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>
51 #define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
52 #define LUSTRE_FPRIVATE(file) ((file)->private_data)
54 #ifdef HAVE_VFS_INTENT_PATCHES
55 static inline struct lookup_intent *ll_nd2it(struct nameidata *nd)
61 struct ll_dentry_data {
64 struct obd_client_handle lld_cwd_och;
65 struct obd_client_handle lld_mnt_och;
66 #ifndef HAVE_VFS_INTENT_PATCHES
67 struct lookup_intent *lld_it;
69 unsigned int lld_sa_generation;
70 cfs_waitq_t lld_waitq;
73 #define ll_d2d(de) ((struct ll_dentry_data*)((de)->d_fsdata))
75 extern struct file_operations ll_pgcache_seq_fops;
77 #define LLI_INODE_MAGIC 0x111d0de5
78 #define LLI_INODE_DEAD 0xdeadd00d
80 /* remote client permission cache */
81 #define REMOTE_PERM_HASHSIZE 16
83 /* llite setxid/access permission for user on remote client */
84 struct ll_remote_perm {
85 struct hlist_node lrp_list;
90 int lrp_access_perm; /* MAY_READ/WRITE/EXEC, this
91 is access permission with
92 lrp_fsuid/lrp_fsgid. */
96 /* MDS has an authority for the Size-on-MDS attributes. */
97 LLIF_MDS_SIZE_LOCK = (1 << 0),
98 /* Epoch close is postponed. */
99 LLIF_EPOCH_PENDING = (1 << 1),
100 /* DONE WRITING is allowed. */
101 LLIF_DONE_WRITING = (1 << 2),
102 /* Sizeon-on-MDS attributes are changed. An attribute update needs to
104 LLIF_SOM_DIRTY = (1 << 3),
105 /* File is contented */
106 LLIF_CONTENDED = (1 << 4),
107 /* Truncate uses server lock for this file */
108 LLIF_SRVLOCK = (1 << 5)
112 struct ll_inode_info {
114 struct semaphore lli_size_sem; /* protect open and change size */
115 void *lli_size_sem_owner;
116 struct semaphore lli_write_sem;
117 char *lli_symlink_name;
120 unsigned long lli_flags;
121 cfs_time_t lli_contention_time;
123 /* this lock protects posix_acl, pending_write_llaps, mmap_cnt */
125 struct list_head lli_pending_write_llaps;
126 struct list_head lli_close_list;
127 /* handle is to be sent to MDS later on done_writing and setattr.
128 * Open handle data are needed for the recovery to reconstruct
129 * the inode state on the MDS. XXX: recovery is not ready yet. */
130 struct obd_client_handle *lli_pending_och;
131 atomic_t lli_mmap_cnt;
133 /* for writepage() only to communicate to fsync */
136 struct posix_acl *lli_posix_acl;
138 /* remote permission hash */
139 struct hlist_head *lli_remote_perms;
140 unsigned long lli_rmtperm_utime;
141 struct semaphore lli_rmtperm_sem;
143 struct list_head lli_dead_list;
145 struct semaphore lli_och_sem; /* Protects access to och pointers
146 and their usage counters */
147 /* We need all three because every inode may be opened in different
149 struct obd_client_handle *lli_mds_read_och;
150 __u64 lli_open_fd_read_count;
151 struct obd_client_handle *lli_mds_write_och;
152 __u64 lli_open_fd_write_count;
153 struct obd_client_handle *lli_mds_exec_och;
154 __u64 lli_open_fd_exec_count;
156 struct inode lli_vfs_inode;
158 /* identifying fields for both metadata and data stacks. */
159 struct lu_fid lli_fid;
160 struct lov_stripe_md *lli_smd;
163 /* open count currently used by capability only, indicate whether
164 * capability needs renewal */
165 atomic_t lli_open_count;
166 struct obd_capa *lli_mds_capa;
167 struct list_head lli_oss_capas;
169 /* metadata stat-ahead */
171 * "opendir_pid" is the token when lookup/revalid -- I am the owner of
174 pid_t lli_opendir_pid;
176 * since parent-child threads can share the same @file struct,
177 * "opendir_key" is the token when dir close for case of parent exit
178 * before child -- it is me should cleanup the dir readahead. */
179 void *lli_opendir_key;
180 struct ll_statahead_info *lli_sai;
184 * Locking to guarantee consistency of non-atomic updates to long long i_size,
185 * consistency between file size and KMS, and consistency within
186 * ->lli_smd->lsm_oinfo[]'s.
188 * Implemented by ->lli_size_sem and ->lsm_sem, nested in that order.
