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 [sun.com URL with a
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 <linux/lustre_acl.h>
42 #ifdef CONFIG_FS_POSIX_ACL
43 # include <linux/fs.h>
45 # include <linux/xattr_acl.h>
46 #endif /* HAVE_XATTR_ACL */
47 #ifdef HAVE_LINUX_POSIX_ACL_XATTR_H
48 # include <linux/posix_acl_xattr.h>
49 #endif /* HAVE_LINUX_POSIX_ACL_XATTR_H */
50 #endif /* CONFIG_FS_POSIX_ACL */
52 #include <lustre_debug.h>
53 #include <lustre_ver.h>
54 #include <lustre_disk.h> /* for s2sbi */
55 #include <lustre_eacl.h>
61 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
62 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
64 #ifdef HAVE_VFS_INTENT_PATCHES
65 static inline struct lookup_intent *ll_nd2it(struct nameidata *nd)
71 struct ll_dentry_data {
74 struct obd_client_handle lld_cwd_och;
75 struct obd_client_handle lld_mnt_och;
76 #ifndef HAVE_VFS_INTENT_PATCHES
77 struct lookup_intent *lld_it;
79 unsigned int lld_sa_generation;
80 cfs_waitq_t lld_waitq;
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 /* llite setxid/access permission for user on remote client */
94 struct ll_remote_perm {
95 struct hlist_node lrp_list;
100 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
101 is access permission with
102 lrp_fsuid/lrp_fsgid. */
106 /* MDS has an authority for the Size-on-MDS attributes. */
107 LLIF_MDS_SIZE_LOCK = (1 << 0),
108 /* Epoch close is postponed. */
109 LLIF_EPOCH_PENDING = (1 << 1),
110 /* DONE WRITING is allowed. */
111 LLIF_DONE_WRITING = (1 << 2),
112 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
114 LLIF_SOM_DIRTY = (1 << 3),
115 /* File is contented */
116 LLIF_CONTENDED = (1 << 4),
117 /* Truncate uses server lock for this file */
118 LLIF_SRVLOCK = (1 << 5)
122 struct ll_inode_info {
124 struct semaphore lli_size_sem; /* protect open and change size */
125 void *lli_size_sem_owner;
126 struct semaphore lli_write_sem;
127 char *lli_symlink_name;
130 unsigned long lli_flags;
131 cfs_time_t lli_contention_time;
133 /* this lock protects posix_acl, pending_write_llaps, mmap_cnt */
135 struct list_head lli_pending_write_llaps;
136 struct list_head lli_close_list;
137 /* handle is to be sent to MDS later on done_writing and setattr.
138 * Open handle data are needed for the recovery to reconstruct
139 * the inode state on the MDS. XXX: recovery is not ready yet. */
140 struct obd_client_handle *lli_pending_och;
141 atomic_t lli_mmap_cnt;
143 /* for writepage() only to communicate to fsync */
146 struct posix_acl *lli_posix_acl;
148 /* remote permission hash */
149 struct hlist_head *lli_remote_perms;
150 unsigned long lli_rmtperm_utime;
151 struct semaphore lli_rmtperm_sem;
153 struct list_head lli_dead_list;
155 struct semaphore lli_och_sem; /* Protects access to och pointers
156 and their usage counters */
157 /* We need all three because every inode may be opened in different
159 struct obd_client_handle *lli_mds_read_och;
160 __u64 lli_open_fd_read_count;
161 struct obd_client_handle *lli_mds_write_och;
162 __u64 lli_open_fd_write_count;
163 struct obd_client_handle *lli_mds_exec_och;
164 __u64 lli_open_fd_exec_count;
166 struct inode lli_vfs_inode;
168 /* identifying fields for both metadata and data stacks. */
169 struct lu_fid lli_fid;
170 struct lov_stripe_md *lli_smd;
173 /* open count currently used by capability only, indicate whether
174 * capability needs renewal */
175 atomic_t lli_open_count;
176 struct obd_capa *lli_mds_capa;
177 struct list_head lli_oss_capas;
179 /* metadata stat-ahead */
181 * "opendir_pid" is the token when lookup/revalid -- I am the owner of
184 pid_t lli_opendir_pid;
186 * since parent-child threads can share the same @file struct,
187 * "opendir_key" is the token when dir close for case of parent exit
188 * before child -- it is me should cleanup the dir readahead. */
189 void *lli_opendir_key;
190 struct ll_statahead_info *lli_sai;
194 * Locking to guarantee consistency of non-atomic updates to long long i_size,
195 * consistency between file size and KMS, and consistency within
196 * ->lli_smd->lsm_oinfo[]'s.
