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 #ifndef DCACHE_LUSTRE_INVALID
56 #define DCACHE_LUSTRE_INVALID 0x100
59 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
60 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
62 #ifdef HAVE_VFS_INTENT_PATCHES
63 static inline struct lookup_intent *ll_nd2it(struct nameidata *nd)
69 struct ll_dentry_data {
72 struct obd_client_handle lld_cwd_och;
73 struct obd_client_handle lld_mnt_och;
74 #ifndef HAVE_VFS_INTENT_PATCHES
75 struct lookup_intent *lld_it;
77 unsigned int lld_sa_generation;
80 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
82 extern struct file_operations ll_pgcache_seq_fops;
84 #define LLI_INODE_MAGIC 0x111d0de5
85 #define LLI_INODE_DEAD 0xdeadd00d
87 /* remote client permission cache */
88 #define REMOTE_PERM_HASHSIZE 16
90 /* llite setxid/access permission for user on remote client */
91 struct ll_remote_perm {
92 cfs_hlist_node_t lrp_list;
97 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
98 is access permission with
99 lrp_fsuid/lrp_fsgid. */
103 /* MDS has an authority for the Size-on-MDS attributes. */
104 LLIF_MDS_SIZE_LOCK = (1 << 0),
105 /* Epoch close is postponed. */
106 LLIF_EPOCH_PENDING = (1 << 1),
107 /* DONE WRITING is allowed. */
108 LLIF_DONE_WRITING = (1 << 2),
109 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
111 LLIF_SOM_DIRTY = (1 << 3),
112 /* File is contented */
113 LLIF_CONTENDED = (1 << 4),
114 /* Truncate uses server lock for this file */
115 LLIF_SRVLOCK = (1 << 5)
119 struct ll_inode_info {
121 cfs_semaphore_t lli_size_sem; /* protect open and change size */
122 void *lli_size_sem_owner;
123 cfs_semaphore_t lli_write_sem;
124 cfs_semaphore_t lli_trunc_sem;
125 char *lli_symlink_name;
128 unsigned long lli_flags;
129 cfs_time_t lli_contention_time;
131 /* this lock protects posix_acl, pending_write_llaps, mmap_cnt */
132 cfs_spinlock_t lli_lock;
133 cfs_list_t lli_close_list;
134 /* handle is to be sent to MDS later on done_writing and setattr.
135 * Open handle data are needed for the recovery to reconstruct
136 * the inode state on the MDS. XXX: recovery is not ready yet. */
137 struct obd_client_handle *lli_pending_och;
139 /* for writepage() only to communicate to fsync */
142 struct posix_acl *lli_posix_acl;
144 /* remote permission hash */
145 cfs_hlist_head_t *lli_remote_perms;
146 unsigned long lli_rmtperm_utime;
147 cfs_semaphore_t lli_rmtperm_sem;
149 cfs_list_t lli_dead_list;
151 cfs_semaphore_t lli_och_sem; /* Protects access to och pointers
152 and their usage counters, also
153 atomicity of check-update of
155 /* We need all three because every inode may be opened in different
157 struct obd_client_handle *lli_mds_read_och;
158 __u64 lli_open_fd_read_count;
159 struct obd_client_handle *lli_mds_write_och;
160 __u64 lli_open_fd_write_count;
161 struct obd_client_handle *lli_mds_exec_och;
162 __u64 lli_open_fd_exec_count;
164 struct inode lli_vfs_inode;
166 /* identifying fields for both metadata and data stacks. */
167 struct lu_fid lli_fid;
168 struct lov_stripe_md *lli_smd;
171 /* open count currently used by capability only, indicate whether
172 * capability needs renewal */
173 cfs_atomic_t lli_open_count;
174 struct obd_capa *lli_mds_capa;
175 cfs_list_t lli_oss_capas;
177 /* metadata stat-ahead */
179 * "opendir_pid" is the token when lookup/revalid -- I am the owner of
182 pid_t lli_opendir_pid;
184 * since parent-child threads can share the same @file struct,
185 * "opendir_key" is the token when dir close for case of parent exit
186 * before child -- it is me should cleanup the dir readahead. */
187 void *lli_opendir_key;
188 struct ll_statahead_info *lli_sai;
189 struct cl_object *lli_clob;
193 * Locking to guarantee consistency of non-atomic updates to long long i_size,
194 * consistency between file size and KMS, and consistency within
195 * ->lli_smd->lsm_oinfo[]'s.
197 * Implemented by ->lli_size_sem and ->lsm_sem, nested in that order.