191 void ll_inode_size_lock(struct inode *inode, int lock_lsm);
192 void ll_inode_size_unlock(struct inode *inode, int unlock_lsm);
194 // FIXME: replace the name of this with LL_I to conform to kernel stuff
195 // static inline struct ll_inode_info *LL_I(struct inode *inode)
196 static inline struct ll_inode_info *ll_i2info(struct inode *inode)
198 return container_of(inode, struct ll_inode_info, lli_vfs_inode);
201 /* default to about 40meg of readahead on a given system. That much tied
202 * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
203 #define SBI_DEFAULT_READAHEAD_MAX (40UL << (20 - CFS_PAGE_SHIFT))
205 /* default to read-ahead full files smaller than 2MB on the second read */
206 #define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - CFS_PAGE_SHIFT))
211 RA_STAT_DISTANT_READPAGE,
212 RA_STAT_MISS_IN_WINDOW,
213 RA_STAT_FAILED_GRAB_PAGE,
214 RA_STAT_FAILED_MATCH,
219 RA_STAT_MAX_IN_FLIGHT,
220 RA_STAT_WRONG_GRAB_PAGE,
224 #define LL_RA_STAT _NR_RA_STAT
225 #define LL_RA_STAT_STRINGS { \
226 [RA_STAT_HIT] = "hits", \
227 [RA_STAT_MISS] = "misses", \
228 [RA_STAT_DISTANT_READPAGE] = "readpage not consecutive", \
229 [RA_STAT_MISS_IN_WINDOW] = "miss inside window", \
230 [RA_STAT_FAILED_GRAB_PAGE] = "failed grab_cache_page", \
231 [RA_STAT_FAILED_MATCH] = "failed lock match", \
232 [RA_STAT_DISCARDED] = "read but discarded", \
233 [RA_STAT_ZERO_LEN] = "zero length file", \
234 [RA_STAT_ZERO_WINDOW] = "zero size window", \
235 [RA_STAT_EOF] = "read-ahead to EOF", \
236 [RA_STAT_MAX_IN_FLIGHT] = "hit max r-a issue", \
237 [RA_STAT_WRONG_GRAB_PAGE] = "wrong page from grab_cache_page",\
241 atomic_t ra_cur_pages;
242 unsigned long ra_max_pages;
243 unsigned long ra_max_read_ahead_whole_pages;
246 /* LL_HIST_MAX=32 causes an overflow */
247 #define LL_HIST_MAX 28
248 #define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
249 #define LL_PROCESS_HIST_MAX 10
250 struct per_process_info {
252 struct obd_histogram pp_r_hist;
253 struct obd_histogram pp_w_hist;
256 /* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
257 struct ll_rw_extents_info {
258 struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
261 #define LL_OFFSET_HIST_MAX 100
262 struct ll_rw_process_info {
265 loff_t rw_range_start;
267 loff_t rw_last_file_pos;
269 size_t rw_smallest_extent;
270 size_t rw_largest_extent;
271 struct file *rw_last_file;
274 enum stats_track_type {
275 STATS_TRACK_ALL = 0, /* track all processes */
276 STATS_TRACK_PID, /* track process with this pid */
277 STATS_TRACK_PPID, /* track processes with this ppid */
278 STATS_TRACK_GID, /* track processes with this gid */
282 /* flags for sbi->ll_flags */
283 #define LL_SBI_NOLCK 0x01 /* DLM locking disabled (directio-only) */
284 #define LL_SBI_CHECKSUM 0x02 /* checksum each page as it's written */
285 #define LL_SBI_FLOCK 0x04
286 #define LL_SBI_USER_XATTR 0x08 /* support user xattr */
287 #define LL_SBI_ACL 0x10 /* support ACL */
288 #define LL_SBI_JOIN 0x20 /* support JOIN */
289 #define LL_SBI_RMT_CLIENT 0x40 /* remote client */
290 #define LL_SBI_MDS_CAPA 0x80 /* support mds capa */
291 #define LL_SBI_OSS_CAPA 0x100 /* support oss capa */
292 #define LL_SBI_LOCALFLOCK 0x200 /* Local flocks support by kernel */
293 #define LL_SBI_LRU_RESIZE 0x400 /* lru resize support */
295 /* default value for ll_sb_info->contention_time */
296 #define SBI_DEFAULT_CONTENTION_SECONDS 60
297 /* default value for lockless_truncate_enable */
298 #define SBI_DEFAULT_LOCKLESS_TRUNCATE_ENABLE 1
299 #define RCE_HASHES 32
301 struct rmtacl_ctl_entry {
302 struct list_head rce_list;
303 pid_t rce_key; /* hash key */
304 int rce_ops; /* acl operation type */
307 struct rmtacl_ctl_table {
309 struct list_head rct_entries[RCE_HASHES];
315 struct list_head ee_list;
316 pid_t ee_key; /* hash key */
317 struct lu_fid ee_fid;
318 int ee_type; /* ACL type for ACCESS or DEFAULT */
319 ext_acl_xattr_header *ee_acl;
324 struct list_head et_entries[EE_HASHES];
327 /* percpu data structure for lustre lru page list */
328 struct ll_pglist_data {
329 spinlock_t llpd_lock; /* lock to protect llpg_list */
330 struct list_head llpd_list; /* all pages (llap_pglist_item) */
331 unsigned long llpd_gen; /* generation # of this list */
332 unsigned long llpd_count; /* How many pages in this list */
333 atomic_t llpd_sample_count;
334 unsigned long llpd_reblnc_count;
335 /* the pages in this list shouldn't be over this number */
336 unsigned long llpd_budget;
338 /* which page the pglist data is in */
339 struct page *llpd_page;
342 unsigned long llpd_hit;
343 unsigned long llpd_miss;
344 unsigned long llpd_cross;
348 struct list_head ll_list;
349 /* this protects pglist(only ll_async_page_max) and ra_info.