198 * Implemented by ->lli_size_sem and ->lsm_sem, nested in that order.
201 void ll_inode_size_lock(struct inode *inode, int lock_lsm);
202 void ll_inode_size_unlock(struct inode *inode, int unlock_lsm);
204 // FIXME: replace the name of this with LL_I to conform to kernel stuff
205 // static inline struct ll_inode_info *LL_I(struct inode *inode)
206 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
208 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
211 /* default to about 40meg of readahead on a given system. That much tied
212 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
213 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - CFS_PAGE_SHIFT))
215 /* default to read-ahead full files smaller than 2MB on the second read */
216 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - CFS_PAGE_SHIFT))
221 RA_STAT_DISTANT_READPAGE,
222 RA_STAT_MISS_IN_WINDOW,
223 RA_STAT_FAILED_GRAB_PAGE,
224 RA_STAT_FAILED_MATCH,
229 RA_STAT_MAX_IN_FLIGHT,
230 RA_STAT_WRONG_GRAB_PAGE,
235 unsigned long ra_cur_pages;
236 unsigned long ra_max_pages;
237 unsigned long ra_max_read_ahead_whole_pages;
238 unsigned long ra_stats[_NR_RA_STAT];
241 /* LL_HIST_MAX=32 causes an overflow */
242 #define LL_HIST_MAX 28
243 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
244 #define LL_PROCESS_HIST_MAX 10
245 struct per_process_info {
247 struct obd_histogram pp_r_hist;
248 struct obd_histogram pp_w_hist;
251 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
252 struct ll_rw_extents_info {
253 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
256 #define LL_OFFSET_HIST_MAX 100
257 struct ll_rw_process_info {
260 loff_t rw_range_start;
262 loff_t rw_last_file_pos;
264 size_t rw_smallest_extent;
265 size_t rw_largest_extent;
266 struct file *rw_last_file;
269 enum stats_track_type {
270 STATS_TRACK_ALL = 0, /* track all processes */
271 STATS_TRACK_PID, /* track process with this pid */
272 STATS_TRACK_PPID, /* track processes with this ppid */
273 STATS_TRACK_GID, /* track processes with this gid */
277 /* flags for sbi->ll_flags */
278 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
279 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
280 #define LL_SBI_FLOCK 0x04
281 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
282 #define LL_SBI_ACL 0x10 /* support ACL */
283 #define LL_SBI_JOIN 0x20 /* support JOIN */
284 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
285 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
286 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
287 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
288 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
290 /* default value for ll_sb_info->contention_time */
291 #define SBI_DEFAULT_CONTENTION_SECONDS 60
292 /* default value for lockless_truncate_enable */
293 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
294 #define RCE_HASHES 32
296 struct rmtacl_ctl_entry {
297 struct list_head rce_list;
298 pid_t rce_key; /* hash key */
299 int rce_ops; /* acl operation type */
302 struct rmtacl_ctl_table {
304 struct list_head rct_entries[RCE_HASHES];
310 struct list_head ee_list;
311 pid_t ee_key; /* hash key */
312 struct lu_fid ee_fid;
313 int ee_type; /* ACL type for ACCESS or DEFAULT */
314 ext_acl_xattr_header *ee_acl;
319 struct list_head et_entries[EE_HASHES];
323 struct list_head ll_list;
324 /* this protects pglist and ra_info. It isn't safe to
325 * grab from interrupt contexts */
327 spinlock_t ll_pp_extent_lock; /* Lock for pp_extent entries */
328 spinlock_t ll_process_lock; /* Lock for ll_rw_process_info */
329 struct obd_uuid ll_sb_uuid;
330 struct obd_export *ll_md_exp;
331 struct obd_export *ll_dt_exp;
332 struct proc_dir_entry* ll_proc_root;