200 void ll_inode_size_lock(struct inode *inode, int lock_lsm);
201 void ll_inode_size_unlock(struct inode *inode, int unlock_lsm);
203 // FIXME: replace the name of this with LL_I to conform to kernel stuff
204 // static inline struct ll_inode_info *LL_I(struct inode *inode)
205 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
207 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
210 /* default to about 40meg of readahead on a given system. That much tied
211 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
212 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - CFS_PAGE_SHIFT))
214 /* default to read-ahead full files smaller than 2MB on the second read */
215 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - CFS_PAGE_SHIFT))
220 RA_STAT_DISTANT_READPAGE,
221 RA_STAT_MISS_IN_WINDOW,
222 RA_STAT_FAILED_GRAB_PAGE,
223 RA_STAT_FAILED_MATCH,
228 RA_STAT_MAX_IN_FLIGHT,
229 RA_STAT_WRONG_GRAB_PAGE,
234 cfs_atomic_t ra_cur_pages;
235 unsigned long ra_max_pages;
236 unsigned long ra_max_pages_per_file;
237 unsigned long ra_max_read_ahead_whole_pages;
240 /* ra_io_arg will be filled in the beginning of ll_readahead with
241 * ras_lock, then the following ll_read_ahead_pages will read RA
242 * pages according to this arg, all the items in this structure are
243 * counted by page index.
246 unsigned long ria_start; /* start offset of read-ahead*/
247 unsigned long ria_end; /* end offset of read-ahead*/
248 /* If stride read pattern is detected, ria_stoff means where
249 * stride read is started. Note: for normal read-ahead, the
250 * value here is meaningless, and also it will not be accessed*/
252 /* ria_length and ria_pages are the length and pages length in the
253 * stride I/O mode. And they will also be used to check whether
254 * it is stride I/O read-ahead in the read-ahead pages*/
255 unsigned long ria_length;
256 unsigned long ria_pages;
259 /* LL_HIST_MAX=32 causes an overflow */
260 #define LL_HIST_MAX 28
261 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
262 #define LL_PROCESS_HIST_MAX 10
263 struct per_process_info {
265 struct obd_histogram pp_r_hist;
266 struct obd_histogram pp_w_hist;
269 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
270 struct ll_rw_extents_info {
271 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
274 #define LL_OFFSET_HIST_MAX 100
275 struct ll_rw_process_info {
278 loff_t rw_range_start;
280 loff_t rw_last_file_pos;
282 size_t rw_smallest_extent;
283 size_t rw_largest_extent;
284 struct ll_file_data *rw_last_file;
287 enum stats_track_type {
288 STATS_TRACK_ALL = 0, /* track all processes */
289 STATS_TRACK_PID, /* track process with this pid */
290 STATS_TRACK_PPID, /* track processes with this ppid */
291 STATS_TRACK_GID, /* track processes with this gid */
295 /* flags for sbi->ll_flags */
296 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
297 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
298 #define LL_SBI_FLOCK 0x04
299 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
300 #define LL_SBI_ACL 0x10 /* support ACL */
301 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
302 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
303 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
304 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
305 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
306 #define LL_SBI_LAZYSTATFS 0x800 /* lazystatfs mount option */
308 /* default value for ll_sb_info->contention_time */
309 #define SBI_DEFAULT_CONTENTION_SECONDS 60
310 /* default value for lockless_truncate_enable */
311 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
312 #define RCE_HASHES 32
314 struct rmtacl_ctl_entry {
316 pid_t rce_key; /* hash key */
317 int rce_ops; /* acl operation type */
320 struct rmtacl_ctl_table {
321 cfs_spinlock_t rct_lock;
322 cfs_list_t rct_entries[RCE_HASHES];
329 pid_t ee_key; /* hash key */
330 struct lu_fid ee_fid;
331 int ee_type; /* ACL type for ACCESS or DEFAULT */
332 ext_acl_xattr_header *ee_acl;
336 cfs_spinlock_t et_lock;
337 cfs_list_t et_entries[EE_HASHES];
342 /* this protects pglist and ra_info. It isn't safe to
343 * grab from interrupt contexts */
344 cfs_spinlock_t ll_lock;
345 cfs_spinlock_t ll_pp_extent_lock; /* Lock for pp_extent entries */
346 cfs_spinlock_t ll_process_lock; /* Lock for ll_rw_process_info */
347 struct obd_uuid ll_sb_uuid;
348 struct obd_export *ll_md_exp;
349 struct obd_export *ll_dt_exp;
350 struct proc_dir_entry* ll_proc_root;
351 struct lu_fid ll_root_fid; /* root object fid */
354 cfs_list_t ll_conn_chain; /* per-conn chain of SBs */
355 struct lustre_client_ocd ll_lco;
357 cfs_list_t ll_orphan_dentry_list; /*please don't ask -p*/
358 struct ll_close_queue *ll_lcq;
360 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
362 unsigned long ll_async_page_max;
363 unsigned long ll_async_page_count;
365 struct lprocfs_stats *ll_ra_stats;
367 struct ll_ra_info ll_ra_info;
368 unsigned int ll_namelen;
369 struct file_operations *ll_fop;
371 /* =0 - hold lock over whole read/write
372 * >0 - max. chunk to be read/written w/o lock re-acquiring */
373 unsigned long ll_max_rw_chunk;
375 struct lu_site *ll_site;
376 struct cl_device *ll_cl;
378 struct ll_rw_extents_info ll_rw_extents_info;
379 int ll_extent_process_count;
380 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
381 unsigned int ll_offset_process_count;
382 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
383 unsigned int ll_rw_offset_entry_count;
384 enum stats_track_type ll_stats_track_type;
385 int ll_stats_track_id;
388 /* metadata stat-ahead */
389 unsigned int ll_sa_max; /* max statahead RPCs */
390 atomic_t ll_sa_total; /* statahead thread started
392 atomic_t ll_sa_wrong; /* statahead thread stopped for
395 dev_t ll_sdev_orig; /* save s_dev before assign for
397 struct rmtacl_ctl_table ll_rct;
398 struct eacl_table ll_et;
401 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
406 struct task_struct *lrr_reader;
407 cfs_list_t lrr_linkage;
411 * per file-descriptor read-ahead data.
