ll_epoch_close(struct inode *inode, struct md_op_data *op_data,
651 struct obd_client_handle **och, unsigned long flags);
652 int ll_sizeonmds_update(struct inode *inode, struct lustre_handle *fh,
654 int ll_inode_getattr(struct inode *inode, struct obdo *obdo);
655 int ll_md_setattr(struct inode *inode, struct md_op_data *op_data,
656 struct md_open_data **mod);
657 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
658 struct lustre_handle *fh);
659 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
660 struct ll_file_data *file, loff_t pos,
661 size_t count, int rw);
662 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
663 struct lookup_intent *it, struct kstat *stat);
664 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
665 struct ll_file_data *ll_file_data_get(void);
666 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
667 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
668 int flags, struct lov_user_md *lum,
670 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
671 struct lov_mds_md **lmm, int *lmm_size,
672 struct ptlrpc_request **request);
673 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
675 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmm,
676 int *lmm_size, struct ptlrpc_request **request);
677 int ll_fsync(struct file *file, struct dentry *dentry, int data);
678 int ll_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
680 int ll_merge_lvb(struct inode *inode);
681 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
682 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
687 * protect race ll_find_aliases vs ll_revalidate_it vs ll_unhash_aliases
689 extern spinlock_t ll_lookup_lock;
690 extern struct dentry_operations ll_d_ops;
691 void ll_intent_drop_lock(struct lookup_intent *);
692 void ll_intent_release(struct lookup_intent *);
693 int ll_drop_dentry(struct dentry *dentry);
694 extern void ll_set_dd(struct dentry *de);
695 int ll_drop_dentry(struct dentry *dentry);
696 void ll_unhash_aliases(struct inode *);
697 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
698 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
699 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
700 int ll_revalidate_it_finish(struct ptlrpc_request *request,
701 struct lookup_intent *it, struct dentry *de);
703 /* llite/llite_lib.c */
704 extern struct super_operations lustre_super_operations;
706 char *ll_read_opt(const char *opt, char *data);
707 void ll_lli_init(struct ll_inode_info *lli);
708 int ll_fill_super(struct super_block *sb);
709 void ll_put_super(struct super_block *sb);
710 void ll_kill_super(struct super_block *sb);
711 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
712 void ll_clear_inode(struct inode *inode);
713 int ll_setattr_raw(struct inode *inode, struct iattr *attr);
714 int ll_setattr(struct dentry *de, struct iattr *attr);
715 #ifndef HAVE_STATFS_DENTRY_PARAM
716 int ll_statfs(struct super_block *sb, struct kstatfs *sfs);
718 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
720 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
721 __u64 max_age, __u32 flags);
722 void ll_update_inode(struct inode *inode, struct lustre_md *md);
723 void ll_read_inode2(struct inode *inode, void *opaque);
724 void ll_delete_inode(struct inode *inode);
725 int ll_iocontrol(struct inode *inode, struct file *file,
726 unsigned int cmd, unsigned long arg);
727 int ll_flush_ctx(struct inode *inode);
728 #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT
729 void ll_umount_begin(struct vfsmount *vfsmnt, int flags);
731 void ll_umount_begin(struct super_block *sb);
733 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
734 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
735 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
736 struct super_block *);
737 void lustre_dump_dentry(struct dentry *, int recur);
738 void lustre_dump_inode(struct inode *);
739 int ll_obd_statfs(struct inode *inode, void *arg);
740 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
741 int ll_process_config(struct lustre_cfg *lcfg);
742 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
743 struct inode *i1, struct inode *i2,
744 const char *name, int namelen,
745 int mode, __u32 opc, void *data);
746 void ll_finish_md_op_data(struct md_op_data *op_data);
748 /* llite/llite_nfs.c */
749 extern struct export_operations lustre_export_operations;
750 __u32 get_uuid2int(const char *name, int len);
751 struct dentry *ll_fh_to_dentry(struct super_block *sb, __u32 *data, int len,
752 int fhtype, int parent);
753 int ll_dentry_to_fh(struct dentry *, __u32 *datap, int *lenp, int need_parent);
755 /* llite/special.c */
756 extern struct inode_operations ll_special_inode_operations;
757 extern struct file_operations ll_special_chr_inode_fops;
758 extern struct file_operations ll_special_chr_file_fops;
759 extern struct file_operations ll_special_blk_inode_fops;
760 extern struct file_operations ll_special_fifo_inode_fops;
761 extern struct file_operations ll_special_fifo_file_fops;
762 extern struct file_operations ll_special_sock_inode_fops;
764 /* llite/symlink.c */
765 extern struct inode_operations ll_fast_symlink_inode_operations;
767 /* llite/llite_close.c */
768 struct ll_close_queue {
770 struct list_head lcq_head;
771 wait_queue_head_t lcq_waitq;
772 struct completion lcq_comp;
776 struct vvp_thread_info {
777 struct ost_lvb vti_lvb;
778 struct cl_2queue vti_queue;
779 struct iovec vti_local_iov;
780 struct ccc_io_args vti_args;
781 struct ra_io_arg vti_ria;
782 struct kiocb vti_kiocb;
785 struct ccc_object *cl_inode2ccc(struct inode *inode);
787 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
789 extern struct lu_context_key vvp_key;
790 struct vvp_thread_info *info;
792 info = lu_context_key_get(&env->le_ctx, &vvp_key);
793 LASSERT(info != NULL);
797 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
798 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
803 read_actor_t cui_actor;
806 struct vvp_fault_io {
808 * Inode modification time that is checked across DLM
812 struct vm_area_struct *ft_vma;
814 * Virtual address at which fault occurred.