333 struct lu_fid ll_root_fid; /* root object fid */
336 struct list_head ll_conn_chain; /* per-conn chain of SBs */
337 struct lustre_client_ocd ll_lco;
339 struct list_head ll_orphan_dentry_list; /*please don't ask -p*/
340 struct ll_close_queue *ll_lcq;
342 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
344 unsigned long ll_async_page_max;
345 unsigned long ll_async_page_count;
346 unsigned long ll_pglist_gen;
347 struct list_head ll_pglist; /* all pages (llap_pglist_item) */
349 unsigned ll_contention_time; /* seconds */
350 unsigned ll_lockless_truncate_enable; /* true/false */
352 struct ll_ra_info ll_ra_info;
353 unsigned int ll_namelen;
354 struct file_operations *ll_fop;
356 #ifdef HAVE_EXPORT___IGET
357 struct list_head ll_deathrow; /* inodes to be destroyed (b1443) */
358 spinlock_t ll_deathrow_lock;
360 /* =0 - hold lock over whole read/write
361 * >0 - max. chunk to be read/written w/o lock re-acquiring */
362 unsigned long ll_max_rw_chunk;
365 struct ll_rw_extents_info ll_rw_extents_info;
366 int ll_extent_process_count;
367 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
368 unsigned int ll_offset_process_count;
369 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
370 unsigned int ll_rw_offset_entry_count;
371 enum stats_track_type ll_stats_track_type;
372 int ll_stats_track_id;
375 /* metadata stat-ahead */
376 unsigned int ll_sa_max; /* max statahead RPCs */
377 unsigned int ll_sa_wrong; /* statahead thread stopped for
379 unsigned int ll_sa_total; /* statahead thread started
381 unsigned long long ll_sa_blocked; /* ls count waiting for
383 unsigned long long ll_sa_cached; /* ls count got in cache */
384 unsigned long long ll_sa_hit; /* hit count */
385 unsigned long long ll_sa_miss; /* miss count */
387 dev_t ll_sdev_orig; /* save s_dev before assign for
389 struct rmtacl_ctl_table ll_rct;
390 struct eacl_table ll_et;
393 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
398 struct task_struct *lrr_reader;
399 struct list_head lrr_linkage;
403 * per file-descriptor read-ahead data.
405 struct ll_readahead_state {
408 * index of the last page that read(2) needed and that wasn't in the
409 * cache. Used by ras_update() to detect seeks.
411 * XXX nikita: if access seeks into cached region, Lustre doesn't see
414 unsigned long ras_last_readpage;
416 * number of pages read after last read-ahead window reset. As window
417 * is reset on each seek, this is effectively a number of consecutive
418 * accesses. Maybe ->ras_accessed_in_window is better name.
420 * XXX nikita: window is also reset (by ras_update()) when Lustre
421 * believes that memory pressure evicts read-ahead pages. In that
422 * case, it probably doesn't make sense to expand window to
423 * PTLRPC_MAX_BRW_PAGES on the third access.
425 unsigned long ras_consecutive_pages;
427 * number of read requests after the last read-ahead window reset
428 * As window is reset on each seek, this is effectively the number
429 * on consecutive read request and is used to trigger read-ahead.
431 unsigned long ras_consecutive_requests;
433 * Parameters of current read-ahead window. Handled by
434 * ras_update(). On the initial access to the file or after a seek,
435 * window is reset to 0. After 3 consecutive accesses, window is
436 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
437 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
439 unsigned long ras_window_start, ras_window_len;
441 * Where next read-ahead should start at. This lies within read-ahead
442 * window. Read-ahead window is read in pieces rather than at once
443 * because: 1. lustre limits total number of pages under read-ahead by
444 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
445 * not covered by DLM lock.