350 * It isn't safe to grab from interrupt contexts. */
352 spinlock_t ll_pp_extent_lock; /* Lock for pp_extent entries */
353 spinlock_t ll_process_lock; /* Lock for ll_rw_process_info */
354 struct obd_uuid ll_sb_uuid;
355 struct obd_export *ll_md_exp;
356 struct obd_export *ll_dt_exp;
357 struct proc_dir_entry* ll_proc_root;
358 struct lu_fid ll_root_fid; /* root object fid */
361 struct list_head ll_conn_chain; /* per-conn chain of SBs */
362 struct lustre_client_ocd ll_lco;
364 struct list_head ll_orphan_dentry_list; /*please don't ask -p*/
365 struct ll_close_queue *ll_lcq;
367 struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
369 /* reblnc lock protects llpd_budget */
370 spinlock_t ll_async_page_reblnc_lock;
371 unsigned long ll_async_page_reblnc_count;
372 unsigned long ll_async_page_sample_max;
373 /* I defined this array here rather than in ll_pglist_data
374 * because it is always accessed by only one cpu. -jay */
375 unsigned long *ll_async_page_sample;
376 unsigned long ll_async_page_max;
377 unsigned long ll_async_page_clock_hand;
378 lcounter_t ll_async_page_count;
379 struct ll_pglist_data **ll_pglist;
381 struct lprocfs_stats *ll_ra_stats;
383 unsigned ll_contention_time; /* seconds */
384 unsigned ll_lockless_truncate_enable; /* true/false */
386 struct ll_ra_info ll_ra_info;
387 unsigned int ll_namelen;
388 struct file_operations *ll_fop;
390 #ifdef HAVE_EXPORT___IGET
391 struct list_head ll_deathrow; /* inodes to be destroyed (b1443) */
392 spinlock_t ll_deathrow_lock;
394 /* =0 - hold lock over whole read/write
395 * >0 - max. chunk to be read/written w/o lock re-acquiring */
396 unsigned long ll_max_rw_chunk;
399 struct ll_rw_extents_info ll_rw_extents_info;
400 int ll_extent_process_count;
401 struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
402 unsigned int ll_offset_process_count;
403 struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
404 unsigned int ll_rw_offset_entry_count;
405 enum stats_track_type ll_stats_track_type;
406 int ll_stats_track_id;
409 /* metadata stat-ahead */
410 unsigned int ll_sa_max; /* max statahead RPCs */
411 unsigned int ll_sa_wrong; /* statahead thread stopped for
413 unsigned int ll_sa_total; /* statahead thread started
415 unsigned long long ll_sa_blocked; /* ls count waiting for
417 unsigned long long ll_sa_cached; /* ls count got in cache */
418 unsigned long long ll_sa_hit; /* hit count */
419 unsigned long long ll_sa_miss; /* miss count */
421 dev_t ll_sdev_orig; /* save s_dev before assign for
423 struct rmtacl_ctl_table ll_rct;
424 struct eacl_table ll_et;
427 #define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
429 #define LL_PGLIST_DATA_CPU(sbi, cpu) ((sbi)->ll_pglist[cpu])
430 #define LL_PGLIST_DATA(sbi) LL_PGLIST_DATA_CPU(sbi, smp_processor_id())
432 static inline struct ll_pglist_data *ll_pglist_cpu_lock(
433 struct ll_sb_info *sbi,
436 spin_lock(&sbi->ll_pglist[cpu]->llpd_lock);
437 return LL_PGLIST_DATA_CPU(sbi, cpu);
440 static inline void ll_pglist_cpu_unlock(struct ll_sb_info *sbi, int cpu)
442 spin_unlock(&sbi->ll_pglist[cpu]->llpd_lock);
445 static inline struct ll_pglist_data *ll_pglist_double_lock(
446 struct ll_sb_info *sbi,
447 int cpu, struct ll_pglist_data **pd_cpu)
449 int current_cpu = get_cpu();
451 if (cpu == current_cpu) {
452 ll_pglist_cpu_lock(sbi, cpu);
453 } else if (current_cpu < cpu) {
454 ll_pglist_cpu_lock(sbi, current_cpu);
455 ll_pglist_cpu_lock(sbi, cpu);
457 ll_pglist_cpu_lock(sbi, cpu);
458 ll_pglist_cpu_lock(sbi, current_cpu);
462 *pd_cpu = LL_PGLIST_DATA_CPU(sbi, cpu);
464 return LL_PGLIST_DATA(sbi);
467 static inline void ll_pglist_double_unlock(struct ll_sb_info *sbi, int cpu)
469 int current_cpu = smp_processor_id();
470 if (cpu == current_cpu) {
471 ll_pglist_cpu_unlock(sbi, cpu);
473 ll_pglist_cpu_unlock(sbi, cpu);
474 ll_pglist_cpu_unlock(sbi, current_cpu);
479 static inline struct ll_pglist_data *ll_pglist_lock(struct ll_sb_info *sbi)
481 ll_pglist_cpu_lock(sbi, get_cpu());
482 return LL_PGLIST_DATA(sbi);
485 static inline void ll_pglist_unlock(struct ll_sb_info *sbi)
487 ll_pglist_cpu_unlock(sbi, smp_processor_id());
494 struct task_struct *lrr_reader;
495 struct list_head lrr_linkage;
499 * per file-descriptor read-ahead data.