816 unsigned long ft_address;
818 * Fault type, as to be supplied to filemap_nopage().
824 * Read-ahead state used by read and page-fault IO contexts.
826 struct ll_ra_read cui_bead;
828 * Set when cui_bead has been initialized.
830 int cui_ra_window_set;
832 * If IO was created directly in low level method like
833 * ->prepare_write(), this field stores the number of method calls
834 * that constitute this IO. This field is decremented by ll_cl_fini(),
835 * and cl_io is destroyed, when it reaches 0. When oneshot IO
836 * completes, this fields is set to -1.
841 * Partially truncated page, that vvp_io_trunc_start() keeps locked
844 struct cl_page *cui_partpage;
848 struct vvp_io vs_ios;
851 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
853 extern struct lu_context_key vvp_session_key;
854 struct vvp_session *ses;
856 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
857 LASSERT(ses != NULL);
861 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
863 return &vvp_env_session(env)->vs_ios;
866 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
867 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
868 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
870 /* llite/llite_mmap.c */
871 typedef struct rb_root rb_root_t;
872 typedef struct rb_node rb_node_t;
874 struct ll_lock_tree_node;
875 struct ll_lock_tree {
877 struct list_head lt_locked_list;
878 struct ll_file_data *lt_fd;
881 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
882 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
883 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
884 __u64 end, ldlm_mode_t mode);
885 void policy_from_vma(ldlm_policy_data_t *policy,
886 struct vm_area_struct *vma, unsigned long addr, size_t count);
887 struct vm_area_struct *our_vma(unsigned long addr, size_t count);
889 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
891 static inline __u64 ll_ts2u64(struct timespec *time)
893 __u64 t = time->tv_sec;
897 /* don't need an addref as the sb_info should be holding one */
898 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
900 return ll_s2sbi(sb)->ll_dt_exp;
903 /* don't need an addref as the sb_info should be holding one */
904 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
906 return ll_s2sbi(sb)->ll_md_exp;
909 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
911 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
917 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
918 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
920 return ll_s2sbi(inode->i_sb);
923 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
925 return ll_s2dtexp(inode->i_sb);
928 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
930 return ll_s2mdexp(inode->i_sb);
933 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
936 LASSERT(inode != NULL);
937 fid = &ll_i2info(inode)->lli_fid;
938 LASSERT(fid_is_igif(fid) || fid_ver(fid) == 0);
942 static inline int ll_mds_max_easize(struct super_block *sb)
944 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
947 static inline __u64 ll_file_maxbytes(struct inode *inode)
949 return ll_i2info(inode)->lli_maxbytes;
953 int ll_setxattr(struct dentry *dentry, const char *name,
954 const void *value, size_t size, int flags);
955 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
956 void *buffer, size_t size);
957 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
958 int ll_removexattr(struct dentry *dentry, const char *name);
960 /* llite/remote_perm.c */
961 extern cfs_mem_cache_t *ll_remote_perm_cachep;
962 extern cfs_mem_cache_t *ll_rmtperm_hash_cachep;
964 struct hlist_head *alloc_rmtperm_hash(void);
965 void free_rmtperm_hash(struct hlist_head *hash);
966 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
967 int lustre_check_remote_perm(struct inode *inode, int mask);
969 /* llite/llite_fid.c */
970 ino_t ll_fid_build_ino(struct ll_sb_info *sbi, struct lu_fid *fid);
971 __u32 ll_fid_build_gen(struct ll_sb_info *sbi, struct lu_fid *fid);
973 /* llite/llite_capa.c */
974 extern cfs_timer_t ll_capa_timer;
976 int ll_capa_thread_start(void);
977 void ll_capa_thread_stop(void);
978 void ll_capa_timer_callback(unsigned long unused);
980 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
981 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
983 void ll_capa_open(struct inode *inode);
984 void ll_capa_close(struct inode *inode);
986 struct obd_capa *ll_mdscapa_get(struct inode *inode);
987 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
989 void ll_truncate_free_capa(struct obd_capa *ocapa);
990 void ll_clear_inode_capas(struct inode *inode);
991 void ll_print_capa_stat(struct ll_sb_info *sbi);
993 /* llite/llite_cl.c */
994 extern struct lu_device_type vvp_device_type;
997 * Common IO arguments for various VFS I/O interfaces.