447 unsigned long ras_next_readahead;
449 * Total number of ll_file_read requests issued, reads originating
450 * due to mmap are not counted in this total. This value is used to
451 * trigger full file read-ahead after multiple reads to a small file.
453 unsigned long ras_requests;
455 * Page index with respect to the current request, these value
456 * will not be accurate when dealing with reads issued via mmap.
458 unsigned long ras_request_index;
460 * list of struct ll_ra_read's one per read(2) call current in
461 * progress against this file descriptor. Used by read-ahead code,
462 * protected by ->ras_lock.
464 struct list_head ras_read_beads;
466 * The following 3 items are used for detecting the stride I/O
468 * In stride I/O mode,
469 * ...............|-----data-----|****gap*****|--------|******|....
470 * offset |-stride_pages-|-stride_gap-|
471 * ras_stride_offset = offset;
472 * ras_stride_length = stride_pages + stride_gap;
473 * ras_stride_pages = stride_pages;
474 * Note: all these three items are counted by pages.
476 unsigned long ras_stride_length;
477 unsigned long ras_stride_pages;
478 pgoff_t ras_stride_offset;
480 * number of consecutive stride request count, and it is similar as
481 * ras_consecutive_requests, but used for stride I/O mode.
482 * Note: only more than 2 consecutive stride request are detected,
483 * stride read-ahead will be enable
485 unsigned long ras_consecutive_stride_requests;
491 extern cfs_mem_cache_t *ll_file_data_slab;
492 struct lustre_handle;
493 struct ll_file_data {
494 struct ll_readahead_state fd_ras;
496 struct lustre_handle fd_cwlockh;
497 unsigned long fd_gid;
498 struct ll_file_dir fd_dir;
502 struct lov_stripe_md;
504 extern spinlock_t inode_lock;
506 extern struct proc_dir_entry *proc_lustre_fs_root;
508 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
510 return &lli->lli_vfs_inode;
514 struct inode *icbd_parent;
515 struct dentry **icbd_childp;
519 __u32 ll_i2suppgid(struct inode *i);
520 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
522 #define LLAP_MAGIC 98764321
524 extern cfs_mem_cache_t *ll_async_page_slab;
525 extern size_t ll_async_page_slab_size;
526 struct ll_async_page {
528 /* only trust these if the page lock is providing exclusion */
529 unsigned int llap_write_queued:1,
530 llap_defer_uptodate:1,
535 llap_lockless_io_page:1;
537 struct page *llap_page;
538 struct list_head llap_pending_write;
539 struct list_head llap_pglist_item;
540 /* checksum for paranoid I/O debugging */
545 * enumeration of llap_from_page() call-sites. Used to export statistics in
546 * /proc/fs/lustre/llite/fsN/dump_page_cache.
549 LLAP_ORIGIN_UNKNOWN = 0,
550 LLAP_ORIGIN_READPAGE,
551 LLAP_ORIGIN_READAHEAD,
552 LLAP_ORIGIN_COMMIT_WRITE,
553 LLAP_ORIGIN_WRITEPAGE,
554 LLAP_ORIGIN_REMOVEPAGE,
555 LLAP_ORIGIN_LOCKLESS_IO,
558 extern char *llap_origins[];
560 #ifdef HAVE_REGISTER_CACHE
561 #define ll_register_cache(cache) register_cache(cache)
562 #define ll_unregister_cache(cache) unregister_cache(cache)
564 #define ll_register_cache(cache) do {} while (0)
565 #define ll_unregister_cache(cache) do {} while (0)
568 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
569 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
570 struct ll_ra_read *ll_ra_read_get(struct file *f);
572 /* llite/lproc_llite.c */
574 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
575 struct super_block *sb, char *osc, char *mdc);
576 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
577 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
578 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
580 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
581 struct super_block *sb, char *osc, char *mdc){return 0;}
582 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
583 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
584 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
586 memset(lvars, 0, sizeof(*lvars));
592 static inline void ll_put_page(struct page *page)
595 page_cache_release(page);
598 extern struct file_operations ll_dir_operations;
599 extern struct inode_operations ll_dir_inode_operations;
600 struct page *ll_get_dir_page(struct inode *dir, __u64 hash, int exact,
601 struct ll_dir_chain *chain);
603 int ll_objects_destroy(struct ptlrpc_request *request,
605 struct inode *ll_iget(struct super_block *sb, ino_t hash,
606 struct lustre_md *lic);
607 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
608 void *data, int flag);
609 #ifndef HAVE_VFS_INTENT_PATCHES