501 struct ll_readahead_state {
504 * index of the last page that read(2) needed and that wasn't in the
505 * cache. Used by ras_update() to detect seeks.
507 * XXX nikita: if access seeks into cached region, Lustre doesn't see
510 unsigned long ras_last_readpage;
512 * number of pages read after last read-ahead window reset. As window
513 * is reset on each seek, this is effectively a number of consecutive
514 * accesses. Maybe ->ras_accessed_in_window is better name.
516 * XXX nikita: window is also reset (by ras_update()) when Lustre
517 * believes that memory pressure evicts read-ahead pages. In that
518 * case, it probably doesn't make sense to expand window to
519 * PTLRPC_MAX_BRW_PAGES on the third access.
521 unsigned long ras_consecutive_pages;
523 * number of read requests after the last read-ahead window reset
524 * As window is reset on each seek, this is effectively the number
525 * on consecutive read request and is used to trigger read-ahead.
527 unsigned long ras_consecutive_requests;
529 * Parameters of current read-ahead window. Handled by
530 * ras_update(). On the initial access to the file or after a seek,
531 * window is reset to 0. After 3 consecutive accesses, window is
532 * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
533 * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
535 unsigned long ras_window_start, ras_window_len;
537 * Where next read-ahead should start at. This lies within read-ahead
538 * window. Read-ahead window is read in pieces rather than at once
539 * because: 1. lustre limits total number of pages under read-ahead by
540 * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
541 * not covered by DLM lock.
543 unsigned long ras_next_readahead;
545 * Total number of ll_file_read requests issued, reads originating
546 * due to mmap are not counted in this total. This value is used to
547 * trigger full file read-ahead after multiple reads to a small file.
549 unsigned long ras_requests;
551 * Page index with respect to the current request, these value
552 * will not be accurate when dealing with reads issued via mmap.
554 unsigned long ras_request_index;
556 * list of struct ll_ra_read's one per read(2) call current in
557 * progress against this file descriptor. Used by read-ahead code,
558 * protected by ->ras_lock.
560 struct list_head ras_read_beads;
562 * The following 3 items are used for detecting the stride I/O
564 * In stride I/O mode,
565 * ...............|-----data-----|****gap*****|--------|******|....
566 * offset |-stride_pages-|-stride_gap-|
567 * ras_stride_offset = offset;
568 * ras_stride_length = stride_pages + stride_gap;
569 * ras_stride_pages = stride_pages;
570 * Note: all these three items are counted by pages.
572 unsigned long ras_stride_length;
573 unsigned long ras_stride_pages;
574 pgoff_t ras_stride_offset;
576 * number of consecutive stride request count, and it is similar as
577 * ras_consecutive_requests, but used for stride I/O mode.
578 * Note: only more than 2 consecutive stride request are detected,
579 * stride read-ahead will be enable
581 unsigned long ras_consecutive_stride_requests;
587 extern cfs_mem_cache_t *ll_file_data_slab;
588 struct lustre_handle;
589 struct ll_file_data {
590 struct ll_readahead_state fd_ras;
592 struct lustre_handle fd_cwlockh;
593 unsigned long fd_gid;
594 struct ll_file_dir fd_dir;
598 struct lov_stripe_md;
600 extern spinlock_t inode_lock;
602 extern struct proc_dir_entry *proc_lustre_fs_root;
604 static inline struct inode *ll_info2i(struct ll_inode_info *lli)
606 return &lli->lli_vfs_inode;
610 struct inode *icbd_parent;
611 struct dentry **icbd_childp;
615 __u32 ll_i2suppgid(struct inode *i);
616 void ll_i2gids(__u32 *suppgids, struct inode *i1,struct inode *i2);
618 #define LLAP_MAGIC 98764321
620 extern cfs_mem_cache_t *ll_async_page_slab;
621 extern size_t ll_async_page_slab_size;
622 struct ll_async_page {
624 /* only trust these if the page lock is providing exclusion */
625 unsigned int llap_write_queued:1,
626 llap_defer_uptodate:1,
631 llap_lockless_io_page:1,
633 unsigned int llap_pglist_cpu:16;
635 struct page *llap_page;
636 struct list_head llap_pending_write;
637 struct list_head llap_pglist_item;
638 /* checksum for paranoid I/O debugging */
643 * enumeration of llap_from_page() call-sites. Used to export statistics in
644 * /proc/fs/lustre/llite/fsN/dump_page_cache.