413 struct ll_readahead_state {
414 cfs_spinlock_t ras_lock;
416 * index of the last page that read(2) needed and that wasn't in the
417 * cache. Used by ras_update() to detect seeks.
419 * XXX nikita: if access seeks into cached region, Lustre doesn't see
422 unsigned long ras_last_readpage;
424 * number of pages read after last read-ahead window reset. As window
425 * is reset on each seek, this is effectively a number of consecutive
426 * accesses. Maybe ->ras_accessed_in_window is better name.
428 * XXX nikita: window is also reset (by ras_update()) when Lustre
429 * believes that memory pressure evicts read-ahead pages. In that
430 * case, it probably doesn't make sense to expand window to
431 * PTLRPC_MAX_BRW_PAGES on the third access.
433 unsigned long ras_consecutive_pages;
435 * number of read requests after the last read-ahead window reset
436 * As window is reset on each seek, this is effectively the number
437 * on consecutive read request and is used to trigger read-ahead.
439 unsigned long ras_consecutive_requests;
441 * Parameters of current read-ahead window. Handled by
442 * ras_update(). On the initial access to the file or after a seek,
443 * window is reset to 0. After 3 consecutive accesses, window is
444 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
445 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
447 unsigned long ras_window_start, ras_window_len;
449 * Where next read-ahead should start at. This lies within read-ahead
450 * window. Read-ahead window is read in pieces rather than at once
451 * because: 1. lustre limits total number of pages under read-ahead by
452 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
453 * not covered by DLM lock.
455 unsigned long ras_next_readahead;
457 * Total number of ll_file_read requests issued, reads originating
458 * due to mmap are not counted in this total. This value is used to
459 * trigger full file read-ahead after multiple reads to a small file.
461 unsigned long ras_requests;
463 * Page index with respect to the current request, these value
464 * will not be accurate when dealing with reads issued via mmap.
466 unsigned long ras_request_index;
468 * list of struct ll_ra_read's one per read(2) call current in
469 * progress against this file descriptor. Used by read-ahead code,
470 * protected by ->ras_lock.
472 cfs_list_t ras_read_beads;
474 * The following 3 items are used for detecting the stride I/O
476 * In stride I/O mode,
477 * ...............|-----data-----|****gap*****|--------|******|....
478 * offset |-stride_pages-|-stride_gap-|
479 * ras_stride_offset = offset;
480 * ras_stride_length = stride_pages + stride_gap;
481 * ras_stride_pages = stride_pages;
482 * Note: all these three items are counted by pages.
484 unsigned long ras_stride_length;
485 unsigned long ras_stride_pages;
486 pgoff_t ras_stride_offset;
488 * number of consecutive stride request count, and it is similar as
489 * ras_consecutive_requests, but used for stride I/O mode.