610 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
613 int ll_lookup_it_finish(struct ptlrpc_request *request,
614 struct lookup_intent *it, void *data);
615 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
618 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
619 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
620 int ll_writepage(struct page *page);
621 void ll_inode_fill_obdo(struct inode *inode, int cmd, struct obdo *oa);
622 int ll_ap_completion(void *data, int cmd, struct obdo *oa, int rc);
623 int llap_shrink_cache(struct ll_sb_info *sbi, int shrink_fraction);
624 struct ll_async_page *llap_from_page(struct page *page, unsigned origin);
625 extern struct cache_definition ll_cache_definition;
626 void ll_removepage(struct page *page);
627 int ll_readpage(struct file *file, struct page *page);
628 struct ll_async_page *llap_cast_private(struct page *page);
629 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
630 void ll_ra_accounting(struct ll_async_page *llap,struct address_space *mapping);
631 void ll_truncate(struct inode *inode);
632 int ll_file_punch(struct inode *, loff_t, int);
633 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
634 void ll_clear_file_contended(struct inode*);
635 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
638 extern struct file_operations ll_file_operations;
639 extern struct file_operations ll_file_operations_flock;
640 extern struct file_operations ll_file_operations_noflock;
641 extern struct inode_operations ll_file_inode_operations;
642 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *);
643 extern int ll_have_md_lock(struct inode *inode, __u64 bits);
644 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
645 struct lustre_handle *lockh);
646 int ll_region_mapped(unsigned long addr, size_t count);
647 int ll_extent_lock(struct ll_file_data *, struct inode *,
648 struct lov_stripe_md *, int mode, ldlm_policy_data_t *,
649 struct lustre_handle *, int ast_flags);
650 int ll_extent_unlock(struct ll_file_data *, struct inode *,
651 struct lov_stripe_md *, int mode, struct lustre_handle *);
652 int ll_file_open(struct inode *inode, struct file *file);
653 int ll_file_release(struct inode *inode, struct file *file);
654 int ll_lsm_getattr(struct obd_export *, struct lov_stripe_md *, struct obdo *);
655 int ll_local_size(struct inode *inode);
656 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
657 struct lov_stripe_md *lsm, lstat_t *st);
658 int ll_glimpse_size(struct inode *inode, int ast_flags);
659 int ll_local_open(struct file *file,
660 struct lookup_intent *it, struct ll_file_data *fd,
661 struct obd_client_handle *och);
662 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
663 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
665 int ll_md_real_close(struct inode *inode, int flags);
666 void ll_epoch_close(struct inode *inode, struct md_op_data *op_data,
667 struct obd_client_handle **och, unsigned long flags);
668 int ll_sizeonmds_update(struct inode *inode, struct md_open_data *data,
669 struct lustre_handle *fh, __u64 ioepoch);
670 int ll_inode_getattr(struct inode *inode, struct obdo *obdo);
671 int ll_md_setattr(struct inode *inode, struct md_op_data *op_data,
672 struct md_open_data **mod);
673 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
674 struct lustre_handle *fh);
675 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid, struct file
676 *file, size_t count, int rw);
677 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
678 struct lookup_intent *it, struct kstat *stat);
679 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
680 struct ll_file_data *ll_file_data_get(void);
681 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
682 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
683 int flags, struct lov_user_md *lum,
685 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
686 struct lov_mds_md **lmm, int *lmm_size,
687 struct ptlrpc_request **request);
688 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
690 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmm,
691 int *lmm_size, struct ptlrpc_request **request);
692 void ll_pin_extent_cb(void *data);
693 int ll_page_removal_cb(void *data, int discard);
694 int ll_extent_lock_cancel_cb(struct ldlm_lock *lock, struct ldlm_lock_desc *new,
695 void *data, int flag);
698 extern struct dentry_operations ll_init_d_ops;
699 extern struct dentry_operations ll_d_ops;
700 extern struct dentry_operations ll_fini_d_ops;
701 void ll_release(struct dentry *de);
702 void ll_intent_drop_lock(struct lookup_intent *);