1000 int cl_sb_init(struct super_block *sb);
1001 int cl_sb_fini(struct super_block *sb);
1002 int cl_inode_init(struct inode *inode, struct lustre_md *md);
1003 void cl_inode_fini(struct inode *inode);
1005 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1006 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1008 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1009 struct ll_readahead_state *ras, unsigned long index,
1011 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1012 int ll_is_file_contended(struct file *file);
1013 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1015 /* llite/llite_rmtacl.c */
1016 #ifdef CONFIG_FS_POSIX_ACL
1017 obd_valid rce_ops2valid(int ops);
1018 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1019 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1020 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1021 void rct_init(struct rmtacl_ctl_table *rct);
1022 void rct_fini(struct rmtacl_ctl_table *rct);
1024 void ee_free(struct eacl_entry *ee);
1025 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1026 ext_acl_xattr_header *header);
1027 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1028 struct lu_fid *fid, int type);
1029 void et_search_free(struct eacl_table *et, pid_t key);
1030 void et_init(struct eacl_table *et);
1031 void et_fini(struct eacl_table *et);
1036 #define LL_SA_RPC_MIN 2
1037 #define LL_SA_RPC_DEF 32
1038 #define LL_SA_RPC_MAX 8192
1040 /* per inode struct, for dir only */
1041 struct ll_statahead_info {
1042 struct inode *sai_inode;
1043 unsigned int sai_generation; /* generation for statahead */
1044 atomic_t sai_refcount; /* when access this struct, hold
1046 unsigned int sai_sent; /* stat requests sent count */
1047 unsigned int sai_replied; /* stat requests which received
1049 unsigned int sai_max; /* max ahead of lookup */
1050 unsigned int sai_index; /* index of statahead entry */
1051 unsigned int sai_index_next; /* index for the next statahead
1052 * entry to be stated */
1053 unsigned int sai_hit; /* hit count */
1054 unsigned int sai_miss; /* miss count:
1055 * for "ls -al" case, it includes
1056 * hidden dentry miss;
1057 * for "ls -l" case, it does not
1058 * include hidden dentry miss.
1059 * "sai_miss_hidden" is used for
1062 unsigned int sai_consecutive_miss; /* consecutive miss */
1063 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1064 * is not a hidden one */
1065 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1066 unsigned int sai_ls_all:1; /* "ls -al", do stat-ahead for
1068 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
1069 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1070 struct list_head sai_entries_sent; /* entries sent out */
1071 struct list_head sai_entries_received; /* entries returned */
1072 struct list_head sai_entries_stated; /* entries stated */
1075 int do_statahead_enter(struct inode *dir, struct dentry **dentry, int lookup);
1076 void ll_statahead_exit(struct dentry *dentry, int result);
1077 void ll_stop_statahead(struct inode *inode, void *key);
1080 void ll_statahead_mark(struct dentry *dentry)
1082 struct ll_inode_info *lli = ll_i2info(dentry->d_parent->d_inode);
1083 struct ll_dentry_data *ldd = ll_d2d(dentry);
1085 /* not the same process, don't mark */
1086 if (lli->lli_opendir_pid != cfs_curproc_pid())
1089 spin_lock(&lli->lli_lock);
1090 if (likely(lli->lli_sai != NULL && ldd != NULL))
1091 ldd->lld_sa_generation = lli->lli_sai->sai_generation;
1092 spin_unlock(&lli->lli_lock);
1096 int ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int lookup)
1098 struct ll_sb_info *sbi = ll_i2sbi(dir);
1099 struct ll_inode_info *lli = ll_i2info(dir);
1100 struct ll_dentry_data *ldd = ll_d2d(*dentryp);
1102 if (sbi->ll_sa_max == 0)
1105 /* not the same process, don't statahead */
1106 if (lli->lli_opendir_pid != cfs_curproc_pid())
1110 * When "ls" a dentry, the system trigger more than once "revalidate" or
1111 * "lookup", for "getattr", for "getxattr", and maybe for others.