490 * Note: only more than 2 consecutive stride request are detected,
491 * stride read-ahead will be enable
493 unsigned long ras_consecutive_stride_requests;
499 extern cfs_mem_cache_t *ll_file_data_slab;
500 struct lustre_handle;
501 struct ll_file_data {
502 struct ll_readahead_state fd_ras;
504 struct ccc_grouplock fd_grouplock;
505 struct ll_file_dir fd_dir;
507 struct file *fd_file;
510 struct lov_stripe_md;
512 extern cfs_spinlock_t inode_lock;
514 extern struct proc_dir_entry *proc_lustre_fs_root;
516 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
518 return &lli->lli_vfs_inode;
522 struct inode *icbd_parent;
523 struct dentry **icbd_childp;
527 __u32 ll_i2suppgid(struct inode *i);
528 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
530 #define LLAP_MAGIC 98764321
532 extern cfs_mem_cache_t *ll_async_page_slab;
533 extern size_t ll_async_page_slab_size;
535 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
536 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
537 struct ll_ra_read *ll_ra_read_get(struct file *f);
539 /* llite/lproc_llite.c */
541 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
542 struct super_block *sb, char *osc, char *mdc);
543 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
544 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
545 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
547 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
548 struct super_block *sb, char *osc, char *mdc){return 0;}
549 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
550 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
551 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
553 memset(lvars, 0, sizeof(*lvars));
559 static inline void ll_put_page(struct page *page)
562 page_cache_release(page);
565 extern struct file_operations ll_dir_operations;
566 extern struct inode_operations ll_dir_inode_operations;
567 struct page *ll_get_dir_page(struct inode *dir, __u64 hash, int exact,
568 struct ll_dir_chain *chain);
570 int ll_get_mdt_idx(struct inode *inode);
572 int ll_objects_destroy(struct ptlrpc_request *request,
574 struct inode *ll_iget(struct super_block *sb, ino_t hash,
575 struct lustre_md *lic);
576 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
577 void *data, int flag);
578 #ifndef HAVE_VFS_INTENT_PATCHES
579 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
582 void ll_lookup_it_alias(struct dentry **de, struct inode *inode, __u32 bits);
583 int ll_lookup_it_finish(struct ptlrpc_request *request,
584 struct lookup_intent *it, void *data,
585 struct inode **alias);
588 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
589 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
590 int ll_writepage(struct page *page, struct writeback_control *wbc);
591 void ll_removepage(struct page *page);
592 int ll_readpage(struct file *file, struct page *page);
593 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
594 void ll_truncate(struct inode *inode);
595 int ll_file_punch(struct inode *, loff_t, int);
596 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
597 void ll_clear_file_contended(struct inode*);
598 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
599 int ll_readahead(const struct lu_env *env, struct cl_io *io, struct ll_readahead_state *ras,
600 struct address_space *mapping, struct cl_page_list *queue, int flags);
603 extern struct file_operations ll_file_operations;
604 extern struct file_operations ll_file_operations_flock;
605 extern struct file_operations ll_file_operations_noflock;
606 extern struct inode_operations ll_file_inode_operations;
607 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *);
608 extern int ll_have_md_lock(struct inode *inode, __u64 bits);
609 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
610 struct lustre_handle *lockh);
611 int __ll_inode_revalidate_it(struct dentry *, struct lookup_intent *, __u64 bits);
612 int ll_revalidate_nd(struct dentry *dentry, struct nameidata *nd);
613 int ll_file_open(struct inode *inode, struct file *file);
614 int ll_file_release(struct inode *inode, struct file *file);
615 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
616 struct lov_stripe_md *lsm, lstat_t *st);
617 void ll_ioepoch_open(struct ll_inode_info *lli, __u64 ioepoch);
618 int ll_local_open(struct file *file,
619 struct lookup_intent *it, struct ll_file_data *fd,
620 struct obd_client_handle *och);
621 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
622 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
624 int ll_md_real_close(struct inode *inode, int flags);
625 void ll_ioepoch_close(struct inode *inode, struct md_op_data *op_data,
626 struct obd_client_handle **och, unsigned long flags);
627 void ll_done_writing_attr(struct inode *inode, struct md_op_data *op_data);
628 int ll_som_update(struct inode *inode, struct md_op_data *op_data);
629 int ll_inode_getattr(struct inode *inode, struct obdo *obdo,
630 __u64 ioepoch, int sync);
631 int ll_md_setattr(struct inode *inode, struct md_op_data *op_data,
632 struct md_open_data **mod);
633 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
634 struct lustre_handle *fh);
635 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
636 struct ll_file_data *file, loff_t pos,
637 size_t count, int rw);
638 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
639 struct lookup_intent *it, struct kstat *stat);
640 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