703 void ll_intent_release(struct lookup_intent *);
704 int ll_drop_dentry(struct dentry *dentry);
705 extern void ll_set_dd(struct dentry *de);
706 int ll_drop_dentry(struct dentry *dentry);
707 void ll_unhash_aliases(struct inode *);
708 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
709 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
710 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
711 int ll_revalidate_it_finish(struct ptlrpc_request *request,
712 struct lookup_intent *it, struct dentry *de);
714 /* llite/llite_lib.c */
715 extern struct super_operations lustre_super_operations;
717 char *ll_read_opt(const char *opt, char *data);
718 void ll_lli_init(struct ll_inode_info *lli);
719 int ll_fill_super(struct super_block *sb);
720 void ll_put_super(struct super_block *sb);
721 void ll_kill_super(struct super_block *sb);
722 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
723 void ll_clear_inode(struct inode *inode);
724 int ll_setattr_raw(struct inode *inode, struct iattr *attr);
725 int ll_setattr(struct dentry *de, struct iattr *attr);
726 #ifndef HAVE_STATFS_DENTRY_PARAM
727 int ll_statfs(struct super_block *sb, struct kstatfs *sfs);
729 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
731 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
732 __u64 max_age, __u32 flags);
733 void ll_update_inode(struct inode *inode, struct lustre_md *md);
734 void ll_read_inode2(struct inode *inode, void *opaque);
735 void ll_delete_inode(struct inode *inode);
736 int ll_iocontrol(struct inode *inode, struct file *file,
737 unsigned int cmd, unsigned long arg);
738 int ll_flush_ctx(struct inode *inode);
739 #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT
740 void ll_umount_begin(struct vfsmount *vfsmnt, int flags);
742 void ll_umount_begin(struct super_block *sb);
744 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
745 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
746 struct super_block *);
747 void lustre_dump_dentry(struct dentry *, int recur);
748 void lustre_dump_inode(struct inode *);
749 struct ll_async_page *llite_pglist_next_llap(struct ll_sb_info *sbi,
750 struct list_head *list);
751 int ll_obd_statfs(struct inode *inode, void *arg);
752 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
753 int ll_process_config(struct lustre_cfg *lcfg);
754 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
755 struct inode *i1, struct inode *i2,
756 const char *name, int namelen,
757 int mode, __u32 opc, void *data);
758 void ll_finish_md_op_data(struct md_op_data *op_data);
760 /* llite/llite_nfs.c */
761 extern struct export_operations lustre_export_operations;
762 __u32 get_uuid2int(const char *name, int len);
763 struct dentry *ll_fh_to_dentry(struct super_block *sb, __u32 *data, int len,
764 int fhtype, int parent);
765 int ll_dentry_to_fh(struct dentry *, __u32 *datap, int *lenp, int need_parent);
767 /* llite/special.c */
768 extern struct inode_operations ll_special_inode_operations;
769 extern struct file_operations ll_special_chr_inode_fops;
770 extern struct file_operations ll_special_chr_file_fops;
771 extern struct file_operations ll_special_blk_inode_fops;
772 extern struct file_operations ll_special_fifo_inode_fops;
773 extern struct file_operations ll_special_fifo_file_fops;
774 extern struct file_operations ll_special_sock_inode_fops;
776 /* llite/symlink.c */
777 extern struct inode_operations ll_fast_symlink_inode_operations;
779 /* llite/llite_close.c */
780 struct ll_close_queue {
782 struct list_head lcq_head;
783 wait_queue_head_t lcq_waitq;
784 struct completion lcq_comp;
788 void llap_write_pending(struct inode *inode, struct ll_async_page *llap);
789 int llap_write_complete(struct inode *inode, struct ll_async_page *llap);
790 int ll_inode_dirty(struct inode *inode, unsigned long flags);
791 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
792 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
793 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
795 /* llite/llite_mmap.c */
796 typedef struct rb_root rb_root_t;
797 typedef struct rb_node rb_node_t;
799 struct ll_lock_tree_node;
800 struct ll_lock_tree {
802 struct list_head lt_locked_list;
803 struct ll_file_data *lt_fd;
806 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
807 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
808 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
809 __u64 end, ldlm_mode_t mode);
810 int ll_tree_lock(struct ll_lock_tree *tree,
811 struct ll_lock_tree_node *first_node,