647 LLAP_ORIGIN_UNKNOWN = 0,
648 LLAP_ORIGIN_READPAGE,
649 LLAP_ORIGIN_READAHEAD,
650 LLAP_ORIGIN_COMMIT_WRITE,
651 LLAP_ORIGIN_WRITEPAGE,
652 LLAP_ORIGIN_REMOVEPAGE,
653 LLAP_ORIGIN_LOCKLESS_IO,
656 extern char *llap_origins[];
658 #ifdef HAVE_REGISTER_CACHE
659 #include <linux/cache_def.h>
660 #define ll_register_cache(cache) register_cache(cache)
661 #define ll_unregister_cache(cache) unregister_cache(cache)
662 #elif defined(HAVE_SHRINKER_CACHE)
663 struct cache_definition {
666 struct shrinker *shrinker;
669 #define ll_register_cache(cache) do { \
670 struct cache_definition *c = (cache); \
671 c->shrinker = set_shrinker(DEFAULT_SEEKS, c->shrink); \
674 #define ll_unregister_cache(cache) do { \
675 remove_shrinker((cache)->shrinker); \
676 (cache)->shrinker = NULL; \
679 #define ll_register_cache(cache) do {} while (0)
680 #define ll_unregister_cache(cache) do {} while (0)
683 void ll_ra_read_in(struct file *f, struct ll_ra_read *rar);
684 void ll_ra_read_ex(struct file *f, struct ll_ra_read *rar);
685 struct ll_ra_read *ll_ra_read_get(struct file *f);
687 /* llite/lproc_llite.c */
689 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
690 struct super_block *sb, char *osc, char *mdc);
691 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
692 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
693 void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
695 static inline int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
696 struct super_block *sb, char *osc, char *mdc){return 0;}
697 static inline void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi) {}
698 static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
699 static void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars)
701 memset(lvars, 0, sizeof(*lvars));
707 static inline void ll_put_page(struct page *page)
710 page_cache_release(page);
713 extern struct file_operations ll_dir_operations;
714 extern struct inode_operations ll_dir_inode_operations;
715 struct page *ll_get_dir_page(struct inode *dir, __u64 hash, int exact,
716 struct ll_dir_chain *chain);
718 int ll_objects_destroy(struct ptlrpc_request *request,
720 struct inode *ll_iget(struct super_block *sb, ino_t hash,
721 struct lustre_md *lic);
722 int ll_md_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
723 void *data, int flag);
724 #ifndef HAVE_VFS_INTENT_PATCHES
725 struct lookup_intent *ll_convert_intent(struct open_intent *oit,
728 int ll_lookup_it_finish(struct ptlrpc_request *request,
729 struct lookup_intent *it, void *data);
730 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
733 int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
734 int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
735 int ll_writepage(struct page *page);
736 void ll_inode_fill_obdo(struct inode *inode, int cmd, struct obdo *oa);
737 int ll_ap_completion(void *data, int cmd, struct obdo *oa, int rc);
738 int llap_shrink_cache(struct ll_sb_info *sbi, int shrink_fraction);
739 struct ll_async_page *llap_from_page(struct page *page, unsigned origin);
740 extern struct cache_definition ll_cache_definition;
741 void ll_removepage(struct page *page);
742 int ll_readpage(struct file *file, struct page *page);
743 struct ll_async_page *llap_cast_private(struct page *page);
744 void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
745 void ll_ra_accounting(struct ll_async_page *llap,struct address_space *mapping);
746 void ll_truncate(struct inode *inode);
747 int ll_file_punch(struct inode *, loff_t, int);
748 ssize_t ll_file_lockless_io(struct file *, char *, size_t, loff_t *, int);
749 void ll_clear_file_contended(struct inode*);
750 int ll_sync_page_range(struct inode *, struct address_space *, loff_t, size_t);
753 extern struct file_operations ll_file_operations;
754 extern struct file_operations ll_file_operations_flock;
755 extern struct file_operations ll_file_operations_noflock;
756 extern struct inode_operations ll_file_inode_operations;
757 extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *);
758 extern int ll_have_md_lock(struct inode *inode, __u64 bits);
759 extern ldlm_mode_t ll_take_md_lock(struct inode *inode, __u64 bits,
760 struct lustre_handle *lockh);
761 int ll_region_mapped(unsigned long addr, size_t count);
762 int ll_extent_lock(struct ll_file_data *, struct inode *,
763 struct lov_stripe_md *, int mode, ldlm_policy_data_t *,
764 struct lustre_handle *, int ast_flags);
765 int ll_extent_unlock(struct ll_file_data *, struct inode *,
766 struct lov_stripe_md *, int mode, struct lustre_handle *);
767 int ll_file_open(struct inode *inode, struct file *file);
768 int ll_file_release(struct inode *inode, struct file *file);
769 int ll_lsm_getattr(struct obd_export *, struct lov_stripe_md *, struct obdo *);
770 int ll_local_size(struct inode *inode);
771 int ll_glimpse_ioctl(struct ll_sb_info *sbi,
772 struct lov_stripe_md *lsm, lstat_t *st);
773 int ll_glimpse_size(struct inode *inode, int ast_flags);
774 int ll_local_open(struct file *file,
775 struct lookup_intent *it, struct ll_file_data *fd,
776 struct obd_client_handle *och);
777 int ll_release_openhandle(struct dentry *, struct lookup_intent *);
778 int ll_md_close(struct obd_export *md_exp, struct inode *inode,
780 int ll_md_real_close(struct inode *inode, int flags);
781 void ll_epoch_close(struct inode *inode, struct md_op_data *op_data,
782 struct obd_client_handle **och, unsigned long flags);
783 int ll_sizeonmds_update(struct inode *inode, struct md_open_data *data,
784 struct lustre_handle *fh, __u64 ioepoch);
785 int ll_inode_getattr(struct inode *inode, struct obdo *obdo);