641 struct ll_file_data *ll_file_data_get(void);
642 #ifndef HAVE_INODE_PERMISION_2ARGS
643 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
645 int ll_inode_permission(struct inode *inode, int mask);
647 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
648 int flags, struct lov_user_md *lum,
650 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
651 struct lov_mds_md **lmm, int *lmm_size,
652 struct ptlrpc_request **request);
653 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
655 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmm,
656 int *lmm_size, struct ptlrpc_request **request);
657 int ll_fsync(struct file *file, struct dentry *dentry, int data);
658 int ll_do_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
660 int ll_merge_lvb(struct inode *inode);
661 int ll_get_grouplock(struct inode *inode, struct file *file, unsigned long arg);
662 int ll_put_grouplock(struct inode *inode, struct file *file, unsigned long arg);
663 int ll_fid2path(struct obd_export *exp, void *arg);
668 * protect race ll_find_aliases vs ll_revalidate_it vs ll_unhash_aliases
670 extern cfs_spinlock_t ll_lookup_lock;
671 extern struct dentry_operations ll_d_ops;
672 void ll_intent_drop_lock(struct lookup_intent *);
673 void ll_intent_release(struct lookup_intent *);
674 int ll_drop_dentry(struct dentry *dentry);
675 extern void ll_set_dd(struct dentry *de);
676 int ll_drop_dentry(struct dentry *dentry);
677 void ll_unhash_aliases(struct inode *);
678 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
679 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
680 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
681 int ll_revalidate_it_finish(struct ptlrpc_request *request,
682 struct lookup_intent *it, struct dentry *de);
684 /* llite/llite_lib.c */
685 extern struct super_operations lustre_super_operations;
687 char *ll_read_opt(const char *opt, char *data);
688 void ll_lli_init(struct ll_inode_info *lli);
689 int ll_fill_super(struct super_block *sb);
690 void ll_put_super(struct super_block *sb);
691 void ll_kill_super(struct super_block *sb);
692 int ll_shrink_cache(int nr_to_scan, gfp_t gfp_mask);
693 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
694 void ll_clear_inode(struct inode *inode);
695 int ll_setattr_raw(struct inode *inode, struct iattr *attr);
696 int ll_setattr(struct dentry *de, struct iattr *attr);
697 #ifndef HAVE_STATFS_DENTRY_PARAM
698 int ll_statfs(struct super_block *sb, struct kstatfs *sfs);
700 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
702 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
703 __u64 max_age, __u32 flags);
704 void ll_update_inode(struct inode *inode, struct lustre_md *md);
705 void ll_read_inode2(struct inode *inode, void *opaque);
706 void ll_delete_inode(struct inode *inode);
707 int ll_iocontrol(struct inode *inode, struct file *file,
708 unsigned int cmd, unsigned long arg);
709 int ll_flush_ctx(struct inode *inode);
710 #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT
711 void ll_umount_begin(struct vfsmount *vfsmnt, int flags);
713 void ll_umount_begin(struct super_block *sb);
715 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
716 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
717 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
718 struct super_block *);
719 void lustre_dump_dentry(struct dentry *, int recur);
720 void lustre_dump_inode(struct inode *);
721 int ll_obd_statfs(struct inode *inode, void *arg);
722 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
723 int ll_process_config(struct lustre_cfg *lcfg);
724 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
725 struct inode *i1, struct inode *i2,
726 const char *name, int namelen,
727 int mode, __u32 opc, void *data);
728 void ll_finish_md_op_data(struct md_op_data *op_data);
730 /* llite/llite_nfs.c */
731 extern struct export_operations lustre_export_operations;
732 __u32 get_uuid2int(const char *name, int len);
734 /* llite/special.c */
735 extern struct inode_operations ll_special_inode_operations;
736 extern struct file_operations ll_special_chr_inode_fops;
737 extern struct file_operations ll_special_chr_file_fops;
738 extern struct file_operations ll_special_blk_inode_fops;
739 extern struct file_operations ll_special_fifo_inode_fops;
740 extern struct file_operations ll_special_fifo_file_fops;
741 extern struct file_operations ll_special_sock_inode_fops;
743 /* llite/symlink.c */
744 extern struct inode_operations ll_fast_symlink_inode_operations;
746 /* llite/llite_close.c */
747 struct ll_close_queue {
748 cfs_spinlock_t lcq_lock;
750 cfs_waitq_t lcq_waitq;
751 cfs_completion_t lcq_comp;
752 cfs_atomic_t lcq_stop;
755 struct ccc_object *cl_inode2ccc(struct inode *inode);
758 void vvp_write_pending (struct ccc_object *club, struct ccc_page *page);
759 void vvp_write_complete(struct ccc_object *club, struct ccc_page *page);
761 /* specific achitecture can implement only part of this list */
762 enum vvp_io_subtype {
765 /** io called from .sendfile */
767 /** io started from splice_{read|write} */
774 enum vvp_io_subtype cui_io_subtype;
778 read_actor_t cui_actor;
782 struct pipe_inode_info *cui_pipe;
783 unsigned int cui_flags;
785 struct vvp_fault_io {
787 * Inode modification time that is checked across DLM
791 struct vm_area_struct *ft_vma;
793 * locked page returned from vvp_io
795 cfs_page_t *ft_vmpage;
796 #ifndef HAVE_VM_OP_FAULT
797 struct vm_nopage_api {
799 * Virtual address at which fault occurred.