812 const char *buf, size_t count, int ast_flags);
813 int ll_tree_unlock(struct ll_lock_tree *tree);
815 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
817 static inline __u64 ll_ts2u64(struct timespec *time)
819 __u64 t = time->tv_sec;
823 /* don't need an addref as the sb_info should be holding one */
824 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
826 return ll_s2sbi(sb)->ll_dt_exp;
829 /* don't need an addref as the sb_info should be holding one */
830 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
832 return ll_s2sbi(sb)->ll_md_exp;
835 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
837 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
843 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
844 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
846 return ll_s2sbi(inode->i_sb);
849 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
851 return ll_s2dtexp(inode->i_sb);
854 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
856 return ll_s2mdexp(inode->i_sb);
859 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
862 LASSERT(inode != NULL);
863 fid = &ll_i2info(inode)->lli_fid;
864 LASSERT(fid_is_igif(fid) || fid_ver(fid) == 0);
868 static inline int ll_mds_max_easize(struct super_block *sb)
870 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
873 static inline __u64 ll_file_maxbytes(struct inode *inode)
875 return ll_i2info(inode)->lli_maxbytes;
879 int ll_setxattr(struct dentry *dentry, const char *name,
880 const void *value, size_t size, int flags);
881 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
882 void *buffer, size_t size);
883 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
884 int ll_removexattr(struct dentry *dentry, const char *name);
886 /* llite/remote_perm.c */
887 extern cfs_mem_cache_t *ll_remote_perm_cachep;
888 extern cfs_mem_cache_t *ll_rmtperm_hash_cachep;
890 struct hlist_head *alloc_rmtperm_hash(void);
891 void free_rmtperm_hash(struct hlist_head *hash);
892 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
893 int lustre_check_remote_perm(struct inode *inode, int mask);
895 /* llite/llite_fid.c */
896 ino_t ll_fid_build_ino(struct ll_sb_info *sbi, struct lu_fid *fid);
898 /* llite/llite_capa.c */
899 extern cfs_timer_t ll_capa_timer;
901 int ll_capa_thread_start(void);
902 void ll_capa_thread_stop(void);
903 void ll_capa_timer_callback(unsigned long unused);
905 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
906 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
908 void ll_capa_open(struct inode *inode);
909 void ll_capa_close(struct inode *inode);
911 struct obd_capa *ll_mdscapa_get(struct inode *inode);
912 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
914 void ll_truncate_free_capa(struct obd_capa *ocapa);
915 void ll_clear_inode_capas(struct inode *inode);
916 void ll_print_capa_stat(struct ll_sb_info *sbi);
918 /* llite/llite_rmtacl.c */
919 #ifdef CONFIG_FS_POSIX_ACL
920 obd_valid rce_ops2valid(int ops);
921 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
922 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
923 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
924 void rct_init(struct rmtacl_ctl_table *rct);
925 void rct_fini(struct rmtacl_ctl_table *rct);
927 void ee_free(struct eacl_entry *ee);
928 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
929 ext_acl_xattr_header *header);
930 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
931 struct lu_fid *fid, int type);
932 void et_search_free(struct eacl_table *et, pid_t key);
933 void et_init(struct eacl_table *et);
934 void et_fini(struct eacl_table *et);
939 #define LL_SA_RPC_MIN 2
940 #define LL_SA_RPC_DEF 32
941 #define LL_SA_RPC_MAX 8192
943 /* per inode struct, for dir only */
944 struct ll_statahead_info {
945 struct inode *sai_inode;
946 struct dentry *sai_first; /* first dentry item */
947 unsigned int sai_generation; /* generation for statahead */
948 atomic_t sai_refcount; /* when access this struct, hold
950 unsigned int sai_sent; /* stat requests sent count */
951 unsigned int sai_replied; /* stat requests which received
953 unsigned int sai_max; /* max ahead of lookup */
954 unsigned int sai_index; /* index of statahead entry */
955 unsigned int sai_index_next; /* index for the next statahead
956 * entry to be stated */
957 unsigned int sai_hit; /* hit count */
958 unsigned int sai_miss; /* miss count:
959 * for "ls -al" case, it includes
960 * hidden dentry miss;
961 * for "ls -l" case, it does not
962 * include hidden dentry miss.