1112 * Under patchless client mode, the operation intent is not accurate,
1113 * it maybe misguide the statahead thread. For example:
1114 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1115 * have the same operation intent -- "IT_GETATTR".
1116 * In fact, one dentry should has only one chance to interact with the
1117 * statahead thread, otherwise the statahead windows will be confused.
1118 * The solution is as following:
1119 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1120 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1121 * will bypass interacting with statahead thread for checking:
1122 * "lld_sa_generation == lli_sai->sai_generation"
1124 if (ldd && lli->lli_sai &&
1125 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1128 return do_statahead_enter(dir, dentryp, lookup);
1131 static void inline ll_dops_init(struct dentry *de, int block)
1133 struct ll_dentry_data *lld = ll_d2d(de);
1135 if (lld == NULL && block != 0) {
1141 lld->lld_sa_generation = 0;
1143 de->d_op = &ll_d_ops;
1146 /* llite ioctl register support rountine */
1153 #define LLIOC_MAX_CMD 256
1156 * Rules to write a callback function:
1159 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1160 * returned to ll_iocontrol_register. Callback functions should use this
1161 * data to check the potential collasion of ioctl cmd. If collasion is
1162 * found, callback function should return LLIOC_CONT.
1163 * @rcp: The result of ioctl command.
1166 * If @magic matches the pointer returned by ll_iocontrol_data, the
1167 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1169 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1170 struct file *file, unsigned int cmd, unsigned long arg,
1171 void *magic, int *rcp);
1173 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1174 unsigned int cmd, unsigned long arg, int *rcp);
1176 /* export functions */
1177 /* Register ioctl block dynamatically for a regular file.
1179 * @cmd: the array of ioctl command set
1180 * @count: number of commands in the @cmd
1181 * @cb: callback function, it will be called if an ioctl command is found to
1182 * belong to the command list @cmd.
1185 * A magic pointer will be returned if success;
1186 * otherwise, NULL will be returned.
1188 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1189 void ll_iocontrol_unregister(void *magic);
1193 /* lclient compat stuff */
1194 #define cl_inode_info ll_inode_info
1195 #define cl_i2info(info) ll_i2info(info)
1196 #define cl_inode_mode(inode) ((inode)->i_mode)
1197 #define cl_i2sbi ll_i2sbi
1198 #define cl_isize_read(inode) i_size_read(inode)
1199 #define cl_isize_write(inode,kms) i_size_write(inode, kms)
1200 #define cl_isize_write_nolock(inode,kms) do {(inode)->i_size=(kms);}while(0)
1202 static inline void cl_isize_lock(struct inode *inode, int lsmlock)
1204 ll_inode_size_lock(inode, lsmlock);
1207 static inline void cl_isize_unlock(struct inode *inode, int lsmlock)
1209 ll_inode_size_unlock(inode, lsmlock);
1212 static inline int cl_merge_lvb(struct inode *inode)
1214 return ll_merge_lvb(inode);
1217 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1218 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1219 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1221 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1223 /** direct write pages */
1224 struct ll_dio_pages {
1225 /** page array to be written. we don't support
1226 * partial pages except the last one. */
1227 struct page **ldp_pages;
1228 /* offset of each page */
1229 loff_t *ldp_offsets;
1230 /** if ldp_offsets is NULL, it means a sequential
1231 * pages to be written, then this is the file offset
1232 * of the * first page. */
1233 loff_t ldp_start_offset;
1234 /** how many bytes are to be written. */
1236 /** # of pages in the array. */
1240 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1241 int rw, struct inode *inode,
1242 struct ll_dio_pages *pv);
1244 #endif /* LLITE_INTERNAL_H */