786 int ll_md_setattr(struct inode *inode, struct md_op_data *op_data,
787 struct md_open_data **mod);
788 void ll_pack_inode2opdata(struct inode *inode, struct md_op_data *op_data,
789 struct lustre_handle *fh);
790 extern void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid, struct file
791 *file, size_t count, int rw);
792 int ll_getattr_it(struct vfsmount *mnt, struct dentry *de,
793 struct lookup_intent *it, struct kstat *stat);
794 int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
795 struct ll_file_data *ll_file_data_get(void);
796 int ll_inode_permission(struct inode *inode, int mask, struct nameidata *nd);
797 int ll_lov_setstripe_ea_info(struct inode *inode, struct file *file,
798 int flags, struct lov_user_md *lum,
800 int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
801 struct lov_mds_md **lmm, int *lmm_size,
802 struct ptlrpc_request **request);
803 int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
805 int ll_dir_getstripe(struct inode *inode, struct lov_mds_md **lmm,
806 int *lmm_size, struct ptlrpc_request **request);
807 int ll_fsync(struct file *file, struct dentry *dentry, int data);
808 int ll_fiemap(struct inode *inode, struct ll_user_fiemap *fiemap,
810 void ll_pin_extent_cb(void *data);
811 int ll_page_removal_cb(void *data, int discard);
812 int ll_extent_lock_cancel_cb(struct ldlm_lock *lock, struct ldlm_lock_desc *new,
813 void *data, int flag);
816 extern struct dentry_operations ll_init_d_ops;
817 extern struct dentry_operations ll_d_ops;
818 extern struct dentry_operations ll_fini_d_ops;
819 void ll_release(struct dentry *de);
820 void ll_intent_drop_lock(struct lookup_intent *);
821 void ll_intent_release(struct lookup_intent *);
822 int ll_drop_dentry(struct dentry *dentry);
823 extern void ll_set_dd(struct dentry *de);
824 int ll_drop_dentry(struct dentry *dentry);
825 void ll_unhash_aliases(struct inode *);
826 void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
827 void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
828 int ll_dcompare(struct dentry *parent, struct qstr *d_name, struct qstr *name);
829 int ll_revalidate_it_finish(struct ptlrpc_request *request,
830 struct lookup_intent *it, struct dentry *de);
832 /* llite/llite_lib.c */
833 extern struct super_operations lustre_super_operations;
835 char *ll_read_opt(const char *opt, char *data);
836 void ll_lli_init(struct ll_inode_info *lli);
837 int ll_fill_super(struct super_block *sb);
838 void ll_put_super(struct super_block *sb);
839 void ll_kill_super(struct super_block *sb);
840 struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
841 void ll_clear_inode(struct inode *inode);
842 int ll_setattr_raw(struct inode *inode, struct iattr *attr);
843 int ll_setattr(struct dentry *de, struct iattr *attr);
844 #ifndef HAVE_STATFS_DENTRY_PARAM
845 int ll_statfs(struct super_block *sb, struct kstatfs *sfs);
847 int ll_statfs(struct dentry *de, struct kstatfs *sfs);
849 int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
850 __u64 max_age, __u32 flags);
851 void ll_update_inode(struct inode *inode, struct lustre_md *md);
852 void ll_read_inode2(struct inode *inode, void *opaque);
853 void ll_delete_inode(struct inode *inode);
854 int ll_iocontrol(struct inode *inode, struct file *file,
855 unsigned int cmd, unsigned long arg);
856 int ll_flush_ctx(struct inode *inode);
857 #ifdef HAVE_UMOUNTBEGIN_VFSMOUNT
858 void ll_umount_begin(struct vfsmount *vfsmnt, int flags);
860 void ll_umount_begin(struct super_block *sb);
862 int ll_remount_fs(struct super_block *sb, int *flags, char *data);
863 int ll_show_options(struct seq_file *seq, struct vfsmount *vfs);
864 int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
865 struct super_block *);
866 void lustre_dump_dentry(struct dentry *, int recur);
867 void lustre_dump_inode(struct inode *);
868 struct ll_async_page *llite_pglist_next_llap(struct list_head *head,
869 struct list_head *list);
870 int ll_obd_statfs(struct inode *inode, void *arg);
871 int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
872 int ll_process_config(struct lustre_cfg *lcfg);
873 struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
874 struct inode *i1, struct inode *i2,
875 const char *name, int namelen,
876 int mode, __u32 opc, void *data);
877 void ll_finish_md_op_data(struct md_op_data *op_data);
879 /* llite/llite_nfs.c */
880 extern struct export_operations lustre_export_operations;
881 __u32 get_uuid2int(const char *name, int len);
882 struct dentry *ll_fh_to_dentry(struct super_block *sb, __u32 *data, int len,
883 int fhtype, int parent);
884 int ll_dentry_to_fh(struct dentry *, __u32 *datap, int *lenp, int need_parent);
886 /* llite/special.c */
887 extern struct inode_operations ll_special_inode_operations;
888 extern struct file_operations ll_special_chr_inode_fops;
889 extern struct file_operations ll_special_chr_file_fops;
890 extern struct file_operations ll_special_blk_inode_fops;
891 extern struct file_operations ll_special_fifo_inode_fops;
892 extern struct file_operations ll_special_fifo_file_fops;
893 extern struct file_operations ll_special_sock_inode_fops;
895 /* llite/symlink.c */
896 extern struct inode_operations ll_fast_symlink_inode_operations;
898 /* llite/llite_close.c */
899 struct ll_close_queue {
901 struct list_head lcq_head;
902 wait_queue_head_t lcq_waitq;
903 struct completion lcq_comp;
907 void llap_write_pending(struct inode *inode, struct ll_async_page *llap);
908 int llap_write_complete(struct inode *inode, struct ll_async_page *llap);
909 int ll_inode_dirty(struct inode *inode, unsigned long flags);