801 unsigned long ft_address;
803 * Fault type, as to be supplied to
809 struct vm_fault_api {
813 struct vm_fault *ft_vmf;
815 * fault API used bitflags for return code.
817 unsigned int ft_flags;
823 * Read-ahead state used by read and page-fault IO contexts.
825 struct ll_ra_read cui_bead;
827 * Set when cui_bead has been initialized.
829 int cui_ra_window_set;
831 * Partially truncated page, that vvp_io_trunc_start() keeps locked
834 struct cl_page *cui_partpage;
838 * IO arguments for various VFS I/O interfaces.
841 /** normal/sendfile/splice */
842 enum vvp_io_subtype via_io_subtype;
846 #ifndef HAVE_FILE_WRITEV
847 struct kiocb *via_iocb;
849 struct iovec *via_iov;
850 unsigned long via_nrsegs;
853 read_actor_t via_actor;
857 struct pipe_inode_info *via_pipe;
858 unsigned int via_flags;
863 struct ll_cl_context {
865 struct cl_io *lcc_io;
866 struct cl_page *lcc_page;
867 struct lu_env *lcc_env;
872 struct vvp_thread_info {
873 struct ost_lvb vti_lvb;
874 struct cl_2queue vti_queue;
875 struct iovec vti_local_iov;
876 struct vvp_io_args vti_args;
877 struct ra_io_arg vti_ria;
878 struct kiocb vti_kiocb;
879 struct ll_cl_context vti_io_ctx;
882 static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
884 extern struct lu_context_key vvp_key;
885 struct vvp_thread_info *info;
887 info = lu_context_key_get(&env->le_ctx, &vvp_key);
888 LASSERT(info != NULL);
892 static inline struct vvp_io_args *vvp_env_args(const struct lu_env *env,
893 enum vvp_io_subtype type)
895 struct vvp_io_args *ret = &vvp_env_info(env)->vti_args;
897 ret->via_io_subtype = type;
903 struct vvp_io vs_ios;
906 static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
908 extern struct lu_context_key vvp_session_key;
909 struct vvp_session *ses;
911 ses = lu_context_key_get(env->le_ses, &vvp_session_key);
912 LASSERT(ses != NULL);
916 static inline struct vvp_io *vvp_env_io(const struct lu_env *env)
918 return &vvp_env_session(env)->vs_ios;
921 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
922 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
923 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
925 /* llite/llite_mmap.c */
926 typedef struct rb_root rb_root_t;
927 typedef struct rb_node rb_node_t;
929 struct ll_lock_tree_node;
930 struct ll_lock_tree {
932 cfs_list_t lt_locked_list;
933 struct ll_file_data *lt_fd;
936 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
937 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
938 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
939 __u64 end, ldlm_mode_t mode);
940 void policy_from_vma(ldlm_policy_data_t *policy,
941 struct vm_area_struct *vma, unsigned long addr, size_t count);
942 struct vm_area_struct *our_vma(unsigned long addr, size_t count);
944 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
946 static inline __u64 ll_ts2u64(struct timespec *time)
948 __u64 t = time->tv_sec;
952 /* don't need an addref as the sb_info should be holding one */
953 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
955 return ll_s2sbi(sb)->ll_dt_exp;
958 /* don't need an addref as the sb_info should be holding one */
959 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
961 return ll_s2sbi(sb)->ll_md_exp;
964 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
966 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
972 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
973 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
975 return ll_s2sbi(inode->i_sb);
978 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
980 return ll_s2dtexp(inode->i_sb);
983 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
985 return ll_s2mdexp(inode->i_sb);
988 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
991 LASSERT(inode != NULL);
992 fid = &ll_i2info(inode)->lli_fid;
993 LASSERT(fid_is_igif(fid) || fid_ver(fid) == 0);
997 static inline int ll_mds_max_easize(struct super_block *sb)
999 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
1002 static inline __u64 ll_file_maxbytes(struct inode *inode)
1004 return ll_i2info(inode)->lli_maxbytes;
1008 int ll_setxattr(struct dentry *dentry, const char *name,
1009 const void *value, size_t size, int flags);
1010 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1011 void *buffer, size_t size);
1012 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1013 int ll_removexattr(struct dentry *dentry, const char *name);
1015 /* llite/remote_perm.c */
1016 extern cfs_mem_cache_t *ll_remote_perm_cachep;
1017 extern cfs_mem_cache_t *ll_rmtperm_hash_cachep;
1019 cfs_hlist_head_t *alloc_rmtperm_hash(void);
1020 void free_rmtperm_hash(cfs_hlist_head_t *hash);
1021 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1022 int lustre_check_remote_perm(struct inode *inode, int mask);
1024 /* llite/llite_capa.c */
1025 extern cfs_timer_t ll_capa_timer;
1027 int ll_capa_thread_start(void);
1028 void ll_capa_thread_stop(void);
1029 void ll_capa_timer_callback(unsigned long unused);
1031 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1032 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1034 void ll_capa_open(struct inode *inode);
1035 void ll_capa_close(struct inode *inode);
1037 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1038 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1040 void ll_truncate_free_capa(struct obd_capa *ocapa);
1041 void ll_clear_inode_capas(struct inode *inode);
1042 void ll_print_capa_stat(struct ll_sb_info *sbi);
1044 /* llite/llite_cl.c */
1045 extern struct lu_device_type vvp_device_type;
1048 * Common IO arguments for various VFS I/O interfaces.