963 * "sai_miss_hidden" is used for
966 unsigned int sai_consecutive_miss; /* consecutive miss */
967 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
968 * is not a hidden one */
969 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
970 unsigned int sai_ls_all:1; /* "ls -al", do stat-ahead for
972 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
973 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
974 struct list_head sai_entries_sent; /* entries sent out */
975 struct list_head sai_entries_received; /* entries returned */
976 struct list_head sai_entries_stated; /* entries stated */
979 int do_statahead_enter(struct inode *dir, struct dentry **dentry, int lookup);
980 int ll_statahead_exit(struct dentry *dentry, int result);
981 void ll_stop_statahead(struct inode *inode, void *key);
984 void ll_d_wakeup(struct dentry *dentry)
986 struct ll_dentry_data *lld = ll_d2d(dentry);
988 LASSERT(dentry->d_op != &ll_init_d_ops);
990 cfs_waitq_broadcast(&lld->lld_waitq);
994 int ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int lookup)
996 struct ll_sb_info *sbi = ll_i2sbi(dir);
997 struct ll_inode_info *lli = ll_i2info(dir);
998 struct ll_dentry_data *ldd = ll_d2d(*dentryp);
1000 if (sbi->ll_sa_max == 0)
1003 /* not the same process, don't statahead */
1004 if (lli->lli_opendir_pid != cfs_curproc_pid())
1008 * When "ls" a dentry, the system trigger more than once "revalidate" or
1009 * "lookup", for "getattr", for "getxattr", and maybe for others.
1010 * Under patchless client mode, the operation intent is not accurate,
1011 * it maybe misguide the statahead thread. For example:
1012 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1013 * have the same operation intent -- "IT_GETATTR".
1014 * In fact, one dentry should has only one chance to interact with the
1015 * statahead thread, otherwise the statahead windows will be confused.
1016 * The solution is as following:
1017 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1018 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1019 * will bypass interacting with statahead thread for checking:
1020 * "lld_sa_generation == lli_sai->sai_generation"
1022 if (ldd && lli->lli_sai &&
1023 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1026 return do_statahead_enter(dir, dentryp, lookup);
1029 /* llite ioctl register support rountine */
1036 #define LLIOC_MAX_CMD 256
1039 * Rules to write a callback function:
1042 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1043 * returned to ll_iocontrol_register. Callback functions should use this
1044 * data to check the potential collasion of ioctl cmd. If collasion is
1045 * found, callback function should return LLIOC_CONT.
1046 * @rcp: The result of ioctl command.
1049 * If @magic matches the pointer returned by ll_iocontrol_data, the
1050 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1052 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1053 struct file *file, unsigned int cmd, unsigned long arg,
1054 void *magic, int *rcp);
1056 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1057 unsigned int cmd, unsigned long arg, int *rcp);
1059 /* export functions */
1060 /* Register ioctl block dynamatically for a regular file.
1062 * @cmd: the array of ioctl command set
1063 * @count: number of commands in the @cmd
1064 * @cb: callback function, it will be called if an ioctl command is found to
1065 * belong to the command list @cmd.
1068 * A magic pointer will be returned if success;
1069 * otherwise, NULL will be returned.
1071 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1072 void ll_iocontrol_unregister(void *magic);
1076 #endif /* LLITE_INTERNAL_H */