910 void ll_queue_done_writing(struct inode *inode, unsigned long flags);
911 void ll_close_thread_shutdown(struct ll_close_queue *lcq);
912 int ll_close_thread_start(struct ll_close_queue **lcq_ret);
914 /* llite/llite_mmap.c */
915 typedef struct rb_root rb_root_t;
916 typedef struct rb_node rb_node_t;
918 struct ll_lock_tree_node;
919 struct ll_lock_tree {
921 struct list_head lt_locked_list;
922 struct ll_file_data *lt_fd;
925 int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
926 int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
927 struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
928 __u64 end, ldlm_mode_t mode);
929 int ll_tree_lock(struct ll_lock_tree *tree,
930 struct ll_lock_tree_node *first_node,
931 const char *buf, size_t count, int ast_flags);
932 int ll_tree_unlock(struct ll_lock_tree *tree);
934 #define ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
936 static inline __u64 ll_ts2u64(struct timespec *time)
938 __u64 t = time->tv_sec;
942 /* don't need an addref as the sb_info should be holding one */
943 static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
945 return ll_s2sbi(sb)->ll_dt_exp;
948 /* don't need an addref as the sb_info should be holding one */
949 static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
951 return ll_s2sbi(sb)->ll_md_exp;
954 static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
956 struct obd_device *obd = sbi->ll_md_exp->exp_obd;
962 // FIXME: replace the name of this with LL_SB to conform to kernel stuff
963 static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
965 return ll_s2sbi(inode->i_sb);
968 static inline struct obd_export *ll_i2dtexp(struct inode *inode)
970 return ll_s2dtexp(inode->i_sb);
973 static inline struct obd_export *ll_i2mdexp(struct inode *inode)
975 return ll_s2mdexp(inode->i_sb);
978 static inline struct lu_fid *ll_inode2fid(struct inode *inode)
981 LASSERT(inode != NULL);
982 fid = &ll_i2info(inode)->lli_fid;
983 LASSERT(fid_is_igif(fid) || fid_ver(fid) == 0);
987 static inline int ll_mds_max_easize(struct super_block *sb)
989 return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize;
992 static inline __u64 ll_file_maxbytes(struct inode *inode)
994 return ll_i2info(inode)->lli_maxbytes;
998 int ll_setxattr(struct dentry *dentry, const char *name,
999 const void *value, size_t size, int flags);
1000 ssize_t ll_getxattr(struct dentry *dentry, const char *name,
1001 void *buffer, size_t size);
1002 ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1003 int ll_removexattr(struct dentry *dentry, const char *name);
1005 /* llite/remote_perm.c */
1006 extern cfs_mem_cache_t *ll_remote_perm_cachep;
1007 extern cfs_mem_cache_t *ll_rmtperm_hash_cachep;
1009 struct hlist_head *alloc_rmtperm_hash(void);
1010 void free_rmtperm_hash(struct hlist_head *hash);
1011 int ll_update_remote_perm(struct inode *inode, struct mdt_remote_perm *perm);
1012 int lustre_check_remote_perm(struct inode *inode, int mask);
1014 /* llite/llite_fid.c */
1015 ino_t ll_fid_build_ino(struct ll_sb_info *sbi, struct lu_fid *fid);
1017 /* llite/llite_capa.c */
1018 extern cfs_timer_t ll_capa_timer;
1020 int ll_capa_thread_start(void);
1021 void ll_capa_thread_stop(void);
1022 void ll_capa_timer_callback(unsigned long unused);
1024 struct obd_capa *ll_add_capa(struct inode *inode, struct obd_capa *ocapa);
1025 int ll_update_capa(struct obd_capa *ocapa, struct lustre_capa *capa);
1027 void ll_capa_open(struct inode *inode);
1028 void ll_capa_close(struct inode *inode);
1030 struct obd_capa *ll_mdscapa_get(struct inode *inode);
1031 struct obd_capa *ll_osscapa_get(struct inode *inode, __u64 opc);
1033 void ll_truncate_free_capa(struct obd_capa *ocapa);
1034 void ll_clear_inode_capas(struct inode *inode);
1035 void ll_print_capa_stat(struct ll_sb_info *sbi);
1037 /* llite/llite_rmtacl.c */
1038 #ifdef CONFIG_FS_POSIX_ACL
1039 obd_valid rce_ops2valid(int ops);
1040 struct rmtacl_ctl_entry *rct_search(struct rmtacl_ctl_table *rct, pid_t key);
1041 int rct_add(struct rmtacl_ctl_table *rct, pid_t key, int ops);
1042 int rct_del(struct rmtacl_ctl_table *rct, pid_t key);
1043 void rct_init(struct rmtacl_ctl_table *rct);
1044 void rct_fini(struct rmtacl_ctl_table *rct);
1046 void ee_free(struct eacl_entry *ee);
1047 int ee_add(struct eacl_table *et, pid_t key, struct lu_fid *fid, int type,
1048 ext_acl_xattr_header *header);
1049 struct eacl_entry *et_search_del(struct eacl_table *et, pid_t key,
1050 struct lu_fid *fid, int type);
1051 void et_search_free(struct eacl_table *et, pid_t key);
1052 void et_init(struct eacl_table *et);
1053 void et_fini(struct eacl_table *et);
1058 #define LL_SA_RPC_MIN 2
1059 #define LL_SA_RPC_DEF 32
1060 #define LL_SA_RPC_MAX 8192
1062 /* per inode struct, for dir only */
1063 struct ll_statahead_info {
1064 struct inode *sai_inode;
1065 struct dentry *sai_first; /* first dentry item */
1066 unsigned int sai_generation; /* generation for statahead */
1067 atomic_t sai_refcount; /* when access this struct, hold
1069 unsigned int sai_sent; /* stat requests sent count */
1070 unsigned int sai_replied; /* stat requests which received
1072 unsigned int sai_max; /* max ahead of lookup */
1073 unsigned int sai_index; /* index of statahead entry */
1074 unsigned int sai_index_next; /* index for the next statahead
1075 * entry to be stated */
1076 unsigned int sai_hit; /* hit count */
1077 unsigned int sai_miss; /* miss count:
1078 * for "ls -al" case, it includes
1079 * hidden dentry miss;
1080 * for "ls -l" case, it does not
1081 * include hidden dentry miss.