1051 int cl_sb_init(struct super_block *sb);
1052 int cl_sb_fini(struct super_block *sb);
1053 int cl_inode_init(struct inode *inode, struct lustre_md *md);
1054 void cl_inode_fini(struct inode *inode);
1056 enum cl_lock_mode vvp_mode_from_vma(struct vm_area_struct *vma);
1057 void ll_io_init(struct cl_io *io, const struct file *file, int write);
1059 void ras_update(struct ll_sb_info *sbi, struct inode *inode,
1060 struct ll_readahead_state *ras, unsigned long index,
1062 void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1063 int ll_is_file_contended(struct file *file);
1064 void ll_ra_stats_inc(struct address_space *mapping, enum ra_stat which);
1066 /* llite/llite_rmtacl.c */
1067 #ifdef CONFIG_FS_POSIX_ACL
1068 obd_valid rce_ops2valid(int ops);
1069 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1070 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1071 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1072 void rct_init(struct rmtacl_ctl_table *rct);
1073 void rct_fini(struct rmtacl_ctl_table *rct);
1075 void ee_free(struct eacl_entry *ee);
1076 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1077 ext_acl_xattr_header *header);
1078 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1079 struct lu_fid *fid, int type);
1080 void et_search_free(struct eacl_table *et, pid_t key);
1081 void et_init(struct eacl_table *et);
1082 void et_fini(struct eacl_table *et);
1087 #define LL_SA_RPC_MIN 2
1088 #define LL_SA_RPC_DEF 32
1089 #define LL_SA_RPC_MAX 8192
1091 /* per inode struct, for dir only */
1092 struct ll_statahead_info {
1093 struct inode *sai_inode;
1094 unsigned int sai_generation; /* generation for statahead */
1095 cfs_atomic_t sai_refcount; /* when access this struct, hold
1097 unsigned int sai_sent; /* stat requests sent count */
1098 unsigned int sai_replied; /* stat requests which received
1100 unsigned int sai_max; /* max ahead of lookup */
1101 unsigned int sai_index; /* index of statahead entry */
1102 unsigned int sai_index_next; /* index for the next statahead
1103 * entry to be stated */
1104 unsigned int sai_hit; /* hit count */
1105 unsigned int sai_miss; /* miss count:
1106 * for "ls -al" case, it includes
1107 * hidden dentry miss;
1108 * for "ls -l" case, it does not
1109 * include hidden dentry miss.
1110 * "sai_miss_hidden" is used for
1113 unsigned int sai_consecutive_miss; /* consecutive miss */
1114 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1115 * is not a hidden one */
1116 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1117 unsigned int sai_ls_all:1; /* "ls -al", do stat-ahead for
1119 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
1120 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1121 cfs_list_t sai_entries_sent; /* entries sent out */
1122 cfs_list_t sai_entries_received; /* entries returned */
1123 cfs_list_t sai_entries_stated; /* entries stated */
1126 int do_statahead_enter(struct inode *dir, struct dentry **dentry, int lookup);
1127 void ll_statahead_exit(struct inode *dir, struct dentry *dentry, int result);
1128 void ll_stop_statahead(struct inode *inode, void *key);
1131 void ll_statahead_mark(struct inode *dir, struct dentry *dentry)
1133 struct ll_inode_info *lli;
1134 struct ll_dentry_data *ldd = ll_d2d(dentry);
1136 /* dentry has been move to other directory, no need mark */
1137 if (unlikely(dir != dentry->d_parent->d_inode))
1140 lli = ll_i2info(dir);
1141 /* not the same process, don't mark */
1142 if (lli->lli_opendir_pid != cfs_curproc_pid())
1145 cfs_spin_lock(&lli->lli_lock);
1146 if (likely(lli->lli_sai != NULL && ldd != NULL))
1147 ldd->lld_sa_generation = lli->lli_sai->sai_generation;
1148 cfs_spin_unlock(&lli->lli_lock);
1152 int ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int lookup)
1154 struct ll_inode_info *lli;
1155 struct ll_dentry_data *ldd = ll_d2d(*dentryp);
1157 if (unlikely(dir == NULL))
1160 if (ll_i2sbi(dir)->ll_sa_max == 0)
1163 lli = ll_i2info(dir);
1164 /* not the same process, don't statahead */
1165 if (lli->lli_opendir_pid != cfs_curproc_pid())
1169 * When "ls" a dentry, the system trigger more than once "revalidate" or
1170 * "lookup", for "getattr", for "getxattr", and maybe for others.
1171 * Under patchless client mode, the operation intent is not accurate,
1172 * it maybe misguide the statahead thread. For example:
1173 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1174 * have the same operation intent -- "IT_GETATTR".
1175 * In fact, one dentry should has only one chance to interact with the
1176 * statahead thread, otherwise the statahead windows will be confused.
1177 * The solution is as following:
1178 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1179 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1180 * will bypass interacting with statahead thread for checking:
1181 * "lld_sa_generation == lli_sai->sai_generation"
1183 if (ldd && lli->lli_sai &&
1184 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1187 return do_statahead_enter(dir, dentryp, lookup);
1190 static void inline ll_dops_init(struct dentry *de, int block)
1192 struct ll_dentry_data *lld = ll_d2d(de);
1194 if (lld == NULL && block != 0) {
1200 lld->lld_sa_generation = 0;
1202 de->d_op = &ll_d_ops;
1205 /* llite ioctl register support rountine */
1212 #define LLIOC_MAX_CMD 256
1215 * Rules to write a callback function:
1218 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1219 * returned to ll_iocontrol_register. Callback functions should use this
1220 * data to check the potential collasion of ioctl cmd. If collasion is
1221 * found, callback function should return LLIOC_CONT.
1222 * @rcp: The result of ioctl command.
1225 * If @magic matches the pointer returned by ll_iocontrol_data, the
1226 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1228 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1229 struct file *file, unsigned int cmd, unsigned long arg,
1230 void *magic, int *rcp);
1232 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1233 unsigned int cmd, unsigned long arg, int *rcp);
1235 /* export functions */
1236 /* Register ioctl block dynamatically for a regular file.
1238 * @cmd: the array of ioctl command set
1239 * @count: number of commands in the @cmd
1240 * @cb: callback function, it will be called if an ioctl command is found to
1241 * belong to the command list @cmd.
1244 * A magic pointer will be returned if success;
1245 * otherwise, NULL will be returned.
1247 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1248 void ll_iocontrol_unregister(void *magic);
1252 /* lclient compat stuff */
1253 #define cl_inode_info ll_inode_info
1254 #define cl_i2info(info) ll_i2info(info)
1255 #define cl_inode_mode(inode) ((inode)->i_mode)
1256 #define cl_i2sbi ll_i2sbi
1258 static inline void cl_isize_lock(struct inode *inode, int lsmlock)
1260 ll_inode_size_lock(inode, lsmlock);
1263 static inline void cl_isize_unlock(struct inode *inode, int lsmlock)
1265 ll_inode_size_unlock(inode, lsmlock);
1268 static inline void cl_isize_write_nolock(struct inode *inode, loff_t kms)
1270 LASSERT_SEM_LOCKED(&ll_i2info(inode)->lli_size_sem);
1271 i_size_write(inode, kms);
1274 static inline void cl_isize_write(struct inode *inode, loff_t kms)
1276 ll_inode_size_lock(inode, 0);
1277 i_size_write(inode, kms);
1278 ll_inode_size_unlock(inode, 0);
1281 #define cl_isize_read(inode) i_size_read(inode)
1283 static inline int cl_merge_lvb(struct inode *inode)
1285 return ll_merge_lvb(inode);
1288 #define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
1289 #define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
1290 #define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
1292 struct obd_capa *cl_capa_lookup(struct inode *inode, enum cl_req_type crt);
1294 /** direct write pages */
1295 struct ll_dio_pages {
1296 /** page array to be written. we don't support
1297 * partial pages except the last one. */
1298 struct page **ldp_pages;
1299 /* offset of each page */
1300 loff_t *ldp_offsets;
1301 /** if ldp_offsets is NULL, it means a sequential
1302 * pages to be written, then this is the file offset
1303 * of the * first page. */
1304 loff_t ldp_start_offset;
1305 /** how many bytes are to be written. */
1307 /** # of pages in the array. */
1311 extern ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1312 int rw, struct inode *inode,
1313 struct ll_dio_pages *pv);
1315 static inline int ll_file_nolock(const struct file *file)
1317 struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1318 struct inode *inode = file->f_dentry->d_inode;
1320 LASSERT(fd != NULL);
1321 return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1322 (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1324 #endif /* LLITE_INTERNAL_H */