1082 * "sai_miss_hidden" is used for
1085 unsigned int sai_consecutive_miss; /* consecutive miss */
1086 unsigned int sai_miss_hidden;/* "ls -al", but first dentry
1087 * is not a hidden one */
1088 unsigned int sai_skip_hidden;/* skipped hidden dentry count */
1089 unsigned int sai_ls_all:1; /* "ls -al", do stat-ahead for
1091 cfs_waitq_t sai_waitq; /* stat-ahead wait queue */
1092 struct ptlrpc_thread sai_thread; /* stat-ahead thread */
1093 struct list_head sai_entries_sent; /* entries sent out */
1094 struct list_head sai_entries_received; /* entries returned */
1095 struct list_head sai_entries_stated; /* entries stated */
1098 int do_statahead_enter(struct inode *dir, struct dentry **dentry, int lookup);
1099 int ll_statahead_exit(struct dentry *dentry, int result);
1100 void ll_stop_statahead(struct inode *inode, void *key);
1103 void ll_d_wakeup(struct dentry *dentry)
1105 struct ll_dentry_data *lld = ll_d2d(dentry);
1107 LASSERT(dentry->d_op != &ll_init_d_ops);
1109 cfs_waitq_broadcast(&lld->lld_waitq);
1113 int ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int lookup)
1115 struct ll_sb_info *sbi = ll_i2sbi(dir);
1116 struct ll_inode_info *lli = ll_i2info(dir);
1117 struct ll_dentry_data *ldd = ll_d2d(*dentryp);
1119 if (sbi->ll_sa_max == 0)
1122 /* not the same process, don't statahead */
1123 if (lli->lli_opendir_pid != cfs_curproc_pid())
1127 * When "ls" a dentry, the system trigger more than once "revalidate" or
1128 * "lookup", for "getattr", for "getxattr", and maybe for others.
1129 * Under patchless client mode, the operation intent is not accurate,
1130 * it maybe misguide the statahead thread. For example:
1131 * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
1132 * have the same operation intent -- "IT_GETATTR".
1133 * In fact, one dentry should has only one chance to interact with the
1134 * statahead thread, otherwise the statahead windows will be confused.
1135 * The solution is as following:
1136 * Assign "lld_sa_generation" with "sai_generation" when a dentry
1137 * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
1138 * will bypass interacting with statahead thread for checking:
1139 * "lld_sa_generation == lli_sai->sai_generation"
1141 if (ldd && lli->lli_sai &&
1142 ldd->lld_sa_generation == lli->lli_sai->sai_generation)
1145 return do_statahead_enter(dir, dentryp, lookup);
1148 /* llite ioctl register support rountine */
1155 #define LLIOC_MAX_CMD 256
1158 * Rules to write a callback function:
1161 * @magic: Dynamic ioctl call routine will feed this vaule with the pointer
1162 * returned to ll_iocontrol_register. Callback functions should use this
1163 * data to check the potential collasion of ioctl cmd. If collasion is
1164 * found, callback function should return LLIOC_CONT.
1165 * @rcp: The result of ioctl command.
1168 * If @magic matches the pointer returned by ll_iocontrol_data, the
1169 * callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1171 typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1172 struct file *file, unsigned int cmd, unsigned long arg,
1173 void *magic, int *rcp);
1175 enum llioc_iter ll_iocontrol_call(struct inode *inode, struct file *file,
1176 unsigned int cmd, unsigned long arg, int *rcp);
1178 /* export functions */
1179 /* Register ioctl block dynamatically for a regular file.
1181 * @cmd: the array of ioctl command set
1182 * @count: number of commands in the @cmd
1183 * @cb: callback function, it will be called if an ioctl command is found to
1184 * belong to the command list @cmd.
1187 * A magic pointer will be returned if success;
1188 * otherwise, NULL will be returned.
1190 void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1191 void ll_iocontrol_unregister(void *magic);
1195 #endif /* LLITE_INTERNAL_H */